Drought, HIV testing, and HIV transmission risk behaviors: a population-based study in 10 high HIV prevalence countries in sub-Saharan Africa

Abstract

Droughts are associated with poor health outcomes and disruption of public health programming. Data on the association between drought and HIV testing and transmission risk behaviors are limited. We combined data from Demographic and Health Surveys from 10 high HIV prevalence sub-Saharan African countries with a high-resolution measure of drought. We estimated the association between drought and recent HIV testing, report of condomless sex, and number of sexual partners in the last year. Respondents exposed to drought were less likely to have an HIV test and more likely to have condomless sex, although effect sizes were small. We found evidence for effect modification by sex and age for the association between drought and HIV testing, such that the negative association between drought and HIV testing was strongest among men (marginal risk ratio [mRR] 0.92, 95% CI 0.89-0.95) and adolescents (mRR 0.90, 95% CI 0.86-0.93). Drought may hinder HIV testing programs in countries with high HIV prevalence.

Introduction

Climate change, including extreme weather events such as droughts, negatively impact human health and pose a challenge for maintaining effective public health prevention strategies. Periods of climate extremes are associated with changes in behavior as people may struggle to survive and experience food insecurity. Climate change may lead to changes in the prevalence, distribution, and transmission of infectious diseases including malaria, dengue, and Lyme disease, as well as various water-borne diseases and fungi [1–3]. Drought in particular has been shown to be associated with childhood illness including fever, diarrhea, and cough [4] as well as coccidiomycosis outbreaks [5]. However, there are significant gaps in understanding of the relationship between droughts and human health and health care access, particularly HIV prevention.

Drought is associated with 11% higher prevalence of HIV infections in Sub-Saharan Africa, which is thought to be due in part to negative income shocks, particularly in countries with high HIV prevalence (>5% prevalence) [6]. However, understanding the effects of drought in key subgroups by gender, age (e.g., adolescent versus adult), and setting (e.g., urban versus rural) remains important for targeting of interventions for vulnerable populations. Adolescents and young adults are key populations at high risk of HIV acquisition in sub-Saharan Africa [7]. One study in Lesotho found that drought was associated with higher HIV prevalence in adolescent girls in rural areas, and HIV transmission risk behaviors in young women 15-25 years old [8]. However, few studies have examined drought and HIV transmission risk behaviors outside of Lesotho. In addition, no studies to date have examined the relationship between drought and HIV testing, a key part of HIV prevention that depends on availability of clinical care and health education around HIV risk and prevention [7]. The health impacts of drought may particularly affect adolescents experiencing economic and educational transitions [9]. Furthermore, gender may affect the association between drought and HIV risk and testing [10,11]. Drought can exacerbate food insecurity [1,12], poor mental health [13–15], gender disempowerment, and interpersonal violence [16], which could lead to HIV testing barriers and HIV prevention measures including condom use and HIV pre-exposure prophylaxis (PrEP).

The primary objective of this study was to describe the associations between drought, HIV testing, and HIV transmission risk behaviors using nationally-representative survey data from 10 countries with high HIV prevalence in sub-Saharan Africa. The secondary objective was to determine if there were differences in these associations by sex, age (adolescent versus adult), and urban versus rural settings.

Methods

Data source and participants

We combined data on 10 countries participating in the Demographic and Health Surveys (DHSs) from 2011-2016 (Cameroon, Kenya, Lesotho, Mozambique, Malawi, Namibia, South Africa, Uganda, Zambia, and Zimbabwe), which are cross-sectional, nationally-representative household-based surveys conducted worldwide. DHSs use a stratified two-stage cluster sampling design selecting enumeration areas (EAs) and households within each EA [17]. All women aged 15-49 in selected households are invited to participate. Men are recruited in a subsample of randomly selected households.

