Public health impacts of drought in high-income countries: a systematic review

The health effects of drought are signicant and widespread, but primarily indirect. As climate change projections indicate future increases in drought events globally, it is essential that we continue to develop the evidence base on the health consequences of drought to inform future public health policy and practice (i.e. the quantication of impacts now and in the future). A systematic review was preformed using: MEDLINE, EMBASE, PsycINFO, Maternity and Infant Care, Global Health, CINAHL and Cochrane. Articles were included in the analysis if they met the following criteria: exposure was specied as drought; outcome was a dened and measured human health outcome; a distinct link had to be made between exposure and outcome; drought was dened as a shortage or deciency that was climate related; population studied was member countries of the Organisation for Economic Co-operation and Development (OECD) to represent HICs; reported original data; published in English from May 2012 to June 2017. health

Introduction largely focused on mental health and respiratory problems with a smaller number also reporting impacts of vector borne, water borne and injury related diseases (Stanke et al. 2013).
This study aims to address the paucity of evidence on the health impacts of drought in England by systematically reviewing the global evidence base, speci cally Organisation for Economic Co-operation and Development (OECD) member states, to inform future drought management strategies and public health responses in HICs.

Data Sources and Search Strategy
This review focused on studies published on or after 1st May 2012 and before 22nd June 2017. It was designed in accordance with PRISMA guidelines (Moher et al. 2009). All studies were identi ed through electronic search strategies only. Electronic databases were searched for relevant published studies. The databases included were Maternity & Infant Care, Global Health, Embase, Medline and PsycINFO which were accessed via OVID as well as the databases CINAHL and the Cochrane Collection. Search strategies were individually adapted for each database. Additional peer reviewed studies that were not available on the databases searched were identi ed using Google Scholar. As this study aimed to map the availability of peer reviewed evidence, no grey literature was included in the review process but instead these were incorporated as supporting text.
To identify relevant exposure, the rst Boolean search was done to explode (by subject heading) and map the keyword "drought". Other words related to drought were searched as keywords only. These terms were combined using the Boolean operator "OR". The second Boolean search was done to explode (by subject heading) and map multiple health outcomes, all of which were then combined using the Boolean operator "OR". The rst and second Boolean searches were then combined using the Boolean operator "AND". To identify the relevant population the 3rd Boolean search was used to explode (by subject heading) and map keywords relevant to OECD countries. The OECD was chosen given the similarity in policy standards and goals with England, whilst remaining geographically and climatically diverse. These terms were combined using the Boolean operator "OR". This search was then combined with the 1st and 2nd Boolean search using the Boolean operator "AND". (Table 1).

Study Selection
After removal of duplicates, two reviewers independently examined article eligibility in a two-stage process. The rst stage involved the initial screening of the abstract and titles of all identi ed studies to determine their relevance to the project. In the second stage, the full text of articles was assessed to determine whether they met the predetermined eligibility criteria. Only studies that met the eligibility criteria were included in the nal systematic review. Any inter-reviewer disagreements about the inclusion of studies were resolved through discussion. Studies that did not meet the criteria but contained relevant information were incorporated as supporting text within the narrative analysis. The number of articles included and excluded at the various stages were noted in accordance with PRISMA guidelines (Moher et al. 2009) (see Fig. 1)

Inclusion criteria Exclusion criteria
Exposure must be speci ed as drought Studies where a known drought indicator was used to measure exposure but did not make a reference to drought in their methodology Outcome must be a de ned and measured human health outcome Studies that examined the effects of water restriction policy implemented during drought periods A distinct link must be made between exposure and outcome Quantitative studies that didn't have a "no-drought" comparator group Drought must be de ned as a shortage or de ciency that is climate related Population studied must be a country in the OECD Any type of study design and/or methodology, Must report original data (i.e. No reviews),

Published in English from May 2012 onwards
Included studies were critically appraised using the Critical Appraisal Skills Programme (CASP) scoring system checklists, for qualitative and cohort studies (CASP 2017) and the AXIS tool for cross sectional studies (Downes et al. 2016). Studies were given a grade of either "good", "fair" or "poor" (see Table 2). No studies were excluded because of quality; 14 studies achieved a "good" score overall, eight a "fair" score, one a "poor" score and one was not scored as it was a case study design.

