Research Article – PLOS Currents Disasters http://currents.plos.org/disasters Thu, 08 Nov 2018 15:48:50 +0000 en-US hourly 1 https://wordpress.org/?v=4.5.3 Does the Humanitarian Sector Use Evidence-informed Standards? A Review of the 2011 Sphere Indicators for Wash, Food Security and Nutrition, and Health Action http://currents.plos.org/disasters/article/does-the-humanitarian-sector-use-evidence-informed-standards-a-review-of-the-2011-sphere-indicators-for-wash-food-security-and-nutrition-and-health-action/ http://currents.plos.org/disasters/article/does-the-humanitarian-sector-use-evidence-informed-standards-a-review-of-the-2011-sphere-indicators-for-wash-food-security-and-nutrition-and-health-action/#respond Tue, 30 Oct 2018 15:30:56 +0000 http://currents.plos.org/disasters/?post_type=article&p=38647 Background: In 1997, the pursuit of greater accountability and effectiveness in humanitarian response prompted a multi-stakeholder collaboration to develop a set of indicators and standards to guide humanitarian practitioners, published later in the form of the Sphere Handbook. Twenty years after the first edition of the Handbook was developed, and in order to guide the 2018 revision, an assessment of the evidence base for current Water, Sanitation and Hygiene (WASH), Food Security and Nutrition, and Health Action indicators, as compared to evidence collated by the 2015 LSHTM Humanitarian Health Evidence Review (HHER), was conducted.

Methodology: In order to assess the utility of the Sphere indicators as a tool with which to monitor and evaluate humanitarian activities, indicators from the WASH, Food Security and Nutrition, and Health Action chapters of the Sphere Handbook were analysed and classified according to the SMART criteria. Each indicator was then assessed based on existing evidence related to the effectiveness of humanitarian health interventions as compiled in the HHER.

Results: Of the 159 Sphere indicators intended to guide humanitarian response, only 2 met all of the SMART criteria. The remaining 157 did not provide any time indication for the measurement of the indicator. Furthermore, only 11 standards (23%) and 14 indicators (8%) are supported in part by 33 studies identified in the HHER. Less than one third of studies captured by HHER that explore interventions related to WASH, nutrition, or health could be linked to existing Sphere indicators.

Conclusion: It is not possible to adequately link the 2011 Sphere indicators and standards to their sources in their current constitution, and they are not sufficiently evidence-informed. In the absence of clear measurement definitions, they do not provide necessarily detailed guidance. While recognising that a number of indicators have emerged as a combination of empirical evidence, expert experience, and “common sense”, a focus on fewer indicators, each better defined, is likely to enhance the practical application of the Sphere Handbook in humanitarian settings.

]]>

Introduction

Evidence-Informed Humanitarian Response & The Sphere Standards

The pursuit of effectiveness and ethical practice in humanitarian health response, coupled with demands for improved accountability, have prompted calls to improve the evidence base for humanitarian interventions 1,2. In turn, greater attention has been paid to the existing evidence base related to the effectiveness of humanitarian interventions in crisis-affected contexts, with recent studies suggesting that a limited quality and quantity of evidence exists to support such interventions 3.

Evidence-informed decision-making in humanitarian crises is now recognised as a priority by humanitarian practitioners 4, and by leading organisations and policy-making institutions, with high-level commitments made by a number of the latter at the 2016 World Humanitarian Summit in Istanbul 5. The pursuit of evidence to support effective humanitarian interventions is not a new phenomenon, and is frequently traced back to the 1990s during which time a noticeable shift occurred from a general acceptance of humanitarian activities as inherently “good”, to greater critique of humanitarian programmes perceived to be ineffectual and inefficient 6. The pursuit of greater accountability and effectiveness in humanitarian response during this period prompted a multi-stakeholder collaboration that would lead to the launch of the Sphere Project in 1997.

The Sphere Project is responsible for the development and periodical update of the Sphere Handbook, which comprises both a Humanitarian Charter and Minimum Standards in Humanitarian Response 7. The first edition of the Sphere handbook (first trial edition in 1998; first final edition in 2000), with the development of best technical standards in particular, was seen as valuable and necessary but engendered criticism. The two main criticisms were that: i) although it had sought to promote a human rights-based approach, it failed to link the minimum standards to human rights principles, with the emphasis placed on technical standards at the expense of other humanitarian concerns such as accountability and protection 8,9,10,11; and ii) “one size does not fit all”, with many contexts requiring an adaptive and creative response 8,10,12,13. Later editions (2004 and 2011) attempted to tackle these issues with debatable success 9,14. Despite these critiques, twenty years since the formation of the Sphere Project, the Sphere Handbook is now widely recognised for its common principles and universal minimum standards for humanitarian response 15.

At its inception, the Sphere Project identified a set of minimum standards related to four key lifesaving sectors: water supply, sanitation and hygiene promotion (WASH); food security and nutrition; shelter, settlement and non-food items; and health action. The minimum standards have been described as ‘evidence-based’ and are intended to reflect ‘sector-wide consensus on best practice in humanitarian response, derived from the principle that all crisis-affected people have a right to a dignified life’ 16. The standards are qualitative in formulation, and specify a minimum level of response that should be achieved in the delivery of humanitarian assistance. Each standard is supported by one or more key indicators, along with key actions and guidance notes, which are intended to act as a measure of progress towards attainment of the associated standard.

Prior to the fourth revision of the Sphere Handbook, a survey commissioned by Enhancing Learning and Research for Humanitarian Assistance (ELRHA) and conducted in collaboration with the Sphere Project was conducted to assess knowledge of the Handbook and the extent of its contemporary use, along with views on its structure and content 17. This survey found that the Handbook remains a useful resource for humanitarian practitioners, and that there remains interest in its improvement. To further inform the revision of the technical standards of the Handbook, and to explore the extent to which one of the humanitarian sectors’ most widely disseminated and well-known resources is evidence-informed, this study assessed the evidence base that supports the current Sphere standards and indicators. As such, two objectives were proposed: 1) to explore both the utility of the Sphere indicators as a tool for the assessment of humanitarian activities; and 2) to examine the evidence-base supporting the Sphere standards and indicators for WASH, Food Security and Nutrition, and Health Action, following comparison with evidence generated by the LSHTM Humanitarian Health Evidence Review (HHER).

Methodology

In order to assess the utility of the Sphere indicators as a tool with which to monitor and evaluate humanitarian activities, indicators from the WASH, Food Security and Nutrition, and Health Action (excluding Health Systems) chapters of the Sphere Handbook were analysed and classified according to the SMART criteria (see Table 1). These three chapters were chosen as they are closely comparable with evidence synthesised in the HHER. Each indicator was classified as follows: SMART criteria met (green); SMART criteria met with supporting guidance notes and / or appendices (yellow); not SMART (red). For the purpose of this analysis, the “R” component was interpreted either as relevant, or realistic. This expanded definition was adopted as a number of the Sphere indicators may be unrealistic, or only possible to measure in very specific settings such as in small, well organised refugee camps. While it is clear that humanitarian assistance is intended to reach all crisis-affected people, it is rarely realistic to fully achieve indicators that use language such as “all users”, “all staff”, “all disaster-affected people”, “all targeted beneficiaries”, “at all times”, and “no cases of health hazards”.

Table 1: SMART Criteria 18,19

Following classification of each of the indicators, evidence for humanitarian health interventions as collated in the 2015 HHER (capturing research published between 1980 and 2015) 3,20, was examined and matched to the relevant Sphere standard or associated indicator from the WASH, Food Security and Nutrition, and Health Action chapters. A detailed overview of the HHER, including the methodology, search strategy, key findings, and limitations, has been published elsewhere 3,20. In summary, the HHER comprised a series of systematic literature reviews of the evidence base for health interventions in humanitarian crises in low and middle income countries for the following health topics: communicable disease control; WASH 21; nutrition; sexual and reproductive health (SRH), including gender-based violence (GBV) 22; mental health and psychosocial support; injury and physical rehabilitation 23; non-communicable disease (NCD) 24; health services; and health systems.

Once matched, the following information was provided for each study that provided evidence to support a Sphere standard or indicator: author, year of publication, study design, study category, and a basic summary of key findings. Study categories were defined as category A if the study reported a statistical association between an intervention and health-related outcomes, and category B if the study measured changes in health-related outcomes, but did not report a statistical association.

Results

Classification of the Sphere Indicators

The WASH, Food Security and Nutrition, and Health Action (excluding health systems) chapters of the Sphere Handbook report a total of 48 standards and 159 associated indicators, comprising 13 minimum standards and 58 indicators for WASH, 18 minimum standards and 63 indicators for Food Security and Nutrition, and 17 minimum standards and 38 indicators for Health Action.

Only two indicators (both for Health Action) of the 159 indicators report a loosely time-bound element: that assessment should take place following completion of a measles campaign for the measurement of measles and vitamin A coverage; and once routine Expanded Programme of Vaccination (EPI) services have been re-established for Diphtheria, Pertussis and Tetanus (DPT) coverage. For ease of further analysis, “Time” was dropped from the SMART criteria, such that all of the remaining indicators was assessed against their Specific, Measurable, Attainable, and Relevant (SMAR) qualities.

There is a notable discrepancy between chapters in the way that indicators are constructed. More than half of the WASH (51.7%; 30 of 58) and the food security and nutrition (57.1%; 36 of 63) indicators can be categorised as “SMAR”, or “SMAR with guidance notes and/or appendices” with a larger proportion of “SMAR with guidance notes and/or appendices” for the food security and nutrition compared to the WASH chapter (19.0% versus 8.6%). In contrast, the health action chapter is supported by a much larger proportion of SMAR or SMAR with guidance notes and/or appendices indicators (81.6%; 31 of 38) (See Fig. 1).

Fig. 1: Classification of the 159 Sphere indicators by chapter

Green=SMAR; Yellow=SMAR with guidance notes and/or appendices; Red=Not SMAR

Assessment of Supporting Evidence from the Humanitarian Health Evidence Review

The HHER included 6 studies related to WASH, 77 studies related to Nutrition, and 236 studies related to Health (151 studies on communicable diseases, 8 on non-communicable diseases, 15 on sexual and reproductive health, and 62 studies on mental health). Very few of these studies provide evidence for the standards and indicators listed in the Sphere Handbook.

The WASH chapter of the Sphere Handbook contains 13 minimum standards and 58 indicators. Of the 6 studies included in the HHER, 4 studies support 3 (5.2%) of the indicators included in the Sphere Handbook and 3 (23.1%) of the standards. Four studies support directly 3 of the indicators associated with two of the water supply standards, and one of the hygiene promotion standards (see Table 2). All 4 studies measured statistical associations between an intervention and diarrhoea (i.e. category A). While the included studies provide some evidence, they do not support fully the associated indicator. For example, a study on soap distribution and diarrhoeal disease does not include all “hygiene items”, while the study on bucket provision and diarrhoeal disease does not specify a size or number of buckets.

Table 2: Evidence Base for Water, Sanitation and Hygiene (WASH) Indicators

The Food Security and Nutrition chapter of the Sphere Handbook contains 18 minimum standards and 63 indicators. Of the 77 studies included in the review, 8 provide evidence for 7 (11.1%) of the 63 indicators, and 5 (27.8%) of the standards (see Table 3). While it should be noted that food security was not included in the HHER, 6 of the included studies support 5 food security indicators. All of the studies provide evidence in direct support of the indicators and 7 measure a statistical association between intervention and health outcome (acute malnutrition, micronutrient deficiencies, underweight, stunting), while the remaining paper measured a change in the prevalence of acute malnutrition but did not report a statistical association (category B). As with the evidence in support of the WASH indicators, the evidence only partially supports the selected Food Security and Nutrition indicators.

Table 3: Evidence Base for Food Security & Nutrition Indicators

The Health Action chapter contains 17 minimum standards and 38 indicators (excluding Health Systems). Of the 236 studies included in the review, 21 support to some extent 4 (10.5%) of the 38 indicators (see Table 4). Two indicators related to measles and routine EPI vaccination are supported by some evidence related to the positive impact of the vaccination itself, but with no supporting evidence regarding the coverage that should be achieved. Of the 21 supporting studies, 12 measure statistical associations between interventions and health outcomes (category A), while 9 measure changes in health outcomes, but do not report statistical associations (category B).

Table 4: Evidence Base for Health Action Indicators

Discussion

Only 2 of the 159 indicators of the WASH, Food Security and Nutrition, and Health Action (excluding health systems) chapters of the Sphere Handbook were assessed as being SMART. The 157 remaining indicators did not provide any time indication for the measurement of the indicator. Excluding time bound information, 97 of 159 indicators (61%) were defined as SMAR and thus can be used as complete measurement tools. According to the Sphere Project, indicators “are ‘signals’ that assist in determining whether a standard has been attained. They provide a way of measuring and communicating the processes and results of key actions”. As such, the way in which many of the current Sphere indicators are phrased represents a weakness in terms of their operational applicability as tools for effective monitoring and evaluation. Furthermore, discrepancies were noted between chapters: a larger proportion of SMAR indicators were found in the Health Action chapter compared to WASH and the Food Security and Nutrition chapters (82% versus 52% and 57% respectively). While described as ‘evidence-based’ 16, only 11 standards (23%) and 14 indicators (8%) are supported in part by 33 studies identified in the HHER. Less than one third of studies captured by HHER that explore interventions related to WASH, nutrition, or health were linked to existing Sphere standards and indicators.

These findings are broadly in keeping with other studies that have sought to both qualify and quantify the evidence base that informs humanitarian interventions 3,18. A 2009 report on priority indicators in complex emergencies included an assessment of the Sphere indicators (based on the 2004 edition) and reported that of 346 indicators, 224 (65%) were not quantifiable, 48 (14%) were quantifiable but could not be supported by a search of the published literature, while 55 (13%) were supported by data 18. In 2017, Blanchet et al. 3, reported that the quantity and quality of evidence for various health interventions in humanitarian crises remains inadequate.

The use of indicators is essential in assessing the impact of an intervention. The Sphere Project tend to focus on process indicators (such as drug doses supplied, clinics supported or staff trained) or outcome indicators (such as clinic attendance) rather than impact indicators (morbidity/mortality reduction), which may raise fewer issues, as many such interventions have a vast literature detailing positive impact (for example measles vaccination or Vitamin A supplementation) 25.

The lack of evidence to support certain humanitarian interventions is largely attributable to challenges associated with the conduct of research in insecure settings, but also to the use of inappropriate methodologies and unsuitable study designs by many researchers attempting to investigate the effectiveness of humanitarian interventions 26. Various solutions, including the introduction of innovative methodologies and tools, may assist in the development of an improved evidence base for humanitarian assistance 27.

In the absence of a strong empirical evidence base, the authors of this paper recommend a thorough process of consultations with experts, with a detailed description of the consultation process and rationale for decisions. Accordingly, the Sphere Project suggests that “evidence-based” includes not only scientific evidence but also expert opinion and sectoral consensus. Although we agree that expert opinion and consensus may establish the foundation for some indicators, we have not adopted the same definition for the purpose of this analysis 28. We believe that greater transparency of the basis for, and origin of, the standards and indicators, and their classification based on their empirical, expert consensus, or “common sense” origins would be beneficial. Furthermore, if the classification of indicators were to be presented in a non-numerically ordered manner (i.e. not as 1, 2, 3 etc. as is currently the case), we would minimise the risk of implying a rank hierarchy of different types of evidence.

The authors of this paper have some concerns related to the number of indicators (159 in total for WASH, Food Security and Nutrition, and Health Action) in the current Sphere Handbook. In 2011, the US Centers for Disease Control (CDC) developed the Measurement of Selected Sphere Indicators (MeSSI) project for humanitarian response 19. Drawing from expert consensus they selected 50 priority Sphere indicators and proposed a definition for each, including a numerator and denominator, along with measurement methods. Unfortunately, this project was discontinued. We recommend that future iterations of the Sphere Handbook should attempt to reduce disparities between the different chapters, and prioritise the most important indicators (core SMART indicators), while providing a compendium of indicators that provides clear guidance on how each are to be measured and calculated. While it has been suggested that the presence of non-measurable indicators in the Sphere Handbook ensures that the issues they address are not forgotten, we would suggest that such “indicators” be reclassified as “principles”.

Nevertheless, it is important to note that since its launch in 1997 the Sphere Project has had a tremendous influence on humanitarian practice by defining minimum standards and situating accountability as a core principle in humanitarian practice. As demonstrated by a survey conducted by the authors of this paper and published in a companion report 17, there is no doubt that the Sphere Project is highly valued and needed by the humanitarian community. Over 70% of the 355 humanitarian professionals who participated in the survey strongly agreed or agreed that the Sphere Handbook: is a useful tool for the monitoring of projects; is a convenient source of information; is a good education tool; is likely to improve the quality of interventions; is a key tool for humanitarian beginners; and acts as a concrete translation of the humanitarian principles into practice. This strong consensus on the utility of the Sphere Handbook lends weight to the need for a clearer and more comprehensive evidence-informed approach.

Following an earlier evaluation of the Sphere Project in 2004, Van Dyke and Waldman expressed a need to revise the standards, clarify the difference between standards and indicators, and a need to rely more heavily on evidence 29. The findings of this study further explore the (in-)appropriateness of the current Sphere standards and indicators, and emphasise the need for an in-depth revision of the current Sphere handbook. In 2018, a revised version of the Sphere Handbook is likely to be published. To ensure that the new version of the handbook takes into account the results of this study, several meetings were organised with members of the Sphere Project. It is expected that the revised 2018 version of the Handbook will review existing evidence for the standards and indicators, and that indicator development will be documented following consultation with experts, with the SMART criteria as a guiding framework. Further operational research is required following the publication of the 2018 Sphere handbook to better understand the usability and relevance of the newly selected set of indicators.

Following the publication of the latest Sphere Handbook in 2011, the humanitarian community has acknowledged a series of new challenges. The number, intensity, and scale of humanitarian emergencies has changed in the last ten years 30. A growing appreciation of a long-established shift towards urbanisation and the challenges such a trend poses to humanitarian response, along with the changing needs of crisis-affected populations, particularly in relation to greater recognition of non-communicable diseases 31, have rendered many existing humanitarian standards and approaches redundant. It is therefore important that the new version of the Sphere Handbook recognises emerging epidemiological patterns and social phenomena (e.g. urban displacement, populations in transit, crises in middle-income countries) in order to remain relevant as a tool to guide humanitarian response.

Finally, it is important to note that the HHER only gathered published, quantitative evidence from crisis-affected settings. As such, not only did the HHER not recognise evidence that has emerged from expert consensus or “common sense” approaches, but it also did not capture evidence from stable settings, some of which has been applied in the development of the current Sphere standards and indicators.

Conclusion

The Sphere Handbook was conceived in 1997 by a group of non-governmental organisations, in part to respond to the overall lack of accountability experienced during, and in the aftermath of, the Rwandan genocide in 1994. The Sphere Project developed from a recognition amongst humanitarian actors of the need for greater accountability and effectiveness in humanitarian response. Twenty years later, these concepts that underpin the Sphere Handbook remain as relevant as ever, in a time when humanitarian actors require up-to-date guidance to assist in their response to established and emerging issues, such as the management of non-communicable diseases in urban, middle-income contexts.

This study has demonstrated that humanitarian initiatives such as the Sphere Project must make better use of existing evidence, and document all decisions made by experts, during the development and revision of humanitarian standards and indicators. The Sphere standards and indicators published in 2011 were not sufficiently robust nor adequately evidence-informed, and do not stand as incontestable guidance, in light of the lack of clear measurement definitions that leave room for interpretation. We hope that our recommendations for greater transparency in the basis for, and origin of, the humanitarian standards and indicators, and a more refined selection of core SMART-compliant indicators, will be addressed in the next edition of the Sphere Handbook, due to be published in 2018. These recommendations, and outputs from the many expert consultations held worldwide, have the potential to reaffirm the relevance of the Sphere Handbook as one of the leading sources of technical guidance for humanitarian response.

Corresponding Author

Karl Blanchet: karl.blanchet@lshtm.ac.uk

Competing Interests

The authors declare no competing interests. The study funding source had no role in the analysis of the data, or the publication of results.

Data Availability

All data used in this review is cited in-text.

Appendix

Appendix I: WASH References (Table 2)

Appendix II: Nutrition References (Table 3)

Appendix III: Health Action References (Table 4)

]]>
http://currents.plos.org/disasters/article/does-the-humanitarian-sector-use-evidence-informed-standards-a-review-of-the-2011-sphere-indicators-for-wash-food-security-and-nutrition-and-health-action/feed/ 0
Drought in the Semiarid Region of Brazil: Exposure, Vulnerabilities and Health Impacts from the Perspectives of Local Actors http://currents.plos.org/disasters/article/drought-in-the-semiarid-region-of-brazil-exposure-vulnerabilities-and-health-impacts-from-the-perspectives-of-local-actors/ http://currents.plos.org/disasters/article/drought-in-the-semiarid-region-of-brazil-exposure-vulnerabilities-and-health-impacts-from-the-perspectives-of-local-actors/#respond Mon, 29 Oct 2018 13:30:12 +0000 http://currents.plos.org/disasters/?post_type=article&p=40402 Introduction: The objective of this study was to understand and assess the perception of communities, organized civil society, health professionals, and decision-makers of several governmental institutions, regarding vulnerabilities and health impacts in drought prone municipalities of Brazil.

Methods: This study was carried out through a qualitative investigation in eight municipalities in the Brazilian Semiarid region. Data collection was done through semi-structure and structure interviews, and discussion with local actors, which included communities groups, health professionals, governmental managers and organized civil society.

Results: The results point to the local actors’ concerns and to the fragility of the health sector in the planning of integrated actions directed towards risks and impacts associated with drought conditions on human health.

Discussion: The lack of a specific knowledge contributes to making invisible the process that determines the impacts of drought on health, leading to an acceptance of drought in those municipalities, reducing the capacity of the health system to respond to droughts.

Keywords: drought, vulnerability, risks, health, perception, Brazilian Semiarid, resilience 

]]>

Introduction

Emerging challenges of environmental changes

Global environmental changes, including climate change are bringing new threats and challenges to environmental and human systems. A report of the Intergovernmental Panel on Climate Change (IPCC) indicates an increase in the global average temperature of the planet and changes in precipitation, with projections to get worse in the future 1,2. These changes will cause serious impacts on climatic and environmental conditions, and consequently, influence the economic, social and health conditions of populations, which will have consequences on human development and on the risk of disaster 2,3,4.

In the Semiarid region of Brazil, the greatest vulnerabilities associated with climate change are the difficulty of access to water in sufficient quantity and quality; the susceptibility to extreme drought events; impacts on food production and food security; and the multiple short and long term impacts on human health 2,3,5,6,7. The impacts of climate change on human health are diverse, and may occur through direct or indirect exposure and through changes on economic and social conditions of affected populations 8,9. Direct effects are deaths and injuries caused by extreme weather events, such as droughts and heavy rains, and extreme heat. Indirect effects are mediated by impacts on the environment, for example, infectious diseases transmitted by water, food and vectors, and respiratory diseases related to air quality in the short term. Non-communicable diseases such as undernutrition and mental stress may occur in the long term 8,9,10,11. The effects on social and economic conditions are associated to declining environmental conditions, which can compromise the access to health services, agricultural and livestock production, unemployment, insufficient income, and population migration. These conditions may cause or amplify health outcomes such as cardiovascular diseases, hypertension, mental illness such as depression, anxiety and stress 9.

The magnitude and extent of these impacts will depend on: 1) what is impacted (water infrastructure, agricultural production, ecosystems that affect the occurrence and transmission of diseases, economic development, employment); 2) who is exposed (individuals, communities) and; 3) existing vulnerabilities (political, demographic and health infrastructures, income, unsustainable practices, individuals physical factors) 4,12,13. The tendency of these effects is to affect poorer regions, which present greater social and economic vulnerabilities, difficulty to cope with and manage the impacts, and insufficient financial condition to implement efficient adaptation and mitigation policies 2,4,10,13,14.

All of these factors result in enormous challenges, and this is particularly the case of the Brazilian Semiarid region, which already presents a difficult situation of severe and recurrent droughts, a high proportion of populations living in poverty and other social and economic vulnerabilities.

Drought impacts on systems and populations

Drought is an extreme climatic event, which is characterised by the reduction of water reserves in a geographic area, precipitation below the normal average and a high rate of evapotranspiration influenced by increased air temperature 15,16,17. The types of drought, meteorological, hydrological and agricultural, reflect differences in both climatic characteristics and land use. These differences depend on local problems and needs related to agriculture, livestock and water resource management, which impact differently in the economic and living conditions of communities 17,18,19. Impacts on access to water and food can compromise systems and services, especially in poor regions, where there is low economic development 1,20,21,22,23.

The impacts of an extreme event tend to be amplified in a situation of multiple vulnerabilities. In the case of drought, especially for a prolonged period, local vulnerabilities include low income, living in rural or remote areas, low educational and socioeconomic level, family food production, other unfavorables environmental conditions, and low level of development 24,25,26,27. These conditions when added to meteorological and/or hydrological droughts contribute to the risk of agricultural drought with the possibility of affecting economic conditions and life of the communities 1,28,29,30.

From studies about the social problems that these all types of drought cause on the development of a region, the term socioeconomic drought was developed 17,18. According to Carvalho,18 socioeconomic drought is determined by the intensity of multiple impacts on a society, especially in its economic development, health and well being.

A clear example of this situation is pointed out by Castro 31 who explains that in past droughts in the Northeast of Brazil (referring to droughts in 1877 and 1932), the climatic condition together with an inadequate socioeconomic and political infrastructure left populations in worse conditions than in normal times. Droughts marked this region with death (of people and animals), hunger and devastation of the land. Faced with this situation people were forced to migrate to other places in search of a better living condition for their families, but this migration process brought other risks, both for the families left behind and for the migrants, which depended on environmental and social conditions of the new place.

Drought impacts continue today to be multiple and intense, having large repercussion on populations, which present higher socioeconomic vulnerabilities to cope with the adverse effects of drought. This situation results in and influences social inequalities and injustices, which in turn results in differences in livelihoods, human rights, health and human well being 28,29,32,33,34.

Drought and vulnerabilities in the semiarid region of Brazil

In Brazil, drought is considered the most frequent climatic event, reaching in greater proportion the area of the Brazilian Semiarid, with intense, prolonged and recurrent droughts 35. The Atlas of Natural Disasters in Brazil reported 19,517 registered events of drought for the period between 1991 and 2012, corresponding to 48% of the total of 39,837 records of all types of natural events in the country. The Northeast region presented most of the events of drought, as well as the highest number of deaths and affected people 36.

Considering the most recent delimitation of the Semiarid region in Brazil, carried out in 2017 by the Ministry of National Integration, this area increased from 1,135 municipalities to 1,262, occupying a territorial area of approximately 1,530 million km2 (18% of the Brazilian territory). The current area covers parts of all nine states of the Northeast region (covering a large part of the states of Rio Grande do Norte, Ceará, Piauí, Pernambuco, Paraíba, Alagoas, Bahia e Sergipe, and a small part of Maranhão), and the northern part of the Southeast region in the state of Minas Gerais. This all area has a population of approximately 26.5 million inhabitants that corresponds to 12% of the population of Brazil 37,38,39. The Brazilian Semiarid is one of the most populated semiarid regions in the world 24.

The projections of temperature increase and precipitation decrease for the next decades will result in days and nights of extreme heat. This situation increases the possibility of extreme droughts and greater impacts to drought prone regions 1,8,40,41. Studies of the evaluation of impacts of climate change on Brazilian regions and its biomes point out to this region as susceptible to having more frequent and intense droughts in future 2,41,42,43,44,45,46. According to Nobre 43 and studies carried out by the Brazilian Panel on Climate Change (PBMC in Portuguese), it is estimated for the Brazilian Semiarid region a reduction of up to 70% in groundwater recharge by 2050, and a reduction of up to 20% in the flow of reservoirs and rivers, which will result in multiple impacts, mainly on agricultural irrigation process 43,45. The availability of water in reservoirs usually corresponds to only 40% of their storage capacity 18. Reduction in water availability coupled with the existing process of aridity and drought, climate change impacts, and unsustainable human practices (forest fires, deforestation, grazing, monoculture, irrigation and groundwater exploitation) tend to intensify the desertification 1,2,47 and salinization processes,48 which have already begun in the Semiarid region of Brazil 25,41,44,46,48,49,50.

The combination of these factors works as a pressure on water supply and demand, a condition that may result in greater environmental damages. These processes, if maintained, can increase economic and social damage, and generate risks that can intensify the health-illness process, which further extends the economic impacts and social exclusion of the population 18,29.

Methods

This study was carried out in eight municipalities of the Brazilian Semiarid region. Its objectives were 1) to identify and evaluate the knowledge and perception of communities, organized civil society, health professionals and managers of several governmental institutions about the vulnerabilities and impacts related to drought and human health; and 2) to learn how the health sector manages drought situations to reduce risks.

The methodology used in this study was based on the analysis of primary qualitative data, through interviews conducted in two municipalities in the state of Rio Grande do Norte (RN) and six in the state of Ceará (CE). We sought to include municipalities with different socio-demographic characteristics that were representative of the municipalities of these two semiarid states (Table 1).

Firstly, a pilot field study was conducted in the municipalities of Rio Grande do Norte state (Acari and Currais Novos) with semi-structured interviews. After evaluating this method, the interviews in the municipalities of Ceará state (Canindé, Itatira, Parambu, Quixadá, Quixeramobim e Tauá) were adapted to a structured type with data collected by a scale of agreement (Likert scale method).

Each participant was informed about the aims of the research and the interview methodology and signed a consent form authorizing recorded interviews. Participants were identified by a code for use on tabulating data, in order to protect the participant’s identity. This research was approved by the Research Ethic Committee of the Oswaldo Cruz Foundation.

Characterization of the field study and interview groups

The characteristics of the interviewed municipalities are shown in Table 1.

Table 1: Indicators of eight municipalities in the Brazilian Semiarid region that participated in the research. Source: PNUD,51 based on IBGE data. Legend: Population – population of municipality; Illiteracy – Proportion of illiteracy population (%); Poverty – proportion of the population living in poverty (%); Water access – proportion of population with access to piped water (%); Under 5 Mortality – Under 5 mortality rate per thousands live births; Life expectancy – life expectancy at birth; MHDI – Municipal Human Development Index; IRIS – Drought Disaster Risk Index 52.

State Municipality Population Illiteracy Poverty Water access Under 5 Mortality Life expectancy MHDI IRIS
RN Acari 10958 18.9 20.8 87.5 23.4 71.7 0.679 33.7
RN Currais Novos 42240 19.1 22.6 88.2 23.3 72.6 0.691 35.4
CE Canindé 74224 27.0 45.2 73.7 25.7 70.9 0.612 44.5
CE Itatira 18865 35.7 53.8 53.9 29.1 69.8 0.562 56.4
CE Parambu 31257 38.0 51.6 63.7 25.8 70.9 0.570 52.3
CE Quixadá 80117 24.9 36.2 72.1 23.8 71.5 0.659 44.0
CE Quixeramobim 71409 26.4 38.4 78.7 21.6 72.3 0.642 44.8
CE Tauá 55530 29.4 41.0 78.2 24.2 71.4 0.633 47.5

The indicators that represent the social (illiteracy), economic (poverty), environmental (water) and health (infant mortality) dimensions were selected because they are important measures of the development of a municipality or region. These indicators are also a measure of inequality. Life expectancy and MHDI are useful in analyzing human development and the improvement of life conditions in the municipalities.

MHDI considers three relevant dimensions: health (opportunity to have a long life), education (opportunity to access knowledge) and income (opportunity to have a decent life) 51. Regarding IRIS, an index developed to measure the municipal disaster risk of drought,52 it corresponds to a set of indicators of vulnerability (measured by level of education and poverty), threat or hazard (measured by the number of damage assessments for drought and incidence of drought) and exposure (obtained by the percentage of population living without access to piped water). The results of the selected municipalities showed a high risk of disaster from droughts.

In total, 53 interviews were conducted between Rio Grande do Norte and Ceará, with 103 participants. In Rio Grande do Norte 18 semi-structured interviews were conducted in November 2015, in the municipalities of Acari and Currais Novos. The three groups of actors who participated in this phase corresponded to a community, health professionals and interviews in mixed groups (Table 2), with the participation of 38 women and 30 men.

Interviews were addressed to a community living close to a dam (Açude Gargalheiras, in Acari). The community participants had a diversity of professions and activities, such as fishermen, farmers, retired people, housewives, community leaders, teachers and students. Regarding health professionals these included community health agents, nurses and health managers. Three group interviews were carried out: one in a Quilombola village (a type of settlement consisting of people of African descent) with the presence of the community leader and the population; the second one in a larger group with participation of community leaders, population, and education and health professionals; and the third with an organized social group, the Union of Rural Workers and Family Farmers.

Table 2: Number of interviews per group interviewed in two municipalities of Rio Grande do Norte: Acari and Currais Novos.

Group interviewed Number of interviews Number of participants
Community 9 9
Health professionals 6 6
Interviews in group 3 53
Total 18 68

In Ceará, 35 structured interviews were conducted in April 2016, in the municipalities of Canindé, Itatira, Parambu, Quixadá, Quixeramobim and Tauá, with a total of 24 women and 11 men. The three groups of respondents corresponded to organized civil society, health professionals and government managers (Table 3).

Health professionals interviewed included community health agents, nurses, doctors and health managers. The health managers were secretaries of health, coordinators of Primary Care and coordinators of Epidemiology, Environmental and Sanitary Surveillance. Interviews also included managers of Planning and Budget, Agriculture and Livestock, and Water Supply and Sewerage System. Regarding the institutions of Organized Civil Society, they included representatives of the Union of Rural Workers and Family Farmers and Cáritas Diocesana de Crateús (a religious organization). These institutions integrate their activities with the support of a regional organization (Articulação do Semiárido) that puts in practice projects of coexistence with the Semiarid, through the support of public policies.

Table 3: Number of interviews per group interviewed in six municipalities of CE state, Canindé, Itatira, Parambu, Quixadá, Quixeramobim and Tauá.

Group interviewed Number of interviews Number of participants
Organized Civil Society 7 7
Health professionals 14 14
Government managers 14 14
Total 35 35

Data collection

The interviews, both semi-structured and structured, addressed issues identified in the scientific literature. The questions for the structured interviews were improved based on the results of the semi-structured interviews conducted in Rio Grande do Norte. The questions were classified by relevant topics to facilitate the discussion. These topics were related to 1) social, environmental and economic vulnerabilities; 2) knowledge about drought and climate change; 3) effects of drought on human health; 4) factors or mechanisms resulting from the drought process that generate the impacts on health; 5) health sector interventions of preparedness, prevention and responses to reduce risks associated to drought; 6) evaluation and integration of actions, involving community participation; and 7) the resilience of population and government, including social programs.

Data Analysis

After collecting and organizing the data, the interviews (both, semi-structured and structured) were analysed by the method of analysis of content 53.

Data were organized by pre-defined thematic categories and sub-categories in order to facilitate the interpretation. Associations between elements was made, selecting the similarities, contrasts and differences; the understanding of the most relevant contents of the groups, in order to compare with the literature; and finally, an interpretative synthesis that answered both, the questions of the study and the formulation of new questions.