We used surveys in areas with high HIV prevalence, using 5% as a threshold as defined by UNAIDS [18,19], that included geolocated information on each EA and took place during or after 2011. (See Supplemental Table 1 for a full list of surveys included in this analysis.) We used the year 2011 as the cutoff because of how the exposure (described in more detail below) was defined. Full information on outcomes of interest and covariates, in addition to variation in drought exposure in the study window, was also required for inclusion.

Measures

Drought

Drought was measured using the Climate Hazards InfraRed Precipitation with Station Data (CHIRPS) [20]. CHIRPS combine satellite images with weather station data to generate precipitation estimates in millimeters at 0.05 decimal degree resolution from 1981 to present. Annual cumulative precipitation for the 12 months preceding the survey date was calculated for each survey date/EA combination. We ranked this quantity of precipitation in millimeters with the prior 29 years and converted this ranking to a percentile. We used a timeframe of 29 years so as to balance the need for a long-time horizon (for assessing deviations) and the need to aggregate data across multiple countries. This use of deviations from long-term precipitation trends is standard in the literature, as it captures weather shocks representing deviations from the norm and therefore do not reflect inherent differences in populations that live in dryer or wetter areas [6,8,16]. We generated a binary categorization of drought, defined as 12-month precipitation prior to survey that was equal to or lower than the 15th percentile of the historical record (over the previous 29 years), reflecting the level of precipitation that impacts Gross Domestic Product and agricultural productivity, defined previously in the literature [6]. The calculation and classifications of drought were specified a priori.

HIV testing and risk outcomes

We considered three outcomes selected prior to analysis, each representing a different dimension of HIV risk [21]. These include: (1) a binary indicator representing whether the respondent reported receiving an HIV test in the 12 months prior to the survey date; (2) a binary indicator representing whether the respondent reported condomless sex during their last sexual encounter (among respondents who reported at least one sexual encounter in the 12 months prior to the survey); and (3) a count of the number of sexual partners the respondent reported (other than the respondent’s spouse, if the respondent was married) over the previous 12 months.

Covariates

We included several sociodemographic variables a priori that have theoretical and empirical associations with HIV testing and risk behaviors [21,22]. These include sex (woman and man), marital status (binary), age (15-19, 20-29, 30-39, 40-49, 50+ years), education (none, primary, secondary, and higher), household-level variables (wealth index [23], urban residence, and household size), and an indicator variable representing the survey calendar month, which accounted for seasonality.

Effect modifiers

We assessed effect modification by sex, urban/rural, and age group (adolescent 15-19 versus adult 20+) [24] to determine whether the associations differed between population subgroups.

Statistical analysis

To assess the associations between drought and HIV testing and risk behaviors, a series of multivariable regression models were fitted for each outcome, first pooled across all surveys and subsequently stratified by survey. In all models, we included survey fixed effects to control for country/time-level differences such as norms, sociodemographic characteristics, and economies. We included robust standard errors clustered at the EA level. We first fitted the pooled models with only country-level fixed effects and the drought indicator variable, and subsequently included the covariates. For binary outcomes (HIV testing and condom use), a logistic distribution was assumed. Because outcomes did not meet the “rare events” assumption, we used the results from these regression models to estimate marginal risk ratios [25]. For the count of sexual partners outcome, we assumed the Poisson distribution and present the exponentiated regression coefficients as incident rate ratios. To assess for potential interactions between drought and the effect modifying variables of interest described above (male/female sex, rural/urban residence, and adolescent/adult age), we included product terms in the regression models and considered an alpha significance level of 0.05 for the interaction term. All analyses were conducted in Stata v14 and R v3.4 [26,27].

Ethics

Demographic and Health Surveys obtain informed and voluntary consent from participants. We obtained permission to use DHS data from the DHS program. Specific approval for this de-identified secondary data analysis was not required.

Results

The analytic sample included 206,205 survey respondents from 10 high HIV burden countries. Figure 1 shows how the study population was selected.

Figure 1.