Results
Results of the systematic review yielded a total of 10,383 citations. Following title and abstract review, 10,184 articles were excluded, and full text screening was undertaken on the remaining 199 articles. An additional 175 articles were excluded due to lack of original primary data as well as an absence of 'drought' indicated as the exposure. In total, 24 articles met the full inclusion criteria with nine papers included as supporting text (see Fig. 1 for a ow diagram illustrating the selection and screening process).
Of the 24 articles, four were from the USA, 16 from Australia, one from Mexico, one from Austria, one from Israel and one examined the health effects of drought in Europe. Drought characterization varied considerably between studies: some studies used a single index of drought; others a combination of measures; others simply stated that drought had occurred during the period of the study. Additionally, some studies examined contemporary drought events whilst others studied historic events retrospectively.
Five major health related themes emerged: (i) water borne disease (ii) vector borne disease (iii) airborne, dust and respiratory related disease, (iv) mental health & wellbeing and (v) other health effects. These themes were used to structure and form the basis for the narrative synthesis. Studies that examined health outcomes related to multiple categories had their results split across categories in the analysis tables and narrative synthesis.
(i) Water Borne Disease

Necrotizing Fasciitis
The evidence for waterborne disease was very limited, consisting of only one case study. Two cases of necrotizing fasciitis were reported in Austria within one month of each other (Hirk et al. 2016). Each case had reported swimming at separate fresh water, non-European Union registered bathing sites. Both sites were previously not known to harbour the bacteria V. cholerae. It was hypothesised that the heatwave and period of summer drought during that period had led to optimal growing conditions for the bacteria in those environments.
(ii) Vector Borne Disease A province in Turkey experienced outbreaks of Tularemia after a period of unusually heavy rainfall which followed a dry 17 year period. The majority of the cases occurred in areas with higher altitudes (Balci et al. 2014).
Multiple studies found drought to be associated with West Nile Fever (WNF) (Aharonson-Raz et al. 2014;Anyamba et al. 2014;Marcantonio et al. 2015). In 2011 in Texas, USA and in 2012 in the central USA, the widespread and persistent nature of drought indicated by negative Normalized Difference Vegetation Index (NDVI) values as well as high temperatures appeared to coincide with large-scale epidemics of WNF in both areas (Anyamba et al. 2014). Summer drought was also found to be a key predictor of WNF outbreaks throughout regions in Europe, Africa and Asia (Marcantonio et al. 2015). Another study reported that periods of spring drought, where precipitation was signi cantly lower than average, preceded epidemic states of WNF observed in Israel in both 1999/2000and 2004/2005(Aharonson-Raz et al. 2014) Taken together, these studies suggest that drought conditions, particularly surrounding periods of spring and summer, increase the likelihood of WNF outbreaks occurring. One study examined differences in environmental factors (wetland con guration and hydrological regime) and their association with higher incidence of WNF during periods of drought in the USA. This study found that drought-affected areas with a high proportion of semi-permanent wetland had over 300% higher annual incidence of WNF incidence than drought-affected areas with a low proportion of wetland (Skaff and Cheruvelil 2016). This study highlights the complexities involved with assessing the relationship between weather patterns, vectors and subsequent health outcomes.
In 1915 a typhus epidemic in Mexico occurred, this was also a year characterised by drought (Rodriguez 2016 In the USA, hospital admission risk for respiratory disease in a population of geographically diverse elderly patients was found to be signi cantly reduced during periods of full drought compared with nondrought periods (Berman et al. 2017).
Respondents of a Community Assessments for Public Health Emergency Response (CASPER) health impact assessment conducted in Calfornia USA reported worsening of respiratory conditions including asthma and allergies upon perceived increased amounts of dust during the drought period (Barreau et al. 2017). An inverse association between Palmer Hydrological Drought Index (PHDI), an index of hydrological drought, and the prevalence of hayfever across the states in the United States was also reported, such that drier states had more hayfever. (Silverberg et al. 2015) (iv) Mental Health & Wellbeing