After organizing by topics, the data were analyzed in an articulation with the theoretical basis with the purpose of discussing the logic of meanings and the perception of the interviewees regarding the interface between drought, local vulnerabilities and impacts on health and on living conditions of the population who live in a semiarid climate.

For the analysis of Ceará data, the interview script consisted of a series of statements, following the Likert scale method, which contained six categories of response (totally agree, partially agree, partially disagree, totally disagree, do not know and does not apply). We first tabulated the responses within the six categories. Most of interviewees after responding to the statements within Likert scale, also expressed themselves on the subject. These statements were noted and considered in the analysis of the results. We transcribed the recordings and organized the data by thematic categorization, in order to facilitate the interpretation for content analysis, (as was done in Rio Grande do Norte).

Perceptions of the interviewees of both states, Rio Grande do Norte and Ceará (communities, organized civil society and government) were organized into four main thematic categories. Then, each of these categories and their corresponding variables were analyzed, including the perceptions of interviewees in both states. Key statements of some interviewees were highlighted.

Results and Discussion

The results presented from the analysis of the eight municipalities of the states of Rio Grande do Norte and Ceará reflect the participants’ perceptions about four key areas: 1) socioeconomic and environmental vulnerabilities; 2) risk factors for diseases; 3) impacts of drought on human health; and 4) drought risk management (prevention, preparedness, response and recovery, including adaptation measures). Perceptions in this 4th key area refer to the 2nd, 5th, 6th and 7th topics related to the questions’ script described in the data collection section. These themes address strategies for preparedness and response for decision-making within the health system. It is important to clarify the difference between vulnerability and risk for the interpretation of responses. Vulnerability was understood as one of the conditions that make systems, populations and individuals more likely to be affected by a risk factor; and risks were understood as factors that can cause diseases 1,54,55,56. A similar scenario is observed between the two states. Table 4 presents the percentage of responses from Rio Grande do Norte.

Table 4. Percentages of mentions in responses on the perception of the interviewees (communities, organized civil society and government) in relation to exposure to drought in the Rio Grande do Norte.

Table 4: Percentages of mentions in responses on the perception of the interviewees (communities, organized civil society and government) in relation to exposure to drought in the Rio Grande do Norte.

Table 4. Percentages of mentions in responses on the perception of the interviewees (communities, organized civil society and government) in relation to exposure to drought in the Rio Grande do Norte.

As for Ceará, Table 5 presents the percentage of agreement (total and partial) referring to the Likert scale, for each one of the variables selected in the questionnaire.

Table 5. Percentage of agreement (totally and partially) of the interviewees in Ceará regarding perceptions of exposure to drought.

Table 5: Percentage of agreement (totally and partially) of the interviewees in Ceará regarding perceptions of exposure to drought.

Social, environmental and economic vulnerabilities present in the municipalities

In general, the vulnerabilities mentioned by participants reflect the unfavorable situation of populations living in the Brazilian Semiarid. It has been reported that despite the improvements in this region due to governmental social programs, there is still poverty especially in rural areas. The living and health conditions of the populations are changed by the magnitude of the drought impacts, mainly due to their low income and lack of access to employment, which increases social inequalities.

Many interviewees mentioned the low economic development in these municipalities and also low agricultural and livestock production for business purposes. These social and economic vulnerabilities coupled with the most mentioned environmental vulnerabilities, such as lack of sanitation, lack or scarcity of water and access to food in quantity and quality, as well as precarious housing, especially in rural areas, can increase the existing social inequalities in the region, as pointed out by several studies 57,58,59,60,61. Some statements express this problem, such as those mentioned by the managers regarding social inequalities and poverty, and by health professionals regarding the precarious housing and sanitation conditions. We highlight some statements: ‘Inequalities exist, service structures and poverty can already be mapped’. ’In the periods of drought people get poorer by affecting their economic conditions and their mental health’. ‘Housing conditions are inadequate, many homes do not have a safely managed sanitation service, nor sewer, and many people live in the same house together’.

Regarding the lack of access to water, most interviewees reported that even with some initiatives by the government to establish measures for access to water, the access is still limited. This situation results in the need for rationing water. Examples of such initiatives include: the construction of dams, tanks, cisterns, fountains, deep wells, and distribution of water by water trucks, during long periods of droughts. It causes water supply shortages in some communities (both through the water tanks and water distributed by truck), increasing their financial expenses due to the need to buy water, which is more expensive in drought periods. It also increases the risk of diseases. The following statements mentioned by health professionals express this situation well: ‘Piped water only comes every 15 days, forcing people to buy water or store it’. ‘When there is no water supply, the population looks for alternatives, which can bring complications to health’. Regarding the lack of financial conditions to buy water, the following divergent statement is highlighted: ‘Most people have condition to buy water and those people who do not, receive water distributed by the government’.

It was reported that even with water scarcity, there is still waste, especially in urban areas, and that sometimes it occurs by inadequate water use by family farming, through cisterns and agricultural productive yards. The interviewees also stated that the practice of flooding is used to irrigate the agricultural production, which requires a large amount of water, arguing that there is no technical training or improvement of sustainable practices that can improve production in drought prone areas.

When water was associated with lack of quality for human consumption, it was considered that there is no adequate treatment of water by the government, and that there is inappropriate storage, mainly by the population. On the other hand, there were divergences pointed out by the interviewees. Some participants mentioned that the water in the wells has poor quality, because it is saline, and others said that it was of good quality. Regarding water distributed by trucks, some interviewees stated that this water is treated and has good quality, but others said that it does not have quality. The following statements by some managers and health professionals express this question: ‘The water in the dam has poor quality and the water in the cistern is good’. ‘Most of the reservoirs are contaminated’. ‘There is water supply, but with inappropriate quality’. ‘The water distributed by the government is treated, but the water that we buy, we do not know if it is supervised’.

In relation to inadequately stored water by the population, a manager explained that: ‘The populations store water and do not let the professionals, who are responsible for control of zoonoses, do the right work in the houses, which can result in people getting sick with arboviruses. There were also reports that in some communities people use the water from reservoirs to bath and do home hygiene, even though the water is dirty and contaminated; as one farmer stated: ‘Any water, dirty or not, is useful for something, and it has to be used’. It is observed through these statements that the lack of access to water, especially fresh water, may result in inadequate practices of using water without quality. These situations reflect that people exposed to drought conditions take advantage of any opportunity of water availability; even knowing that water is not suitable for human consumption.

In relation to limited access to food in quality and quantity it was pointed out, mainly by health professionals that this condition is more related to social inequalities and limited knowledge about food quality. The following statements express this: ‘Access to quality food depends on each family priority. It is necessary to educate and make people aware about food quality’. On the other hand, most organized civil society has reported that the government’s social programs have significantly improved access to food, although some families, who really need it, do not receive these benefits and sometimes they go hungry. In general, most health professionals said that there were more obese children than undernourished.

Regarding the lack of financial condition to buy food, the responses showed a similar perception. They stated that in spite of the increase in food prices and the lack of employment, people do not starve as they used to during long drought periods before. This is due to a social program to help families in need (called Bolsa Família, a government program of income transfer, which has some conditionality that strengthens access to basic social rights such as education, health and social assistance) and also because of benefits from other social programs, which have improved family access to food (despite inadequate eating and food purchasing habits). This social program aims to reduce poverty, especially in the poorest regions, and to help families to overcome existing vulnerabilities 62. Other government benefits which subsidize improvement in the development of the Brazilian Semiarid region are: Cistern Construction; Food Acquisition Program; National Family Farming Strengthening Program; National School Feeding Program; Government support to loss of agricultural produce due to drought; Drought assistance; Rural Activities Program; and National Program of Inclusion of the Young, which helps in the professional formation of young people of low income.

Regarding to lack of access to health services, despite high percentage of agreement, it is important to refer to the observation that there are health services, including specialists and activities by the Family Health, but sometimes there are difficulties accessing these services. Statements from interviewees from the organized civil society were highlighted: ‘There is no lack of health services, but there is a need to improve the infrastructure of these services in terms to guaranteeing better access to whose living in remote communities, far from the health district’s headquarters’. ‘There is also a need to improve the implementation of the Family Health program for all families, because this program does not yet include 100% of the rural families, who need assistance’. We also highlight the statements from group interviews: ‘Service provision here in the community only occurs every two months and only 15 people can be seen, one per family’. ‘In early years there were more children deaths, recently the access to the hospital is better, avoiding deaths’. It was reported by the organized civil society that despite the difficulty of access to health care services, the Unified Health System (in Portuguse, Sistema Único de Saúde – SUS, the universal health system in Brazil) represents the only means of health care for a large part of the population. It is considered a health reference for them. In the group interviews, it was mentioned that although the SUS is for everyone, in rural areas coverage is not complete.

Ebi and Bowen 61 point out that health services should not be interrupted, at the time when their demand may increase, as for example, in periods of drought. The statements of a health manager express this situation well: ‘Due to the lack of piped water, having water only in deposits or tanks, the health officers cannot act, increasing concerns about disease’. ‘When there is no water, the dental service is affected; and there is no water for the vaccine room, compromising hand hygiene’. ‘Health units have had to close due to lack of water’.

In relation to the lack of access to quality education for all, we highlight important observations. The first concern refers to quality education, especially in rural areas. Most of the organized civil society indicated that education needs to be improved. It needs to be expanded to all grades, in order to avoid student displacement to others cities, often distant, and having nothing to eat. The second concern, pointed out by both organized civil society and managers, was about the need for education to be focused on environmental issues, especially on local issues related to the semiarid region, as demonstrated by the following statements: ‘It is necessary to have an education focused on a rural area vision, including training and values, not an education only with urban vision, as the traditional education is’. ‘Teachers need to learn and have specific training to work on the reality of the students whose living in the rural areas’. Regarding to the high illiteracy rate in the semiarid region, the interviewees claimed that there is more illiteracy in the elderly, and that it is not a high rate. Currently there is a government program called Education for young people, adults and elderly, allowing access and opportunity of education for all.

We also highlight an important manager’s comment about the need for access to schooling: ‘We need to strengthen schools, because when there is a lack of food, it affects the students’ performance. The school lunch is the only food supply, and sometimes the schools are closed because there is no water’. This perception also shows how it is important that the basic services have uninterrupted access to water. In the case of schools, in addition to the provision of food, it allows for better children cognitive development, and also a great value to parents by being reassured that their children will have an opportunity to eat.

Regarding questions about land degradation it is observed that many tendencies are confirmed by indicators, such as production of firewood, deforestation, inadequate crop practices, and burning vegetation to prepare the soil for crops. Concerning the production of firewood, the majority claimed that there are communities that still use wood stoves, produce firewood to sell, charcoal preparation for sale, and the use of coal for ceramics production in some localities. Regarding deforestation, most of interviewees confirmed that this practice still occurs, but that it has been decreasing with the Family Agricultural program, and with the inspections by the Government Environmental Agency. An important observation was made by the organized civil society in this statement: ‘When addressing environmental degradation it is unfair to compare a small family production farmer with a big businessman’. The perception on these issues is also consistent with the following statements related to the increase of drought, both in frequency and intensity, and related to climate change. Almost 100% of the interviewees agreed and attributed as causes of environmental changes anthropogenic activities, such as soil degradation, deforestation and natural resources exploitation. We highlight an interesting observation made by a health professional: ‘I am 46 years old and the situation is getting worse, the rains are scarcer, rivers used to fill when it rained, currently it does not happen’. All the issues identified above, have been identified in the literature 1,2,12,41,44,46,47,49,63.

Regarding the groups considered most vulnerable to the impacts of drought, all the interviewees agreed with the identification of the elderly, children and pregnant women, as pointed out by the literature,24,25,26 but emphasised that the most vulnerable group are rural farmers, especially, elderly farmers. In relation to health care services, several interviewees expressed that health care should be the same for all, and not differentiated between specific social or population groups.

It is important to highlight the occurrence of migration mentioned by health professionals and managers, such as in the following statements: ‘Lack of employment and income leads the process of migration from rural to urban areas, causing depression and concerns.’ ‘Drought causes migration, many families go to other states and sometimes come back because it did not work for them; others go and cannot come back, living marginalized’. ‘There is no agriculture, so people leave the countryside for the cities to search for a better life’. ‘With these five years of drought, the production of subsistence has become unfeasible; those who have a better level of education migrate to the cities, because they no longer want to work in the fields’. ‘The rural exodus is increasing and the peripheries [referring to the urban zone] are growing more and more’. What is perceived in these different statements is the reflection of both the magnitude of drought impacts in the current period [referring to the 2011-2016 drought period] and the fragility of the local infrastructure, still inadequate to keep farmers in the rural drought prone areas. Another perception referred to these statements is the persistence of inequalities in this region, which may influence the migration process contributing to people maintain or entering the cycle of poverty, with a greater possibility of being exposed to other risks and vulnerabilities in the places to which they are migrating.

Risk factors for diseases in drought periods

In general, the perception about risk factors during droughts periods, presented a strong association with the most prevalent diseases during drought.

The following statements from health professionals express well the perception that lack of access to water or the scarcity of water, and water without quality, are direct risk factors of diarrhoea and gastrointestinal infections. Also these risk factors can indirectly generate other risk factors that determine illness. Examples include difficulty in maintaining hygiene, lack of food production, food contamination, difficulty finding employment and income, and the concerns regarding all of these factors, in their life 11,29,61,64,65. For example, regarding contaminated water: ‘The major complaint and population dissatisfaction is the lack of treated water’. ‘Lack of piped water causes concern about its origin because it can cause illness’. ‘There was no looting because of the drought, but there were 30 deaths because of contaminated water’, noting that in the past, desperately poor people looted stores for food. The health professionals also mentioned: ‘In the periods of drought, dirty water brings Hepatitis A’. ‘We had outbreaks of Hepatitis A in 2012’. As for hygiene issues it was mentioned: ‘With the lack of water, it is impossible to maintain hygienic practices, both individually and in the house’. ‘People depend on the water truck, which is of low quality limiting the use for body and housing hygiene’. ‘With lack of water, there is no way to have proper hygiene’. Regarding food and nutrition: ‘Due to lack of water serious health problems can arise related to insufficient food and even to personal hygiene’. ‘The lack of water reduces the supply of food because there is no more production and the prices of food increase’. From a health manager perspective: ‘The greatest fear and concern of people is the lack of water because they can loose their animals due the lack of grazing, and because of diseases’.

As for the inexistence or difficulty of obtaining an income and the opportunity of employment, the following statements express the severity of this risk factor, associated principally with psychosocial problems: ‘People suffer with unemployment’. ‘The cause of the lack of employment is that there is no economic development in the municipality and region’. ‘Drought causes unemployment that leads to family disruption’. These statements can be summarized in the following conclusion by a health Professional: ‘Drought is socially worsened by the precarious political and economic structures. If there were a political project and more opportunity for employment and income, the impacts of drought would be less’.

Some health problems, such as respiratory diseases, allergies, cardiac problems, flu, diarrhea and dry skin were attributed to prolonged exposure to the sun, hot weather and low air humidity. This perception is corroborated in some studies 11,66,67. To this concern we highlight statements that have been mentioned by two group interviews: ‘At night people who have good conditions sleep better than those who don’t, because dryness and heat can affect more those who cannot afford to buy a fan’. ‘The very strong sun is causing headache and tiredness, leaving people sick’.

Although concerns about saline water was mentioned by a few interviewees, it would important to consider further investigations, given the association of this risk factor with some diseases mentioned by the interviewees, such as hypertension, skin problems and renal problems, as well as the increase of hemodialysis procedures in some drought prone municipalities. This perception can be seen in the following statements: ‘Search for water is increasing, compromising the groundwater and water sources. The water is drawn from the local crevices, which is salty’. ‘The water is salty, which can bring skin problems’. ‘There are renal problems due to the lack of water and having salty water’. ‘There is hypertension, and renal and heart problems due to salty water’.

Most frequent diseases during drought period

The most common illnesses perceived by the interviewees during drought periods were diarrhea, respiratory diseases (respiratory problems and flu), dengue and gastrointestinal infections. Some statements related to this perception stand out: ‘Lack of water or contaminated water cause belly pain and diarrhea’. ‘The water from the water truck is for drinking, but it causes pain in the belly’. Although diarrhea had been cited as the most frequently illness, it was reported to be very common both during periods of drought, as well as during rain periods. This statement calls the attention to factors related to the lack of quality water for human consumption, food safety due to contaminated water and food, and the lack of water for personal hygiene. Sanitation is an important element in reducing several gastrointestinal infections including diarrhea, which can also be intensified in with higher temperatures. Several studies corroborate with this perception 11,29,64,65,68,69. Regarding respiratory problems, the following statements from health professionals reported the perception related to prolonged sun and heat exposure, and to low air humidity: ‘It is during the hottest and least humid period of the drought, that respiratory problems appear’. ‘Around where I live the most frequent problems in the population are respiratory’. ‘There is a lot of hospitalization from diarrhoea and lung problems with cough, due to low air humidity’.

Although most of the interviewees agreed that there are cases of dengue during drought periods, it was mentioned that this disease also increased during the rainy season. A relevant factor to be considered, according to literature, is the persistence of Aedes aegypti eggs in the dry season, due to the easy adaptation of the mosquito to the human environment 70. This situation tends to be aggravated in places with structural problems of water supply, facilitating the vector production through natural or artificial reservoirs.

As for physical fatigue and headache, it was mentioned that these problems are associated with work exposure to the sun and the hot and dry climate. The cases of dehydration were associated with some disease that causes diarrhea. It also was mentioned that these illnesses are well controlled by health system, because all cases when identified are immediately referred for treatment with oral rehydration and also to the Food and Nutrition Program. Regarding skin and allergic problems, most interviewees stated that skin problems are associated to hot and dry weather. Some have reported that some families use the water from the reservoirs to bathe and wash clothes, even knowing that the water is dirty, which could also result in allergic problems.

Most interviewees agreed that there are no cases of undernutrition, but there are cases of nutritional deficiencies in low-income populations. This occurs even among the families who receive financial benefits from the government, although this situation has improved greatly, as shown in the study by Rasella et al. 71. It was mentioned that cases of nutritional deficiencies occur due to some factors such as, limited knowledge about food quality; preference for the most practical and cheapest food; or consumption of those foods influenced by the media. As previously noted, there were more cases of obesity than undernutrition, which warrants further investigation. We also note that the literature points out important effects caused by nutritional deficiencies, such as infections. For example, when infections are added to other vulnerability factors, such as food shortage and poverty, it can bring about serious consequences, mainly to children and pregnant women 71,72.

As for depression, most organized civil society and health professionals agreed there were increased cases of depression during drought periods, even though they had never noted this association before. Some have reported knowing cases of depression or anxiety, including suicide in older farmers for several reasons such as loss of agricultural production, loss of animals, idleness, worries about not being able to sustain their family, sadness and anguish. However they were not certain of the association between depression and suicide with problems caused by drought. The following statements express this perception: ‘We don’t know if it is depression, but we see sadness in rural men’. ‘The lack of employment and income brings depression and worry to the rural man’. ‘I know cases of depression in farmers who see their animals dying, but I had never thought that it could be associated with drought’. ‘Depression has become part of population’s life who live in rural areas, it is very common here’. ‘There are many cases of suicide in elderly people, but we do not know if the reason is associated to drought’. ‘There are cases of suicide due to depression and concern’; ‘There are farmers who committed suicide because they had lost everything’. ‘There are cases of suicide due to economic problems’. Other interviewees referred that the cases of depression have a hereditary origin, because there were some cases of depression and suicide in the same family. All of this calls for a better investigation in this region, because the literature evidences the association of these mental health problems in drought prone areas, as for example, in the United States and Australia 73,74,75.

Concerns about the possibility of loss of agricultural produce and lack of food production to support the family and feed animals, and lack of income were also mentioned as significant problems for mental health. Both health professionals and a group interview show this perception according to the following statements: ‘Lack of access to water causes psychological problems, anxiety and concern about the elderly and the impacts of drought’. ‘More money is being spent to buy water than food, and this is causing psychological problems’. ‘Lack of employment and income affects the mental health of people who live in the rural areas’. These concerns were also attributed as a cause of high blood pressure.

As for alcoholism and violence, it was reported by most of the interviewees that there is no association with drought, stating that the cause is due to low income or financial difficulties. However, the literature identifies this risk 77.

Regarding to pain in the lower back and arms, it is important to note that the state of Rio Grande do Norte pointed this out due to the need for people to carry water from wells and fountains, often for long distances. According to those interviewed in Ceará, this problem occurs more in older people as a consequence of previous droughts, but today few people need to carry water because the conditions of water supply and access have improved.

Accidents with venomous animals and leishmaniasis were hardly mentioned, despite the latter being informed by the literature 78. Regarding trachoma, although the answers pointed out little knowledge about the subject, especially, whether there is an association with drought, it was reported that in that same period of the survey, a study was being carried out in schools to diagnose trachoma, which resulted in some positive cases. This bacterial infectious diseases deserves a better investigation in that area, because the literature points out that the main risk factors are associated with lack of hygiene and lack of sanitation. This is a prevalent disease in Brazil, with medium and high endemic levels in the Northeast, mainly in the poor municipalities with low socioeconomics status,79,80 characteristics present in the semiarid region.

The majority of interviewees’ perceptions about the impacts of drought on human health agree with the literature in identifying risk factors arising from damages in the systems and services and the vulnerabilities present in the region, as shown in Table 6. The symbol (*) highlighted in the table refers to the diseases, injuries and mechanisms that affect the process of health determinants that were mentioned by the interviewees.

Table 6: Possible impacts of drought on human health, through damages in the essential basic systems and services, and mechanisms for social determinants of health. Source: Adapted from references 11,27,29,61,68,73,77 * Diseases, injuries and mechanisms for health determination mentioned by the interviewees.

Damages in systems and services Mechanisms of social determinants of health Impacts on human health
Availability and safety of water Water shortage*. Implication in irrigation for agricultural production and in livestock and fishing increasing the possibility of food shortages*. Consequences of water quality (non-potable water, saline water)*. Contamination of water by various means, such as algal blooms, bacteria, fungi, viruses. Contamination of food*. Damages to the functioning of the health services, with consequences to the provision of some sanitary procedures*. Consequences on the water supply and distribution system (for piped water, water trucks, cisterns, artesian wells, dams and other alternative sources)*. Need for household water storage, which may compromise water quality*. Difficulty in maintaining personal, food and home hygiene*. Rising water prices due to scarcity and high purchase demand*. Consequences of urban sanitation and sewage services. Change in vectors, hosts and reservoir cycles. Gastrointestinal infectious diseases (diarrhea*, hepatitis A and other infections). Dehydration. Parasitic infectious diseases (worms*). Bacterial infectious diseases (trachoma, gastroenteritis). Dermatological infectious diseases*. Diseases transmitted by vectors and zoonoses (dengue*, zika*, chikungunya*, leishmaniasis, leptospirosis). Non-communicable diseases (hypertension, renal and mental problems)*. Infectious diseases transmitted by physical contact (flu*, conjunctivitis).
Availability and safety of food Deficiency in agricultural, livestock and fishery production causing food shortages*. Difficulty in the sustainability of family agriculture, livestock and fishery*. Consequences in food quality and safety*. Food contamination*. Rising food prices*. Decreased access to food, especially to healthy food*. Nutritional deficiencies*. Anemia. Malnutrition and its complications (low physical and cognitive development, deficiency of the immune system). Infections from food contaminated by viruses, bacteria, fungi, parasites (diarrhea*, cholera, hepatitis A*, worms*, other infections). Chronic non-communicable diseases* (hypertension, obesity)
Air quality Low humidity*. Increased temperature (heat)*. Dust*. Contamination of the air by particles from fires, and toxins accumulated in soil and water. Acute respiratory diseases (flu*, sinusitis, rhinitis, bronchitis, pneumonia). Allergic respiratory diseases (asthma, allergic rhinitis)*. Diseases caused by fungi, viruses, bacteria
Cleaning, hygiene and sanitation Difficulty in cleaning and hygiene (personal, household, water truck supply, food, health service equipment) due to lack of water*. Consequences of sanitation services, urban cleaning, health services* and other basic services. Dermatological infectious diseases*. Parasitic diseases (worms*). Infectious diseases transmitted by viruses, bacteria, fungi (flu*, conjunctivitis, pneumonia, gastrointestinal infections*, hepatitis A*, trachoma).
Social and economic factors Loss and damage in economic, livestock and subsistence plantations due to the difficulty in accessing water*. Loss or lack of employment and income*. Migration of populations seeking improvement in their quality of life, needing to face other social changes and cultural changes*, and changes in the epidemiological profile of the receiving areas. Displacement of the spouse to other municipalities in search of employment to supply family needs, which cause disruption and changes in the family structure and dynamics*. Loss of social identity. Uncertainty and concerns for the future*. Psychological disorders (anxiety*, stress*, behavioral change generating other problems such as violence, alcoholism). Depression*. Suicide. Chronic non-communicable diseases (heart, hypertension)*. Increased demand of health services and other social problems in the places where people migrate to.
Health care services Risk of interruption of health care procedures due to lack of water or contamination due to lack of hygienic conditions (dressing wounds, immunization, dentistry, and hospital services etc.)*. Increased demand for care and supplies of health services*. Risk of impacts in energy supply, impairing the use of health equipment, refrigeration of medicines and vaccines, and the health care of some hospital services. Communicable and non-communicable diseases*. Mental disorders*. Allergic respiratory diseases*. Nutritional deficiencies*. Lack of or reduction of health care due to lack of working conditions, which may worsen the health conditions of the population*.

Risk management in drought (prevention, preparedness, response and recovery)

Interviewees in both states showed awareness about drought increasing in intensity and frequency, citing deforestation as a cause. A high percentage had knowledge about the increasing effects of climate change on drought, although most of those interviewed did not know how this process occurred. The following statements refer to these perceptions: ‘Drought is increasing because of deforestation, and degraded soil that does not hold water’. ‘Drought is increasing because of the natural and environmental factors influenced by anthropogenic actions on deforestation and pollution’. Among the actions related to drought risk reduction management, the largest percentage referred to having information on disinfection of water stored in cisterns, and that the population does not perform the procedures correctly. It was also reported that there are actions to guarantee access to health care with equity by the Family Health Strategy and there is intersectoral integration for action planning. Despite this result, most responses stated that the actions are not specifically planned for drought and health management. Some actions are integrated in specific issues, for example, the control of dengue, which requires integration between epidemiological and environmental surveillance, specifically with endemic diseases vector control agents. The following statements from health professionals express this: ‘We live with the problem, but there is still no focus to address it, we only stay in the direct medical care to the patient’. ‘We establish goals and our focus is related to health problems associated from other causes and not from drought’; ‘We made evaluations, but they are not adequate for the current reality of drought’. ‘There is no drought-related health planning policy’. ‘I have been working on health for eight years, there is a trained team, but I have never participated in any drought-oriented planning’. ‘There is a lack of integration between levels of the government [municipal, state, federal]’. ‘There is a need for more integration between sectors and more professional training in drought’.

Regarding activities carried out with community participation, none of the interviewees reported on exchange of knowledge and decisions among the population and professionals, regarding drought. Such initiatives would help increase resilience in the population and government 56,81,82. The interviewees informed that integration between the health sector and the population occurs only in specific issues related to government programs, for example campaigns required by the federal or state levels of government, but these do not address drought. Example of campaigns were: guidance on using sodium hypochlorite to treat water for human consumption; breast cancer prevention; combating dengue; vaccination campaigns; and actions related to national programs of control of high blood pressure, diabetes and monitoring of acute diarrheal diseases. The following statements express this perception: ‘The health sector does not do activities with the population to discuss about drought, because there is no knowledge about it’. ‘We only call the population in relation to breast cancer prevention campaign and for other campaigns demanded by the federal level’. ‘There is a group of people receiving health care, but there is no discussion about drought’. ‘In the past [referring to the period before the drought] we discussed about problems such as hypertension and diabetes, but not problems focused on drought’.

Considering droughts before the current one (from 2012 to 2016), most of the managers interviewed agreed that the public policies have improved and strengthened through changes in the government program of access of water and food, access to health services, education and sanitation, as well as the implementation of other social policies. Most of the interviewees reported that current public policies, especially the income transfer program Bolsa Família, has been contributing to better access to food, as shown in the following statements: ‘People do not go hungry like they did in past droughts, and there are no more looting for food because the program Bolsa Família helps families to have better access to food’. ‘The drought of 1980s was lower than now, but the impacts were greater, for example, there was distribution of poor quality food, there was no water policy, there were many diseases and there were many cases of looting. Nowadays, with social policies, there are more opportunities with the implementation of cisterns and the stimulation of production and trade have increased greatly’. However, most of the interviewees form civil society stated that water access policies need to be consolidated in permanent and preventive policies rather than emergency policies as they currently are. The following statement expresses this perception: ‘It is necessary to take actions in the Semiarid region, not as an emergency policy, but as public policy with financial resources’. Regarding to the policies of social programs, this statement stands out: ‘These policies need more supervision and better structure, both to avoid inadequate use of its benefits and to ensure that all families that really need these benefits are registered in the program. There are many families in extreme poverty that are not registered’.

Although most of the interviewees pointed out that educational measures are developed both in health units and in schools, through the School Health Program, they informed that they had never worked specifically on the issue of drought. The following statement of a civil society interviewee expresses this concern: ‘There is a project of continuing education that should enable secondary school teachers to focus on the local reality’, referring to the rural areas, including those affected by drought. ‘Many educational policies are developed only in the districts (referring to urban area), it should be decentralized to the rural communities’.

Regarding the implementation of adaptive measures for implementing a resilient and sustainable agriculture, all interviewees of organized civil society reported that the government sells a seed that is more resistant to pests, but the farmers prefer the seed called “crioula” (autochthonous seed, defined by them as the original seed of the land), which is more resilient to the weather and type of soil (i.e. the Caatinga biome), and has the opportunity to sustain future resilient plantations. The follow statement corroborate with this perception: ‘The seed we buy from the government dos not give pests, it must be because it has pesticides, but we have to buy every year, sometimes we can use it twice; for this reason we prefer the “crioula” seed, which can be planted several times’.

In relation to training the population with provision of courses to generate employment and income it was mentioned that the government together with the civil society offer several courses, but that often the population does not participate for lack of information. Regarding the training of health professionals on the management of risks and vulnerabilities associated with drought, most of them said that there is training in several issues, which may be associated with drought, but in relation to drought specifically, they had never received any training. Health professionals made the following statements: ‘It is necessary to have a training policy for health professionals to know how to relate diseases to the drought season’. ‘It is necessary to have more lectures, orientations, qualification and training of professionals to improve their knowledge to be able to orient the population’.

The results related to risk reduction management of drought (with measures of prevention, preparedness, response and recovery, including adaptation measures) by health professionals showed lack of organization and preparedness to reduce risks and vulnerabilities associated to drought. The information and communication process necessary for understanding and addressing the social determinants of health (taking into account the people, their beliefs, their cultures, their ways, their place where they live and their needs) is also neglected; not only diseases related to drought. Figure 1 summarizes the perception of populations and government regarding drought, in two states of Rio Grande do Norte and Ceará. It is highlighted in this figure the summarized results in around key issues, such as inequalities in the social vulnerability, poverty in the economic vulnerabilities, water in the environmental vulnerabilities; and the actions that were identified as necessary to increase the resilience of populations and health systems exposed to drought.

ASynth11Oct

Fig. 1: Synthesis of the perception of the government and populations exposed to drought in the states of Rio Grande do Norte and Ceará.

Conclusion

The results obtained from this research shows, in general, the fragility in the social and political infrastructure necessary to improve the living conditions of populations, particularly those that depend on family agriculture in drought prone areas. The conditions of social, economic and environmental vulnerabilities presented in the region can be amplified by the drought process, and can aggravate the impacts resulting drought events, thus disadvantaging the population of this region, as shown in the Fig. 1.

The general perception of the interviewees of the health sector shows an agreement with what is found in the literature on the relationship between drought and health conditions. However, the lack of specific knowledge regarding the relation between drought and health, both by health professionals and managers, and the lack of intra and intersectoral integration, prevents the appropriate planning of actions aimed at managing risks and impacts of drought on health. This fragility of the health system tends to disregard the process that determines drought conditions, with consequences in the recognition of this climatological event as a public health problem. Lack of comprehensive recognition of the issue may result in the possibility of greater negative impacts on health conditions and, consequently, on the life of these populations.

With the projection that droughts in this Brazilian Semiarid region would be aggravated in the future by the global process of climate change, there is a possibility to further increase unfavorable conditions for the populations. This would occur mainly due to climatic variability in the region, added to the social and economic vulnerabilities present in the area. From this point of view, there should be a greater concern to subsidize the formulation of public policies aimed at health management on risks associated to drought. The construction of the risk reduction management process, taking into consideration prevention, preparedness, response and recovery, including adaptation strategies, with actions involving community participation would help in increasing resilience by building capacity of the government and of the populations living in this region.

Despite efforts by the government in relation to implementing programs to minimize the impacts associated to lack access of water, food and income, additional mitigation measures are required. Considering the results of this research, it is necessary to implement educational programs in several areas, such as: change in sanitary practices mainly related to the treatment of water for human consumption; food safety; improvement on the use and reuse of water; training of the health sector to ensure appropriate planning for the region focusing on the reality of its drought conditions, in order to provide adequate and equitable health care; and training the population based on their reality of living with semiarid a climate, taking into account their human rights. Other major change processes would be necessary to ensure greater socioeconomic stability in the Semiarid region, such as, basic sanitation, which is extremely necessary to improve the health conditions of the populations; permanent measures of access to water and food; access to education, information and technical knowledge for all; and opportunity to generate employment and income.

Specifically for the health sector, a health professional statement summarizes what we presented in this research: ‘I found interesting this initiative about alerting on drought impacts, because we are already punished and suffer so much from droughts that we have already become accustomed to it, not noticing its impacts’. In view of the results found, some important measures need to be established in order to increase the capacity of management in preparedness and response to droughts. These measures relate to: establishing an intra and intersectoral dialogue to discuss the issue of drought; integrated training with other sectors in risk reduction management associated to droughts; increase knowledge of the risks and vulnerabilities present in each area where the Family Health Strategy operates; improvement in the knowledge of what diseases can be accentuated in drought conditions; development of activities with community participation; improvement in the health information data system, adding variables that could indicate the increase of diseases caused by drought conditions; investigation of the injuries and diseases highlighted by the interviewees, as well as, of the situations of local social and environmental vulnerabilities, in order to subsidize strategies for the prevention of risks and diseases.

The greater perception of the communities living in the Brazilian semiarid refers to the importance of knowing how to live with drought conditions and be resilient, instead of feeling victims or vulnerable to the impacts of drought. In the words of a person from the organized civil society: ‘It is necessary to work with sustainable practices of coexistence with the Semiarid’. This is the main lesson of adaptation, which should be learned by all.

Competing Interests

The authors have declared that no competing interests exist.

Data Availability

All relevant data are within the paper. However, to ensure full data availability, the raw data for this paper may be accessed at the Brazilian Observatory of Climate and Health: https://www.climasaude.icict.fiocruz.br/novo/ftp.html. The data can also be accessed through the figshare repository: https://doi.org/10.6084/m9.figshare.7268519.v1 and https://doi.org/10.6084/m9.figshare.7268513.v1.