General
Multiple studies have found drought to be associated with adverse mental health outcomes. Respondents of a health impact assessment conducted in Calfornia USA reported that the drought had a negative impact on their household's peace of mind or that a drought-related experience in the past 30 days lead to acute mental stress (Barreau et al. 2017). Some respondents also reported worsening of preexisting mental conditions of someone in their household since the drought began (Barreau et al. 2017).
Another study found that those living in drought affected rural areas were more likely to have mental health problems and have a lower average mental health wellbeing score than those living in rural areas that were not drought affected (Edwards et al. 2015).
Drought appears to have a greater effect on the mental health outcomes for rural communities in particular, with studies nding that a long period of recent cumulative unbroken drought was associated with an increase in poor mental health outcomes for people living in rural but not urban areas. (Friel et al. 2014;O'Brien et al. 2014). Of those within rural communities, farmers in particular appear to be most affected by drought and have the greatest risk of adverse mental health outcomes. Several qualitative studies explored insights and determinants of emotional distress experienced by farmers during periods of drought (50-54). Drought was listed as a major stressor by farmers in South Australia, alongside other drought-speci c related stressors such as poor crop yields, unsatisfactory conditions for livestock and overwhelming workloads (Fennell et al. 2016). Farmers in Australia also described wind erosion that occurred as a result of chronically dry conditions as having a negative impact on their emotional wellbeing (Ellis and Albrecht 2017). Another study saw that when farm viability is threatened and farmland degraded through drought or economic pressures, male farmers experience shame and an 'undoing' of masculine subjectivity that constitutes a suppression of their farming identity (Bryant and Garnham 2015). A life chart analysis of Australian farmers who had died by suicide identi ed drought to be an environmental factor that re ected on the pathways of the majority of suicide cases. Drought was thought to have acted as a stressor by contributing to nancial di culties of the farmers (Kunde et al. 2017). In contrast, when the concerns of wine grape growers in South Australia were examined, it was found that they thought of drought as a factor that helped to mitigate other stressful micropolitical issues of the wine industry they were facing, rather than as a source of stress itself (Bryant and Garnham 2013).
A qualitative study in Australia suggested that multiple stakeholders in rural communities, not just farmers, experience profound levels of nancial and emotional stress and anxiety during periods of drought (Ng et al. 2015). Another qualitative study also reported that it was not farmers that experienced poor mental wellbeing as a result of the farming pressures of drought as they were able to adapt. Instead, drought appeared to have an impact on the mental wellbeing of those living in farming communities due to the degradation of green and natural spaces such as the rivers, parks, and home gardens (Sherval and Askew 2012).
One quantitative study in Australia found that farmers exposed to drought were more likely to experience mental health issues and have a lower wellbeing score than other rural employment sectors who had been exposed to drought (Edwards et al. 2015). Another study also found that that Australian farmers' overall life satisfaction, a measure of wellbeing, was signi cantly lower than that of the general rural community during a period of drought (Obrien et al. 2012). This was con rmed in another study where farmers in rural Australia were more likely to report higher levels of adverse mental health outcomes than those working in other occupations rurally (Brew et al. 2016). However, farmers also reported higher levels of drought stress and this difference in drought stress was not signi cantly associated with the change in mental health outcomes (Brew et al. 2016). Overall there appears to be strong evidence that farmers and rural communties in general are more likely to experience adverse mental health outcomes than other rural residents during periods of drought.
Few studies assessed the difference in impacts by gender, apart from a small number of Australian studies of farming commuities. The ndings in these studies was mixed but provide some evidence that amongst farming communities of Australia, women may be more resilient to drought.. Australian women in one study acknowledged that they had a few tough years associated with drought but that they were resilient enough to manage the challenges associated with it (Harvey 2014). No association was found between exposure to drought and mental health outcomes in rural Australian women aged 45 to 61, both generally or in potentially vulnerable groups including those who were more geographically isolated, poorer, less educated or pre-exiting chronic disease or poor mental health (Powers et al. 2015). Another study found no difference in self reported health or wellbeing outcomes between Australian women exposed to adverse climate events, primarily drought, and those who weren't (Powers et al. 2012). Rural Australian females were also found to be at decreased risk of suicide with increasing duration of drought for all age groups, whereas increased drought duration was found to be associated with increased risk of suicide in rural males aged 30-45 and 10-29 (Hanigan et al. 2012). However, another study found no difference between the levels of distress reported by male and female farmers South Australian farmers (Gunn et al. 2012).
Similarly, few studies assessed the role of age as a mediator of the relationship between drought and health outcomes. One study found that younger farmers in the 25-54 age group experienced signi cantly higher levels of distress than those in the 55-64 age group but not those aged 65-74 years (Gunn et al. 2012). Living in drought conditions appeared to negatively affect the wellbeing of Australian women as they aged with some reporting the impact it had on their ability to retire and coping with the symptoms of menopause by adding to depressive feelings ). There appears to be no clear association between age and sex on the mental health outcome of drought stricken rural residents. This may be because the relationship is complex or because few studies have assessed this.
The effect of drought on psychological distress was found to be greater in people who had missed a meal compared to those who hadn't (Friel et al. 2014). This association was statistically signi cant for all drought exposures except that of recent long dry periods of drought. This suggests that the extent to which one's mental health is affected by drought depends on the degree to which the drought has affected their ability to sustain aspects of their livelihood.
Finally, a qualitative study discussed the potential of adverse mental health outcomes in elderly gardeners because of water restriction policies implemented during drought periods so that maintaining their residential gardens became challenging (Adams et al. 2014). The results suggest however that the gardeners were mostly resilient and found way adjust to the policy by using other water sources to maintain their gardens.
(v) Other Health Effects In Australia, 50% of all bush re related loss of life between 1901-2011 occurred during days when the 3 pm Fire Weather Severity Index (FFDI), an indicator of drought exceeded 100, conditions which are deemed 'Catastrophic' (Blanchi et al. 2014). This highlights the role of bush re as a concurrent risk to drought and suggests that res occurring during more severe periods of drought result in increased incidence of re related mortality.
No signi cant differences in risk of cardiovascular admissions or mortality in an elderly population in the USA were seen between during full 'moderate,' 'severe,' 'extreme,' or 'exceptional' drought periods compared with non-drought periods (Berman et al. 2017). During periods of high-severity worsening drought, a signi cant increase in risk of mortality and a non-signi cant increase in risk of cardiovascular disease, compared with non-drought periods was found (Berman et al. 2017). There was no association seen between drought severity and risk of respiratory admissions. Risks for mortality and cardiovascular admissions during drought periods compared with non-drought periods were elevated in counties with less frequent drought suggesting that those who weren't used to experiencing drought were more likely to experience adverse health outcomes (Berman et al. 2017).
Households during the 2015 Californian drought generally reported that it had negatively impacted the health of their household members, speci cally the worsening of a chronic disease since the onset of drought with some reporting that they had sought additional medical care as a result (Barreau et al. 2017).