Corresponding Author

Aderita Sena: aderitasena@gmail.com

]]>
http://currents.plos.org/disasters/article/drought-in-the-semiarid-region-of-brazil-exposure-vulnerabilities-and-health-impacts-from-the-perspectives-of-local-actors/feed/ 0
Impact of the 2013 Floods on the Incidence of Malaria in Almanagil Locality, Gezira State, Sudan http://currents.plos.org/disasters/article/impact-of-the-2013-floods-on-the-incidence-of-malaria-in-almanagil-locality-gezira-state-sudan/ http://currents.plos.org/disasters/article/impact-of-the-2013-floods-on-the-incidence-of-malaria-in-almanagil-locality-gezira-state-sudan/#respond Mon, 08 Oct 2018 15:00:44 +0000 http://currents.plos.org/disasters/?post_type=article&p=37568 Background: Heavy rain hit Sudan in August 2013 with subsequent flash floods in different parts of the country. This study investigated the impact of the flooding on incidence of malaria in Almanagil Locality in central Sudan.

Methods: This observational retrospective study compared malaria data sets during rainfall seasons in the Almanagil Locality in the year of flooding (2013) with those of corresponding rainfall seasons of previous two non-flood years (2011 and 2012).

Results: A marked increase of new malaria cases and incidence rate was observed in the 13 sentinel malaria notification sites in the locality  (IR increased from 6.09 per 100,000 person­days in 2011 [95 % CI: 5.93-6.26] and 6.48 in 2012 [95 % CI: 6.31-6.65] to 8.24 in 2013 [95 % CI: 8.05-8.43] ; P< 0.0001), with a peaking of the incidence rate in the under-5-years age group (IR for this age group jumped from 9.80 per 100,000 person­days in 2011 [95 % CI: 9.29­10.32] and 10.00 in 2012 [95 % CI: 9.52­10.49] to 15.02 in 2013 [95 % CI: 14.41­15.64]). A noticeable increase in the slide positivity rate (P< 0.0001) was observed in the 12-week period of 2013 (SPR = 20.86% [95 % CI: 20.40 ­21.32%]) compared with the same periods in 2011 (SPR = 8.72% [95 % CI: 8.36 ­9.08%]) and 2012 (SPR = 12.62% [95 % CI: 12.24 ­13.01%]), with a more marked rise of the SPR in the under-5-year age group. Hospital data showed increase in both the inpatient and outpatient incidence proportions in the study period of 2013 compared to those of the years 2011 and 2012. Hospital OPD incidence proportion in 2013 was 19.7% (95% CI: 19.24­20.18%) compared to 12.85% (95% CI: 12.48­13.23%) in 2011, and 12.16% (95% CI: 11.82­12.51%) in 2012. The < 5 year old groups were responsible for the overall rise in the proportion of malaria cases in 2013 , particularly the < 1 year old group which more than doubled in the 2013 period compared to both 2011 and 2012 periods (Age­specific proportion of the outpatient malaria cases of the < 1 year old group in 2013 was19.5% [95% CI: 18.5­20.6%]  compared to 7.7% [95% CI: 6.9­8.6%] in 2011 and 8.1% [95% CI: 7.3­8.9%] in 2012. Incidence proportion of severe malaria cases (inpatients) increased to 22.5 % (95 % CI: 21.5 to 23.6 %) in the study period of 2013 compared to 19.8 % (95 % CI: 18.6 to 21.0 %) in 2011 and 18.4 % (95 % CI: 17.4 to 19.5) in 2012. The increase in the proportion of severe malaria cases was mainly due to a higher proportion of children < 5 years of age and especially to a higher proportion of children < 1 year of age.

Conclusion: The study revealed a significant increase in the incidence rate of malaria in Almanagil Locality following the flash flood of August 2013. The flooding had the highest impact on the malaria incidence of the under-5-years age group, and particularly of the under-1-year age group.

Keywords: Flood, Flooding, Malaria, Disaster, Sudan, Gezira, Almanagil

]]>

Background

Besides the rise in global temperature, climate change also results in an increased frequency of extreme weather events such as floods and droughts 1, 2, 3 . Floods are the most common natural disasters globally and have led to extensive morbidity and mortality throughout the world especially in low-resource countries 4, 5, 6 . Impacts of floods on the health of populations depend not just on the magnitude of the flood, but also on geographic, socio-economic and human factors including the vulnerability and adaptive capacity of the population affected 6, 7, 8 . Climate change, resulting in rising temperatures, changing precipitation patterns and increase in extreme weather events, can affect infectious disease outbreaks by altering vector population, density and survival rates, and pathogen reproduction and development rate 2, 9, 10 . This in turn will influence human exposure to bites from infected vectors 8 .

Malaria has been identified as one of the diseases most sensitive to climatic factors 10, 11, 12 . Conflicting results exist on the impact of extreme weather events on malaria transmission 12, 13, 14, 15 . Malaria epidemics attributed to prolonged precipitation and flooding have been reported in endemic areas world-wide 10, 12, 13, 15, 16, 17, 18, 19, 20, 21 . However, other studies showed a decrease in the malaria burden or did not provide evidence in support of causal relationship between malaria transmission and heavy rainfall or flooding 22, 23, 24, 25 . Although the two opinions seem to be contradictory, both viewpoints can be true. Heavy precipitations and floods may initially decrease vector populations by eliminating existing mosquito-breeding sites and, hence, lower malaria transmission 11, 12, 13, 21, 26. However, as the heavy rainfall stops and floods waters gradually recede, stagnant pools are created, providing ideal habitats for mosquitoes and resulting in an increase in the vector population and an upsurge of malaria transmission in the following weeks 10, 11, 19, 20, 21. Moreover, these differences in the impact of heavy rainfall and flooding may also be the result of the specific geographic setting of the affected area such as topography, distance to water bodies, altitude, climate and land type 12, 14, 15 . The lag time, which may vary by geographic and climatic conditions, is usually around 4 to 8 weeks between the flooding and the onset of a malaria outbreak 9, 16, 18, 26, 27, 28, 29 .

Malaria is considered a leading cause of mortality and morbidity around the world, with a global 2013 estimate of 198 million cases and about 584000 deaths. Africa accounts for an estimated 90 % of all deaths and children aged under five represented 78 % of all deaths 30 . Malaria in Sudan is endemic 31 and continues to be a major public health problem. The whole population is at varying degrees of risk. Climate models estimate that 75 % of the population of Sudan are at risk of endemic malaria and 25 % are at risk of epidemic malaria 29 . Plasmodium falciparum is the main parasite and Anopheles arabiensis is the primary vector 32 . A National Malaria Control Program (NMCP) has been established under the Sudan Federal Ministry of Health (FMOH). The country endorsed the international Roll Back Malaria initiative in 1998 emphasizing efforts towards more attention on early detection, prompt treatment and various prevention measures 33 . Collaboration of the NMCP with the United Nations Development Program (UNDP), the World Health Organization (WHO) and national Non-Governmental Organizations to fight malaria in Sudan, resulted in reducing the number of malaria cases from more than four millions in 2000 to less than one million in 2010, and 75% reduction in mortality due to malaria between 2001 and 2010 34, 35 .

According to the WHO criteria for risk of malaria transmission and the 2012 Malaria Indicators Survey (MIS), Gezira State is considered hypo­endemic 36 , entailing that the state is more liable to malaria outbreaks. Indoor ­household residual spraying (IRS) and distribution of insecticide treated nets (ITNs) were implemented in Gezira State. The 2012 MIS showed that approximately 95% of the population was protected by IRS, but only 34% of the households owned at least one ITN and only 5% of the households members slept under an ITN 36 .

In 2013, the worst flash floods in 25 years hit Sudan 37 . Continual heavy rains from early August 2013 and consequent flash floods caused extensive damage and loss of life in 15 states. Reports from the Humanitarian Aid Commission estimated 499,900 people countrywide were affected by the heavy rainfall and floods across Sudan since the onset of the events in early August 2013. Assessments have shown that the floods destroyed or damaged over 85,385 houses in the states resulting in the displacement of a large part of the affected population, disrupting the healthcare system, the provision of drinking water and access to sanitation 38 . Gezira State was one of the most affected states: 52,975 affected people, 5,946 houses destroyed and 5198 houses damaged 39, 40 . There were concerns about epidemics of communicable diseases including vector-borne diseases such as malaria 41 .

The objective of the study was to estimate the malaria incidence attributable to the 2013 flood in Almanagil Locality, Gezira State in central Sudan.

Methods

Study Area

Gezira State is located in the east-central region of Sudan, is crossed by the Blue Nile and is irrigated by two canals of the Gezira and Managil agriculture schemes. Almanagil Locality was in 2013 one of the 8 localities that constituted the Gezira State and was situated in the south western part of the state. Almanagil Town, the capital of the locality, is 62 km away from Wadmedani, the capital of Gezira State, and 156 km from Khartoum, the capital of Sudan (Figure 1).

Figure 1

Fig. 1: The location of Almanagil Locality in Gezira state in Sudan in 2013.

The climate is characterized by an average daily temperature of 32° C during summer and 22° C during winter. The rainy season starts in July/August and ends by October, with an estimated annual rainfall of 140 to 225 mm and a relative humidity of 30% to 38% 42 . The estimated population of the locality in 2013 was 1,050,000 persons, living in 6 towns, 416 villages and 356 agricultural labourer settlements and consisted of 90% rural households 43 .

Malaria case definition

The malaria case definition remained unchanged throughout the 3 years of study and was based on the national protocol for diagnosis and treatment of malaria: “A malaria case is confirmed by demonstration of asexual forms (trophozoite stage) of the parasite in a thick or thin peripheral blood film or by rapid diagnostic test (RDT) in the presence of fever” 36 .

Data collection

The sources for data collection are twofold: the sentinel surveillance system and the routine health management information system operated by the Gezira State Ministry of Health (MOH). Data on confirmed malaria cases was collected from the sentinel malaria notification sites (SMNSs) of Almanagil Locality. The data was extracted from the weekly reports of 13 SMNSs (9 hospitals and 4 health centres) forwarded to the Department of Malaria Control Programme of the Gezira State MOH for the years 2011, 2012 and 2013. The SMNSs are facilities equipped with laboratories and trained clinical and laboratory staff capable of performing microscopy and RDTs for malaria. Hence the malaria incidence rate (IR) and the slide positivity rate (SPR) are calculated from this data.

Hospital data of outpatient malaria cases was collected by the 9 hospitals of Almanagil Locality. Data of inpatient malaria cases was collected by 8 of the 9 hospitals (the ninth hospital was not yet fully operational in 2013 with only outpatient services). The data was extracted from the monthly reports forwarded to the Statistics Department of the Gezira State MOH for the years 2011, 2012 and 2013. The data consisted of the monthly numbers of all out- and inpatient diagnoses, including the confirmed malaria cases.

As heavy rain and subsequent flash flood occurred in the beginning of August and taking into account that the effect of these climatic factors on malaria is not immediate, a four-week lag effect was assumed 11, 27, 29 . A new generation of infective vectors needs about 30 days to develop: 15 days for the preimaginal development of vector Anopheles, 4-7 days for the gonadotropic cycle for parous/nulliparous female mosquitoes and 12 days for the sporogonic cycle for the Plasmodium falciparum parasites in the vector mosquitoes 44 . As a result, the study compared data of the months September, October and November (hospital data registered by Gezira State Ministry of Health) and the corresponding 12 weeks from 36th week to 47th week (data registered by 13 SMNSs) of the year of the flooding (2013) with data of the same periods in the two preceding years without flooding (2011 and 2012).

Estimates of the general population were obtained from the Preventive Medicine Department and under 5 years population from the Vaccination Department in Almanagil Locality.

To determine the impact of flooding on malaria burden in Almanagil Locality the following outcomes of interest were selected: (1) the malaria incidence rate expressed as the number of new malaria cases per person-time, (2) the SPR or test positivity rate (TPR) defined as the number of laboratory-confirmed cases (microscopy or RDT) per 100 suspected cases, (3) incidence proportion of uncomplicated malaria cases as the ratio of the number of malaria-related outpatient visits to the total of all-cause outpatient visits, (4) incidence proportion of severe malaria cases measured as the ratio of the number of malaria-related hospitalisations to the total hospital admissions.

Data analysis

The confirmed malaria cases and SPR data were extracted from the SMNSs’ reports of the weeks 36 to 47 of each year representing the months September, October and November of each year under study. The registered data was stratified by age (under 5 years of age and above 5 years old) and gender. The confirmed malaria cases were extracted from the hospital reports of the months September, October and November of each year under study. The registered data was available in a stratified form by age groups (< 1 year, 1-4 years, 5-14 years, 15-24 years, 25-44 years and 45 years and older) and gender. As the time interval of the SMNS and hospital reports is not quite the same, i.e. 12 weeks or 84 days versus 3 months or 91 days, the malaria incidence rate is expressed as the number of new cases per 100,000 persons per day.

Data was exported to Microsoft Excel® 2007 spreadsheets and statistical analysis was performed using MedCalc® version 18. P ≤ 0.05 was considered significant for all tests.

Ethical considerations

Ethical clearance for the study was given by the Gezira State MOH.

Results

Malaria incidence rate

Analysis of the confirmed malaria cases detected through passive case surveillance in the 13 SMNSs in Almanagil Locality during the 36th to 47th week of the 3 years under study revealed that the malaria incidence rate was highest in 2013 (Figure 2). A marked IR increase to 8.24 per 100,000 person­-days (95% CI: 8.05-8.43) was noticed in the year of the flood in comparison to the two non­-flood years 2011 (IR: 6.09; 95% CI: 5.93-6.26; P < 0.0001) and 2012 (IR: 6.48; 95% CI: 6.31-6.65; P < 0.0001) as shown in Table 1.

Table 1: Numbers and incidence rates with 95% confidence intervals of new malaria cases recorded in the SMNSs of Almanagil Locality in weeks 36-47 of years 2011-2013.

Year 2011 Year 2012 Year 2013
Population of Almanagil Locality 990,247 1,019,953 1,049,659
Number of new SMNS cases 5,069 5,549 7,262
Incidence rate per 100,000 person-days (95% CI) 6.09 (5.93-6.26) 6.48 (6.31-6.65) 8.24* (8.05-8.43)
SMNS: Sentinel malaria notification site. * P < 0.0001.

Fig. 2: Malaria incidence rates with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013. * P < 0.0001.

Fig. 2: Malaria incidence rates with 95% confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013. * P < 0.0001.

The malaria IR of both age groups (<5 years and > 5 years) increased in the 12-­week period of the flood year (P < 0.0001) compared to the corresponding period in the two non-­flood years. IR for the above­-5-­year age group increased from 5.31 per 100,000 person­-days in 2011 (95% CI: 5.14­-5.49) and 5.64 in 2012 (95% CI: 5.46-­5.82) to 6.80 in 2013 (95% CI: 6.62-­7.00). IR for the under-­5-­year age group jumped from 9.80 per 100,000 person-­days in 2011 (95% CI: 9.29-­10.32) and 10.00 in 2012 (95% CI: 9.52­-10.49) to 15.02 in 2013 (95% CI: 14.41­-15.64) as shown in Table 2 and Figure 3.

Table 2: Number of new malaria cases and incidence rates by age group with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013.

Year 2011 Year 2012 Year 2013
Age group Under 5 years Above 5 years Under 5 years Above 5 years Under 5 years Above 5 years
Total number of SMNS cases 11,032 33,645 11,917 37,073 10,603 41,170
Number of malaria cases 1,419 3,650 1,645 3,904 2,307 4,955
Age category proportion from total malaria cases 28.0% 72.0% 29.6% 70.4% 31.8% 68.2%
Population by age group 172,431 817,816 195,840 824,113 182,866 866,793
Age-specific incidence rate of malaria per 100,000 person-days (95% CI) 9.80 (9.29-10.32) 5.31 (5.14-5.49) 10.00 (9.52-10.49) 5.64 (5.46-5.82) 15.02* (14.41-15.64) 6.80* (6.62-7.00)
* P < 0.0001.

Fig 3

Fig. 3: Age-specific malaria incidence rates with 95% confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013. * P < 0.0001.

Slide Positivity Rate

Analysis of the blood smears examined at the SMNSs showed a noticeable increase (P < 0.0001) in the SPR in the 12­-week period of the flood year (SPR = 20.86% [95% CI: 20.40-­21.32%]) in comparison to the corresponding periods in 2011 (SPR = 8.72% [95% CI: 8.36-­9.08%]) and 2012 (SPR = 12.62% [95% CI: 12.24­-13.01%]) as shown in Table 3 and Figure 4.

Table 3: Number of positive blood smears and SPR with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013.

Period BSE Number of positives SPR 95% Confidence interval
All 12 Weeks 2011 23,623 2,059 8.72% (8.36 -9.08%)
All 12 Weeks 2012 28,903 3,647 12.62% (12.24 -13.01%)
All 12 Weeks 2013 29,900 6,236 20.86%* (20.40 -21.32%)
SPR = Slide positivity rate, BSE = Blood smears examined. * P < 0.0001.

Fig. 4: Slide positivity rate with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013. * P < 0.0001.

Fig. 4: Slide positivity rate with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013. * P < 0.0001.

Persons exposed to the flooding in 2013 in Almanagil Locality had respectively 2.39 (95% CI: 2.27-2.51) and 1.65 times the risk of having been infected with malaria parasites compared to persons who were not exposed to flooding in 2011 and 2012 (P < 0.0001). The SPR of both age groups (< 5 years and > 5 years) increased in the 12­-week period of 2013 (P < 0.0001), particularly in the under­-5-­year age group, in comparison to the corresponding period in the non-­flooding years. SPR for the above­-5-­year age group increased from 7.79% in 2011 (95% CI: 7.41-8.18%) and 12.25% in 2012 (95% CI: 11.83-12.69%) to 19.94% in 2013 (95% CI: 19.44-20.46%). IR for the under­-5-­year age group increased from 11.94% in 2011 (95% CI: 11.09-12.84%) and 13.86% in 2012 (95% CI: 13.04-14.71%) to 24.30% in 2013 (95% CI: 23.2-25.38%) as shown in Table 4 and Figure 5.

Table 4: Number of positive blood smears and slide positivity rate by age groups with 95% confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36-47 of years 2011-2013.

Table 4. Number of positive blood smears and slide positivity rate by age groups with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36-47 of years 2011-2013.

Fig. 5: Slide positivity rate by age groups with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013. * P < 0.0001.

Fig. 5: Slide positivity rate by age groups with 95 % confidence intervals recorded in the SMNSs of Almanagil Locality in weeks 36 to 47 of years 2011 to 2013. * P < 0.0001.

Malaria incidence proportion in outpatient departments (OPD) of public hospitals

A noticeable increase (P < 0.0001) in the proportion of outpatient malaria cases in public hospitals of Almanagil Locality was observed for the months September, October, November 2013 in comparison to the same period of 2011 and 2012. Hospital OPD incidence proportion in the 2013 period was 19.7% (95% CI: 19.24-­20.18%) compared to 12.85% (95% CI: 12.48-­13.23%) in 2011, and 12.16% (95% CI: 11.82-­12.51%) in 2012 (Table 5). Persons exposed to the flooding in 2013 and consulting the OPD of the public hospitals in Almanagil Locality had respectively 1.53 and 1.62 times the risk of contracting malaria compared to persons who were not exposed to flooding in 2011 and 2012.

Table 5: Number and percentage of malaria cases out of all outpatient visits with 95 % confidence intervals in public hospitals of Almanagil Locality in the months September, October and November of the years 2011 to 2013.

Number of Malaria OPD cases Total all-cause OPD cases Malaria cases / 100 all-cause OPD cases (95% Confidence Interval)
Year 2011 (Sep-Nov) 3,991 31,060 12.85% (12.48-13.23%)
Year 2012 (Sep-Nov) 4,211 34,624 12.16% (11.82-12.51%)
Year 2013 (Sep-Nov) 5,447 27,647 19.7% (19.24-20.18%)*
* P < 0.0001.

A more detailed age­-specific analysis was possible as hospital data were stratified by six age groups. Analysis of the age-­specific proportion of the outpatient malaria cases revealed that the age groups responsible for the overall rise in the proportion of malaria cases in 2013 were the two under­-5-­year age groups (< 1 year and 1-­4 years). Considering < 5 years of age as one group, the proportion of outpatient malaria cases to the total number of outpatients in 2013 study period was as high as 38.5% (95% CI: 37.2-39.8%) compared to 24.3 % (95% CI: 23.0-25.6 %; P < 0.0001) in the 2011 study period and 25.2% (95% CI: 23.9-26.6%; P < 0.0001) in the 2012 study period (Table 6).

Table 6: Percentage of malaria cases out of all outpatient visits by age group with 95 % confidence intervals in public hospitals of Almanagil Locality in the months September to November of the years 2011 to 2013.

Year 2011 Year 2012 Year 2013
Age group (in years) Number Proportion in % (95% CI) Number Proportion in % (95% CI) Number Proportion in % (95% CI)
Under 1 308 7.7% (6.9-8.6%) 340 8.1% (7.3-8.9%) 1,062 19.5% (18.5-20.6%)*
1 – 4 661 16.6% (15.4-17.7%) 722 17.1% (16.0-18.3%) 1,033 19.0% (17.9-20.0%)*
.
< 5 969 24.3% (23.0-25.6%) 1,062 25.2% (23.9-26.5%) 2,095 38.5% (37.2-39.8%)*
5 -14 846 21.2% (20.0-22.5%) 898 21.3% (20.1-22.6%) 958 17.6% (16.6-18.6%)
15 – 24 833 20.9% (19.6-22.2%) 802 19.0% (17.9-20.3%) 880 16.1% (15.2-17.2%)
25 – 44 798 20.0% (18.8-21.3%) 769 18.3% (17.1-19.5%) 773 18.3% (17.1-19.5%)
45+ 545 13.6% (12.6-14.8%) 680 16.1% (15.1-17.3%) 741 13.6% (12.7-14.5%)
Total 3,991 100% 4,211 100% 5,447 100%
* P < 0.0001.

Analysis of the < 5 years of age group showed that the proportion of the hospital OPD malaria cases in the < 1 year old group more than doubled in the year of the flooding compared to the two non-flood years (Figure 6). Children < 1 year of age exposed to the flooding in 2013 had respectively 2.5 (95% CI: 2.2-2.9) and 2.4 (95% CI: 2.1-2.7) times the risk of catching malaria compared to children of < 1 year who were not exposed to flooding in 2011 and 2012 (P < 0.001).

fig 6

Fig. 6: Comparison of the percentages of malaria cases out of all hospital outpatient visits with 95 % confidence intervals between <1 year old group and 1-4 years old age group during September to November in years 2011 to 2013. * P < 0.001.

Malaria incidence proportion of severe malaria cases in public hospitals

The proportion of admitted malaria cases to the total hospital admissions in the 8 public hospitals of Almanagil Locality increased to 22.5 % (95% CI: 21.5-23.6%) in the study period of 2013 compared to 19.8 % (95% CI: 18.6-21.0%; P = 0.001) in 2011 and 18.4 % (95% CI: 17.4-19.5%; P < 0.0001) in 2012 as shown in Table 7.

Table 7: Inpatient malaria cases by number and percentage of total public hospital admissions with 95 % confidence intervals in Almanagil Locality during September to November in years 2011 to 2013.

Period Number of malaria admissions Total hospital admissions Proportion in % (95% confidence Interval)
Year 2011 (Sep-Nov) 832 4,198 19.82% (18.64-21.05%)
Year 2012 (Sep-Nov) 984 5,335 18.44% (17.43-19.51%)
Year 2013 (Sep-Nov) 1,307 5,805 22.51% (21.46-23.61%)*
* P < 0.0001.

Analysis of age-specific proportion of the inpatient malaria cases (Table 8) showed that the increase in the proportion of severe malaria cases was due to a higher proportion of children < 5 years of age (P = 0.05) and especially to a higher proportion of children < 1 year of age (P = 0.01).

Table 8: Percentage of malaria cases out of all hospital admissions by age group with 95 % confidence interval of malaria in public hospitals of Almanagil Locality in the months September to November of the years 2011 to 2013.

Year 2011 Year 2012 Year 2013
Age group (in years) Number Proportion in % (95% CI) Number Proportion in % (95% CI) Number Proportion in % (95% CI)
Under 1 101 12.1% (10.0-14.6%) 119 12.1% (10.2-14.3%) 209 16.0% (14.1-18.1%)*
1 – 4 187 22.5% (19.8-25.4%) 211 21.4% (12.0-24.1%) 298 22.8% (20.6-25.1%)
.
< 5 288 34.6% (31.5-37.9%) 330 33.5% (30.6-36.6%) 507 38.8% (36.2-41.5%)#
5 -14 131 15.7% (13.4-18.4%) 162 16.5% (14.3-18.9%) 216 16.5% (14.6-18.6%)
15 – 24 129 15.5% (13.2-18.1%) 128 13.0% (11.0-15.2%) 153 11.7% (10.1-13.6%)
25 – 44 145 17.4% (15.0-20.2%) 181 18.4% (16.1-20.9%) 209 16.0% (14.1-18.1%)
45+ 139 16.7% (14.3-19.5%) 183 18.6% (16.3-21.1%) 222 17.0% (15.0-19.1%)
Total 832 100% 984 100% 1,307 100%
* P = 0.01, # P = 0.05

Discussion

According to the Malaria Programme Review 2001-2012 report of Sudan’s NMCP, the national malaria incidence rate decreased from 38/100,000/day in 2000 to 9.9/100,000/day in 2011, based on both laboratory-confirmed and clinically diagnosed malaria cases reported by all outpatient departments 36 . Our study in Almanagil Locality revealed a malaria incidence rate in the non-flood years 2011 and 2012 of respectively 6.1/100,000/day and 6.48/100,000/day based only on confirmed malaria cases reported by the SMNSs (outpatient department of 4 health centres and 9 hospitals).This lower malaria incidence rate may be caused by the fact that only laboratory confirmed cases are included in the study.

Sudan is frequently exposed to flash floods due to torrential rainfall or by overflow of the river Nile during the rainy season 45, 46, 47, 48 . Floods may play a major role in the emergence of malaria epidemics 2, 7, 12, 16 . Surveys showed that floods in endemic areas in Sudan were often associated with malaria outbreaks over and above the annual increase of malaria cases normally expected in the rainy season 36, 45, 46, 47, 48 . Comprehending the interrelationship between malaria and flooding is essential for predicting epidemics, adequate preventive measures and appropriate response in order to minimize the morbidity of the affected populations 10, 11, 15, 49 . Malaria epidemics attributed to flooding depend on a number of factors such as the malaria endemicity and topography of the affected area, the severity of the flooding (damage to private houses and public infrastructure), the ecological change caused by the flood (propagation of malaria infected mosquitoes), displacement and vulnerability of the affected population and level and accessibility of healthcare services before and after the flood 36, 47, 50, 51 .

The main sources of information on malaria incidence are disease surveillance systems and health information systems operated by ministries of health. In acute emergencies such as floods, surveillance is generally based on data collected by healthcare workers and reported at regular interval through health centres and hospitals and should be representative of the entire disaster area 49 . It provides early warning of an epidemic and trends at geographical and temporal levels 49 . The study used two sources of information to collect the malaria data: a sentinel-based (SMNSs) surveillance method advocated by the NMCP and the routine health information system of the Gezira State Ministry of Health. The SMNSs of the Almanagil Locality were selected to gather information representative of the entire locality based on a number of criteria, including the coverage of hospitals and health centres and geographical areas of earlier epidemics and natural disasters 52 . The malaria incidence in this study was measured by passive case detection based on healthcare facility-based data. This assumes completeness of reporting, accurate laboratory confirmation of all malaria cases and all patients present to healthcare facilities 53, 54 . In order to minimize the impact of incomplete reporting, changes in healthcare utilization and errors in laboratory confirmation of malaria cases, it is recommended to focus on confirmed malaria cases, to assess trends in slide or test positivity rate (SPR/TPR) and to determine the proportion of cases that malaria make up out of all cause outpatient visits and hospital admissions 53, 54 . For surveillance purposes in a well-defined cohort (SMNSs), the SPR/TPR can provide a rapid and inexpensive method of estimating temporal changes in malaria incidence after a flood 27, 49, 53 . They are less affected by change in reporting rates, diagnostic practices and healthcare facility utilization rate and consider only laboratory confirmed cases of malaria 27, 54 . Like SPR/TPR, the proportions of outpatient and inpatient malaria cases are less sensitive to changes in reporting rates and healthcare facility utilization rate and they can reflect the burden that malaria places on the healthcare system after the flood 49, 54 . Change in the proportions of outpatient and inpatient malaria cases as well as SPR/TPR can reflect relative change in the malaria incidence over time, they however cannot estimate the actual incidence of malaria in a target population at national or regional level 54, 55 . Identifying flood-related changes in malaria incidence needs the access to reliable baseline data. An effective surveillance system is essential for providing the baseline data in order to identify flood-related changes in malaria incidence 56 .The malaria morbidity indicators in the study were extracted from surveillance and routine health information systems of the Gezira State Ministry of Health. As the analysis has been confined to SMNSs that recorded and reported consistently over time, the size of the catchment population and coverage of service utilization were stable over the evaluation period, as only laboratory-confirmed malaria cases were included in the study and no shortage in supply of microscopy supplies or RDTs in the SMNSs were observed 36 , our health facility-based data may provide a reliable indication of the malaria incidence rates over time 53, 54 .

The 2013 flooding in Almanagil Locality was associated with a malaria epidemic resulting in a marked and concurrent increase in the malaria incidence rate, SPR and proportion of outpatient visits and admissions in public healthcare facilities in comparison with the same period in the two previous years without floods. Age group analysis demonstrated that young children (< 5 years) are at the greatest risk from malaria in the post-flooding period because of an immature immune system and possibly under- or malnutrition, consistent with previous published reports 16, 57, 58, 59, 60, 61, 62 . In this study, the age-stratified hospital data showed that the post-flood increase in the proportion of uncomplicated and severe malaria cases in young children is mainly due to a large increase in the under-1-year age group.

Malaria epidemics following flooding are a well-known phenomenon in endemic areas worldwide 19, 29, 49, 50, 51 . Most studies attribute the increase of malaria incidence to rainfall and floods resulting in the formation of new breeding sites for the Anopheles mosquitoes and in providing favourable conditions for the mosquito development and survival, particularly high humidity 9, 10, 11, 19, 20, 21 . This is what most probably occurred in Almanagil Locality after the 2013 flood, given the presence of several factors that favour development of standing water pools and subsequent rise in the mosquito population.

Besides changes in recording completeness, healthcare utilization and diagnostic accuracy, other potential contextual factors that can confound the true malaria incidence rate include IRS, ITNs, Intermittent Presumptive Treatment of Pregnant Women (IPTp), and case management coverage 53 . IRS was implemented annually since 2000 in Gezira State. According to the NMCP, IRS in the irrigation schemes in Gezira State was consistent with the national standard method of implementation. The percentage of households covered in IRS campaigns was 97.7 % and the percentage of population protected was 98.9 % in Almanagil Locality in 2012 36 . According to the malaria indicators survey (MIS) 2009 and 2012, the percentage of households in Gezira State with at least one ITN decreased from 48 % in 2009 to 34.3 % in 2012 and the percentage of household members who slept under ITN decreased from 13.7 % in 2009 to 5 % in 2012 36. Given that the NMCP started a distribution of ITNs for target risk groups in Gezira State in 2012, it is unlikely that the use of ITNs would decrease further in 2013 36 . Since 2010 the IPTp was no longer in the national strategy and was limited to effective case management and prevention through ITNs 36 . Malaria case management guidelines and coverage in the public health facilities were unchanged over the observation period 36 . In summary, since no major changes in the possible confounding factors occurred, this study provides a strong evidence that flooding can lead to a significant increase in malaria incidence based on malaria data collected in public healthcare facilities.

Limitations

The estimation of malaria morbidity based on healthcare facility-related data may be too low or too high due to the potential weakness of the quality of data used to measure the malaria incidence, and hence should be analysed and interpreted with caution 53, 54 . Possible causes include: reporting completeness, accuracy of confirmed malaria cases by microscopy or RDTs, asymptomatic and subpatent malaria infections, the extent to which patients seek treatment in public and private healthcare services or are treated at home 53, 54, 63 . In order to minimize the influence of these sources of error and bias we used the strategy recommended by the World Health Organization: focusing on confirmed malaria cases, monitoring trends in SPR/TPR (microscopic or RDTs) and monitoring malaria outpatient visits and hospital admissions. Accuracy and completeness of data could not be fully verified as retrospective data were used. The surveillance sentinel sites of the public health centres and hospitals reported consistently over time. However, a spot check of reporting completeness of Almanagil Locality health centres not included in the SMNSs showed that data were not consistently reported to the Gezira State Ministry of Health. Although the private health sector is limited in Almanagil Locality, data from the private health sector were not available during the study period. This could lead to an underestimation of malaria incidence.

Conclusion

Overall, these results suggest that flooding following heavy precipitation has great potential to increase malaria burden in the affected population. It appears that the increase in malaria transmission occurs in the recovery phase of the flood disaster after a lag period of approximately 4 to 8 weeks. This initial delay between the flood and the post-flood malaria outbreak may provide the opportunity for vector control measures (IRS, ITNs/ Long-Lasting Insecticidal Nets [LLINs], larvicidal programmes) together with early case detection and management to mitigate the post-flood epidemic.

Corresponding Author

Yasir E A Elsanousi, MBBCh, DTM&H, MeHM, EMDM. Email: yasir3@yahoo.com

Competing Interest Statement

The authors have declared that no competing interests exist.

Data Availability Statement

All relevant data are within the manuscript and the public repository Figshare at https://figshare.com/articles/S1_Dataset_rar/5458135. The DOI is: 10.6084/m9.figshare.5458135. For more information, please contact the corresponding author: Yasir E A Elsanousi, yasir3@yahoo.com.

List of abbreviations

BSE, Blood Smears Examined

CI, Confidence Interval

FMOH, Sudanese Federal Ministry of Health

IPTp, Intermittent Presumptive Treatment of Pregnant Women

IR, Incidence Rate

IRS, Indoor-household Residual Spraying

ITNs, Insecticide Treated Nets

LLINs, Long-Lasting Insecticidal Nets

MIS, Malaria Indicators Survey

MOH, Ministry of Health

NMCP, National Malaria Control Program

OPD, Outpatient Department

RDT, Rapid Diagnostic Test

SPR, Slide Positivity Rate

SMNS, Sentinel Malaria Notification Site

TPR, Test Positivity Rate

UNDP, United Nations Development Program

WHO, World Health Organization

]]>
http://currents.plos.org/disasters/article/impact-of-the-2013-floods-on-the-incidence-of-malaria-in-almanagil-locality-gezira-state-sudan/feed/ 0
Behavioral, Normative and Control Beliefs about Earthquake Preparedness: A Deductive Content Analysis Study http://currents.plos.org/disasters/article/behavioral-normative-and-control-beliefs-about-earthquake-preparedness-a-deductive-content-analysis-study/ http://currents.plos.org/disasters/article/behavioral-normative-and-control-beliefs-about-earthquake-preparedness-a-deductive-content-analysis-study/#respond Fri, 28 Sep 2018 12:35:52 +0000 http://currents.plos.org/disasters/?post_type=article&p=32519 Introduction: Despite efforts to increase earthquake preparedness (EP), the level of earthquake preparedness in Tehran is low, even when people acknowledge the risk they face. This problem has its roots in the beliefs of Tehran inhabitants about EP. This study is aimed to elicit the salient beliefs about earthquake preparedness among Tehran citizens.

Method: This is a deductive content analysis research. The theory of planned behavior (TPB) has been applied as the theoretical framework of the study. 132 semi-structured interviews have been conducted with Tehran heads of households and the obtained data have been analyzed.

Results: The interviews showed that the belief in the usefulness of the EP and the belief that “the EP can cause anxiety among family members” were the salient behavioral beliefs (the ones influencing the attitude towards a behavior). The main normative belief (which influences the subjective norms), was “my family doesn’t disagree with the EP” although most of the interviewees did not know about their families’ views. Finally, the main control belief (which is the basis of perceived behavioral control), was that “education can facitilates EP”.

Conclusion: Tehran inhabitants preparedness behaviors can be influenced by their behavioral, normative and control beliefs about preparedness. Recognition of these beliefs may assist policy makers and executives to develop a better understanding of the origins of the preparedness behaviors. Any interventions to change these behaviors should be made based on such knowledge.

Key words: Earthquake; preparedness; salient beliefs; theory of planned behavior

]]>

Introduction

Tehran, the capital of Iran, is located at the foot of the Alborz Mountains Range which is a part of the Alpine-Himalayan Orogenic Belt. The city has great seismic potential with its numerous active faults. Tehran has not experienced a severe earthquake in the past 150 years, but a number of massive earthquakes has been recorded in its history1,2,3. Apart from its geological position, Tehran’s rapid population growth of over 12 million people has made it more vulnerable to earthquakes4. According to the studies, Tehran’s probable earthquake might result in serious losses and damages5. Nevertheless, earthquake preparedness of Tehran inhabitants is extremely low6. Of course, this is not the case merely in Tehran. The earthquake preparedness level in some of the cities prone to earthquake in some countries also is low7,8 while more or less the people are aware of the risk. Thus, people’s knowledge of the risk does not necessarily result in the people’s preparedness9.

According to the studies conducted on disaster preparedness, several factors can affect the preparedness including risk perception10,11,12,13 preparedness perception14,15,16 critical awareness12,17,18 optimistic and normalization biases19,20 self-efficacy14,21,22,23 collective efficacy24 fatalism13,22,25,26,27 locus of control13,23,27 anxiety7,12,27 previous disaster experience11,13,28,29 societal norms30 sense of community31 community participation and empowerment32,33 social trust34 perceived responsibility11,15 responsibility towards others18 coping style14,21,35,36 and available resources32,37.

Few studies have tried to identify the capability of people’s beliefs in influencing the earthquake preparedness behaviors. The main problem in most of these studies is a lack of theoretical underpinning. Since applying a theory to develop evidence-based knowledge is a requirement, this study makes an effort to apply a suitable theory to identify the people’s beliefs which may influence the preparedness behaviors. Understanding of belief-preparedness relationships can facilitate the development of preparedness behaviors in community members. To explain the earthquake preparedness behaviors, various theories and models can be employed including: Theory of Planned Behavior (TPB)38,39,40,41 Protection Motivation Theory (PMT)16,42 Person Relative to Event Theory (PrE)21,43,44,45,46 Protective Action Decision Model (PADM)47,48 and Social-Cognitive Preparation Model12 (Here, only the most used relevant models and theories are cited. There are other models and theories in the social psychology domain that are not cited here). Among these, the theory of planned behavior (TPB), as far as it is concerned, has not been applied to explain the earthquake preparedness behaviors49. The TPB is a belief-based theory. As we wanted to find the beliefs concerning earthquake preparedness behaviors, the TPB has been applied as the theoretical framework of the study. The aim of this study was to elicit the salient beliefs relevant to earthquake preparedness among Tehran inhabitants.

According to Ajzen, who has theorized TPB, “human behavior is guided by three kinds of considerations: beliefs about the likely outcomes of the behavior and the evaluations of these outcomes (behavioral beliefs), beliefs about the normative expectations of others and motivation to comply with these expectations (normative beliefs), and beliefs about the presence of factors that may facilitate or impede performance of the behavior and the perceived power of these factors (control beliefs). In their respective aggregates, behavioral beliefs produce a favorable or unfavorable attitude toward the behavior; normative beliefs result in perceived social pressure or subjective norm; and control beliefs give rise to perceived behavioral control. In combination, attitude toward the behavior, subjective norm, and perception of behavioral control lead to the formation of a behavioral intention. As a general rule, the more favorable the attitude and subjective norm, and the greater the perceived control, the stronger should be the person’s intention to perform the behavior in question. Finally, given a sufficient degree of actual control over the behavior, people are expected to carry out their intentions when the opportunity arises”50. Figure 1 schematically illustrates the theory of planned behavior51. In order to properly plan for disaster situations, it is vital for policymakers and emergency responders to understand the attitudes, concerns, and reactions of individuals and families caught in a disaster. Moreover, this information can guide the development of disaster preparedness educational materials or programs for the general public52. TPB can provide a framework for these activities.

Theory of planned behavior

Fig. 1: The theory of planned behaviour.

Methods

This is a deductive content analysis research. The deductive approach is based on previously theoretically derived categories and the initial coding starts with a theory. Based on this theory, the researchers embark on identifying the key concepts as the initial coding categories of the analysis, bringing them in connection with the text53. This explorative study is meant to elicit the commonly held beliefs and identify the content of behavioral, normative and control beliefs that are shared by the target population about earthquake preparedness based on the TPB. The data have been collected through interviews with Tehran citizens.

Sampling

Tehran consists of 22 districts and 134 sub-districts. 132 households of different socioeconomic backgrounds were selected from 22 districts of the city. First, three sub-districts were selected randomly in each district. Second, the two households were purposefully selected following a ‘maximum variation sampling’ approach within each selected sub-district. Finally, the heads of the households were selected as respondents, since they are considered to be the main decision makers. More specifically, the research team contacted the sub-districts council to support this study. The members working in the respective sub-districts council were asked as key informants to select the households in each socioeconomic group. The requirement for selecting the participating households was that they had stayed in Tehran for at least 10 years. Table 1 shows the participants’ characteristics.

Table 1. Participants’ Characteristics (N=132)
Percent Mean SD
Male 83
Age 44.34 12.41
Married 87.9
Years of education 12.5 3.7
Employed 75
Years of living in Tehran 25.7 5.2
Had home ownership 51
Families with more than two members 86.4
Had disaster experience 25
Prepared (DPBscore ≥ 5) 10.2

Materials

Since the TPB is especially applied to behaviors, the behavior elements of the Public Readiness Index (PRI)54were adopted which is scored on a scale from 0 to 7, for defining the earthquake preparedness behavior (Table 2).

Table 2. The behavior elements of the Public Readiness Index (PRI)
Question
1 Have you actually prepared a disaster supply kit with emergency supplies like water, food and medicine that is kept in a designated place in your home?
2 Have you actually prepared a small kit with emergency supplies that you keep at home, in your car or where you work to take with you if you had to leave quickly?
3 Have you actually made a specific plan for how you and your family would communicate in an emergency situation if you were separated?
4 Have you actually established a specific meeting place to reunite in the event you and your family cannot return home or are evacuated?
5 Have you actually practiced or drilled on what to do in an emergency at home?
6 Have you actually volunteered to help prepare for or respond to a major emergency?
7 Have you actually taken first aid training such as CPR in the past five years?

Moreover, a semi-structured interview approach was adopted to explore the participants’ beliefs about the earthquake preparedness. Although the interviewees could freely talk about their beliefs on the earthquake preparedness, a set of questions based on the TPB had to be answered too55 . If needed, the participants were consulted about the questions. The questions were already designed by an expert panel (Table 3).

Table 3. Interview Questions
Question
What do you think about the advantages of earthquake preparedness?
What do you think about the disadvantages of earthquake preparedness?
Is there anything else you would like to say about the advantages and disadvantages of earthquake preparedness?
What does your family think about earthquake preparedness?
What do your friends think about earthquake preparedness?
Is there anything else you would like to say about other persons’ ideas regarding earthquake preparedness?
What factors or circumstances can contribute to your earthquake preparedness?
What factors or circumstances make the earthquake preparedness difficult or impossible for you?
Are there any other issues striking to your mind regarding the earthquake preparedness?

Procedure

The study was conducted from January to February 2015. Two preliminary interviews were conducted to test the quality of the questions. The interview guideline was developed based on previous studies56 and expert opinions. The participants were interviewed by trained interviewers in their homes. They were asked whether they would participate in a study on “peoples’ attitudes towards earthquake preparedness”. Written consents were received from the participants and no identifying data was collected. All interviews were conducted in persian language. They were recorded electronically and transcribed verbatim. Each interview took 30 minutes in average. The study was approved by the Research Ethics Committee of Tehran University of Medical Sciences.

Analysis

A deductive content analysis was used to analyze the data. First, a structured categorization matrix based on the TPB was developed. Then, all the data were reviewed for content and coded for correspondence with the identified categories57. To increase the validity of the analysis, the data coding was done by two researchers independently. The data were analyzed based on main concepts which have been extracted from the theory.The data collection continued until the study reached saturation, in which no new information was provided. The MAXQDA software was used to process, order and compare the codes.

Results

The data extracted from the interviews were divided into three main categories: behavioral beliefs, normative beliefs and control beliefs. Behavioral beliefs included the advantages and disadvantages of the expected behaviors in earthquake preparedness. Normative beliefs were related to the approval or disapproval of the behaviors by the people whose views are important for the interviewees. Control beliefs were facilitating factors or barriers of the earthquake preparedness behavior from the interviewees’ point of view. The content of behavioral, normative and control beliefs was identified from the transcripts by two researchers who were familiar with TPB (Table 4).

Table 4. Codes of behavioral, normative and control beliefs that extracted from the data
Codes Quotation Count
Behavioral beliefs Preparedness reduces casualties and losses of my family 80
Preparedness causes anxiety in my family 55
Preparedness makes my family members aware of what to do in an earthquake 40
Preparedness is useless 30
Preparedness protects my family as long as the relief workers have not arrived. 20
Preparedness helps my family members find each other after the disaster 20
Preparedness helps my family prove their identity and their ownership of the properties and assets 15
Normative beliefs My family doesn’t disagree with the earthquake preparedness 55
My family approves the earthquake preparedness 40
My friends approve the earthquake preparedness 30
Some of my friends approve and some of them disapprove the earthquake preparedness 30
My family and those around me have no opinion on the earthquake preparedness 20
Control beliefs Educating makes it easier to prepare my family for the earthquake 122
Daily involvements are obstacles for preparing my family 55
Indolence and nonchalance are obstacles for preparing my family 40
Financial restrictions are obstacles for preparing my family 40
Earthquake is an act of God and I cannot influence its consequences by earthquake preparedness 20
Living in the apartment is an obstacle for preparing my family 15
My family’s preparedness depends on the other families’ preparedness 15
There is no barrier for preparing my family 10

Behavioral beliefs

Although 80.8% of the interviewees were not prepared for earthquakes, they believed that earthquake preparedness is useful. When they were asked to comment on the advantages of earthquakes preparedness, they often pointed out the reduction in casualties and damages. One of the interviewees stated: “the most important advantage of readiness is that our lives would be less in danger. Besides, if we are prepared, our financial losses will be reduced.”

Some of the interviewees believed that the earthquake preparedness might minimize the confusion in case of an earthquake. One of them stated: “if we get prepared, we are aware what to do in an earthquake.”

The statements such as those mentioned do not directly highlight the ultimate advantage of the earthquake preparedness, but they evidently show that the interviewees expected positive consequences from such preparedness. Some people mentioned some particular issues in this case such as protecting the family until relief workers arrive, finding the family members after the earthquake, the proof of identity, the proof of ownership of properties and assets and the like. On the other hand, for various reasons, some of the interviewees thought that the earthquake preparedness was useless. The most important reason behind such a claim is thinking about the worst possible scenario. One of the interviewees said: “In the current situation of Tehran, how would preparedness be beneficial? If a strong earthquake occurs, Tehran will be reduced to rubble. I do not think anyone could do anything.”

Thinking about the worst possible scenario, namely reducing into rubble, was common among almost all of the participants. Even those who believed earthquake preparedness was useful felt doubt about it. One of the interviewees, a professor of geomorphology, admitting the preparedness advantage, stated: “when our house is ruined, how could we access the tools and facilities prepared? So, it is not clear that the providing of these things would be advantageous.”

There were few people who believed that although earthquake preparedness was good, there was no point in preparing for something which might never happen. One of the interviewees stated: “It is so good to get prepared, but why should one get prepared for something which might never happen or for something which causes so extensive damages that it makes preparedness useless.”

In this statement, the interviewee made reference to two of the widespread beliefs about earthquakes: first, “earthquake might never happen, so preparedness is not necessary”, and second: “if a devastating earthquake happens, it will ruin everything including what we have prepared”. In fact, the first belief refers to “no need to the preparedness because of the earthquake uncertainty” and the second one refers to “the preparedness uselessness in the case of the worst possible earthquake scenario”. Believing that “earthquake preparedness is not necessary” is about the preparedness. This is not a behavioral one, since it does not refer to any positive or negative expectations from the earthquake preparedness. Then, “the earthquake preparedness is not useful” can be considered as a behavioral belief. Whereas the main benefit of the earthquake preparedness is considered as reducing the casualties and damages, this belief would mean “the earthquake preparedness does not reduce the casualties and damages”.

Some of the participants believed that the earthquake preparedness would cause anxiety among family members. One of them stated: “One of its disadvantages is that it provokes anxiety in their family over the earthquake. This can highly affect the living very much and may lead to an excessive stress.”

This clearly refers to the negative expectation about the preparedness. In other words, this can be categorized under behavioral beliefs. No other case was found in the data concerning the preparedness disadvantages.

Normative beliefs

Most of the interviewees didn’t know the ideas of their family and friends about the earthquake preparation. It seems that it was not so important for them to be asked about it. One of them stated: “I don’t know the views of my family and friends, because we do not talk about these things at all.”

A lack of knowledge about those whose views are important to the interviewees is a key finding of the current study. However, most of these people thought their family members and friends do not disagree with the earthquake preparedness. Nevertheless, “not disagree” cannot quite be equivalent to “agree”. This normative belief more or less means “the people around me neither agree nor disagree with the preparedness”. Even those claiming that their families approved earthquake preparedness , had not frankly asked about the families’ opinion. It was evident that the interviewees did not feel a sense of pressure on the part of their acquaintances in this regard. One of them pointed out: “Most of my family members and friends do not even think of it. I can say that almost all of them are indifferent to it.”

This in fact implies a normative belief which can be formulated as follows: “the people around me have no opinion about the earthquake preparedness”. Although this belief does not disapprove the preparedness, therein no point in approving it.

Control beliefs

Most of the interviewees believed that “the public education” could help their families for earthquake preparedness. Most of them believed that informing and educating (especially via radio and television) could facilitate the earthquake preparedness. This statement is an equivalent to the control belief that “if I am informed, it would be easier for me to prepare my family for earthquakes”.

Most of the interviewees believed that “daily involvements” did not leave any room to get involved in the earthquake preparedness. One of the participants remarked: “The peoples are so busy, so they have no time to think about these things.”

“Indolence and nonchalance” were also believed to be obstacles to preparedness. This belief clearly shows an internal barrier influencing the person’s behavior. One of the interviewees stated: “I think the main obstacle is indolence and nonchalance not restricted to this matter. An indolent nonchalant person deals with everything in the same way.”

“Financial constraint” was reported by some of the interviewees as a barrier. They believed that preparedness incurred expenses which might be difficult to bear. One of them remarked: “preparedness incurs some expenses. With our low incomes, it is not even at the bottom of our to-do list.”

The other interviewee believed that it was not a family responsibility to pay for the preparedness. He pointed out: “if the lives of the people are really worthy for the government, it should provide us with preparedness packages for free or with very low prices.”

This statement means that the government is responsible for the family’s preparedness, because families have extra expenses more critical than the preparedness costs.

Some of the interviewees believed that if it is “destined” to die, it happens, no matter to be prepared or not. One of them remarked: “if it is destined to die, it happens. Earthquake or the like is a cover. It is not a matter of getting prepared or not.”

Most people who believed in preparedness also believed in “the destiny”. One of them remarked: “we should try our best, leave the rest to God. We cannot control everything.”

The statements like these imply the control belief “earthquake is an act of God and I cannot influence its consequences by the earthquake preparedness”. It seemed that the people with such beliefs regard the main source of control somewhere out of reach.

Some of the participants thought they could not be prepared in the “apartments”. This mental image is rooted in the belief that preparedness kits should be kept in a safe place, while in an apartment completely destroyed in an earthquake, it is not possible. This belief implied again the worst possible scenario. One of the interviewees believed: “there is no safe place in an apartment for keeping the preparedness kits. If we had a house, we could keep the equipments in the stockroom in the corner of the courtyard. Now in the absence of such a house, we cannot prepare for an earthquake even if we desire.”

These statements are indicative of the control belief “living in an apartment is an obstacle for preparing my family”.

The beliefs such as “the community readiness leads to my family’s preparedness” and “the whole building preparedness is a necessity for my family preparedness” are equivalent to the control belief “my family’s preparedness depends on the other families’”. Few people believed that there was no barrier for preparing their families.

The interviews clearly showed that there were some salient behavioral, normative and control beliefs about the earthquake preparedness which can lead to the participants’ relevant behaviors. Namely, some of these beliefs may be motivating and some non-motivating in the process of preparation. Table 5 summarizes the salient beliefs and their probable implications.

Table 5. Behavioral, normative and control beliefs and their probable motivating and non-motivating implications for the earthquake preparedness
Motivating Non-motivating
Behavioral beliefs It reduces the casualties and damages It causes anxiety
It helps people not be confused It is useless
It manages families until the relief workers arrive
It helps to organize families after the disaster
Normative beliefs The family and friends approve it The family and friends are indifferent to it
Some of the family members and friends disapprove it
Control beliefs Training facilitates it Daily involvements are obstacles
There is no obstacles Indolence and nonchalance are obstacles
Financial constraints make it difficult
It is not destined by human’s will
Living in an apartment makes it difficult
The family preparedness is dependent on the community preparedness

Discussion

Few studies have tried to identify the belief-behavior relationships in the earthquake preparedness. In the study by Becker and his colleagues58, “the positive outcome expectancy” restated as the general phrase “preparing helps me in a disaster” is similar to the belief in our investigation: “the earthquake preparedness is useful for my family”. Its general form may show that the positive expectations are not that clear to the people. To state the matter differently, some people are doubtful about the benefits of earthquake preparedness, because of their beliefs about the earthquake hazard and their scant knowledge, mainly via mass media, about Tehran’s structural vulnerability. It means that the behavioral belief of “uselessness of preparation” is possibly formed by their exaggerated vulnerability assessment of the town structures and the false beliefs about the hazard. However, it is important to distinguish between the “usefulness” of preparedness in general and its “helpfulness” in particular. It seems that the interviewees who believed that preparedness was useful, but doubted about its helpfulness in Tehran, detected such a difference. At the present stage, this is only a speculation that needs further investigation. Moreover, emotions, feelings and social norms are likely to influence shaping these beliefs58.

Taking the worst possible scenario, stated or implied in many of the interviews, may be associated with their unrealistic assessment of the earthquake and its consequences. Two beliefs seriously challenging the behavioral belief about the preparedness effectiveness are: 1) in earthquakes, all buildings collapse and 2) it is not possible to access preparedness kits when an earthquake happens. These beliefs have made some interviewees believe that “the earthquake preparedness in apartments is useless”. The assumption behind such a belief is that earthquake destroys all of the apartments. In the previous studies, the implications of the house ownership for preparedness have been mentioned59, while the effect of house type on preparation is probably a new finding not included in those studies.

The behavioral belief that “actions for the earthquake preparedness can create anxiety among family members” is apparently a new finding too. In the previous studies, the source of anxiety was the hazard itself, not the hazard preparedness12,60.

The lack of information about family members’ views shows the absence of any talks about the preparedness in the families. Some of the interviewees expressed this issue explicitly. “Not talking” about the earthquake preparedness in the families could be considered as “indifference”. This normative belief can be related to the technical term “critical awareness”61 which in this particular case is the amount of thinking or talking about earthquakes and its risks12.

In the previous studies in Tehran, public education, especially the role of television in informing, was highlighted62. This emphasis indicates that the people believed that their lack of preparedness is a result of their lack of knowledge. This control belief is also considered by some people from other countries63. It seems that the people’s emphasis on educating specially via television programs can be interpreted as shifting responsibility towards formal institutions. In other words, although “the public education via television facilitates getting prepared for people” may be a true belief, it would show that how people justify themselves not being prepared12. In addition, too much emphasis on informing and educating can be misleading, since it may be thought that people who know how to reduce the risk and its consequences inevitably prepare themselves. This assumption was not supported by several studies14,36,64 . On the contrary, informing and educating may sometimes cause a decrease in the perception of risk and preparedness level12 or at least they would not have a significant effect on the risk perception6. The control belief “the daily involvements are obstacles for preparedness” is often regarded as “lack of time”. Tehran inhabitants, like the people from some other countries65 , considered this factor as a barrier to preparedness. Lack of time for a particular thing usually means that thing is not of priority. This is clearly expressed by some of the interviewees in this study. This belief may be formed due to the people’s poor perception of risk. As in the previous studies, including a study conducted in Tehran6 showed that those with a greater perception of risk were more prepared. “Indolence and nonchalance” might be associated with the poor perception of risk. On the other hand, this belief may show the role of personality characteristics in the disaster preparedness. More researches are needed to be conducted on this issue.

Financial constraints usually expressed as “not having enough money”. It is a control belief emphasized by the people in similar researches63. The control belief “I have not enough money, so I cannot prepare my family” also shifts the responsibility from the families towards the government. This finding is consistent with the findings of the previous researches11,14.

Fatalism which almost in all of the cases was rephrased as the religious belief in “destiny” implies the concept that disaster and its consequences are “being out of control”. Even those considering the earthquake preparedness as necessary, accepted the role of an element not controlled by the willpower of human. Some of the studies have shown that fatalism can reduce the level of earthquake preparedness22. Although fatalism has been seen more in people with lower incomes63, it seems this control belief relates to the religious beliefs, whatever they would be, more than any other factors67.

Limitation

The main limitation of the study is that, even though data saturation was reached, since these beliefs are expressed by those who are not prepared for the earthquake, the researchers cannot fully make sure that the same results would be yielded among the more prepared groups. Therefore, this study needs to be replicated to see whether the current results are found elsewhere. If so, a follow-up quantitative study is needed to explore the actual percentage of the people holding these beliefs. Another limitation was sampling procedures used in this research. A purposive sample is not representative, although a maximum variation sample aims to be in certain situations, like in this study, more representative than a random sample.

Conclusion

In the present study, the salient behavioral, normative and control beliefs of Tehran inhabitants about the earthquake preparedness are defined and discussed based on TPB. The fidings are indicative of the fact that the Tehranis’ preparedness behaviors can be influenced by these beliefs. Recognition of these beliefs may assist policy makers and executives to develop a better understanding of the origins of the attitudes, the subjective norms, and the perceived barriers to the preparedness. In the other words, with such an understanding, they may determine the factors that influence the public preparedness behaviors. Any interventions to change these behaviors should be made based on this knowledge.

Corresponding Author

Mehdi Najafi is the corresponding author and can be contacted via najafirc@gmail.com.

Competing Interest

The authors have declared that no competing interests exist.

Source of funding

The authors received no specific funding for this work.

Data Availability

All relevant data are reported in the paper.

]]>
http://currents.plos.org/disasters/article/behavioral-normative-and-control-beliefs-about-earthquake-preparedness-a-deductive-content-analysis-study/feed/ 0
The Impact of Hurricane Sandy on HIV Testing Rates: An Interrupted Time Series Analysis, January 1, 2011‒December 31, 2013 http://currents.plos.org/disasters/article/the-impact-of-hurricane-sandy-on-hiv-testing-rates-an-interrupted-time-series-analysis-january-1-2011%e2%80%92december-31-2013/ http://currents.plos.org/disasters/article/the-impact-of-hurricane-sandy-on-hiv-testing-rates-an-interrupted-time-series-analysis-january-1-2011%e2%80%92december-31-2013/#respond Thu, 13 Sep 2018 16:08:09 +0000 http://currents.plos.org/disasters/?post_type=article&p=40431 Background: Hurricane Sandy made landfall on the eastern coast of the United States on October 29, 2012 resulting in 117 deaths and 71.4 billion dollars in damage. Persons with undiagnosed HIV infection might experience delays in diagnosis testing, status confirmation, or access to care due to service disruption in storm-affected areas. The objective of this study is to describe the impact of Hurricane Sandy on HIV testing rates in affected areas and estimate the magnitude and duration of disruption in HIV testing associated with storm damage intensity.

Methods: Using MarketScan data from January 2011‒December 2013, this study examined weekly time series of HIV testing rates among privately insured enrollees not previously diagnosed with HIV; 95 weeks pre- and 58 weeks post-storm. Interrupted time series (ITS) analyses were estimated by storm impact rank (using FEMA’s Final Impact Rank mapped to Core Based Statistical Areas) to determine the extent that Hurricane Sandy affected weekly rates of HIV testing immediately and the duration of that effect after the storm.

Results: HIV testing rates declined significantly across storm impact rank areas. The mean decline in rates detected ranged between -5% (95% CI: -9.3, -1.5) in low impact areas and -24% (95% CI: -28.5, -18.9) in very high impact areas. We estimated at least 9,736 (95% CI: 7,540, 11,925) testing opportunities were missed among privately insured persons following Hurricane Sandy. Testing rates returned to baseline in low impact areas by 6 weeks post event (December 9, 2012); by 15 weeks post event (February 10, 2013) in moderate impact areas; and by 17 weeks after the event (February 24, 2013) in high and very high impact areas.

Conclusions: Hurricane Sandy resulted in a detectable and immediate decline in HIV testing rates across storm-affected areas. Greater storm damage was associated with greater magnitude and duration of testing disruption. Disruption of basic health services, like HIV testing and treatment, following large natural and man-made disasters is a public health concern.  Disruption in testing services availability for any length of time is detrimental to the efforts of the current HIV prevention model, where status confirmation is essential to control disease spread.           

]]>

Introduction

The devastation caused by Hurricane Sandy in late October of 2012 resulted in an estimated 117 deaths and over $70 billion in physical damage1, 2. Three hundred and seventy-five thousand housing units were destroyed in New Jersey and New York and 8.5 million people lost power across 21 states. More than 1,000 patients were evacuated from metro-area hospitals due to unsafe conditions such as untreated waste; 2.75 billion gallons flowed from treatment plants and contaminated the public water systems3, 4. The storm destroyed homes, displaced loved ones, and deprived many individuals of essential necessities. Vulnerable populations, or groups with medical or other functional and access needs, were disproportionally impacted and were particularly at risk for being exposed to post-disaster distress. Individuals infected with Human Immunodeficiency Virus (HIV) and their contacts are disproportionally affected during storm impacts as they lose access to testing and treatment due to disruptions in healthcare services.

The Centers for Disease Control and Prevention (CDC) recommends everyone ages 13 to 64 be tested for HIV at least once as part of routine health checks, and those with increased risk factors should get tested more frequently5. CDC also recommends healthcare providers test individuals who engage in high risk behaviors for acquiring and transmitting HIV1. However, the ability to follow recommendations and reach a local testing site might be impeded for persons who experience dislocation, damaged homes, and economic losses immediately after a disaster. The breakdown in security, damage to infrastructure, building loss and displacement of medical personnel associated with the hurricane could also result in inaccessibility of HIV testing and healthcare services. Disruptions in diagnostic testing, access to preventive measures like condoms and antiretroviral therapies, might be unavailable during major disasters contributing to increased risky behaviors and potential HIV transmission6. In addition, not knowing one’s status might contribute to the likelihood of spreading the disease through unsafe sex or progression of the disease to Acquired Immunodeficiency Syndrome (AIDS) and other associated opportunistic infections5.

According to the National Health Interview Survey, as of 2010, only 45% of people in the U.S. ages 18 to 64 were tested for HIV7. Treatment success requires early diagnosis, achievable through testing as it is estimated that 13% of the 1 million people living with HIV are unaware of their infection7. The goal of the National AIDS Strategy is to increase the percentage of people living with HIV who know their serostatus to at least 90 percent8, 9. Natural or man-made disasters that disrupt HIV testing and treatment can be a barrier to reaching the goal set by the National AIDS Strategy.

Our study examined the impact of Hurricane Sandy on HIV testing in affected states by Storm Impact Rank (SIR). We used Interrupted Time Series (ITS) analyses to estimate the impact of Hurricane Sandy on HIV testing rates among privately insured individuals not previously diagnosed with HIV. We estimated, the immediate impact as well as the duration of impact and used these estimates to calculate the number of missed testing opportunities attributable to the storm.

Methods

Study Population

The study population was identified as privately insured enrollees at risk for HIV from January 1, 2011‒December 31, 2013 in the eastern region of the United States. Data extracted from the Truven Health MarketScan database was used to conduct these analyses of HIV testing rates; Truven Health MarketSan is a health insurance claims database compiled from enrollees with employer-sponsored private health insurance including employees, retirees, and dependents. Medicare-eligible retirees with employer-provided Medicare Supplemental plans and Consolidated Omnibus Budget Reconciliation Act (COBRA) enrollees are included in the MarketScan database. Data are representative of the United States’ privately insured population10. The database includes paid claims as well as detailed patient information. The specific databases within MarketScan used to obtain the target population for this study were the Commercial Claims and Encounters Database (CCAE); and Medicare Supplemental and Coordination of Benefits Database (MDCR)10. The data use agreements only allow data sharing if it is aggregated and fully de-identified as presented in this paper. No interviews were conducted and no informed consent was obtained as human subjects protection procedures were not required. Our research design was reviewed by the CDC Office of Public Health Emergency Preparedness and Response Human Subjects Protection Coordinator and found to be exempt, as this study does not constitute human subjects research, Human Subjects Research #2016083102.

Enrollment detail data were used to identify enrollees residing in the Hurricane Sandy Impact Area (HSIA). While media coverage focused on impacts in New York and New Jersey, the Federal Emergency Management Agency (FEMA) declared impact area included 21 states: Connecticut, Delaware, Indiana, Illinois, Kentucky, Maine, Maryland, Massachusetts, Michigan, New Hampshire, New Jersey, New York, North Carolina, Ohio, Pennsylvania, Rhode Island, Tennessee, Vermont, Virginia, Washington D.C. and West Virginia. All enrollees residing in the HSIA not identified as having an HIV diagnosis met study eligibility.

Data Sources and Measurement

The Federal Emergency Management Agency Modeling Task Force (FEMA MOTF) Hurricane Sandy impact analysis data included storm intensity data (precipitation, buildings lost, households exposed to the surge, etc.) by county. The FEMA MOTF data was compiled by experts in hazard assessment and derived from federal agencies, universities, national labs, state, and local entities. These data were used to develop estimates of impact after the event11. The four categories of Final Impact Rank were: a) low, b) moderate, c) high and d) very high. For purposes of this study, SIR was derived by cross-walking the final impact rank assigned to counties to Core Based Statistical Areas (CBSAs) and by calculating a median final impact rank for a given CBSA weighted by the county population estimate for 2012. Rural areas were excluded from this analysis because SIR could not be assigned accurately for combined rural areas in affected states. SIR designated from low to very high was indicated according to the following color scheme: 1) Green (low), 2) Yellow (moderate), 3) Red (high), and 4) Purple (very high) for the CBSAs (see Figure 1).

HIV Sandy Paper 1_Figure 1

HIV Testing Rates per 1,000 Enrollees

HIV testing rates per 1,000 enrollees were calculated as the number of enrollee claims for HIV tests per week divided by total enrollment without previous HIV diagnosis per week and multiplied by 1,000 for all CBSAs in the HSIA. The SIR resulting from analyses of FEMA final impact ranks was linked by CBSA to HIV testing rates. Weekly testing rate time series were then compiled by SIR.

Statistical Methods

Descriptive statistics of mean, median, standard deviation, and range were calculated by storm impact rank for HIV testing rates per 1,000 enrollees before and after the event date. There were 153 weeks in each time series with 95 weeks pre-event and 58 weeks post-event. All statistical analyses were conducted using the statistical software package SAS v9.3.

In order to estimate the impact of Hurricane Sandy on HIV testing rates, ITS models of weekly rates were estimated via maximum likelihood for each SIR. Note that we defined the “event” that interrupted the time series as the week of October 29, 2012, the date Hurricane Sandy made landfall in the eastern United States. ITS models were specified as linear segmented regression with autoregressive errors. Details on the methods employed are provided in the Technical Appendix.

ITS models produce estimates of the immediate impact of the storm on HIV testing rates as well as the impact of the storm on the trend in testing rates12, 13.The parameter estimates produced by the ITS models were used to calculate the percent change in testing rates attributable to the storm and associated confidence intervals immediately and at one, four, eight, and twelve weeks post-event. The estimated absolute change in HIV testing rates attributable to the storm was used to calculate the estimated absolute impact of the event in terms of missed tests, as well as the duration of weeks that testing rates for each SIR was below baseline. Return to baseline was defined as the week that the percent change in testing rate between actual and baseline was no longer statistically significant. (See Technical Appendix)

Results

Table 1 presents descriptive statistics for weekly HIV testing rates among privately insured enrollees in the HSIA by SIR and the mean HIV testing rates, pre-event (January 1, 2011 ─ October 27, 2012, 95 weeks) and post-event (October 28, 2012 ─ December 31, 2013, 58 weeks). HIV testing rates did not appear to differ greatly when simply considering pre- versus post-event. Figure 2 displays weekly testing rates by SIR during the study period from January 2011‒December 2013. Seasonal variation is evidenced by the periodic fluctuations in weekly testing rates as well as several downward spikes prior to Hurricane Sandy (identified in Figure 2 as a vertical reference line labeled “Event”). A positive relationship between the HSIA and HIV testing rates was apparent prior to the event. The standard deviations (and thereby variance) in testing rates do appear to increase post-event. Sharp downward spikes are observed in Figure 2 occurring at the time of the event. The magnitude of these spikes tend to increase with storm impact rank.

Estimated ITS effects, displayed in Table 2, revealed that HIV testing rates declined significantly in storm-affected areas at each SIR. Relative changes in HIV testing rates were largest immediately following the event, declining towards baseline over time. Similarly, the largest negative effect was observed in the highest SIR immediately following the event (-24%), while the smallest negative effect was observed in the lowest impact rank (-5%). Testing rates returned to baseline in low impact areas by six weeks post-event, December 9, 2012. In moderate impact areas, testing rates returned to baseline by 15 weeks post-event, February 10, 2013. In high and very high impact areas, testing rates returned to baseline by 17 weeks after the event, February 24, 2013. Estimated numbers of missed testing opportunities are displayed in Table 3. By the time testing returned to baseline across SIRs, 9,736 (95% CI: 7,540, 11,925) testing opportunities were missed. The largest amount of missed testing opportunities occurred in very high impact areas, accounting for 6,811 (95% CI: 5,644, 7,978) (see Technical Appendix for detailed results of the ITS regressions).