Summary of the Evidence
The results of this systematic review highlight multiple health outcomes associated with drought in HICs.
Most studies were conducted in either Australia or the USA and, as expected, there were no studies relating to nutrition related health effects and only one study relating to water borne disease. The results of this study strengthen the hypothesis that the risk of these health effects occurring during drought in HICs are low due to factors such as the level of social support, and resilient water and sanitation systems.
There is good evidence that suggests that impacts on the mental health of rural communities, particularly the rural farming population, is the greatest health burden during periods of drought in HICs. Rural communities are already at a disadvantage with regards to access to health services (AIHW 2017), however it has been shown that farmers are even less likely to access health services, including both primary care general practitioners and mental health professionals, than their non-farming rural counterparts (Brew et al. 2016). Targeted services are crucial to address the needs of rural populations and increase access to health services, especially during periods of drought.
Multiple respiratory health outcomes have been linked to drought; however, this review found that the available evidence was limited and con icting. While one very comprehensive study found overall hospital respiratory admissions to be reduced during drought, these ndings con ict with the larger body of evidence that suggests greater incidence of respiratory conditions during times of drought, including allergenic type conditions such as asthma, wheezing and hayfever, dust related conditions such as coughing and coccidioidomycosis as well as silo gas exposure. This inconsistency in evidence could be due to the lack of sensitivity of the hospital data, which does not capture the incidence of less severe respiratory problems (i.e. not requiring attendance/admission) or could be due to how respiratory outcomes are classi ed, aggregated or coded for. Additional observational studies are needed to determine the nature of the association between drought, incidence of respiratory outcomes and health system usage.
This review also identi ed good evidence that the vector borne disease WNV is another potential health outcome that should be of immediate cause for concern for HICs during drought periods, especially as several WNV transmitting mosquitoes can now be found in many part of Europe and the USA, and several recent outbreaks have already occurred in these regions (ECDPC 2018;Medlock and Leach 2015). This is a particular concern, given some of the evidence emerging showing that drought conditions encourage the proliferation of vectors at remaining pooled bodies of water, whilst also increasing vector competence (Johnson and Sukhdeo 2013). As drought is predicted to become more frequent and, given the strong evidence that associates WNF outbreaks and epidemics with periods of drought, WNF outbreaks may also increase in frequency.
Epidemic typhus is another vector borne disease, spread by body lice, that also appears to be correlated with drought periods. The single study within the OECD identi ed in this review was carried out in Mexico, one of the only upper-middle income countries in the OECD. Whilst epidemic typhus has been limited to developing countries in recent years, policy makers in places such as Europe as well as the USA should still be alert for potential outbreaks given that body lice infestation still exists within homeless populations in these places (Raoult et al. 2001).
Although few cases of water-borne infections were identi ed, the evidence supports the avoidance of public swimming in non-government regulated bodies of water during periods of drought, both from an injury and water borne disease perspective.
Data linking mortality is also limited, although there appears that severe drought increases likelihood of re related mortality, suicide death in rural working males as well as mortality in elderly populations. This evidence gap may re ect the absence of an association between drought and increased mortality, due to the protections a resilient infrastructure provides; however, it could also re ect the methodological challenges with de ning and measuring the exposures and impacts of drought, and the di culty with attributing a small number of health outcomes with a diffuse and/or distal exposure.
The existing evidence suggests a trend that communities that experience drought less frequently, may face more severe health effects than those who incur drought more often and have adapted and gained resilience (Berman et al. 2017;Friel et al. 2014). This is a similar phenomenon that has been seen in studies of temperature and heatwaves (Curriero et al. 2002). Therefore, there is heightened concern for communities and countries globally that may experience drought for the rst time.
Despite the gaps, this review provides su cient evidence to support the development of multi stakeholder National Drought Plans to streamline policy implementation efforts. A health needs assessment for drought would aid the capturing of public health impacts whilst supporting and informing the public health response. The identi cation of cost-effective interventions that include disease surveillance, vector population control and mental health support for affected farming communities would further increase resilience. Other potential interventions include the testing of unregulated water, public educational water safety campaigns (Baker-Austin et al. 2013) and the introduction of re bans during drought to mitigate re-related mortality and morbidity (Hoang et al. 2013).
Drought risk is predicted to increase in light of climate change and consideration of future scenarios will need to consider additional future risks, such as health system resilience to ensure adequate adaptive capacity and the safe delivery of healthcare (Balbus et al. 2016;Ferenc 2016;PAHO 2017). It is also important that food health and safety standards are continuously updated and monitored to re ect the changing pro le of risk associated with climate change (Kirezieva et al. 2015;Markland et al. 2017).