HIV Sandy Paper 1_Table 1

HIV Sandy Paper 1_Table 2

HIV Sandy Paper 1_Table 3

HIV Sandy Paper 1_Figure 2

Discussion

The results of the ITS analysis indicated an immediate decline in HIV testing rates after Hurricane Sandy among privately insured in storm affected areas. Greater disruption in HIV testing was noted in higher impact areas. The relative impact on HIV testing rates was most extreme immediately after the event, returning to baseline over time. Areas with greater storm impact had progressively longer periods of HIV testing disruption. While causality between Hurricane Sandy and declined HIV testing is not established unequivocally, storm impacted areas exhibited declined HIV testing rates and larger declines were associated with higher storm impact rank. We explored the relationship with a time series analysis in order to control for seasonality and autocorrelation and to isolate the independent effect of Hurricane Sandy on HIV testing rates.

Our findings supports earlier studies on the effects of Hurricane Katrina, indicating that areas affected by severe weather events are susceptible to disruption in access to healthcare14. Specifically, access to HIV testing and prevention healthcare services was interrupted until five months after the devastation occurred15. Severe weather-related disasters often disrupt important health maintenance and health care delivery services. This study found a similar disruptive effect on HIV testing, which became secondary to recovering from the storm and restoring the community.

HIV can persist for up to ten years before symptoms are detected, and when symptoms are revealed, it might be too late to avoid immune system break down and full-blown AIDS. Interruption in availability of HIV testing services could be costly in terms of inadvertent spread of the disease, increased risk of opportunistic infections, late diagnosis, and progression to AIDS among persons unaware of positive status. Shifting from the public health policies of the 1990s and early 2000s, which emphasized evidence based interventions and behavior modification education, the current prevention strategy is a “test and treat” model16. The test and treat model of HIV prevention emphasizes identification of HIV-infected persons early in the disease stage, timely initiation of care, and uninterrupted administration of antiretroviral medications to achieve viral load suppression sufficient enough to prevent secondary transmission16. Lack of availability of HIV testing eliminates the foundation of this prevention model.

Thus, these results have implications for public health practitioners in high HIV disease burden areas. State and local health jurisdictions should consider assessing the risk of vulnerable populations at risk for HIV infection and working with stakeholder partners to ensure sufficient staff and infrastructure could be maintained in advance of natural disasters to prevent disruptions in services. Moreover, public health practitioners should consider prioritizing restoration of HIV testing as soon as feasible in areas that receive the most damage after an event. In the event of a natural or a man-made disaster, public health and health care organizations serving populations at risk for HIV should have emergency response protocols to educate individuals at risk for HIV17. Looking ahead, professionals engaged in preparedness planning for events that might disrupt regular health services, especially those in high HIV burden areas, should be aware of the possibility of HIV testing disruption and have plans in place to avoid long-term gaps in availability of these vital services.

Strengths and Limitations

One strength of this study is the use of Interrupted time series (ITS), a robust modeling technique that controls for prior trends and seasonality while estimating the independent impact of an event12. ITS specified as segmented regression with autoregressive error is a very strong quasi-experimental research design that produced estimates of immediate as well as extended effects of the event in question12, 18. While one potential limitation of ITS is confounding by a co-occurring event, the authors did not identify any co-occurring event that might confound these estimated HIV testing rates.

Marketscan is a rich private health insurance claims database that is representative of the privately insured population in the U.S.10. Because county identifiers were not available, these analyses were limited to non-rural areas, specifically CBSAs. The storm impact ranks could not be accurately assigned to combined rural areas in each state in the absence of county identifiers. Thus, those who lived in rural areas, were uninsured, or were publically insured were excluded from this study. CBSA-specific effects were also not addressed in this analysis as ITS was conducted by storm impact rank. Therefore, these results might understate the disruption that actually occurred among uninsured, rural, and publically insured populations and ignores CBSA-specific effects.

Conclusions

Our results suggest that public health preparedness professionals in areas with high HIV disease prevalence should collaborate with HIV testing and treatment providers to enhance readiness in advance of natural or man-made disasters that could disrupt preventative healthcare services. Public health emergency preparedness professionals should consider reviewing their epidemiology data to determine if the populations they serve are at increased risk for HIV infection during an event and work with testing venues and treatment providers to establish testing and treatment provision plans. In high HIV burden areas, an emergency plan for continuity of HIV testing and results services should be a priority. This research supports emergency preparedness in states, local jurisdictions and communities at risk for HIV by highlighting a neglected concern, the potential of HIV testing service interruption during catastrophic events, and suggests advanced planning to resume availability of status determination. These findings can inform public health preparedness policies and practice during a large-scale emergency capable of disrupting normal health care access at the state, local, or community level due to high HIV/AIDS morbidity and disruption of health care services. This research can also inform pre-event planning for disease surveillance in advance of natural disasters to prevent long-lasting disruptions in HIV testing and help prioritize routine healthcare services in areas following a natural or manufactured disaster.

In the Hurricane Sandy impact area, about 70% of the population across all age groups, have private health insurance. Our results are representative of that population, since MarketScan is representative of the privately insured population across all age groups.

Because the approximately 30% of the population that does not have private insurance is typically less healthy19, we would expect HIV testing among that population to be more severely impacted. Specifically, HIV testing would experience a sharper decline immediately following Hurricane Sandy and HIV testing would take longer to return to baseline among this (the publically insured and uninsured) population. Future studies would include the effect of Hurricane Sandy on HIV testing rates in the publically insured and uninsured.

Technical Appendix

Statistical Methods: Interrupted Time Series

Claims for HIV tests were extracted from the Truven Health MarketScan database and compiled on a weekly basis for Core-Based Statistical Areas (CBSAs) from January 1, 2011‒December 31, 2013. Enrollees were considered to have a HIV diagnosis if claims were processed on their behalf in the preceding claims year and referenced the following ICD-9-CM codes: ‘042’, ‘0420’, ‘0421’, ‘0422’, ‘0429’, ‘07953’, ‘79571’, and ‘V08’20. Thus, enrollees with claims containing HIV diagnoses between January 1, 2010 and December 31, 2010 were excluded from the 2011 database. The remaining enrollees were retained in the reference population for 2011, the population at risk for HIV. Outpatient claims were examined to identify all enrollees without prior HIV diagnosis that had claims for HIV tests ordered using Current Procedural Terminology (CPT) codes (‘86689’, ‘86701’, ‘86703’, ‘87389’, ‘87390’, ‘87534’, and ‘87535’). These CPT codes have been widely used in previously published studies to identify claims for HIV tests performed on privately-insured enrollees to detect HIV and STIs21, 22, 23, 24, 25.

The ITS was modeled using the following linear segmented regression with autoregressive errors, which represents the baseline level and trend of the outcome variable before the event and changes in the level and trend after the event:

Here, Yt represents the dependent variable at a point in time, weekly HIV testing rates. Timet is the continuous variable representing time in weeks since the beginning of the study period. The intervention function was specified as a step function. Eventt is an indicator variable set to zero prior to the date of the event and becoming 1 the week of the event and for the duration of the time series after the event. Timeaftereventt is a continuous variable counting the days elapsed since the event at time t and set to zero prior to the event. The regression error, et, is comprised of a random error component as well as an autoregressive error component to adjust for autocorrelation and seasonality.

The estimated factual case is represented as:

The estimated counterfactual case is represented as:

Equation (2) was estimated for each storm impact rank in the Sandy Impact Area. The parameter estimates from equation (2) were used to calculate equation (3) for each storm impact rank.

Note that the parameter estimate β0^ is the baseline level of the outcome variable at time zero, or the intercept. β1^ is the estimated baseline trend of the outcome variable, or the weekly deviation from the baseline level prior to the event. ^β2 is the estimated absolute change in level, or intercept, of the HIV testing rate that occurs immediately following the event and ^β3 +^β4 is the estimated absolute change in trend that occurs after the event. The estimated relative, or percentage, change in the baseline level due to the event is (^β0 +^β2 )/^β0 , while the estimated relative change in trend is (^β1 +^β3 +^β4 )/^β1 . Thus, ^β0 +^β2 is the estimated post-event baseline of the outcome variable and ^β1 +^β3 +^β4 is the estimated post-event trend in the outcome variable. The estimated absolute impact of the event measured one week post-event is equation (2) minus equation (3) evaluated at t=97, or ^β1 +^β3 *1+(^β4 )*1) and the estimated relative impact of the event measured one week post-event is (^β2 +^β3 *1+(^β4 ) *1)/(^β0 +^β1 *97).

Thus, absolute impact is the difference between the estimated values given the event occurred and the estimated value given the event did not occur at a particular time post-event. Relative impact is expressed as the percentage change in HIV testing rates in the factual case compared to the counterfactual case. Each time series was tested for stationarity via the augmented Dickey-Fuller test and autocorrelation via generalized Durbin-Watson tests13. Autoregressive error models addressed autocorrelation and seasonality. Models were estimated in SAS using “proc autoreg” with autoregressive error terms identified via backward elimination12, 13, 18.

Estimated parameters and standard errors were used to produce estimates and associated 95% confidence intervals (CI) of the relative changes in level and trend as well as the estimated absolute impact of the event in terms of missed tests immediately and at one, four, eight, and twelve weeks post-event as well as at the time that each storm impact rank returned to baseline. Specifically, predicted trends for the factuals and counterfactuals with confidence intervals were retrieved from the regressions for each storm impact rank and used to calculate the relative change and impact at each time point post-event26.

Results: Table and Figure

Table A presents results from interrupted time series regressions by storm impact rank. Parameter estimates from these regressions were used to estimate the predicted factual and counterfactual trends as well as the confidence intervals by storm impact rank that are presented in Figure A. Note that while the estimated impact on the intercept is not always highly significant, the estimated impact on trend is significant for a quadratic trend at p<0.05 across storm impact rank. This produces ITS effects that are significant across storm impact rank.

Data Availability

All relevant data are reported in this manuscript. Data sources include Truven Health Analytics Commercial and Medicare supplemental Claims and Encounters Database, Federal Emergency Management Agency Modeling Task Force (FEMA MOTF) database and Area Health Resource Files (AHRF) data. AHRF and FEMA MOTF data are publicly available and accessible. Data from Truven Health Analytics Commercial and Medicare Supplemental Claims and Encounters Database are designed to address the requirements of the Health Insurance Portability and Accountability Act of 1996 (HIPAA). The MarketScan Research Databases meet the criteria for a limited-use data set and contain none of the data elements prohibited by HIPAA for such data sets. Formal Data Use Agreements (DUAs) are in place with every entity that do not allow public disclosure of detailed datasets used in this study. The DUA’s are in place to protect both patient and hospital sensitive data from public disclosure as the data is classified as a “Limited Data Set” under the Health Insurance Portability and Accountability Act of 1996 (HIPAA) which is to be protected as personal health information (PHI). DUA’s only allow data sharing if it is aggregated and fully de-identified as presented in this paper. Posting detailed data online or providing access to the detailed data that allows replication of this study, or other use of the data to non-public health personnel constitutes a legal violation of these DUA’s.

For requests regarding this data, please contact: Market Scan Information Systems, Inc., 815 Camarillo Springs rd. Suite B, Camarillo, CA. 93012

Phone: 800-MKT-SCAN (658-7226); Fax: 855-MKT-SCAN (658-7226); Email: www.MarketScan.com or www.mDesking.com

Competing Interests

The authors have declared that no competing interests exist.

Corresponding Author

Linda I. Ekperi, DrPH, MPH

Applied Science and Evaluation Branch, Division of State and Local Readiness, Centers for Disease Control and Prevention

1600 Clifton Rd., N.E., MS-D44, Atlanta, Ga 30333

Phone: 404-718-6657

Email: wpf7@cdc.gov

]]>
http://currents.plos.org/disasters/article/the-impact-of-hurricane-sandy-on-hiv-testing-rates-an-interrupted-time-series-analysis-january-1-2011%e2%80%92december-31-2013/feed/ 0
Vulnerabilities Associated with Post-disaster Declines in HIV-testing: Decomposing the Impact of Hurricane Sandy http://currents.plos.org/disasters/article/vulnerabilities-associated-with-post-disaster-declines-in-hiv-testing-decomposing-the-impact-of-hurricane-sandy/ http://currents.plos.org/disasters/article/vulnerabilities-associated-with-post-disaster-declines-in-hiv-testing-decomposing-the-impact-of-hurricane-sandy/#respond Tue, 21 Aug 2018 14:00:44 +0000 http://currents.plos.org/disasters/?post_type=article&p=39450 Introduction: Using Interrupted Time Series Analysis and generalized estimating equations, this study identifies factors that influence the size and significance of Hurricane Sandy’s estimated impact on HIV testing in 90 core-based statistical areas from January 1, 2011 to December 31, 2013.

Methods: Generalized estimating equations were used to examine the effects of sociodemographic and storm-related variables on relative change in HIV testing resulting from Interrupted Time Series analyses.

Results: There is a significant negative relationship between HIV prevalence and the relative change in testing at all time periods. A one unit increase in HIV prevalence corresponds to a 35% decrease in relative testing the week of the storm and a 14% decrease in relative testing at week twelve. Building loss was also negatively associated with relative change for all time points. For example, a one unit increase in building loss at week 0 corresponds with an 8% decrease in the relative change in testing (p=0.0001) and a 2% at week twelve (p=0.001).

Discussion: Our results demonstrate that HIV testing can be negatively affected during public health emergencies. Communities with high percentages of building loss and significant HIV disease burden should prioritize resumption of testing to support HIV prevention.

]]>

Introduction

Emergency preparedness plans aim to anticipate and mitigate the direct and indirect effects of large scale disasters on the public’s health. Often, these plans deal with the direct impact of a natural disaster or its indirect effects on disease transmission. However, the prospect of natural disasters, such as severe weather events, also has broader implications for prevention strategies.

When Hurricane Sandy made landfall in October of 2012, the storm caused damage to homes and businesses, displaced survivors and crippled vital infrastructure systems such as power, transportation, water treatment and healthcare. 375, 000 housing units were destroyed in New Jersey and New York; 8.5 million people lost power across 21 states; and more than 1,000 patients had to be evacuated from metro area hospitals due to unsafe conditions such as untreated waste which flowed from the treatment plants and impacted the public water systems. 1, 2 At least sixty-seven people in the U.S. died as a direct result of the storm and many others faced infrastructure loss and other conditions that can affect disease prevention strategies and long-term health.3

The National HIV/AIDS Prevention Strategy identifies increasing the proportion of HIV-infected persons who know their status and linkage to care as targets for reducing the burden of the disease. Approximately 1.2 million people were living with HIV in the US in 2011. Among those, 14% had undiagnosed infection.4 Persons who are not aware of their HIV status contribute to a third of the ongoing transmission in the U.S.5 Diagnosis is a critical first step in managing HIV. Diagnosis helps to connect people to treatment, treatment can reduce viral load, increase immune function and reduce the risk of transmission. CDC recommends that adolescents and adults are tested for HIV at least once and that those with an increased risk for HIV infection are tested at least annually.6

Ekperi and colleagues7 have identified that Hurricane Sandy had a negative impact on HIV testing rates using Interrupted Time Series (ITS) analyses. Natural disasters impact individual, social and environmental contexts that can contribute to these testing declines. For example, disasters may increase individual risk behaviors (e.g., increase in sex risk and/or injection behaviors), lead to changes in sexual networks, result in population changes (e.g., migration), and impact a community’s self-efficacy. Disasters also impact a community’s ability to mobilize resources and result in infrastructure damage. These post-event changes pose a threat to established HIV testing and prevention strategies and can contribute to disease spread.8 Therefore, it is essential to examine the impact of this storm on HIV testing.

Pre-event factors can also impact testing declines. HIV testing is associated with age, gender, race, marital status and access to healthcare. For example, Merchant et al.9 found that individuals who were male, white, married, or never married/partnered and those with private insurance were more likely to have never been tested for HIV. Similarly, although having a provider recommend HIV testing increases the odds that someone will get tested,10 Anderson et. al.11 found that physician perceptions of a patient can affect a doctor’s likelihood of recommending HIV testing. Patients that are perceived to be lower risk (e.g., whites) are less likely to be referred for HIV testing.11 Taking these pre- and post-storm factors into account, this study identifies factors that influence the size and significance of the estimated impact that Hurricane Sandy had on HIV testing using data on storm and community characteristics. These impact estimates were produced using methods similar to those used in previous work 7 and are presented in the Appendix.

Methods

Data Sources

Using Truven Health Analytics Commercial and Medicare supplemental Claims and Encounters Database, this study analyzed HIV testing rates in Sandy-affected regions from 2011 to 2013 by core based statistical area (CBSA). The database links paid claims and encounter data to detailed patient information across provider sites and types and comes from a selection of large employers, health plans, and government and public organizations.12 The database also contains person-specific clinical utilization and expenditures, enrollment for inpatient, outpatient, and prescription drug services, and services not covered in a health insurance contract.12 Interrupted Time Series (ITS) analyses, specified as segmented regressions with autoregressive errors, were estimated to capture the impact of Hurricane Sandy on weekly rates of HIV testing per 1,000 enrollees in CBSAs located within FEMA disaster-designated counties.

Weekly outpatient claims data from 2011 to 2013 were used to identify those who tested for HIV. Current Procedural Terminology (CPT) codes (‘86689’, ‘86701’, ‘86703’, ‘87389’, ‘87390’, ‘87534’, and ‘87535’) representing HIV testing procedures were extracted, along with enrollee ID and the enrollment detail information. Enrollment detail data contained demographic information by geographic location, age, and gender. The data captured for the region included all those who were tested and billed by private insurance companies in the states that were hit the hardest by Hurricane Sandy: Connecticut, Delaware, Indiana, Illinois, Kentucky, Maine, Maryland, Massachusetts, Michigan, New Hampshire, New Jersey, New York, North Carolina, Ohio, Pennsylvania, Rhode Island, Tennessee, Vermont, Virginia, Washington D.C. and West Virginia.

HIV testing rates were examined among enrollees not previously diagnosed with HIV to determine if Hurricane Sandy caused any disruption in HIV testing and relative and absolute effects were estimated using Interrupted Time Series (ITS) analyses detailed in the Appendix. Accordingly, the number of people tested for HIV was divided by the number of those enrolled members per capita in the affected CBSAs and multiplied by 1000. The HIV weekly testing rate was calculated as the number of HIV tests administered identified by CPT codes in each CBSA divided by the number of enrollees in the CBSA for each week in the study period. We used CBSAs due to the unavailability of county identifiers in the claims data, which led to the inability to accurately link storm-related factors to rural areas located outside CBSAs. Relative change due to the event was defined as the percentage difference between the estimated value given the event occurred and the estimated value given the event did not occur evaluated at a particular time after the event. . Note that the estimated effect is scaled to be relative to 1 with minimum of zero. Thus, a mean relative impact of 0.92 indicates an 8% decline in the testing rate given Hurricane Sandy occurred.

In order to estimate the effect of storm-related factors that contributed to HIV-testing declines, Federal Emergency Management Agency Modeling Task Force (FEMA MOTF) Hurricane Sandy impact analysis data on storm intensity (e.g., precipitation, buildings lost) by county was incorporated. The FEMA MOTF data are compiled by a task force of FEMA experts in hazard assessment and are used to develop estimates of impacts after events.13 These county-level data were crosswalked to CBSA by calculating the weighted mean levels of storm-related factors for each CBSA using data from the counties included in that CBSA and weighting by county population estimate. Rural areas were excluded because these areas are not included in CBSAs.

To estimate the impact of sociodemographic as well as healthcare access factors that may have contributed to HIV testing decline, Area Health Resource Files (AHRF) data were crosswalked from county to CBSA by calculating the weighted mean value for each CBSA using data from the counties included in that CBSA and weighting by county population estimate. AHRF include county, state and national files that offer a broad range of health resources and socioeconomic indicators, such as numbers of health facilities, health professionals, and health training programs and measures of resource scarcity. In addition, AHRF contains geographic codes and descriptors which enabled us to link the AHRF data to the claims and FEMA data.

2.2 Measures

Because AHRF data in particular provides many variables that aim to measure similar constructs, we ran simple bivariate correlations in order to begin assessing which to include in the model. Variables were selected based on the correlation results as well as prior evidence of significant impact based on literature review. Note that AHRF variables were only available for a subset of the 90 CBSAs included in the Sandy Impact Area. We tested a number of variables for inclusion in this model that were theoretically relevant. For example, income and per capita medical doctors (MDs) were tested but did not approach significance in any model. Percent minority and unemployment rate approached or reached significance at 0.05, so they were included in the model. Several additional measures, including precipitation, population exposed to storm surge, the natural log of median per capita income, hospital beds per capita, and MD’s per capita were also examined as independent predictors of HIV testing rate in our models but were not included due to lack of significance.

HIV prevalence. HIV prevalence rates among enrollees in each CBSA for 2012 were calculated using MarketScan data. The number of members with HIV diagnoses in a CBSA was divided by the total number of members with and without HIV per CBSA and multiplied by 100,000.

Household damage. Total count of households with damage claims after Hurricane Sandy was divided by the total number of households in that CBSA in 2010 – the nearest year for which data was available – and multiplied by 100 to generate percentage of households damaged. (FEMA data).

Building loss. FEMA data included percentage of buildings lost due to Hurricane Sandy.

Housing Unit Density per square mile. Housing unit density per square mile based on 2010 census data – the time period with available data that most closely corresponded to Hurricane Sandy. (AHRF data)

Percentage minority population. Percentage of nonwhite or Hispanic/Latinos individuals in the population. The original data source for this variable was the 2010 Census.

Unemployment rate. The unemployment rate for 2010 was linked from AHRF and is based on 2010 Census data.

2.3 Statistical Methods

Estimates of the impact of Hurricane Sandy on weekly HIV testing rates per 1,000 enrollees not previously diagnosed with HIV by CBSA were calculated from parameter estimates retained from ITS analyses for each of 90 CBSAs included in the Sandy impact area.7, 14 Truven Health Analytics (Marketscan) data were used during the period between January 1, 2011 and December 31, 2013 in the northeastern region of the United States. ITS analyses were specified as segmented regressions with autoregressive errors. Note that ITS is a very robust quasi-experimental method for estimating the impact of an event. 15 See Appendix for additional information on ITS analysis.

Generalized Estimating Equations

We utilized a robust Generalized Estimating Equations (GEE) regression model to examine the effect of the sociodemographic and storm-related measures in the FEMA data and AHRF files on the relative change in HIV testing rates provided by ITS analysis. We estimated two GEE regression models at each of the five time points post-event; 0 weeks, 1 week, 4 weeks, 8 weeks, and 12 weeks. In Model 1, we utilized only FEMA data and HIV prevalence rates because these variables were available for the full sample of 90 CBSAs. Model 1 regressed the relative change in CBSA-specific HIV testing at each time point based on HIV prevalence rates, the percentage of households damaged, and the percentage of buildings lost in order to identify and examine the relationships of explanatory variables in the model at each time period and across time periods.

In Model 2, we included the AHRF variables, which reduced the sample size to 75 CBSAs. Model 2 regressed the relative change in CBSA-specific HIV testing at each time point based on HIV prevalence rates, the percent of households damaged, the percent of buildings lost, housing density per square mile, the percent of minority population, and the unemployment rate.

Results

Table 1 presents summary statistics for the dependent (HIV testing rate per 1000 enrollees not previously diagnosed with HIV by CBSA) and independent variables. Summary statistics of the estimated relative impact of Hurricane Sandy on HIV testing rates is presented for five time periods post-event. While the relative impact does decline across the five time periods, at week 12 the mean and median relative impact are still over 4%. The standard deviation of the relative impact declines across the time periods, indicating reduced variability in the relative impact across time periods.

Table 1

In Model 1, building loss is consistently negatively related to relative impact across the five time periods. When housing unit density per square mile is added in Model 2, HIV prevalence rate is negatively and significantly associated with change in HIV testing rates for all time periods. The relationship between HIV prevalence and change in testing becomes weaker over time – ranging from -0.0045 (p=.001) at Week 0 to -0.0016 (p=0.013) at Week 12. Percentage of households with damage is significant at all time points except Week 12, but is positively associated with HIV testing. The strength of this relationship similarly declines over time. Unlike household damage, the percent of buildings lost has a negative and significant relationship with change in HIV testing. The strength of this relationship declines over time, but it remains significant (Estimate = -0.107; p = .001) at Week 12. Finally, housing unit density per square mile is positively and significantly associated with change in HIV testing rates at all time periods.

Table 2 presents the relative marginal effects of the storm-related and sociodemographic factors on the relative impact to HIV testing rates for each model. Model 1 presents the 90 CBSA results with FEMA variables and HIV prevalence. Marginal effects represent the relative change in the dependent variable associated with a marginal, or incremental, change in the independent variable. Building loss was the only variable in this model to show significance at the 0.05 level. For example, at the time of the event, a 1% increase in building loss corresponds to 2% decrease in the relative change in testing at time 0, (or 98% of the baseline level). A 1% increase in building loss likewise corresponds to a 1% decrease in relative testing at week 4 and week 8.

Table 2 final

Model 2 (Table 2) depicts relative marginal effects with the AHRF variables included.

In model 2, there is a significant negative relationship between HIV prevalence and the relative change in testing at all time periods. A 1% increase in HIV prevalence corresponds to a 35% decrease in relative testing the week of the storm and 14% decrease in relative testing by week twelve. Building loss was also negatively associated with relative change for all time points except for week twelve. For example, a 1% increase in building loss at week 0 corresponds with an 8% decrease in the relative change in testing (p=0.0001). At week 12, 1% increase in building loss corresponds to a 2% decrease in the relative change in testing (p=0.001).

In model 2, percentage of household damage maintained a significant and positive relationship with the relative change in testing across all time periods except for week twelve. A 1% increase in the percentage of household damage resulted in a 5% (p<0.0001) increase in testing at week zero. Percentage of household damage was significant at each time point except for week 12. Housing unit density per square mile and unemployment rate were also positively and significantly associated with the relative change in testing at all time periods. Percentage minority population was the only variable that was not significant across any of the five time periods, though it did approach significance at week 12 (p=0.0519).

Table 3

Conclusion

In this study we have identified the effects of storm and sociodemographic factors that most contributed to declines in HIV testing after Hurricane Sandy. Our findings suggest that areas with higher levels of building loss are likely to experience a sharper decline in testing. Similarly, areas with higher HIV prevalence rates are also more likely to experience significant declines in HIV testing. Percentage of household damage maintained a significant and positive relationship with the change in HIV testing. Pre-storm variables such as housing unit density and the unemployment rate in a CBSA were also significantly and positively associated with HIV testing. These associations and their directions are consistent with the literature mentioned previously and may be due to differences in physician and patient perceptions of risk based on race and income and on differences in the availability of and access to HIV testing services in more urban versus less urban CBSAs.9-11

Discussion

This study also reveals an important distinction between household damage and building loss. Percentage of household damage – the total count of households with damage divided by the total number of households in a CBSA – was positively and significantly associated with change in HIV testing while percentage of buildings lost was negatively and significantly associated with change in HIV testing. This suggests that percentage of buildings lost (e.g., healthcare facility loss), but not household damage, may be a better measure of HIV testing service availability and correlates to the overall storm impact that can contribute to testing declines.

This study uses ITS, a robust modeling technique that controls for prior trends and seasonality while estimating the impact of an event. This study also expands on previous analysis of HIV testing rates after Hurricane Sandy to unpack the pre- and post-storm variables that most significantly impact HIV testing. While the one potential limitation of ITS is confounding by a co-occurring event, the authors were not able to identify any co-occurring event that might have confounded the estimated effect of Hurricane Sandy.

This analysis was limited to CBSAs which did not allow estimation of similar effects on HIV testing rates in rural areas. Future studies need to address the effects of hurricanes on HIV testing rates in rural areas as well. This analysis was also restricted to private claims data. Future studies need to incorporate Medicare and Medicaid claims to better understand the impact on the broader population. Finally, while this study focuses exclusively on HIV testing rates, it is likely that that many other types of primary care monitoring/testing may be affected by disasters. Future research needs to assess the impact of a disaster on other primary care prevention activities such as breast screening, immunizations, blood sugar and blood pressure monitoring.

Appendix

Statistical Methods: Interrupted Time Series Analysis

Descriptive statistics of mean, median, standard deviation, and range were calculated among the 90 CBSAs for the HIV testing rate per 1,000 enrollees before and after the event date by storm impact rank. Each of the time series was examined visually and trend and seasonal statistics were analyzed. There were 153 weeks in each time series with 95 weeks pre-event and 58 weeks post-event. Statistical analyses was conducted using the statistical software package SAS v9.3.

In order to estimate the impact of Sandy on HIV testing rates, ITS models of weekly rates were estimated via maximum likelihood for each of 90 CBSAs in the Sandy-affected area. ITS models were specified as linear segmented regression with autoregressive errors. Time series were tested for stationarity via the augmented Dickey-Fuller test and autocorrelation via generalized Durbin-Watson tests. Autoregressive error models were used to address autocorrelation and seasonality where present (Wagner, et al. 2002; Jandoc, et al. 2015). Models were estimated in SAS using proc autoreg. with autoregressive error terms identified via backward elimination.

The ITS was modeled using the following linear segmented regression with autoregressive errors, which represents the baseline level and trend of the outcome variable before the event and changes in the level and trend after the event for each of the six outcomes:

1

Here, Yt represents the dependent variable, weekly HIV testing rate, at a point in time. Timet is the continuous variable representing time in weeks since the beginning of the study period. The intervention function (B2*event(t)+B3*timeafterevent(t)) was specified as a step function. Eventt is an indicator variable set to zero prior to the date of the event and becoming 1 the week of the event and for the duration of the time series after the event. Timeaftereventt is a continuous variable counting the days elapsed since the event at time t and set to zero prior to the event. The regression error, et, is comprised of a random error component as well as an autoregressive error component to adjust for autocorrelation and seasonality.

The estimated factual case is represented as:

2

The estimated counterfactual case is represented as:

3

Equation (2) was estimated for each CBSA in the Sandy Impact area. The parameter estimates from equation (2) were used to calculate equation (3) for each CBSA.

Note that the parameter estimate is the baseline level of the outcome variable at time zero, or the intercept. is the estimated baseline trend of the outcome variable, or the weekly deviation from the baseline level prior to the event. is the estimated absolute change in level, or intercept, of the HIV testing rate that occurs immediately following the event and is the estimated absolute change in trend that occurs after the event. The estimated relative, or percentage, change in the baseline level due to the event is , while the estimated relative change in trend is . Thus, is the estimated post-event baseline of the outcome variable and is the estimated post-event trend in the outcome variable. The estimated absolute impact of the event measured one week post-event is equation (2) minus equation (3) evaluated at t=97, or and the estimated relative impact of the event measured one week post-event is . Thus, absolute impact is the difference between the estimated values given the event occurred and the estimated value given the event did not occur at a particular time post-event. Relative impact is expressed as the percentage change in HIV testing rates in the factual case compared to the counterfactual case.

Estimates of the parameters and standard errors retrieved from equation (2) were used to produce estimates and associated 95% confidence intervals of the estimated relative changes in level and trend as well as the estimated relative impact of the event immediately and at one, four, eight, and twelve weeks post-event. Based on the work of Zhang, et al. 2009, the 95% confidence intervals of the estimated relative changes for baseline, trend, and the various time points post-event were calculated via the multivariate delta method. Specifically, parameter estimates from equation (2) for each CBSA were retrieved and used to estimate the expected value and the variance of the relative change at each time point post-event. Summary statistics for these estimated relative changes were calculated by storm impact rank at 0 weeks post-event, at 1 week post-event, at 4 weeks post-event, at 8 weeks post-event and at 12 weeks post-event and weighted for significance (Table A).

Data Availability:

All relevant data are reported in this manuscript. Data sources include Truven Health Analytics Commercial and Medicare supplemental Claims and Encounters Database, Federal Emergency Management Agency Modeling Task Force (FEMA MOTF) database and Area Health Resource Files (AHRF) data. AHRF and FEMA MOTF data are publically available and accessible. Data from Truven Health Analytics Commercial and Medicare Supplemental Claims and Encounters Database are designed to address the requirements of the Health Insurance Portability and Accountability Act of 1996 (HIPAA). The MarketScan Research Databases meet the criteria for a limited-use data set and contain none of the data elements prohibited by HIPAA for such data sets. Formal Data Use Agreements (DUAs) are in place with every entity that do not allow public disclosure of detailed datasets used in this study. The DUA’s are in place to protect both patient and hospital sensitive data from public disclosure as the data is classified as a “Limited Data Set” under the Health Insurance Portability and Accountability Act of 1996 (HIPAA) which is to be protected as personal health information (PHI). DUA’s only allow data sharing if it is aggregated and fully de-identified as presented in this paper. Posting detailed data online or providing access to the detailed data that allows replication of this study, or other use of the data to non-public health personnel constitutes a legal violation of these DUA’s.

For requests regarding this data, please contact:

Market Scan Information Systems, Inc.

815 Camarillo Springs rd. Suite B

Camarillo, CA. 93012

Phone: 800-MKT-SCAN (658-7226)

Fax: 855-MKT-SCAN (658-7226)

www.MarketScan.com

www.mDesking.com

Conflicts of Interest: None

Corresponding Author: Erin Thomas is the corresponding author for this article and can be reached via kqn4@cdc.gov.

Table A

Estimated effort of Hurricane Sandy on weekly HIV testing rates among privately insured enrollees by storm impact rank*, January 2011 through December 2013: Interrupted Time Series** effects

]]>
http://currents.plos.org/disasters/article/vulnerabilities-associated-with-post-disaster-declines-in-hiv-testing-decomposing-the-impact-of-hurricane-sandy/feed/ 0
Earthquake in Western Iran: Renovation Kills http://currents.plos.org/disasters/article/earthquake-in-western-iran-renovation-kills-2/ http://currents.plos.org/disasters/article/earthquake-in-western-iran-renovation-kills-2/#respond Thu, 16 Aug 2018 13:30:45 +0000 http://currents.plos.org/disasters/?post_type=article&p=41101 Introduction: Earthquake is the most important cause of death from natural disasters in Iran. This paper brings attention to the main causes of loss of life due to the Kermanshah province earthquake (Nov 12 2017), and provides a wakeup call about the unsafe nature of buildings there. Methods: This study is based on official reports review and a field assessment in the areas affected by the earthquake in western Iran. Results: Although buildings in this area are mainly old structures, strangely, more than 70% of the destroyed buildings in this earthquake were under 5 years of age, newly built or renovated buildings according to mandated building codes. Discussion: Mandated building codes and construction rules and regulations are not respected even for the newly constructed or reconstructed structures buildings. Keywords: Earthquake, Iran, construct, reconstruct, Building codes

]]>

BRIEF INCIDENT REPORT

At 21:48 local time, on Nov. 12, 2017, western Kermanshah Province in Iran was shaken by an earthquake of moment magnitude Mw7.3 for about 30 s1 . The epicenter has been reported at the 34.88°N and 45.84°E coordination near the Iran–Iraq border and the depth at 23 km. Based on the active fault map of Iran, this earthquake can be triggered by the movement of the Zagros Mountain Front Fault in Pol-e-Zahab Region11 . In addition, more than 500 aftershocks, greatest of which with magnitude of 4.9 have been recorded by the Iranian Seismological Center.

Due to this earthquake, 620 people died2 , 8,000 injured, 70,000 displaced, and over 12,000 buildings damaged1 . The total exposed population was 4,700,000 including 75% urban and 25% rural3,4 . Most of the fatalities are reported in Sarpol-e Zahab, Qasr-e-Shirin, and Thalath Babajani. The highest number (559 death) was in the city of Sarpol-e Zahab2 around 200 of which from a small area called Fouladi.

Most of the local buildings were one-story adobe, covered with heavy flat roofs. Nevertheless, in larger villages and cities, more-storey brick-based and cement-based buildings were seen. Many of which were newly constructed or reconstructed structures that built according to developed and mandated codes after the bam earthquake of 25 Dec 2003 in Iran. However, around 80% of destructions occurred in buildings with no frame, both in the urban and rural areas4 , around 70% of which are visually estimated as buildings of less than 5 years of age, among them two newly constructed hospitals.

Although buildings with no frame in the rural areas of the Kermanshah Province are 15% more than urban areas, number of residents at the buildings with no frame are 60% more in the urban areas than rural3 . In addition, from the initial assessments, major fatalities are from the newly constructed or reconstructed structures buildings. Therefore, it is more important than ever to ask, in spite of the implemented building codes, why would newly constructed or reconstructed buildings be destroyed and cause devastation and fatalities of this magnitude.

Indeed, further evaluation of this analysis is required; however, the large number of fatalities, injuries, and displacements due to this earthquake is a wakeup call to the unsafe constructed or reconstructed. Poorly constructed or reconstructed buildings in this area can be due to several factors, including lack of quality control on the architectural and structural designs, owners’ employment of non-expert workers and use of non-standard materials, vaguely defined construction regulations, and owners’ right to hire architects, structural engineers, supervisors and contractors.

Due to insufficient payments to supervisors and lack of restriction of law enforcement on them, surveillance is generally weak in structural designs. On the other hand, architectures are under surveillance of both owners and the city. Monetary deals between owners and supervisors are leading to unsafe structural designs. Law enforcement is another challenge in construction since duties of the institutes are not carefully distributed in the written laws and regulations. Considering these preventive factors, we suggest that some changes in the construction surveillance system and the owners’ rights are required for future building code enforcements.

CORRESPONDING AUTHOR

Mehdi Zare, Professor of Engineering Seismology, International Institute of Earthquake Engineering and Seismology (IIEES), and Associate member of Geology Division, Department of Basic Sciences; Academy of Sciences of the I.R. Iran. E-mail: mzare@iiees.ac.ir

COMPETING INTERESTS

The authors have declared that no competing interests exist.

Data Availability

The data underlying this study have been uploaded to figshare and are accessible using the following DOI: 10.6084/m9.figshare.6949988

]]>
http://currents.plos.org/disasters/article/earthquake-in-western-iran-renovation-kills-2/feed/ 0
Knowing What We Know – Reflections on the Development of Technical Guidance for Loss Data for the Sendai Framework for Disaster Risk Reduction http://currents.plos.org/disasters/article/knowing-what-we-know-reflections-on-the-development-of-technical-guidance-for-loss-data-for-the-sendai-framework-for-disaster-risk-reduction/ http://currents.plos.org/disasters/article/knowing-what-we-know-reflections-on-the-development-of-technical-guidance-for-loss-data-for-the-sendai-framework-for-disaster-risk-reduction/#respond Thu, 02 Aug 2018 15:00:05 +0000 http://currents.plos.org/disasters/?post_type=article&p=36974 Introduction: To report on activities aligned with the Sendai Framework for Disaster Risk Reduction 2015-2030, national governments will use the Sendai Monitor platform to track progress using a series of indicators that inform seven Global Targets originally agreed in 2015. In February 2017, the UN General Assembly adopted a set of 38 agreed indicators based on work led by an open-ended intergovernmental expert working group (OIEWG) on indicators and terminology relating to disaster risk reduction. In January 2018 the United Nations Office for Disaster Risk Reduction released technical guidance documents in advance of the launch of the Sendai Monitor in March 2018. Methods: This paper discusses several challenges to recording and reporting on loss data under the Sendai Framework. Additional insights to elaborate on discussion build upon commentary and examples raised during a workshop held on developing loss data that was hosted by the United Nations Office of Disaster Risk Reduction (UNISDR), the Integrated Research on Disaster Risk (IRDR) programme, and Public Health England (PHE) from February 15-17 2017 at the Royal Society in London, United Kingdom. The meeting’s purpose was to refine technical guidance notes concerning Global Targets A, B, C, and D, which had been drafted in coordination with the work of the OIEWG. The workshop was attended by representatives from UN Agencies, UN Member States, international scientific bodies, academic bodies, the government of the United Kingdom and the private sector. Results: Global Targets A, B, C and D of the Sendai Framework have common and specific complexities which require acknowledgement and support in recording, reporting and using disaster loss data. Discussions during the February 2017 loss data workshop highlighted a number of complexities and the need for common standards and principles for loss data. Individual target complexities include attribution of health impacts, assessing impacts, consistently calculating economic losses and measuring disruption to critical infrastructure. Discussion: Transparent monitoring is critical to ensure political will, financial efforts and effective evidence support the global shift towards more sustainable development. Data involves common challenges which can undermine accuracy and understanding of reporting across the frameworks that outline the United Nations’ 2030 Agenda. Disaster loss data adds further challenges which require support and innovation to ensure stakeholders across sectors in all sectors have appropriate technical guidance that can support useful loss data management processes. The February 2017 workshop highlighted systemic challenges with working with loss data and highlighted several pertinent pathways to progress on the breadth and reliability of disaster loss data across different settings.

]]>

Notice of Corrections

24 September 2018: PLOS Currents -. Correction: Knowing What We Know – Reflections on the Development of Technical Guidance for Loss Data for the Sendai Framework for Disaster Risk Reduction. PLOS Currents Disasters. 2018 Sep 24 . Edition 1. doi: 10.1371/currents.dis.e597ed667989b083254fefcac8853875.

1. Introduction

The Sendai Framework for Disaster Risk Reduction 2015-2030 offers national governments an opportunity – to enhance their capacities to deal with disaster risk at all scales and across all sectors. It encompasses all hazards and disaster scenarios, including: small and large scale; frequent and infrequent; sudden- and slow-onset; caused by natural or man-made hazards; as well as related environmental, technological and biological hazards and risks.1 Along with complementary instruments of the United Nations’ 2030 agenda, such as the Sustainable Development Goals (SDGs) and the Paris Climate Agreement, the Sendai Framework offers UN Member States measures of progress.2 Seven agreed Global Targets focus on reducing: mortality, persons affected, economic loss, damage to critical infrastructure and disruption of basic services due to disasters; and improving local and national strategic disaster plans, international cooperation, and multi-hazard early warning systems and disaster-risk information and assessments.1

The development of the indicators against which to measure the Sendai Framework targets took place between September 2015 and November 2016. Development and refinement was led and undertaken by an open-ended intergovernmental expert working group (OIEWG) on indicators and terminology relating to disaster risk reduction. Formal meetings were held between 28 – 30 September 2015, 10 – 11 February 2016, and 15 – 18 November 2016. On 2 February 2017, the UN General Assembly adopted the OIEWG’s consensus on 38 indicators for use across the seven Global Targets.3,4 As of March 1 2018, the infrastructure for Member State reporting – the “Sendai Framework Monitor” – will provide the platform for measuring progress as part of the Sendai Framework Monitoring Process.5

There are significant challenges in the collection, recording and reporting of data. The Sendai Framework Readiness Review 2017 compiled the monitoring capabilities of 87 UN Member States, revealing significant heterogeneity between, and within, countries in their capacity to report against the approved indicators.6 This paper explores the utility of disaster loss data and examines the processes necessary to ensure monitoring under the Sendai Framework. Discussion in this paper also builds upon examples and commentary raised during a workshop on disaster loss data hosted by the United Nations Office for Disaster Risk Reduction (UNISDR), the Integrated Research on Disaster Risk (IRDR) programme, and Public Health England (PHE) between February 15 – 17 2017 at the Royal Society in London, United Kingdom.7 The workshop was attended by 44 participants in total, with representatives from UN Agencies, UN Member States, international scientific bodies, academic bodies, the government of the United Kingdom and the private sector (see Appendix 1).7

2. A place for indicators

The Millennium Development Goals (MDGs) Data Catalogue exemplified what was possible to better understand global problems and progress. Across the eight MDGs, data became available for directing funding and supporting political pressure that fed into local, regional and global progress outlined in the UN’s 2015 Millennium Development Goals Report.8,9 The need to support and account for progress catalysed better recording, reporting, and reviewing of information. Reliable data and evidence is crucial to effective policy making.10 By 2015, with the phasing out of the MDGs and the close of the Millennium Development Era, two challenges were apparent. There was invisibility and inequality in data across the 48 indicators that informed the eight MDGs. Datasets were incomplete geographically, temporally, and in terms of socio-demographic disaggregation.11

Building on the success and limitations of the Millennium Development Era, the United Nations 2030 agenda includes ambitious aims for data and monitoring. The alone include 18 goals, reported upon by more than 230 indicators, set for countries to achieve by 2030.12 The Sendai Framework for Disaster Risk Reduction 2015-2030 will map progress through seven Global Targets and 38 indicators.3 Ambition extends throughout the 2030 agenda via a call to “leave no one behind”. This phrase recognises the rights and dignity of all individuals in all countries and the need to target support towards those deemed “furthest behind” due to marginalisation or neglect of efforts.13 To understand who is “left behind”, it is critical to record and report disaggregated data. This requires robust information systems, accepted standards and technical guidance. However numerous countries cannot take reliably and specifically informed actions based on disaster loss data, due to unobtainable or aggregated information.6 Critical elements of domestic capacity include ability to collect, record, and report information at all levels. Acknowledging the fiscal and other limitations that exist, coordinating bodies such as UNISDR, can lead on two efficiency-promoting actions:

(1) Ensure complementarity between indicators across global instruments.

(2) Develop clear and contextually relevant technical guidance for data collection.11

Carrying out the former – ensuring complementary – has been apparent in the efforts of independent organisations and the independent expert working groups tasked with indicator development for the Sendai Framework.4,14,15 Ensuring in data efforts is central to the Sendai Framework Monitor, an online reporting mechanism, and coordinating progress assessments for the SDGs and the Sendai Framework.16 For instance, common goals are clear between several Sendai Framework Global Targets and SDG Goal 11 – “Make cities and human settlements inclusive, safe, resilient and sustainable”. Targets A and B require reporting on mortality, missing individuals, and affected persons for disasters; Targets C and D, which address damage and disruption to the build environment and critical infrastructure. All of these elements are critical to ensuring urban resilience, while Target E supports assessments of national and local disaster-risk strategies critical to urban resilience.7,12

Without accurate reporting, accepted and comprehensive indicators lose value. If the methods for recording and reporting on data are unclear, or perceived to give an inaccurate picture of reality, then opportunity for learning and progress may be lost.17 The Inter-agency and Expert Group (IAEG) on SDG Indicators and the OIEWG discussed and drafted recommendations of technical guidance which? informed the indicator refinement process, but these were? not included at the stage of acceptance for individual indicators.3,18 In January 2018 UNISDR released a collection of technical guidance notes for data and methodology to support the first cycle of monitoring using the Sendai Framework Monitor.19 However, further refinement is possible, particularly with respect to ensuring available capacity is appropriately harnessed and supported. The following analysis examines crucial issues present in assessing progress on Sendai Framework Targets A, B, C and D, using commentary and examples from the loss data workshop to elaborate.

3. Evaluating the state of loss data

Insights into disaster loss data in the context of the Sendai Framework are available in the aforementioned OIEWG Report, sector-specific examinations, and examinations of coherence with other instruments of the 2030 agenda.3,4,14,20 Analysis in this paper focuses on addressing comments raised during the loss data workshop on gaps in technical guidance and areas for further work to ensure countries can report against Sendai Framework Global Targets A, B, C and D.

Table 1: Sendai Framework Targets A-D

Targets A, B, C and D for the Sendai Framework for Disaster Risk Reduction 2015-2030.

Target Description
Target A Substantially reduce global disaster mortality by 2030, aiming to lower average per 100,000 global mortality rate in the decade 2020-2030 compared to the period 2005-2015.
Target B Substantially reduce the number of affected people globally by 2030, aiming to lower average global figure per 100,000 in the decade 2020 -2030 compared to the period 2005-2015.
Target C Reduce direct disaster economic loss in relation to global gross domestic product (GDP) by 2030.
Target D Substantially reduce disaster damage to critical infrastructure and disruption of basic services, among them health and educational facilities, including through developing their resilience by 2030.

The Sendai Framework calls for country self-monitoring to assess progress.1 This implies that internationally comparative methods are not required and disaster loss data recording can take place using existing databases. This can ensure country capacities for recording and reporting are not overstretched and there is a focus on progress at the country level. Assessing the impact of hazardous events requires collaboration across national governments. Ministries devoted to health, business and the environment are a sample of those who will be key to accurately the extent and impact of biological and technological hazards. In the case of international comparisons, differences in technical capacities mean that the dispersion of results and potential presumptions of greater-than-calculated loss will vary. Nevertheless, loss data workshop participants agreed on the significant value of common standards and principles for loss data, including those applying across the Global Targets and corresponding indicators.

Coherent data principles for local and national reporting structures ensure that foundations for reporting on loss data are similar. The Sustainable Development Solutions Network Thematic Research Network on Data and Statistics (SDSN TReNDS) offers nine core principles to improve data quality and set the foundations for new data partnerships.21 In “Counting on the World” (2017), the following principles are proposed to support useful and usable contributions to the measurement of sustainable development:21

  1. Data quality and integrity: Ensuring clear standards support the entire process of data design, collection, analysis and dissemination.
  2. Data disaggregation: Informing, with appropriate safeguards in place, that data is disaggregated across dimensions including as geography, wealth, disability, sex, gender and age.
  3. Data timeliness: Using standards and technology to reduce time between initial design of data collection and publication of statistics.
  4. Data transparency and openness: Making all data on public matters or funded publically, including that produced by the private sector, open by default (with exemptions for genuine security or privacy concerns).
  5. Data usability and curation: Designing data architecture that is user-oriented and user-friendly.
  6. Protection and privacy: Developing and enforcing clear frameworks to regulate access and use of data.
  7. Data governance and independence: Strengthening and protecting data quality through national statistical offices that are functionally autonomous from other government agencies.
  8. Data resources and capacity: Investing in human capital, physical assets and technology to support governmental, intermediary and independent data systems.
  9. (Human) data rights: Protecting human rights at the core of any mechanisms or entities set up to mobilize the data revolution for sustainable development.

If implemented, the above principles would ensure that recording and reporting of loss data has a common direction and cause. However, particular elements of loss data present more unique challenges. The following sub-sections examine issues and pathways to progress for reporting on Sendai Framework Global Targets A, B, C and D.

Target A

Loss of life severely disrupts the households and communities and is particularly felt by highly vulnerable, low-income groups in the context of disasters.22 Yet measuring mortality is challenging. The World Health Organization (WHO) regularly receives cause-of-death statistics from about 100 Member States, yet two-thirds (38 million) of 56 million annual deaths are still not registered.23 Workshop participants noted that the disruption associated with disasters adds to the challenge of registering mortality. Furthermore, Target A is also informed by an indicator of missing persons. Yet across different settings there is limited comparability and coordination on this matter. For example, in the United Kingdom, a person cannot be registered as missing and declared dead until 7 years afterwards, whereas in Italy, it is at least 10 years.24,25

Comprehensive attribution of mortality to disasters is complex. Alongside direct trauma or ill-health from infectious disease during health emergencies, there are many indirect impact pathways.26 Several examples were raised during the loss data workshop, including how during slow-onset hazards, such as droughts, health effects may be mediated through the disruption to basic healthcare services and spread of communicable diseases.27 Participants agreed that technical guidance and work to improve loss data capacity should harness the available evidence to identify common and applicable causes of death from different types of hazards. This also raises the question of how to best classify hazards, as different taxonomies exist across different settings and there is no commonly accepted standard. Furthermore, time periods between the exposure to hazards and death can vary widely. Disruption of care for chronic conditions and onset of persistent stress can lead to greater disease burden or even death that may not occur for months or years after a disaster.28

Engaging broader systems for assessing mortality may offer another avenue to support disaster loss data management for health. The Global Burden of Disease (GBD) study, led by the Institute for Health Metrics and Evaluation, offers a platform to better assess disaster-related mortality using advanced modelling approaches.29 The GBD study is the most comprehensive worldwide epidemiological study in existence, with a description of mortality from a variety of causes at global, national and regional levels. The extraction of baseline health measurements for some of the SDGs from the GBD is already being explored.30 In addition, the World Health Organization’s ‘Global Reference List of 100 Core Health Indicators’ collates comprehensive reported information and aims to contribute to greater alignment between countries on the reporting of health trends.31 Whereas initiatives such as “The Lancet Countdown: Tracking Progress on Health and Climate Change” provide insights about an array on influential outcomes from and causes of disruption to health and health care.32

Target B

Each year between 2006 and 2016 an estimated 224 million people were affected by disasters attributed to natural hazards alone.33 Better understanding of impacts upon livelihoods is critical to reducing welfare impacts, especially in light of World Bank estimates that losses from shocks to economic activity from disasters amounts to US$520billion.22 Along with the acceptance of the indicators for the Sendai Framework Global Targets, United Nations Resolution 71/276 also accepted the definition affected people to include individuals that have sustained injuries or illness, whose houses have been damaged or destroyed, or those who have experienced disruption to their livelihoods as a result of a disaster event.3 As with Target A, concerns around attribution apply. Target B encompasses scenarios where cascading effects from hazards can develop into significant impacts. A simple assessment approach is critical, as measurement involves drawing information from a wide range of sectors.

Discussions at the February 2017 loss data workshop focused on establishing examples of harnessing existing systems of measurement for persons affected. Similar to Target A, data on injured and ill people can come from existing health indicators that are adapted to target disaster specific impacts. However, clear clarification is essential for periods of time use for measurement and the inclusion of secondary illness and injury. Mental health issues, amongst the most acute health impacts associated with disasters, are a specific area requiring definition within ill and injured person calculations. Geographic information systems (GIS) and remote sensing techniques can assess impacts to the physical environment, such as dwellings and local infrastructure, however local authorities and international standards needs to also account for degrees of damage to informal settlements. Further discourse noted the value of establishing proxies for assessing impacts to affected persons. Such methods are used by actors including the World Bank Groups’s Global Facility for Disaster Reduction and Recovery (GFDRR), which has employed post-disaster needs-assessment techniques using sector-specific data for employment, agriculture, health, transport, and communication to calculate the impact of disasters on human well-being.34 Moreover, the UN Food and Agricultural Organization (FAO) has previously estimated the livelihood impact of disasters using data on agriculture, food security and nutrition.35

Target C

In agreed indicators, “economic loss” encompasses value in the following categories: agricultural, productive, housing, critical infrastructure, and cultural heritage. The term “direct”, based on guidance of the OIEWG, refers to losses in assets. Despite progress during the implementation of the Hyogo Framework in building physical resilience to disasters, economic losses remain substantive. The GFDRR estimates that global annual losses attributed to disasters amount to over $300bn in asset stock. 22 This definition omits the substantial losses in productivity and well-being which lead to economic impact, however the complexity of necessary assessment protocols was avoided to ensure that indicator calculation was practical and feasible.3 Measurements for assessment of indirect economic losses are less developed and not included in the Sendai Framework. But understanding the cascading impacts of disasters on economic welfare and productivity is critical, particular as drivers of hazard risks changes over time.3,26

Economic loss assessments by member states will engage a broad cross-section of actors. These include international institutions (e.g. the World Bank and UNISDR), private sector companies expertise (e.g. insurance and catastrophe risk modelling industries) and national governmental bodies.19,22 Loss data workshop participants noted collaboration in this area can build on and support existing cooperation between public actors and risk transfer supply chains composed of catastrophe-modelling firms, primary insurance companies, and reinsurance providers.36 The World Bank’s Disaster Risk Financing and Insurance Program (DRFIP) is one example of a public-private partnership. DRFIP aims to reduce economic disruption by supporting prompt government responses to disasters that does not compromise sovereign fiscal balances.37

Reliable and consistent economic-loss calculations practices are critical for disaster loss data. At the loss data workshop, discussions highlighted the value of improving transparency in methods between private actors (e.g. catastrophe modelling companies) and accounting for geographical and temporal price fluctuations. When reliable information is absent proxies may be useful, but come with the caveat that non-private price indices are used as often as possible; an example of this is reconstruction inputs such as building materials. Noted challenges extend to the application of affected ratios (i.e. amount of damage due to a hazard) that may give binary, categorised (segmented), or continuous (percentage) values in damage ratios. At different periods following a hazard impact, reporting practices should also reflect need. Such rapid assessment protocols for soon after hazard impact, and another, more accurate, record of direct economic loss several c.1 year after a hazardous event. Estimating losses to cultural heritage, which inform indicator C6, are a unique and context specific challenge. While available guidance proposes assignment for non-movable and movable cultural heritage assets, their value is difficult to disentangle from local connection and (if applicable) tourism related income.19 Stakeholders also made it clear that cultural heritage issues associated with the natural environment further adds to this challenge.

Target D

Critical infrastructure comprises ‘physical structures, facilities, networks and other assets which provide services that are essential to the social and economic functioning of a community or society’.3 These factors play a critical role in how communities and systems will cope with hazardous events. If severe disruption takes place, emergency capacities are critical to mitigate disruption to essential services such as health care, education and transportation. Maintaining societal functions and productive capacities offsets financial and welfare risks in the short and medium term due to damage and disruption attributed to hazardous events. Strengthening key facilities is a key principle of locally led and internationally coordinated programs such as the Comprehensive Safe Hospitals Framework and the Worldwide Initiative on Safe Schools.38,39

Disruption to basic services may not require damage or destruction to infrastructure. Technological hazards include those which disrupt information systems, such as threats to cybersecurity. Across countries, computers are critical to continuity across basic service sectors. In March 2017, the global spread of “WannaCry” ransomware revealed the vulnerability of the UK’s National Health Service, causing disruption to clinical care and trust in the security of health record systems.40 The June 2017 “Petya” malware spread in Ukraine infected several elements of state infrastructure including energy, finance and government ministries.41

Target D has similarities with Target C that echo those between Target A and Target B. UNISDR technical guidance for monitoring and reporting recommends calculating indicators D-1 to D-4 using the same data and critical infrastructure units and facilities as C-3 and C-5. Common metadata formats are also recommended across C-5/D-4 and C-3/D-8.19 Furthermore, clear definitions are key to consistency in reporting on Target D. For instance, loss data workshop participants noted the challenges of measuring disruption due to slow-onset and small-scale disasters. Contrary to recommendations, damage and disruption to infrastructural assets and services can be disaggregated according to the institutional level e.g. primary or secondary health facilities, rather than based upon size. Such classifications are in line with practices in public sector risk assessment and private sector catastrophe modelling used to inform insurance products.42,43,44

4. Conclusion

Transparent monitoring is essential to ensure that political and financial efforts to implement the 2030 agenda align with accepted goals and foster an effective evidence base. It is critical to offer support to the institutions tasked with doing so. Technical guidance literature is an essential part of this, which the February 2017 loss data workshop and recent publication of UNISDR technical guidance have revealed. Across Sendai Framework Global Targets A-D, there remain specific issues that the academic community can support with innovating methods for improving estimation and accurate record of the impacts of disasters. The February 2017 disaster loss data workshop provided a pointed moment to reflect and exemplify these issues among a diverse range of stakeholders, from different regions and sectors.

UN member states have accepted the Sendai Framework Global Targets and their component indicators. To move forwards on implementation and monitoring of progress, country statistical offices and stakeholders need support to develop reliable loss data recording, reporting and analysis capacities that provide useful and usable information. However, disaster loss data has problematic characteristics common to all data. With further specific reliability and coverage challenges that, if not acknowledged and addressed, can undermine comprehensive understanding of what has happened, what could happen and what to do about it.

Competing Interests Statement

The authors have declared that no competing interests exist. Prof. Virginia Murray serves on the Editorial Board of PLOS Currents: Disasters.

Data Availability Statement

All relevant data are within the paper. No quantitative data was recorded for the purposes of this paper. Workshop discussions and findings were synthesised from rapporteur reports delivered during and after the workshop. Attendee affiliations outlined in Appendix 1.

Corresponding Author

Lorcan Clarke – clarkel5@tcd.ie

Appendix 1: Outline of Workshop on Disaster Loss Data – Held 15-17 February 2017 in London, UK.

In light of the value of developing the technical guidance for indicators and to build on previous efforts, UNISDR, IRDR, and PHE coordinated a three-day Loss Data Workshop in London, United Kingdom from 15 to 17 February 2017. The aim of the workshop was to advance and support production of technical guidance loss data informing Sendai Framework Global Targets A, B, C, and D.7 This guidance seeks to inform how UN Member States can collect, record and report under the Sendai Framework for Disaster Risk Reduction. The workshop occurred following after ratification by the UN General Assembly of the OIEWG recommended indicators to report on Sendai Framework’s Global Targets on 2 February 2017, and during the lead-in period to the 5th Global Platform for Disaster Risk Reduction in May 2017.4 The workshop was attended by 44 participants in total, with representatives from UN Agencies, UN Member States, international scientific bodies, academic bodies, the government of the United Kingdom and the private sector. For this consultation, workshop organisers invited stakeholders across public, private and multilateral agencies. Discussions took place under the Chatham House rule. This ensured anonymity was given to participants’ contributions and provided an open environment for pragmatic discussion. 45 Participants were split across several working groups that had assigned rapporteurs to report back to all attendees from breakout sessions.

Organizations represented at the loss data workshop included the following (sorted by type): UN Member States: Japan, Fiji, Ecuador, Indonesia, Zambia, Zimbabwe. International Scientific Bodies: Joint Research Centre of the European Commission, Pacific Community, Committee on Data for International Council for Science. Government of the United Kingdom: Cabinet Office, Public Health England, Office for National Statistics, Department for International Development, Environment Agency. Academic Sector: UK Research Council, King’s College London, University of Bristol, University College London, London South Bank University.

Rapporteurs were selected by organisers from multi-disciplinary backgrounds and varying levels of experience in academia, industry, or non-profit organisations. These diverse perspectives benefited proceedings through rapporteurs’ ability to actively contribute to and provide context for discussion notes among the working groups assigned to targets. For instance, those with a background in public health worked on Target A, while those with experience in economics or engineering meant were assigned to Targets C and D. Beyond this, rapporteurs were tasked with the preparation of workshop documents, integrating comments and recommendations into revised technical-guidance notes for the Global Targets under review. Additionally, the rapporteurs produced an overarching “essential reading” document to promote clarity for attendees and users of the technical guidance. These documents were then further developed by UNISDR in advance of the Global Platform for Disaster Risk Reduction in May 2017. Where further efforts were made to develop “essential reading” documents.

During the workshop preparation process it emerged that the role of rapporteurs is not set out in accepted and accessible guidance documents. The role of a rapporteur is context specific and can encompass various meanings. At international institutions, such as the European Parliament, rapporteurs are appointed to lead investigations and report back to the assigning body.46 The United Nations “Special Rapporteurs” take on a similar role, for example via appointment by the Office of the High Commissioner for Human Rights to investigate specific relevant issues. 47 However openly published guidance for rapporteurs appointed to assist in a meeting or workshop is not available. This is concerning due to the ubiquity of the role and its utility within multilateral systems of governance. The resulting lack of clarity, for how meeting reports are produced and published, then inhibits understanding of how openly available documents of international institutions come to exist. Although this paper does not attempt to define the role of a rapporteur in this context, it highlights a need for guidance to ensure defined good practice is applied in future events.

]]>
http://currents.plos.org/disasters/article/knowing-what-we-know-reflections-on-the-development-of-technical-guidance-for-loss-data-for-the-sendai-framework-for-disaster-risk-reduction/feed/ 0
Attachment, Bushfire Preparedness, Planning, and Response among Animal Guardians: A South Australian Case Study http://currents.plos.org/disasters/article/bushfire-preparedness-planning-and-response-among-animal-guardians-a-south-australian-case-study/ http://currents.plos.org/disasters/article/bushfire-preparedness-planning-and-response-among-animal-guardians-a-south-australian-case-study/#respond Thu, 02 Aug 2018 11:17:05 +0000 http://currents.plos.org/disasters/?post_type=article&p=38509 Abstract Background: Animal ownership has been identified as a risk factor for human survivability of natural disasters. Animal guardians have been reported to react or act in ways that may put their own safety and that of emergency services personnel at risk when faced with a natural disaster. Recent research has suggested that this risk factor could be reconfigured as a protective factor, whereby desires to save animals from natural disaster harm could motivate increased planning and preparedness behaviours amongst animal guardians. However, there has been no research to determine if bushfire planning and response behaviours differ between pet owners with low and high attachment; and how the relationship may differ in relation to small or large animals. Methods and procedure: We investigated the relationship between people’s emotional attachment to different types of pets and their preparation and actions during the Pinery bushfire in South Australia in November 2015. Thirty-four people who were impacted by the fire participated in an online survey. Data were collected about their preparedness, planning and response behaviours as well as their animal attachment (high or low). Results: We identified 10 characteristics (behaviours, attributes, skills and beliefs) associated with high animal attachment scores, and eight associated with low animal attachment scores. Discussion: Our discussion of the differences in demographics, preparedness, planning and response characteristics of participants with high and low animal attachment confirms research suggesting that animal guardians take risks to save their animals during disasters. Our findings also support recent propositions that animal attachment and ownership could be used to increase the natural disaster preparedness and survivability of animal guardians. However, making sure that animal attachment functions as a protective factor requires active and effective intervention through education, behaviour change and social marketing strategies. Whilst our study is high in ecological validity, future research with larger samples sizes is required to determine the generalisability of our findings to animal owners and guardians in other locations, facing fires with other characteristics, especially for owners and guardians with low levels of attachment.

]]>

Introduction

Animal ownership has been identified as a risk factor for human survivability of natural disasters. Animal guardians have been reported to react or act in ways that may put their own safety and that of emergency services personnel at risk when faced with a natural disaster 1,2,3,4. Recent research has suggested that this risk factor could be reconfigured as a protective factor, whereby desires to save animals from natural disaster harm could motivate increased planning and preparedness behaviours amongst animal guardians. Whilst the primary beneficiaries are humans and animals, improving the natural disaster survivability of animal guardians can also reduce demands on responders, emergency services and evacuation centres 5,4 ).

Most of the research on the impact of animals on human responses to natural disasters has focussed on the small animals that are most readily identified as pets, such as dogs and cats6,7,8,9 However, there are other larger animals with whom humans develop high levels of attachment, but who rarely share the same domestic spaces as humans. Often, small and large companion animals are present in a single household. A survey of 606 participants living in regional South Australia, for example, found that 74 percent of households were responsible for pets, of which 12 percent were responsible for horses or ponies and 7 percent were responsible for alpacas10. Another survey of horse owners in Australia determined that ‘the average numbers of [other] animals owned were two dogs, two cats, eight birds, two reptiles, 188 sheep, 27 goats or 45 ‘other’ animals’11 Attempting to evacuate from a fire threat with a horse is vastly different compared to a dog or a cat, and some households may be trying to evacuate both small and large companion animals – and not necessarily from the same property. Large companion animals are usually more difficult or even dangerous to handle and transport without specialised skills and equipment12 and emergency responders in Australia have noted that horse owners can be particularly difficult as they will do anything for their horses.3

Some of the reasons for the evacuation failure of small animal guardians have been practical, such as a lack of sufficient animal carriers or leashes7. From an emotional perspective, ‘animal attachment’ is often used to explain why people refuse to leave without their pets, or enter hazardous areas to ‘rescue’ them 7. It seems plausible that the degree of human risk taking to save animals would increase with the strength of the attachment. However, the implications of animal attachment for human responses to bushfires are not straightforward 13 and high animal attachment has not been found to be a predictor of evacuation14.

Still, there has been no research to determine if bushfire planning and response behaviours differ between pet owners with low and high attachment; and how the relationship may differ in relation to small or large animals. In relation to the preparedness stage of a natural disaster, there is a need to understand how attachment levels might be associated with animal identification, having evacuation kits, confidence administering first aid and perceptions of animals’ behaviours in response to fires. In relation to planning, there is a need to better understand how animal attachment characterises written plans and attention to fire warnings. In relation to response to actual fire threats, there is a need for ecologically valid data on how animal attachment impacts ultimate actions taken on catastrophic days, evacuating other people’s animals, evacuation without one’s own animals, perceptions of unnecessary evacuations and the driving rationale for research into the impact of animal ownership on human natural disaster survivability – risks people take to save animal lives.

We investigated the relationship between people’s emotional attachment to different types of pets and their preparation and actions during the Pinery bushfire in South Australia in November 2015.

The Pinery fire and affected areas

As illustrated in Figure 1, the Pinery fire burnt 85,000 hectares in cereal cropping/grassland country known as the northern Adelaide Plains, located 80 km north of the Adelaide metropolitan area. The fire caused two human fatalities with 31 injured, destroyed 97 houses and destroyed or severely damaged 546 non-residential structures and 412 vehicles and pieces of machinery were. More than 50,000 poultry and 18,000 livestock (mostly sheep) also perished along with 120,000 tonnes of cereal crops worth up to A$30 million15 .

Townships and populated localities within or largely within the firescar include Alma (population 75), Barabba (117), Fischer (62), Grace Plains (54), Hamley Bridge (717), Linwood (48), Magdala (53), Morn Hill (17), Pinkerton Plains (76), Stockport (263), Templers (125), Wasleys (722) and Woolsheds (29), with a total population of about 236016.

Nearby townships and localities and those partly within the firescar include Daveyston (75), Freeling (2214), Greenock (1087), Mallala (894), Nain (29), Owen (511), Pinery (102), Redbanks (182), Roseworthy (994), Stockyard Creek (20) and Tarlee (302), total population 641016 . Local Governments for the affected and adjacent areas include Mallalla District Council, Clare and Gilbert Valleys Council, Light Regional Council and Wakefield Regional Council.

The socioeconomic profile of the affected and surrounding townships and localities is somewhat lower than for South Australia as a whole, with 17.1 percent of the population holding a bachelor’s degree or higher and 10 percent employed in professional occupations, compared to 32.9 and 20.2 percent for South Australia respectively16. Agriculture and manufacturing are the main industrial base (together employing 26 percent of workers, compared with 12.5 per cent for South Australia)16 although as a peri-urban area, a considerable percentage of the working population is likely to commute to Adelaide or other small regional centres.

Method

Survey design

An online survey was targeted at pet owners affected by the Pinery fire, ie persons located within or near the fire. The survey was adapted from an extensive survey tool developed for use in post-fire event community taskforce research overseen by the Bushfire and Natural Hazards CRC17. Questions asking about specific preparedness actions relating to animals were informed by the literature on the topic18, as well as the experience and advice of industry partners from Horse SA, the SA CFS and RSCPA Qld. As a widely used and well-validated measure19,20,21 , the 23 item Lexington Attachment to Pets Scale (LAPS)22 was included to measure respondents’ attachment to their pets.