Conclusion
The health effects of drought in HICs identi ed in this study are signi cant, with mental health of rural and farming populations in HICs being the greatest burden during periods of drought. The risk of WNF outbreaks during drought periods is also of concern especially given the increasing geographical spread of its mosquito vector. Overall, evidence surrounding other health effects is still limited and further research is needed to explore potential associations with drought. Considering these results, policy makers in HICs globally should view the formulation and implementation of both preparedness and response strategies for drought, both as a current potential risk and a requirement for climate adaption. A crucial aspect of this process will be communicating with the wider stakeholders involved. Health policy makers in particular should ensure they plan not only for the health effects identi ed in this study but to consider wider health monitoring and preparedness strategies to ensure health system resilience to future risks, such as vector-transmitted infections that may become more commonplace as a result of climate -------------AND--------------Worsening drought (which was strati ed into low and high severity which were then compared to nondrought periods). This was a subset of full drought periods where the drought conditions are the same or worse than previous day

Free text comments
Many participants commented on how long of a period they had spent in drought. Living in drought conditions was reported to impact on the women's ability to age and retire in the way that they wanted to. It appears that living in drought has an impact on a woman's ability to cope and manage the symptoms associated with menopause and adding to depressive feelings. One of the important contributing environmental factors re ected also on the pathways of majority of suicide cases in both groups was drought. There were thought to have acted as a stressor by contributing to nancial di culties of the farmers.
It is important to note that 94% (n = 17) of all farmers were found to have a diagnosis of mental disorder at time of death.

Households in South
Tulare were most likely to report worsening of a chronic disease since the drought began. Of these, 20% reported that they had sought additional medical care