Most of the 130 questions were “tick box” format, including (up to) six questions with five-point rating scales and excepting seven questions requiring single word responses for each dog or horse owned or agisted. Twelve of the tick box options were multiple response, indicated by the instruction “tick all that apply” while the remainder required a single mutually exclusive response, eg “yes”, “no”, “don’t know” or “other”. If the “other” option was selected, respondents were able to provide optional explanatory detail. “Other” responses were coded to existing codes where appropriate. There were no unanticipated responses requiring additional codes. Not all questions were applicable to all respondents and respondents were automatically skipped to the next applicable question based on previous responses. The average time taken to complete the survey was 35 minutes.

The survey was constructed using the Qualtrics online survey platform www.qualtrics.com. Survey data were downloaded from the Qualtrics website in SPSS format and analysed in SPSS Version 23.

Recruitment

Residents of the Pinery area were invited to participate via a mailout of flyers to approximately 400 addresses in the affected areas which were targeted by the local postmasters who had personal knowledge and experience of affected streets, roads and households. The flyer explained the purpose of the research and directed householders to the online survey. Posters explaining the research with tear-off tags bearing the survey URL were displayed in local shops, post offices, community noticeboards and the Bushfire Recovery Centre located in the nearby town of Gawler. Short articles inviting affected residents to participate in the survey and including the survey URL were included in the local newsletter distributed by the Light Electoral Office (and endorsed by the Member for the electorate of Light), the Pinery Fire Recovery newsletter distributed by the SA and press releases sent to local newspapers and the four local governments with areas affected by the fire. The nearby Gawler Country Fire Service (CFS) also distributed flyers and promoted the survey through its various public activities, while one of the researchers gave a talk about the aims of the survey at a Fire Season Ready Family Expo held in Hamley Bridge, a township affected by the fire (see Figure 1). The project (and directions to the survey website) was discussed on ABC North and West and the survey information and link placed on the Central Queensland University media page. HorseSA promoted the survey to horse owners in the area through its mailing list and local social or interest clubs were tagged on Facebook with information about the survey and directions to the link. Participants were invited to provide their contact details if they wanted to be in a draw to win an Aud$50 voucher from Bonnetts Saddlery or PETStock.

Response rate

Despite the extensive recruitment efforts, responses to the survey were limited, with an initial response rate of ten percent, based on 58 responses from a population of approximately 600 pet-owning households. The estimated number of households with pets is based on the 920 households in the directly affected areas16 and an estimated pet ownership rate of 70% of households for a rural area in South Australia; the AMA23 cites 68% for total South Australia in 2016. After removing blank responses and responses indicating no consent, the number of cases was 34, for a response rate of six percent.

Analysis

Descriptive statistics (means, sums and percentages) were used to summarise the data. Differences between groups were compared using t tests for continuous data and Chi-Square (χ2) tests for categorical data. Differences in proportions for two independent groups were analysed using z tests. Correlations were measured using Spearman’s Rho and Kendall’s Tau where appropriate.

The data have been weighted to reflect the demographic composition of the population usually resident within and adjacent to the firescar using 2016 Census age and sex data16 .

The LAPS was scored by summing the scores for the 23 items on a 5 point Likert scale, after reverse scoring the two negatively worded items (allowing a possible range of 23- 115) and then converted to a percentage. Where there were missing values for some items, the percentage was based on the total number of completed items, with a minimum of three completed items.

The distribution of attachment to all pets in general was skewed toward the high end of the range with a mean score of 73 on the percentage scale (n=27). Due to the small number of cases and the location of a natural break between 60 and 69, 60 was selected as the cut off point demarcating high and low or lesser levels of attachment.

Figure 3 presents the mean number of animals owned by survey respondents using the total survey population as the denominator as well as the total number of owners of each species as the denominator. This means for example that on average, any Pinery resident is likely to own 2.6 horses, but the average number of horses owned by Pinery horse owners only, is 6.4. Horses dominate the large animal pet group on both measures, while the most commonly owned small animals are dogs, cats and chickens. The marked discrepancies between the means for rodents and ferrets, and birds, goats and sheep, shows that although these are not as frequently owned as dogs, cats and horses, those few respondents who do own them tend to have a large number of them in comparison with dogs, cats and horses.

Inclusion of animals in survival plans

Most participants (70% of respondents, n=24) had a bushfire survival plan of some kind (written or mental). All of these bushfire plans included animals. Fifty-one percent (n=12) included all animals in the household whilst 46% (n=11) included small animals only (excluding poultry). Those who included all animals had both large and small animals; the single exception had small animals only.

Respondents who had a written plan (n=10) had the largest number of animals with a mean of 19, compared to a mean of nine for those who had a plan in mind (n=14). Respondents who did not have a plan (n=5) also had a mean of nine animals.

A comparison of those respondents who did not include their pets in the plan (n=10) with those who did (n=24), showed that they tended to be much younger (85% were aged under 44, compared with 31% of those who did include their pets). They were more likely to be in casual or part-time work (86% vs 15%) and less likely to have vulnerable family members (children aged under 5, a person with a disability or a person over 60 requiring a carer) (7% vs 29%). There was no difference between these two groups in their ownership of large vs small animals.

Animal attachment scores

Almost one third (29%, n=8) of respondents who answered the attachment questions had low attachment levels ranging between 25 and 60 (X̅=45), and 69% (n=18) were classed as having high attachment (61 and above, X̅=92). An independent groups t test confirmed a statistically significant difference in attachment scores between these two groups; t(10.1-) = 7.4, p<.001, equal variances not assumed, (CI 32.5, 60.5).

The numbers are small with some missing data and so the results should be interpreted with caution, but with these caveats in mind, some indicative patterns emerge from Table 2. The species are listed in rank order of attachment with horses at the top and chickens at the bottom. The attachment levels for dogs, horses and cats are similar and it is possible that the true mean scores for cats and horses may differ from the numbers presented here due to the small numbers of completed responses.

The gender balance was heavily tipped towards males in the low attachment group (89% were male, n=8), biased toward females in the high attachment group (31% male, n=18) and also biased toward males amongst participants who did not respond to the attachment measures (75% male, n=8). Respondents with low attachment were much younger than the other two groups, with 89% (7 of 8) aged between 35-44, compared with 29% for the high attachment group (n=18) and 2% for the no attachment responses group (n=8). Seventeen percent (3 of 18) of highly attached respondents were aged 75 or older and there were no respondents in this age range in the other two groups.

Highly attached animal owners were more likely than low attachment animal owners to have a large number of animals in their care with a mean of nine animals in their care (ranging from two to 41) and a variety of small and large animals (48% had both large and small animals). Low attachment animal owners had a mean of three animals (ranging from two to nine) and only one of the eight had large animals. Respondents who did not complete any of the attachment questions had the most animals [mean of 23, (excluding one outlier with more than 30 rodents or ferrets)] and none had small animals only – all had both large and small animals.

An important aim of this study was to determine if and how preparations and responses to bushfire differ between pet owners with low and high attachment. Behaviours, attitudes, beliefs and skills are discussed below arranged around the stages of preparedness, planning and response. Most of the analyses address horses and other pets (mostly small animals) separately, given the different management needs of each group. Horses’ size, handling, behaviour and transport requirements and their management and evacuation require different responses from those of small animals (Thompson 2013). Other large animals (“lovestock”, i.e. pet pigs, sheep, goats, cattle etc) were not well represented in the sample so no clear patterns for this group could be established from the limited available data.

PREPAREDNESS

Animal identification

With the exception of cats, the attachment score for each species (i.e. dogs, horses and other pets as a group) is used as the attachment measure in the following analyses of animal identification, rather than the summary maximum attachment for all species. Maximum attachment for all species owned including cats is used as a proxy for cat attachment due to the low response rate for the cat attachment questions.

All respondents reported some form of ID, regardless of attachment level. Dogs owned by highly attached respondents more than twice as likely to have their dogs microchipped and have other forms of identification than low attachment respondents but there was no real difference between the two groups in the rate of council dog registration (z=0.77, p=0.44). Council registration was the most common form of identification, followed by microchipping, but rates of microchipping were low for both attachment groups. The use of other types of collar ID was very infrequent and only used by highly attached respondents (see Figure 4). Two respondents commented that their dogs were registered but that the dogs chew each other’s tags off. One of these respondents used microchipping and the other reported that one of her two dogs had a council tag but the other did not. Respondents who did not complete the attachment measures were most likely to have their dogs microchipped but least likely to have council registration tags for them.

Only one cat owner had a low attachment score and this respondent did not complete the identification questions (note that of the 22 cat owners, 3 completed the attachment questions and 21 completed the identification questions). The data presented therefore reflect the cat owners who completed the identification questions, regardless of attachment level.

Although 44% cat owners (n=9) had their cats microchipped, half (51%, n=11) had no form of identification for their cats at all. Almost none (one percent) had a collar ID and 13% (n=3) had other forms of identification such as photos or papers (Figure 5 ).

None of the respondents with horses had low attachment and so Figure 6 shows the ID choices for horse owners who are highly attached and those with no attachment score. Six of the nine (66%) highly attached horse owners’ horses were clearly branded. Five percent their horse(s) microchipped and two (18%) had other forms of identification such as photographs or papers. All respondents with horses described their horses as pets or retired, or for pleasure and/or recreation, rather than for breeding and/or racing. Respondents with no attachment score for their horses were much more likely to have horses with unclear brands and much less likely to have other forms of ID such as photos or registration papers. In total, almost all (99%) of horse owners had some form of identification for their horses.

Of the twenty-three respondents with animals other than dogs, cats or horses, three had low attachment scores. Twelve respondents without attachment scores completed the identification questions and their responses are included in this analysis.

Figure 7 shows that pets other than dogs, cat and horses were much more likely to have no ID and to be species that are difficult to individually identify, such as fish, rodents, frogs and reptiles and chickens. The species owned by the three respondents with low attachment were goats, sheep, birds, chickens and ducks, although the sheep had ear tags. The branded animals were a cow and a goat. In sum, species other than dogs, cats or horses are unlikely to have any form of ID, including some pet sheep, pigs, goats and cattle.

Evacuation kits

Counterintuitively, highly attached respondents were much less likely to have their small animal evacuation kits packed and ready to go (13%, n=18 vs 42%, n=5 for the low attachment group) but this rate should be viewed in the context of the small number of cases in the low attachment group and the type of pet. The low attachment group was more likely to have pet species for which evacuation kits (apart from pet food) were not relevant. Species such as fish, reptiles, rodents or budgies are generally already confined to portable tanks or cages and do not require leads, collars, ropes or items not already in their cages. Yet while 64% of the highly-attached respondents (12 of 18 respondents) had everything they would need, it was not actually packed and ready to go. Conversely, all respondents without attachment scores had small animal kits packed and ready to go (n=6).

The sole respondent with low attachment and a horse reported that they did not have a kit packed but had everything they would need. More than a third (41%, n=4) of highly attached respondents (n=9) did not have any items needed for an equine evacuation kit at all, while another 43% (n=4) said they had everything they would need but not packed and ready to go. All of the respondents with no attachment scores (n=5) had kits packed and ready to go.

As evacuation kits were generally reported as not being relevant to their types of pets, only three respondents from the low attachment group completed the question about items included in their small animal evacuation kit, precluding a statistically valid comparison with the other groups. However, the patterns within the high attachment group reveal two distinct clusters of items more and less likely to be included in the kit. The first cluster appears to be the basic essentials – all (n=8) packed enough cages or containers for all pets, 89% (7 of 8) of the high attachment group packed food for at least three days, 89% packed food bowls, feeders or dispensers, 88% packed leads or harnesses with identification tags and 75% packed water. The second cluster of less frequently packed items were items possibly perceived as non-essential, or impractical to include in a kit. These were vaccination certificates (27%) and dietary, medical or behavioural information (20%). A similar pattern was reported by respondents without attachment scores (n=6) with the exception that none packed food to last for three days.

There were insufficient responses to the question on items in horse evacuation kits to allow comparison (no respondents from the low attachment group and three from the high attachment group). All three high attachment respondents reported the inclusion of most items listed in the survey question in their horse evacuation kit (enough halters and leadropes, means for tying horses and water and feed buckets, woollen blankets and towels and wire cutters or a sharp knife, first aid items, and tools for identifying horses such as paints or tags). Only one of the three included a hose and identification in the form of ownership papers or other documents. All respondents in the no attachment score group (n=6) included all listed items except for identification in the form of ownership documents (this item was not selected by any of these respondents).

Confidence administering first aid

Half (50%, n=8) respondents with higher attachment (n=17) were moderately, very or extremely confident in small animal first aid and the rest were somewhat confident, but all of the six respondents in the low attachment group who answered this question were at least moderately confident. All of the five respondents in the no attachment score group were very confident. That is, level of confidence in giving first aid to small animals was inversely associated with level of attachment (τc = -.4, p =0 .02). Further, the no attachment score group (n=6)was significantly more confident in giving small animals first aid than either of the attachment groups, based on the residual values with Bonferoni adjustments (χ2 =20.9(6), p=0.002).

None of the low attachment respondents completed the question on confidence with basic first aid for horses, but three of the eight highly attached respondents who completed the question reported that they were not at all confident (43%). Thirteen per cent reported that they felt very or extremely confident. All of the respondents with no attachment score were very confident (n=6). A Chi Square test showed a significant difference in confidence levels between the high attachment group and the no attachment score group (χ2 = 7.82(3), p=.03).

Perceptions of animals’ instincts in fires

The small number of cases (n=6) in the low attachment group means that caution is advised in interpreting the comparison with the high attachment group (n=17), but it is worth noting that 88% of the low attachment group thought that animals have intuition about fire, while only 16% (3 of 17) of the high attachment group held this view. Two thirds (66%) of the high attachment group though that animals do not have any fire intuition and 19% did not know. Respondents with no attachment score (n=6) were also unanimous in their agreement that animals do have intuition about fire. The comments indicated that respondents from all groups felt that even if animals generally have intuition about fire, they recognised that there may be individual variation – i.e. some animals do have instinct or intuition and others do not, and that even where animal do have that instinct, they may be prevented from acting on it:

“My dogs were all locked in runs so unable to go anywhere”;

“My horse just stood still and watched the fire spread across the paddock”;

“I think most would be just as lost as people”;

“Generally yes if they can escape – however individual personalities may impact e.g. my most nervous dog would have to depend on the other dogs to lead her to safety – she would just cringe somewhere or run in any direction.”;

“They will find shelter if they can escape”;

“Some do but fences etc limit it”;

“Yes, but may not be able to escape due to fencing”.

Planning

Written plans

Highly attached respondents (n=18) were actually less likely than low attachment respondents to have a written plan (5%, vs 30% of the low attachment group and 77% of the group with no attachment score [n=8]). Just over half (56%) of the high attachment group had no plan at all while the other two groups had either written plans or plans in mind. Forty four percent of the highly attached group had either a written plan or a plan in mind, compared with all of the low attachment group and 97% of the no attachment score group.

Familiarity with the area is known to influence propensity to plan (Smith, Taylor and Thompson (2015). Most respondents were long term residents; 70% of both attachment level groups had lived in the area for more than five years. Highly attached respondents were much more likely than lesser attached respondents to have lived in the area between 11 and 20 years (44% vs 2%) while respondents with no attachment score had lived in the area the longest, with 84% living there between 11 and 20 years.

Attention to fire warnings

All respondents (regardless of attachment level) indicated that they pay attention to fire warnings.

RESPONSE

Actions on catastrophic days

The different attachment level groups acted differently on catastrophic days: most respondents listen for fire warnings although at 55% the low attachment group was least likely to do this (Table 3). Approximately half (51%) of the high attachment group also said they actually do nothing, even if they listen to the warnings. Highly attached respondents were much more likely to stay at home or close to home (37%) compared to low attachment respondents (0%) but virtually all respondents with no attachment reported stayed close to home. About one third (32%) of highly attached respondents reported getting their evacuation kit ready, but no low attachment respondents reported this action.

Evacuating other people’s animals

Four per cent (one case of 17) of the highly attached respondents evacuated pets that they did not have primary responsibility for, while 42% (n=2) of respondents in the low attachment group (n=6) did so. None of the no attachment score group evacuated others’ pets (n=5). Although approaching statistical significance, a Chi Square test showed no differences between the three groups (χ2=5.62(2), p=0.06).

Highly attached respondents were also less likely to evacuate other people’s horses (9%, n=1 of 17) compared with two of the five respondents in the low attachment group and none of the no attachment score group. A Chi Square test showed that the differences between the three groups are not statistically significant (χ2=5.36(2), p=0.07).

Evacuation without animals

One respondent in the low attachment group responding to this question (n=3) evacuated without their pets. None of the 17 high attachment respondents who answered this question evacuated without their pets, although three clarified their experience with the following comments:

“I had to leave the chicken”;

“I wasn’t home”;

“My partner evacuated without our pets”.

This last comment suggests that the respondent herself would not have evacuated without her pets if she had had the option and also that individual members within households may have different views on appropriate actions in bushfire situations.

None of the low attachment respondents answered the question about evacuating without their horses. Of the seven high attachment respondents who answered this question, four left their horses behind, two did evacuate with their horse/s, and one evacuated with some horses but had to leave two behind.

Unnecessary evacuation

No respondents in the sample felt they had wasted their time or money evacuating before finding out that their property was not under threat, but only one of the three low attachment respondents and two of the seven high attachment respondents who answered the question felt they had still done the right thing. None of the low attachment respondents felt that evacuating had been a good practice run for the future or felt relieved that they had evacuated, but two of the seven highly attached respondents thought it had been a good practice run and five of the seven (74%) felt relieved.

Risk to personal safety

No respondents in the low attachment group answered the question “did your actions to protect your small animals involve any personal risk to your own safety?” Only eight respondents in the high attachment group answered the question, of whom two reported that they risked their own life while acting to protect their small animals. By contrast, all of the six no attachment score respondents reported risking their lives to save their small animals. These different response patterns were significantly different (χ2 (2)=7.88, Fisher’s exact p=0.01. Only three of the high attachment respondents who risked their lives rated the level of risk; one judged it as high, one as medium and the last as low.

A somewhat different pattern emerges for respondents acting to protect their horses – six of eight highly attached respondents took risks to protect their horses, and four of six rated the risk as high. None of the low attachment respondents took personal risks to their own safety to protect their horses. All of the no attachment score respondents took high risks to protect their horses. There was no statistically significant difference between these two groups in their actions to protect horses (χ2 (2)=2.86, p=0.24).

Discussion

Participants

The age and gender distribution of our sample is typical for both postal and online surveys 41,42 . The pet ownership rate for dogs and cats follows the usual pattern of being the most commonly owned species 23,24, although given the rural location, the rate for horses and chickens was comparatively higher than reported elsewhere. Excepting the greater representation of horses and the inclusion of other large animals, the distribution of pets by species is similar to the distribution reported by Taylor et al (2015:19)4 for respondents in both urban and rural areas.

All pet owners with plans included their pets in their plans – but not all pet owners had plans. At 30 percent, the percentage of respondents with no plan is much higher than the eight percent reported by Trigg, Smith and Thompson 13 for South Australian communities directly and indirectly affected by three large bushfires in January 2014 and the 19% reported by McLennan, Patton and Wright43 . The pattern of most respondents having only a mental plan (42%) rather than written (29%) is similar to Trigg, Smith and Thompson’s findings13 (65% and 19% respectively), but six times higher than a national average of 5% found across several post-bushfire community surveys. The propensity for a mental plan rather than written was also held by the South Australian livestock producers surveyed by Smith, Thompson and Taylor3.

Whilst modest in size, this study of 34 pet owners affected by the Pinery fires identified 10 characteristics (behaviours, attributes, skills and beliefs) associated with high animal attachment scores, and eight which were associated with low animal attachment scores. As shown in Table 4, participants with a high animal attachment score were more likely than low attachment respondents to: be female, be older, own a horse, dog or cat; have more than the average number of animals, have dogs who are microchipped, stay home or close to home on potentially catastrophic fire risk days, evacuate with their horses, and take personal risks to their own safety to save horses but not small animals. Participants with low animal attachment scores were more likely to be male, younger, own animals other than dogs, cats or horses, not have an evacuation kit, be confident giving first aid to small animals and horses, perceive that animals have intuition or instincts about responding to fire, evacuate small animals and horses for which they did not have primary responsibility and not see their evacuation as good practice or feel relieved that they had evacuated unnecessarily.

The finding that highly attached respondents (who were also most likely to have horses) were less likely to have plans reflects the findings of Smith, Thompson and Taylor’s3study of emergency responders’ experiences of large animal rescue. ERs reported that most large animal owners had no or insufficient plans and that horse owners, in particular, are highly attached, which can equate to extreme emotions interfering with decision making and action during an emergency.

Gender may be a moderating factor in the relationship between attachment and level of confidence in giving first aid to both large and small animals, based on the well known relationship between gender and confidence25,2627,28,29 However, an ordinal logistic regression analysis showed that although attachment level was a significant predictor of confidence in giving first aid (proportional odds ratio of 27.5, 95% CI,0.554-6.08,Wald (X2(1)=5.54, p=0.019), gender was not (odds ratio = 1:0.85, 95% CI,4.28 – 0.59, Wald X2(1)=2.2,p=0.137).

Characteristics of participants with high animal attachment scores

This study found that high animal attachment scores were associated with ownership of dogs, cats and horses. Most attachment studies focus on dogs and to a lesser extent cats. Comparative quantification of human attachment to horses and other small animals have been neglected in the literature, with most existing research addressing the attachment of horses to humans in the human-horse dyad30 or how humans interact with horses44. This finding confirms previous arguments that human-horse relationships share similar emotional attachment to companion animal relationships, despite horses not sharing domestic spaces of humans31.

Attachment levels for dogs, cats and horses were on par with other recent studies of pet owners experiencing natural disasters. Brackenridge et al14 reported an equivalent mean LAPS attachment score of 76 which compares favourably with the scores ranging from 70 for cats, 77 for horses and 71 for dogs in the present study. Taylor et al4 reported an average attachment score of 9.76 on a ten point scale for all species, or 97.6 in percentage terms compared with 73 in the present study, but this difference may reflect different study populations or instruments used to measure attachment.

Our study also found higher animal attachment scores reported by those with more than the average number of animals.

We found that older age groups and females were more likely to have high animal attachment scores. This is consistent with a study by Brackenridge et al14 who also used the LAPS and found a gender difference, as well as Bagley and Gonsman32 who reported a positive relationship between attachment and age. However, this finding could be peculiar to our sample, as Herzog’s review33 of twelve studies addressing gender differences in pet attachment found a small gender difference overall (invariably females feeling stronger attachment than males) although many studies found no difference32,34,35.

Unexpectedly, we found that those with high animal attachment scores were less likely to have prepared written bushfire survival plans. While this finding appears counterintuitive, it may be that high levels of attachment are associated with other factors negatively affecting propensity to plan, such as cognitive dissonance between currently enjoying and maintaining pets’ wellbeing and thinking about uncontrollable risks to their wellbeing, or gendered power structures within households determining if and how plans are made. Smith, Thompson and Taylor3 reported the comment of an emergency responder who also agisted other people’s horses: “many [horse owners] are [not] able to deal with the emotions involved in planning for their best friend ….. Some refuse to think about it and others over react”. More detailed research is needed to unpack the possible psychological aspects of relationships between planning and attachment, and between planning and gendered household decision making.

The association between high attachment and low rates of planning does not necessarily undermine the idea that animal attachment can be used from a social marketing perspective to motivate preparedness 12,36,45. The question becomes how to acknowledge or account for the factors in an indirect relationship. Another important question is how to motivate pet owners with low attachment to prepare and act to include their pets and how to target them, given they are less likely to have ID for their animals, especially dog registration. Lower levels of attachment may also be associated with low rates of using veterinary care for their animals between disasters, which would preclude veterinary surgeries as an avenue for contact. Similarly, they may also be less likely to frequent pet supply stores or be involved in organised animal related clubs such as breed societies or dog obedience. More research is necessary to compare other animal-related leisure, shopping, or social behaviours and attitudes of highly attached and low attachment pet owners. Beverland, Farrelly and Lim37 argue for example that there are two main types of motivation for pet ownership – pets as companions to love, versus pets as toys, status markers and brands. Further, the different motivations of these two groups, which may be analogous to our highly attached and low attachment groups, affect their appreciation of the pet and the purchase of pet related paraphernalia. A more directive or prescriptive approach rather than relying on a strategy of providing information and raising awareness may be necessary for the low attachment subgroup of pet owners – if they can be reached. The prospect of prosecution under the Animal Welfare Act 1985 (SA) for failure to take reasonable action to protect or evacuate pets (admittedly difficult to police and prove either before or after a natural hazard) may also motivate some pet owners to give greater consideration to their pets.

Those with high animal attachment were more likely to remain at home or stay close to home on potentially catastrophic fire risk days. In relation to horse owners, further research may determine if this was because they were more likely to attempt to evacuate with horses, which is less practical than evacuating with smaller animals (Smith et al., 2014). Furthermore, it was not clear if attempts to evacuate with horses were directly related to those with higher animal attachment taking personal risks to their own safety to protect horses.

Evacuation drills or simulations for horses in specific areas have been reported to be effective overseas in identifying potential difficulties able to be addressed before an emergency occurs. While it may be logistically difficult to arrange on a community wide scale, it is possible for organisations such as pony clubs and adult riding clubs to conduct simulated evacuations. HorseSA reports that the protocol for a similar drill exercise has been available on the HorseSA website for more than three years but has never been downloaded (Fiedler, Pers. Comm. December 6, 2016). Perhaps such drills should be mandatory for all animal based organisations. Pet or horse owners who are not members of any animal related organisations can still be invited to participate in drills or given information via social media or other local community advertising such as posters and flyers from councils sent to post boxes.

Those with high animal attachment scores seemed more likely to engage in behaviours associated with responsible animal guardianship, such as preparing evacuation kits for small animals and horses as well as attaching council registration discs attached to their dogs’ collars and having microchipped dogs.

The identification rate for cats at 65% was low, even for highly attached cat owners, compared to the rate for dogs. The only formal or centralised means of identification for cats is via microchipping, which applied to only 42% of cats in this study – and these were cats owned by highly attached owners. The microchipping rate of cats in the general population of South Australia is currently unknown (but note that microchipping of dogs and cats in South Australia will be compulsory as from July 1, 201838. The Australian Veterinary Association39 estimates that the microchipping rate for cats is 72% nationally (including those states without compulsory microchipping). Further, dogs have much higher rates of other forms of identification, mainly due to council dog registration requirements. In addition to microchipping, council registration for cats could be supported on the grounds of identification in emergency situations (including although not necessarily natural disasters).

Horse racing and equestrian sports’ requirements for identification and council dog registration requirements have benefitted emergency response and reunion after natural disasters. Other species may also benefit from stringent, centralised means of identification – particularly cats, but other small animals including birds, rabbits, ferrets, reptiles, tortoises and even fish can also be microchipped. At present, public awareness that these species can be microchipped is low. An animal identification promotion campaign may be useful. Such a campaign could be targeted to children (as well as the general adult population) to leverage their pester power and emotional ties to pets, even where parents’ attachment is low. Pressure from children may be particularly useful for species such as guinea pigs and rabbits, which are typically children’s pets. The ability to identify lost or stolen animals is likely to be the main motivation for microchipping such species but as with dogs and horses, will also be of benefit in the event of natural disasters. Note that microchipping is already mandatory for cats in all Australian States except South Australia, the Northern Territory and Tasmania.

Characteristics of participants with low animal attachment scores

Those with low animal attachment scores were more likely to be young and male. They were also more likely to own animals other than dogs, cats or horses. This pattern is consistent with the literature discussed above.

People with low animal attachment scores were more likely to be confident administering first aid, although most pet and horse owners reported that they felt very confident in administering first aid regardless of their level of attachment. However, this self-report survey cannot reveal if this confidence represents objective abilities in administering first aid.

Almost half (42%) of highly attached respondents reported that they were not at all confident treating horses. To confirm first aid abilities for all horse and pet owners, online information and other resources on first aid for animals during or after typical bushfires and other natural hazards can be developed. Such resources can be packaged for dissemination via social clubs or community organisations as well as groups such as dog obedience clubs, cat fancy clubs, riding clubs or pony clubs.

Those with high animal attachment scores were more likely to believe that animals have instinctive behaviours towards fires. This would seem to explain why they are less likely to evacuate with horses than those with high animal attachment scores. However, there is a need to understand if and how animal attachment might be related to beliefs about animal’s own fire-related instincts of ‘sixth sense’ for natural disaster survival. In particular, it might be necessary to systematically compare guardians’ beliefs with ethological knowledge about animal behaviour when facing bushfire threats or responding to other environmental hazards. The findings could have a significant impact on the ways in which owners of horses and other animals plan and prepare for and respond to natural disaster threat. Indeed, research on another fire in South Australia cases of horse guardians taking their cues to respond to a fire threat from their horses’ reactions and behaviours46.

Those with low animal attachment scores were more likely to evacuate pets and horses for which they did not have primary responsibility. The context for these evacuations requires further investigation, especially as the same group of people (with low animal attachment) are more likely to not have an evacuation kit and may have little or no experience with species such as horses, given their propensity to have small animals only (only one of eight respondents with low attachment had horses for example). This uneasy combination of apparent willingness to save other people’s animals without preparations may explain why this group was more likely to neither see their evacuation as good practice, nor feel relieved when their evacuation had been unnecessary (ie when in retrospect the fire threat did not seem to warrant evacuation). As noted elsewhere in relation to three major South Australian fires in South Australia,46 precautions may be required to ensure that evacuations are praised even if they may be described as unwarranted, lest they discourage pre-emptive evacuation in response to future fire threats.

Limitations and further research

The main limitation in this study was the small sample size and low response rate, with a possible bias toward more highly attached pet owners, who are more likely than low attachment pet owners to respond to a survey about pets. Whilst the strength of this small survey lies in the ecological validity of the Sampson Flat fire event, greater statistical power in future studies with a larger sample size may confirm statistical significance of some of the apparent differences evident in the current study and indicate their generalizability to other animal owners in regional areas of Australia.

Another limitation is the use of the LAPS for respondents who have many animals of the same or different species. Some respondents failed to complete all or any of the 23 item LAPS in this study, particularly cat and horse owners. The response rates for cats and horses may be an artefact of the survey structure, as the LAPS for cats and horses followed the measure for dogs, which were filled out by most dog owners. The LAPS may be too onerous or tedious for respondents with more than one or two pets to complete and thus may not be the best scale to measure attachment in rural areas where respondents are more likely to have multiple animals. Further, some of the LAPS items, such as “ I believe that ____ should have the same rights and privileges as family members” may not readily apply to species such as chickens, reptiles or fish, for most people.

Another issue requiring further attention is whether respondents who do not fill out attachment questions have similar, more or less attachment than respondents who do answer these questions, given that they formed 22% (n=8) of the 32 cases and also have more animals than either the high or low attachment groups. It is possible of course that the more animals a respondent has, the less likely they are to complete the 23 item LAPS due to the considerable effort required. Although the high attachment group also had large numbers of animals, the patterns of responses for this group sometimes resembled those of the low attachment group, while at other times were similar to the high attachment group.

Conclusion

Despite our modest sample size, we were able to distinguish participants with high and low animal scores according to a total of 18 characteristics. Participants with a high animal attachment score were more likely to be female, older, own a horse, dog or cat; have more than the average number of animals, prepare small animal and horse evacuation kits, have dogs who are microchipped and wear a council registration disc on their collar; prepare written plans, stay home or close to home on potentially catastrophic fire risk days, evacuate with their horses, and take personal risks to their own safety to save animals. Participants with low animal attachment scores were more likely to be male, younger, own animals other than dogs/cats/horses, not have an evacuation kit, be confident giving first aid to small animals and horses, perceive that animals have intuition or instincts about responding to fire, evacuate small animals and horses for which they did not have primary responsibility and not see their evacuation as good practice or feel relieved that they had evacuated unnecessarily. Whilst our study is high in ecological validity, future research with larger samples sizes is required to determine the generalizability of our findings to animal owners and guardians in other locations, facing fires with other characteristics, especially with regard to owners and guardians with low levels of attachment.

Competing Interests Statement

The authors have declared that no competing interests exist.

Data Availability Statement

The data underlying this study have been uploaded to figshare and are accessible using the following DOI: 10.6084/m9.figshare.6297629.

]]>
http://currents.plos.org/disasters/article/bushfire-preparedness-planning-and-response-among-animal-guardians-a-south-australian-case-study/feed/ 0
Community Perspective on Policy Options for Resettlement Management: A Case Study of Risk Reduction in Bududa, Eastern Uganda http://currents.plos.org/disasters/article/community-perspective-on-policy-options-for-resettlement-management-a-case-study-of-risk-reduction-in-bududa-eastern-uganda/ http://currents.plos.org/disasters/article/community-perspective-on-policy-options-for-resettlement-management-a-case-study-of-risk-reduction-in-bududa-eastern-uganda/#respond Thu, 26 Jul 2018 14:08:50 +0000 http://currents.plos.org/disasters/?post_type=article&p=35190 Introduction: Despite existing policy actions on Disaster Risk Reduction (DRR), many community members in Bududa still continue to settle in high-risk areas re-zoned for nonsettlement. There seems to be an apparent information asymmetry on expectations between the community and Government. The challenge then is ‘how to consult communities and seek their opinion in an adequately representative unbiased way’. This paper sets out to explore policy options on resettlement management as a DRR approach and how engaging with communities in a public discourse using the Deliberative Polling (DP) approach; to obtain their opinions and insights on these policy issues, revealed underlying challenges to policy implementation.

Methods: A qualitative study was conducted in Bududa in eastern Uganda with fourteen group discussions; comprising 12-15 randomly assigned participants of mixed socio-economic variables. Trained research assistants and moderators collected data. All discussions were audio taped, transcribed verbatim before analysis. Data were analyzed using latent content analysis by identifying codes from which sub-themes were generated and grouped into main themes on policy options for resettlement management.

Results and Discussion: We used Deliberative Polling, an innovative approach to public policy consultation and found that although the community is in agreement with most government policy options under resettlement management, they lacked an understanding of the rationale underlying these policy options leading to challenges in implementation. The community members seemed uncertain and had mistrust in government’s ability to implement the policies especially on issues of compensation for land lost.

Key Words: Policy, Deliberative Polling, Climate change, risk-reduction, landslides, Uganda

]]>

Introduction

Disaster incidents are on the increase globally in frequency, intensity and duration especially in the advent of climate change/variability manifested as floods, landslides, drought and glacial runoffs among others 1. This has been worsened by the unpredictable nature of these events. Climate variability is attributed directly or indirectly to human activity that alters the composition of the global atmosphere 2.

In Uganda, landslides are one of the devastating effects that have been faced due to climate variability. Landslides usually occur in hilly terrain and are triggered by persistent rainfall 3,4. Bududa district of Uganda has a history of landslides, attributed to the hilly terrain of Mt Elgon. Bududa receives an average precipitation of above 1, 500 millimeters (mm) per year (meaning it basically rains every day) triggering landslide occurrence 3. This is worsened by the ever-increasing population which puts much pressure on the land 3. The current population is about 227,400 inhabitants with a population density of 906.7 persons per square kilometre, four (04) times higher than the National average, making it the most densely populated region in Uganda 6. The occupants exploit the slopes of Mt Elgon for settlement and agriculture often causing land degradation 7,5

The most devastating landslide in Uganda occurred on 1st March 2010 in Bududa District8,9,10,11.The landslide was triggered by heavy rains that lasted over three months. The landslide buried three villages in Bududa, killing over 400 and displacing an estimated 5,000 people. The landslide led to an immediate breakdown of water and sanitation systems predisposing affected people to disease outbreaks such as cholera8,11.

Following the landslide, several policy recommendations and options were issued in the Uganda National Policy for Disaster Preparedness and Management (2010) 12. The options included resettlement of affected people, re-zoning of the high risk areas for no settlement, compensation of victims, voluntary relocation and establishment of early warning systems many of which were enforced by the relevant authorities 12

In Bududa, resettlement was applied as a long-term risk reduction solution. It involved resettlement of people away from high-risk areas 12,13. Resettlement of affected persons from Bududa was implemented by the Government of Uganda 12. Affected communities were relocated to Kiryandongo district, in western Uganda majorly because of the availability of vast lands 10. This also included the displaced persons who were temporarily taking refuge in Internally Displaced Peoples (IDP) camps at Bulucheke sub-County headquarters in Bududa district 10. However, despite the availability of vast land, It is important to appreciate the significant contextual backgrounds and differences between these two districts; socially, culturally and economically 22. Communities in Bududa are used to settling in the highlands as compared to those in Kiryandongo who are used to settling in low lands. The cultural practices of these two peoples are also different; in Bududa the annual male circumcision”Imbalu”, a rite of passage signifying transition of the young boys to manhood is celebrated and held to such high esteem because it constructs the Bagisu identity, while those in Kiryandongo do not practice these cultural practices, making it difficult for the Bagisu to fit in 22. Economically, the people from Bududa are farmers owing mainly to the highly fertile volcanic soils while Kiryandongo has less fertile soils and are mainly mixed farmers but predominantly herdsmen 22.

Currently, after years of implementing the policy on resettlement of people in Bududa, this policy has not yielded the required outputs. Many Bududa community members still continue to settle in high-risk areas rezoned for non-settlement and many previously relocated to Kiryandongo have returned to the same affected landslide stricken area 14.This poses the question: why has the resettlement policy failed in this vulnerable community?

One of the reasons for the failure of this policy could be the ineffective consultation of the affected communities prior to the implementation of the policy. The Uganda National Policy for Disaster Preparedness and Management was developed by the government through a process of conducting consultations at all levels using local leaders at community level, through the district leaders to stakeholders at national level 12. However, it seems that there is information asymmetry between the local community and government expectations regarding resettlement as a risk reduction policy. This has contributed to an apparent complacency about the proposed policy measures in these communities.

In many countries, public consultation during the policy making process does not adequately involve the communities right from the initial stages 15. Governments often use subjective assessments of situations to craft policies for risk reduction. Although the bottom-up approach may seem effective in policy formulation, only a selected few in leadership positions are consulted upon 16

In the bottom-up approach currently being used in Uganda, the community members do not have the opportunity to carefully think through the issues, be educated upon and make an informed decision hence the community members lack the right information on issues affecting them. There is a need to bridge this gap in information asymmetry by devising better ways of public consultation.

The challenge then is ‘how to consult the communities and seek their opinion in an adequately representative unbiased way’. In order to counteract this challenge, we used a Deliberative PollingÒ (DP) approach. DP offers an innovative tool in which a representative sample of the community can be consulted in depth on key issues. It provides representative and informed opinion data, both quantitative and qualitative, about the views of the public once they have really considered the issues 16

Deliberative PollingÒ(DP) in essence assesses the representative opinions of a population 17. The premise of Deliberative Polling is that when policy options are important for a community, then public consultations about them should be representative of the population and thoughtfully based on the best information available 16,18.

The first ever successful DP in Africa was conducted in Bududa district in the Mt Elgon region on 7th-8th July 2014. In this paper we examine why there are unsuccessful efforts by the government to effectively communicate the rationale underlying the current policy on resettlement. The reasons were derived from consultation with the community using the DP approach.

Ò Deliberative Polling® is a registered trade mark of James S. Fishkin. The trade mark is for quality control and benefits the Stanford Center for Deliberative Democracy.

Methodology

Study Location

This paper focuses on the DP proceedings from Bududa district. Bududa district is located in Eastern Uganda, bordering Kenya to the east, Manafwa district to the south, Mbale to the West and Sironko the north. The district is mostly mountainous with an average altitude of 5,900 ft above sea level. The area has been prone to landslides that have been catastrophic . The population is mainly Lumasaba speaking 19.

Study Design

The entire design of the DP process involved both Quantitative and Qualitative methods. The Qualitative methods assessed whether there was a policy change in attitudes regarding the policy options both before and after the plenary sessions while the Quantitative methods (handled in a separate research paper) assessed to what extent and their levels of significance. This paper focuses on the discussion before the plenary session. The group discussions were conducted on the 7th and 8th July 2014 in Bududa District. The plenary session is a session where all participants convene and pose questions from the group discussions to a group of experts and policy makers. It provides a platform for transparency, accountability, knowledge dissemination and learning.

For this paper, the study design was a case study. According to Thomas (2011), “Case studies are analyses of persons, events, decisions, periods, projects, policies, institutions, or other systems that are studied holistically by one or more methods 20. In this instance, the case is the Bududa participants who came for the DP and we put them into group discussions. This is because it is from these group discussions that participants’ opinions regarding the various policy options were captured.

Participants and sampling

Participants for this study were recruited by random selection of households and random selection of those within the households 18. The DP participants were originally selected through a three-stage sampling technique. During the first stage, 7 sub-Counties from Bududa district were randomly selected: three sub-Counties from the high-risk areas, two from moderate risk and one from low-risk areas. The sub-Counties were simple randomly selected. In the second selection stage, three parishes from each sub-County were selected using simple random sampling technique and the sample size for the district was then allocated to the 21 parishes proportionate to their population sizes. In the third and final stage, participants aged 18-75 years were randomly selected from the parishes. A list of the households, and their adult occupants in each of the selected parishes was compiled by community scouts identified in the respective parishes and guided participant selection. The selection of the sub-Counties was guided by Bududa District Disaster Management Plan 2013, which stipulates that ten sub-Counties are high risk, five are medium risk and one Sub -County; low risk of landslides. One sub-County of low risk was automatically selected and the remaining fifteen sub-Counties were subjected to a ratio of 1:2 hence two sub-Counties for moderate risk and five sub counties for high risk sub-Counties respectively.

The total sample size for this study was 208 participants. In conducting the Deliberative Poll, the random sample first completed a baseline survey in order to collect information about community perception and ranking of importance on the specific policy proposals from stakeholders. The survey respondents were then invited to participate in a DP meeting to deliberate face to face on their understanding and concerns regarding proposed policy options. Originally, the participants of the group discussions were sampled using simple random sampling at household level in the different communities for a quantitative survey. It is from the DP meeting that we purposively selected participants for the group discussions. A total of 14 group discussions of 12-15 participants each were conducted. Figure 1 shows the schematic illustration of the DP process in Bududa.

Fig 1-Schematic illustration of the DP process in Bududa (Source Authors’ own)-Revised_001

Fig. 1: Schematic illustration of the DP process in Bududa.

Data collection methods and procedures

During the deliberations, participants focused on the pros and cons of the policy proposals and arrived at key questions they wished to pose in the plenary session of experts.

The guide used to moderate the discussion focused on the pros and cons of the policy proposals and arrived at key questions they wished to pose in the plenary session of experts. The guide used to moderate the discussion focused on the policy options around: resettlement management, as an option that can be taken to reduce the damage of landslides. Under resettlement management, the proposals of discussion included re-zoning high risk areas for no settlement, compensation for relocation, resettlement in newly built peri-urban centers, temporary resettlement after a disaster, building an early warning system, supporting local disaster management committees, use of sirens in the early warning system and use of text messages in the early warning system.

Selection and training of Moderators

Fifteen research assistants were recruited and trained to facilitate group discussions. The selected research assistants had a minimum of a bachelor’s degree and prior experience in research specifically qualitative research-interview skills. They were knowledgeable in Lumasaba the local language commonly spoken in the district. They were equipped with digital audio recorders to record the group discussions. The training of moderators was jointly conducted by experts from the ResilientAfrica Network, Stanford University and a faculty member from Makerere University School of Public Health.

Data analysis

Data collected through the group discussions were transcribed verbatim and those in the local languages translated without altering the meaning. A content analysis approach was used as described by Graneheim and Lundman (2004) 21. Analysis was done in two stages, first, the manifest content analysis (what the text says, deals with the content aspect and describes the visible, obvious components) and then the latent content analysis (what the text talks about, deals with relationship aspects and involves an interpretation of the underlying meaning of the text). The transcripts were read and re-read by the authors who then assigned codes and came up with a coding structure (Open coding). Data meaning units were then aligned under their respective codes. This was followed by axial and selective coding to develop higher codes and categories. Categories were reviewed further to develop overarching themes.

Ethical considerations

Ethical approval was obtained from the Makerere University School of Public Health Higher Degrees, Research and Ethics Committee and approval from the Uganda National Council of Science and Technology (UNCST) [study number SS 3532]. Permission to carry out the research was further sought from Bududa District Administration. Verbal consents were obtained from the participants and a request was made to audio-record the discussions. Study objectives, benefits and risks were explained to our respondents. In addition, respondents had the opportunity to ask questions or clarification before consent for the discussion to proceed. All information obtained during the study was treated as confidential.

Results

In this section we describe the thematic structure of our analysis, showing the main themes and sub-themes regarding policy options for resettlement management. Our key themes were relocation from high to low risk areas, relocation to relatives, compensation for resettlement and risk communication.

Participants

We conducted fourteen group discussions where participants were assigned randomly to groups comprising 12-15 participants of mixed gender (58.7% male and 41.3% female); 90% were married,57.7% primary education,10.4% had no education;86.6% were farmers, and the average number of children per woman was 6.3 as shown in Table 1.

Table 1: Demographic characteristics of study participants
Variable Number Percentage (N=208)
Sex:
Male 122 58.7%
Female 86 41.3%
Marital Status:
Married 187 90.0%
Single 10 5.0%
Separate/Divorced 3 1.5%
Widowed 7 3.5%
Highest Level of Education:
None 22 10.5%
Primary 120 57.7%
O Level 58 27.9%
A Level 2 1.0%
Tertiary 6 2.9%
Occupation:
Farmer 180 86.6%
Professional/technical/managerial 6 3.0%
Entrepreneur(business owner) 7 3.5%
Merchant 2 1.0%
Teacher 4 2.0%
Student 4 2.0%
Other 4 2.0%
Average Number of Children: 6.39

The analysis identified within the three policy options for resettlement management as: (i) resettlement with support for livelihood and in the same community, (ii) Modalities of compensation; (iii) risk communication as early warning favourable to save life. These themes are described in detail in the next section of the article. Figure 2 shows the thematic structure of the research findings based on Gioia (2013) 23.

Fig 2-Thematic structure of the research findings (Gioia ,Corley et al 2013)-Revised_001

Fig. 2: Thematic structure of the research findings (Gioia ,Corley et al 2013)

Relocation from high to low risk areas

The first main theme was resettlement from high to low-risk areas. Low-risk areas would involve being relocated from the high risk mountainous areas characterized by cliffs, steep colluvial deposits and scars due to previous landslides to low risk areas which are low lying, receive less rain to trigger landslides and relatively safer. These low risk areas mainly have trading centres and are inhabited mainly by relatives. In all the group discussions held, the issue of relocation created mixed feelings. They discussed whether relocation would be temporary or permanent. Participants seemed to weigh the risks of staying in the risky areas and the benefits of relocation.

In general relocation of people who live in hilly risky areas to lower less risky areas was acceptable. This would help them go back to check on their gardens and do some farm work. This was mostly on condition that they would access their gardens or be relocated in areas where they can do farming,”

I also say that people like us from Bukalasi, say that, when it rains a lot of our lives are always in danger and always leave us very worried. Therefore, I suggest that you people should fight for us a lot. That government should help us to take us in places like this one so that we can avoid being worried all the time when it rains,” (group 3).

Saving life and property outweighed clinging to their land. They reported of the many landslides in that area where many people including their relatives lost their lives as illustrated by the quote below:

As I talk, in 2010 there was a heavy down pour in the morning and the mudslide covered up my brother’s house and killed five children on spot. It’s good for people to be resettled to other parts of the district, “(Group 6).

Participants were affirmative and wanted relocation due to the situation they were in especially those who were in very high risk places such as Buwilimbi parish, Bukibokolo sub-County where they would lose their lives in case of landslides. They noted that preventive relocation would save the government in spending more money to manage disasters. Hence landslide risk significantly affects people’s willingness to resettle. People living in the riskiest areas such as Bukibokolo sub-County in Bududa had a strong willingness to resettle.

That said, most participants were in favour of relocation so long as the government was willing to support them. An assurance of compensation would facilitate their acceptance of relocation. Thus relocation was tagged to compensation of some kind.

I think it helps, because if problems befall you, at least you have a starting point, and if the government gives you this assistance, relocation becomes easier. So I’m in support of the idea of relocation,(Group 5).

The participants often highlighted that the reasons as to why people resist relocations is because in most cases, they are promised assistance but in the end they get nothing. They were not sure what the funds promised for relocations were used for. Referring to the past experience of those who were relocated in 2011 to Kiryandongo District, people suffered more and some of their land was sold.

Now like for us people from Nametsi, we are in area where there were landslides but they carried people and took them to Kiryandongo but for us we refused to go. Now they are yearning to come back and yet some of them have sold off all their land. Where will they start from? Because the conditions are not favourable where they were taken” (group 4)

The destination for resettlement was discussed extensively and this influences the willingness to be relocated. Most groups preferred to be resettled anywhere in greater Masabaland where they share the same language and culture. They were not in favour of being resettled elsewhere. Social cohesion would be fostered better and they would be able to perform their cultural rituals especially the male circumcision ”Imbalu” which is a rite of passage to adulthood for this particular ethnic group. Hence, such opinions indicate that an individual’s sense of identity and belonging profoundly discourages people from relocating.

For me, I was saying that relocating in Masaba land, within Mbale is better than relocating out (of Masaba land). Because we shall understand lugisu (the local language), we speak, work together and perform our cultural rituals together. And if a problem has befallen us like I may be in Bubulo, I can just walk and come even if I do not have money I can come home,” (Group 4)

However,a few people, were specific and mentioned that the relocation should be in Bududa low lying areas such as Busanza, Manjiya and Namatyale. They alluded to the fact that Bududa has very fertile soils compared to other districts in greater Mbale.

On the issue of being relocated is good but we the people of Bududa, we want to be relocated within, they should not take us to another sub-county or other district because there is a lot of land Bududa like in the low lands of Nalwanza”(Group 1)

Group participants had many questions related to the type of support provided by the government and how long it would last. Concerns about the ownership of their land in the high risky areas when they leave were raised. They were not sure whether they would still remain with their land after relocation.

“My question is like this, Government thinks of relocating people who are in risky areas, to remove them and take them to other places; if they take them away, who will have authority over the places they have left behind,” (Group 4)

Relocations to trading centres: Most participants felt that relocation from risky areas to trading centers was a good idea because trading centers are spacious, safe and accessible to social amenities than rural areas (e.g. clinics, shops). They noted that the services from government and NGOs would reach relocated community better and faster, and all would benefit.

I support it because, it will bring development in the trading centres and besides, when government sends assistance, it will be easily accessed by many since they will be in one place,”(Group 5)

A few of the participants disagreed to be relocated to trading centres because they were of the view that trading centres would be crowded and easily breed diseases such as Cholera that were typical of where they camped during the landslide disasters. Negative influences and behaviours would be acquired from the different people put together especially if there is scarcity of food and other needs.

Government should be able to meet our demands but not just making us crowd in one area .Now like the way we are here, the first thing is theft and secondly diseases.So it is better for government provide us with a place where we can temporarily be in case a disaster is about to happen so that we do not crowd in centres and get diseases, “(Group 1)

Related to the above was the fear of the concentration of so many people from different areas with different behaviours and habits such as those who abuse substances like alcohol and drugs. Such would breed quarrels, conflicts and insecurity. These examples given were a reflection of what happened in the previous landslides when they were relocated to some trading centres.

There is something burning in my heart, we were here in 2010-2011 but we suffered due to the crowding of people in one area. There were many diseases and the people whom we left behind who were in good places started stealing our things that we left at our homes. If we are to be relocated it has to be within Bukalasi because there are places that are safe it will be easy for us to go and check on our gardens,” (Group 8)

Most of the participants, being subsistence farmers, felt that in trading centres they would not have land to graze their cows and goats and to do farming. However those who supported being relocated to trading centres were putting their lives by relocating to a relatively safe area and only using the risky area to farm. Trading centres were preferred because they would be near their ancestral homes than relocation outside the greater Mbale region.

I think that if the trading centres are near our original homes, we shall be going to farm and come back to the centres,so that even if the landslides occur, only the food can be affected but not life,” (Group 5).

Relocation to relatives

Although extended families are common, most participants did not favour the idea of relocating to relatives. They felt that they would temporarily stay with their relatives but not for long. They noted that the social support systems have weakened and the hospitality would be abused given the high number household members they have including their domestic animals and other property. They felt they would be a burden to their relatives.

“I also concur with the last member’s idea even if it is a brother’s home you shared the same breast or even share the same mother, if you go there with your children it reaches a time and he chases you away,” (Group 1).

Relocating to relatives or friends with their families is something that they were not comfortable with. Most felt they could only stay for a short time until the situation stabilizes. Anticipated family conflicts were some of the inhibiting factors in relocation to relatives and friends. Others mentioned that some of their relatives are poorer than they are and so may not be of much help. A few mentioned that since it would be for a short time to relocate to relatives and friends, they can endure that instead of losing life and property due to landslides.

It is good to shift to the relative’s even if you are to quarrel than losing all your entire life and family , it’s good to go and endure and after the disaster you can easily come back home, “(Group 6 )

A few reported they would be itinerant migrants whereby during the rainy season they migrate to a safe place and go back to their homes during the dry season.

So for me, I was suggesting in times of dry seasons, those people should remain there and cultivate their crops but in rainy seasons they should get small rooms for their shelter elsewhere, when rainy season stopped, they go back, that is my opinion, I don’t know whether it helps,” (Group 3).

Compensation for resettlement

Another theme on policy options for resettlement was compensation. In several group discussions participants seemed relatively uncertain regarding compensation by government after relocation from high to low risk areas. The land was valued highly valued and thus compensation was seen as not a feasible option. This led to negative attitude towards compensation for resettlement. They were unclear about the modalities of pay off and this resulted into a prolonged discussions. In fact, they had more questions than answers “ As we were asking, when you compensate people and leave the hilly areas, does that place remain for the government or for the local people? They were not sure about ownership of the land the moment they are paid off. They preferred to be compensated with the same amount of land they had before the resettlement.

Here we need to agree, but let me ask; when I am being paid, am I paid to leave that place permanently[my land] and it remains free? And the money I am paid, am I allowed to buy a place of my choice or what is it”?(Group 2)

That said, they were positive that it was a good idea if the government paid for their land they left behind in the risky areas. But they were cautious and preferred to get a place to be relocated before they are compensated.

One particular issue that emerged from most groups was the inequitable compensation by government to affected persons which in a way influenced relocation negatively.

I may have my coffee and I am earning much from it and when I equate it and see that what they are giving in a year is not equal to what government is giving me, I can refuse to relocate. We also have bananas, yams and many others. These help us a lot in our homes,” (Group 2)

Thus, government commitment in compensation was considered crucial for people to accept relocation. Moreover, they articulated the perceived benefits of relocation such as being alive and safe, with less difficulties than staying in the risky areas prone to landslides. However, they noted other negative consequences on people’s lives such loosing cultural and social ties which may not be cost during compensation.

Risk communication

An early warning system is a response to an assessment of the risk and it involves monitoring, forecasting, warning, dissemination and communication of warning using a range of media and communication channels. Communities and other key actors should know how to respond promptly to avoid loss of life and adverse effects on livelihoods. Group participants were asked about the systems that can be instituted to warn the residents early enough before landslides strike.

In several group discussions, instituting early warning system was desired by participants. They felt it was a good measure to put in place so that people are aware when a problem is about to happen instead of being caught up by the disaster leading to loss of lives and property

For me I support it because it will have helped us so much because you may be in the house and maybe not aware that at this moment a landslide is taking place. But if it (early warning system) sounds like an alarm, or when it (early warning system) sounds like an ambulance, you just know that we have got a problem and we start moving away from that area and relocate to another area,” (Group 9)

The participants however, reported that traditionally warning systems were in place such as the traditional drum beats that were used to alert people in case of danger, community work or even for festive events as illustrated by the quote below.

These things were in place like long ago if somebody died they could easily drum,there was a particular drumming which showed that it is circumcision and there was also a particular drumming which could also alert people to come for drinking. When it drummed, someone could easily tell that there is local brew (alcohol) at this person’s place. So when those things are put in place, one knows if there is a particular kind of drumming, it signifies landslide. People shall always be aware,” (Group 3).

Drawing on such experiences of community warning systems helps inform and lay the groundwork for the future early warning systems because the early warning systems are able to use both indigenous knowledge and modern knowledge. The community noted that those with their indigenous knowledge know when the risky months are; usually May, September and October when there is a lot of rains.

So we know all these periods in our heads but we still suddenly find when it has slid, so it doesn’t help us much especially if you are near such risky areas”. The participants who were from the high-risk areas prone to landslides gave their real life experience with landslides and so supported the establishment of early warning systems.

I support it because for us who live on the hills, now like for me I sleep in between escarpments, there is a hill on this side and one on the other side. There is a time when a landslide occurred across there, we stood on that hill to look at the people at the other end and we were listening as people were shouting, it was dark, what we saw were only torches and we also continued to make alarms so much and the people from the other end continued running and yet this alarming of ours does not help so much. That is why I was saying that if that early warning system is put in place, it will have helped us,” (Group 14)

The bells and sirens were desired by most groups and that they could be instituted in areas that are risky. The bells and sirens were thought to be good because they were audible enough. However some participants were not sure how these early warning systems could work given that disasters such as floods or landslides happen suddenly. Moreover, they needed to understand the type of sound of the bell that would signify danger.

It’s true these early warning systems are good , but we don’t know the devil’s plans or those of God. Because when these disasters are going to happen, they don’t inform people, now how will they inform people that a disaster in form of a landslide is going to happen?”(Group 7)

A few were skeptical about the bells and sirens as the landslide might occur at night when people are asleep and so they may not hear the alarm. Some of the houses are iron roofed and when there is a storm the sound of the siren may not be heard. Or if it is heard people may be in disarray and end up running to where the danger is. They alluded to the fact that the way landslides occur is a process; that it does not happen during the heavy downpour rather it is when the rain is slowing down that the mud slides begin to move down as a mass. So, they felt that the bells and sirens may warn when the landslide is forming and this may be too late as some people would be swallowed on the way to safety. Hence the risk communication system may not be dependable or may not be effectively communicated especially to the most vulnerable populations. In one of the landslide affected area of Nametsi they reported the landslide happened at night.

When a landslide is happening, it doesn’t do so during a heavy downpour. It times when the rain is slowing down then the mud slides down. The early warning systems may warn when it happening and when you run out it can find you on the way. For me I support the issue of relocation to safer places than depending on the warning systems,”(group 5)

I have a doubt with regard to that issue because these landslides may occur in the night and usually it happens during heavy stormy rains when the clouds are very heavy and dark in iron sheet roofed houses which are very nosy. For example, in Nametsi the landslide occurred in the night when people were sleeping and even those who moved in the low land, the landslides buried them as well. So our thoughts are really troubled,” (Group 11)

The participants wanted to know more about the early warning system that would be established in order to act responsibly. They categorically indicated that alarms with no guidance on where to run to will not save the situation in such emergencies.

One of the options mentioned by a few group discussions was the use of short phone messages popularly known as ‘SMS’. This option was found only feasible to a few community members that owned mobile telephone sets. While this communication system had potential to reach out to many people at the same time, a number of issues were raised that rendered it ineffective. Participants raised the problem of low literacy (ability to read and write) as most know only how to receive and make a call by pressing some familiar iconic buttons; phone ownership density where only few people especially men owned mobile phones; low network coverage as well as low battery. At night, most of the phones are switched off for charging rendering SMS unreliable for disaster response.

‘’…the idea of sending messages on phone is good for like me who have a phone. Once the message is sent and received; that message on phone, you can go to your brother who is in the neighbor hood and inform them about what is bound to happen. It is a very good idea,’’ (Group 1)

‘’..the massages sent by phone are good but the problem is not all people have phones and another problem is that some of us may have phones but do not know how to read so even if a message is sent I will not be able to read it…..They just told us that you press here like this, (illustrates), (laughter), and you put on the ear,’’ (Group 1).

‘’the way me I see, not all of us have cell phones, and me I really see the best option of reminding us of any problem is the siren but not text messages since we have no phones’’(Group 3).

Discussions

The results show that while the community is in agreement with most of the policy options proposed by the government such as relocation from high to low risk areas including trading centres and to relatives and risk communication including early warning systems, others were in disagreement with the above policy options for the reasons highlighted in the results section. Still,others were uncertain about the proposed policy options e.g compensation for resettlement. The community members also still have mistrust in the ability of the government to fulfill the promises. Moreover, the community members do not understand the rationale underlying some of the policy options proposed by government let alone an understanding of the proposed policy itself.

Regarding the policy on resettlement, most participants were in favour of relocation as long as the government was willing to support the affected persons through the process. This is in line with Bankoff who noted that communities that are affected by hazards tend to respond by way of helping one another, by providing shelter, food and other necessities with those who have lost their livelihood 24 .

Although in principal, mobility is often understood as a potentially beneficial strategy for vulnerable households, to cope with and reduce exposure to hazards, the exacerbated climatic shocks that have resulted from climate change have rendered resettlement as a core risk reduction strategy as is in the case of Bududa 3. In Uganda, resettlement was done as part of the risk reduction strategy for Bududa District 12.

For its success, the benefits of resettlement have to be clarified to the affected community. Much as some institutions define resettlement as a physical movement , in this community of Bududa, resettlement has not been defined well,hence the question from the community on whether resettlement was permanent or temporary. This speaks to the need for better communication about risk reduction programs. Other questions still unanswered were whether the community members would still own the land in the area where they have been resettled from.

Benefits such as the intention to lessen site-specific vulnerabilities for example in areas like Bududa that are prone to recurrent landslides must be re-laid to the community as was the case in 2008, when a landslide severely affected 85 households in a densely populated and low-income community of Cochabamba city,Bolivia 37. The proposed solution was the relocation of the affected communities from the high risk zone areas 25. Resettlement has also been implemented in areas of civil stiff to reduce risk faced by persons during wars 26

Resettlement as a coping strategy can contribute to income diversification enhancing capacity of households and communities to cope with the adverse effects of environmental and climate change stresses. It also can be a long-term adaptation strategy. The intention is that resettled people will be better off over time as a result of resettlement – according to their own assessment and external expert review 27

In the case study of Vietnam, the outcomes of relocation and resettlement were mixed and it was demonstrated that resettlement programs have the potential to increase resilience and security of vulnerable households 28. However, the question sometimes remains, what do the affected communities perceive as benefits for resettlement? In the case of Bududa district in Uganda, it is unlikely the community perceives the benefits of resettlement such as improved access to public services; protection of the community from environmental shocks and stresses and improved living conditions. This was negated by previous resettlement experiences of 2011 to Kiryandongo. All the groups reported more perceived risks than benefits.

While the resettlement processes have many benefits, resettlement has also been shown to have challenges such as the increased distances thus people need more time to travel to their agricultural fields 28. This has an implication on the consultations that must take place between the state agencies and communities to identify and address the several factors that contribute to the failure of resettlement efforts.

In Bududa community the issue of the type and place of relocation/resettlement came out strongly with some communities preferring to be relocated in the trading centre. This speaks to the adequacy of the relocation site as has been documented elsewhere 29. Officials have to consider the adequacy of relocation site during their planning since the choice of the relocation site could either enable or hinder the resettlement efforts.

In Bududa, the community members preferred to be relocated anywhere in Masabaland in greater Mbale District where they share the same language and culture. Social cohesion would be fostered better and they would be able to perform their cultural rituals especially male circumcision which is a rite of passage to adulthood. This highlights the complexities and enormous challenge in finding suitable sites for relocating disaster-affected communities.That said, with unchecked population growth in the Mt Elgon region intra resettlement may carry a short term relief which is unsustainable in the long run if the livelihood source remains signifcantly agro and eco-system based.

Policy makers need to be aware that unsuitable new sites can lead to lost livelihoods, lost sense of community and social capital, cultural alienation, poverty, and people abandoning the new sites and returning to the location of their original community 8,11,37. The economic, social, and environmental costs of relocation should be carefully assessed before the decision to relocate and where is finalized.

In the case of Bududa district in Uganda, following the March 2010 catastrophic landslide, residents were temporarily resettled in IDP camps that led to challenges of poor sanitation, overcrowding and environmental degradation 8. Wisner and colleagues note that choosing inappropriate land for resettlement i.e. if it’s not close to sources of employment, distancing the new site from vital resources etc can lead to the failure of the resettlement efforts 13.

According to Putro (2012), following a large scale mudflow that happened in Sidaorja, Indonesia in 2006, the villagers’ decision-making process on where to resettle was guided by job patterns: 1) workers tended to choose locations near the city center; 2) farmers preferred to move as a group, maintaining their social network with other community members; and 3) traders, self-employed workers, and others lost their jobs and were forced to live in severe hardship because of the relocation 27. This is in line with our study findings in Bududa where the community members preferred to be resettled in areas that were close to their farmland so that they can continue with livelihood activities. Representative community consultations would have brought out these concerns and possibly aided the success of the resettlement policy in Bududa 30

An assurance of compensation is one of the ways that would facilitate community acceptance of relocation. In Bududa, the state agencies have never come out to clearly communicate the compensation terms for relocation of the affected victims. Compensation has been noted to be a major factor in relocation plans by the International Finance Corporation (IFC) 31 Lack of adequate information on compensation and terms of resettlement compromised the trust in the government policies. In times of compensation during resettlement, it is the right of the community to have fair and transparent compensation process.

IFC recommends that the compensation provided should be equal to or above what is required by law and in agreement with host communities on the methodology for calculating compensation. Although compensating for the loss of social capital can be challenging, IFC considers it a key aspect of compensation. Where possible, compensation is provided in forms other than cash so that long-term goals and livelihood improvements can be achieved 31. Given that in Bududa, resettlement meant that the community members had to incur some losses, compensation has to be taken seriously and consulted on.

Another issue that policy makers had to be aware of during this policy implementation in Bududa was the fact that resettlement could lead to some social disruptions such as men losing their social status and or political positions in cases where populations had to be dispersed 32. Such fears that were not identified prior might have contributed to the failure of the policy.

Added to this, most victims, having lost most of their assets in the landslide are literally left with nothing and therefore cannot be in a position to support them to relocate. They will need to be provided with some relief items- beddings, soap, cooking oil, sugar 33. This implies that relocation may be an expensive venture to a victim and this is worth considering by technical persons and policy makers who enforce these policies.

It has also been noted that resettlement is more likely to be successful when communities fully participate in well-planned adequately financed programmes that include elements such as land-for-land compensation, livelihood generation, food security. In other words there is increased chance of success when resettlement is conceived as a sustainable development programme that includes Disaster Risk Reduction (DRR) 13

As we note that resettlement must be part of a holistic risk reduction plan, in highly vulnerable communities, there is need for effective early warning systems that can generate and disseminate timely and meaningful warning information to enable individuals, communities and organizations threatened by a hazard to prepare and to act appropriately and in sufficient time to reduce the possibility of harm or loss 34.

Interestingly, majority of the participants did not trust use of text messaging for early warning because they did not think the mobile phone system was sufficiently reliable. In fact, majority of the members in the group discussions resisted it because of the various reasons as mentioned in the results section. However, use of Short Messaging Service (SMS) for early warning can be a potentially feasible option when the challenges associated with its use are overcome.

It was noted that sirens as an early warning system technology was desired by the people and it means that this acceptability can be leveraged to initiate and scale up the Early Warning Systems (EWS). It is important for the implementing agencies to adhere to standards of cultural sensitivity, acceptability and suitability of the EWS in order to assure sustainability in building EWS. Choosing a warning communication technology is dependent on considering who the recipients are, their location, their activity, the systems they rely on to receive local news and information, any special needs they may have and how they understand and respond to warnings 35

Rogers and Tsirkunov (2010) noted that one critical step is the willingness to act on a warning and take appropriate individual and collective measures to protect lives and poverty 36. So it’s important to have effective warning systems that can engage its expected beneficiaries by raising awareness and knowledge of risks and ensuring that the actions taken are realistic 36. In line with this our study demonstrated that collective traditional warning systems were noted such as drum beats that were used to alert people in case of danger, community work or for festive events.

Methodological discussion

The credibility of this study lies in the fact that we used group discussions and a stratified random sampling strategy. It is rare for qualitative work to be conducted with random samples and almost unprecedented for qualitative work to be done where the number of group discussions together is enough to add up to a credible representative sample of the population. We ought to note, however, that while participants from the group discussions were obtained from the 3 zones (low, medium and high risk zones to landslides), participants were heterogeneously composed. Therefore, it was not possible to conduct analysis by the 3 zones. Non-the-less this form composition generated rich and diverse discussion about policy options for resettlement management.

Conclusions

From the consultations using the deliberative poll method with the community, it can be generally agreed that resettlement is a highly complex issue. Policy makers have to be aware that resettlement and economic displacement of people can have significant adverse impacts on their future life, social fabric and livelihoods. If consultations are not adequately conducted, success determining issues are swept under the carpet. Ineffective consultations can leave the affected community feeling aggravated,hence do not adhere to the “agreed” position because some facts are only known to the technical people at the district level and policy makers in the District.

We recommend that for disaster risk reduction policies, in order to increase community acceptability and successful implementation of the proposed policies there is a need to increase community engagement during the policy formulation process. Deliberative polling presents a new and inclusive community consultation process of obtaining community perspectives for successful policy implementation.

Data Availabilty

The supplementary raw and analyzed data that support the findings of this study are available in figshare with the identifier data DOI: 10.6084/m9.figshare.5501326 https://doi.org/10.6084/m9.figshare.5501326

Conflict of Interest

The authors declare that no conflicts of interest exist.

Funding

This work was supported by the United States Agency for International Development under Makerere University School of Public Health’s Resilient Africa Network (RAN) project. The contents of this work are solely the responsibility of the authors and do not necessarily represent the official views of USAID. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Corresponding Authors

Grace Mongo Bua (gracebuamongo@gmail.com; gbua@ranlab.org; gracefridaypreston@gmail.com)

]]>
http://currents.plos.org/disasters/article/community-perspective-on-policy-options-for-resettlement-management-a-case-study-of-risk-reduction-in-bududa-eastern-uganda/feed/ 0