Flow diagram depicting analytic sample determination.

Table 1 shows sociodemographic characteristics and outcomes of respondents in the analytic sample. The sample was approximately one third (34.1%) male, half (46.2%) were married, and more than half (61.0%) lived in rural areas. Under half (42.5%) of respondents were tested for HIV in the 12 months before the survey, 23.6% respondents reported using a condom during their last sexual encounter (among those who had previously had at least one sexual encounter), and the mean number of sexual partners in the past 12 months (other than spouse for married respondents) was 0.87 (standard deviation [SD] 1.25). The distribution of drought by country is shown in Supplemental Table 1. Drought was least prevalent in Cameroon (0.3% of respondents) and most prevalent in Namibia (86.7% of respondents).

Table 1.

Descriptive statistics of survey respondents included in the analysis (n=206,205)

Variable	% (n)
Male	34.1 (70,407)
Married	46.2 (95,255)
Age
15-19	22.1 (45,470)
20-24	18.2 (37,436)
25-29	15.8 (32,557)
30-34	13.7 (28,216)
35-39	11.4 (23,580)
40-44	8.9 (18,337)
45-49	6.9 (14,297)
50+	3.1 (6,312)
Education
None	9.3 (19,265)
Primary	43.3 (89,202)
Secondary	40.1 (83,632)
Higher	6.8 (14,106)
Wealth quintile
Poorest	17.3 (35,746)
Poorer	18.6 (38,254)
Middle	19.6 (40,391)
Richer	21.3 (43,902)
Richest	23.2 (47,912)
Household size
1-3	23.0 (47,353)
4-5	29.8 (61,527)
6-7	24.2 (49,933)
8+	23.0 (47,392)
Rural residence	61.0 (125,772)
Outcomes
HIV testing in past 12 months	42.5 (87,697)
Condomless sex during last sexual encounter	76.4 (118,453)
Number of sexual partners in the past 12 months, mean (SD)	0.87 (1.25)

Figure 2 shows the effect estimates for the association between drought and the three outcomes of interest (recent HIV testing, condom use, and number of sexual partners) for each country individually and pooled. In pooled analyses with survey fixed effects, respondents living in areas recently affected by drought were less likely to have been tested for HIV in the previous 12 months (marginal risk ratio [RR] = 0.97, 95% confidence interval [CI] 0.95-0.99) and less likely to have used a condom during the previous sexual encounter (marginal RR = 0.96, 95% CI 0.93-0.99). We did not find evidence for an association between drought and the respondent’s reported number of sexual partners (incidence rate ratio [IRR] = 1.06, 95% CI 0.98-1.12). At the country level, three countries demonstrated significant negative associations between drought and recent HIV testing, four countries demonstrated significant negative associations between drought and condom use, and one country demonstrated a significant positive association between drought and sexual partners. These analyses also revealed scenarios where drought was positively associated with recent HIV testing and negatively associated with condomless sex, both in Uganda.

Figure 2.

Country-level and pooled associations between drought and HIV testing and risk outcomes. All models control for sex, marital status, age, education (none, primary, secondary, and higher), wealth index, urban residence, household size, and survey month. Standard errors are clustered at the EA level.

We found evidence for multiplicative effect modification between sex and drought, urban/rural residence and drought, and age and drought. Risk ratios stratified by effect modifiers are presented in Table 2. Drought was negatively associated with recent HIV testing among men (marginal RR = 0.92, 95% CI 0.89-0.95) but not women (marginal RR = 1.00, 95% CI 0.97-1.02); this heterogeneity was statistically significant (interaction p < 0.001). We did not find evidence for effect heterogeneity by sex for the relationship between drought and condomless sex, nor for the association between drought and number of sexual partners; however, we did find that drought was positively associated with condomless sex among men (marginal RR = 1.02, 95% CI 1.01-1.04). Drought was negatively associated with recent HIV testing among respondents living in urban areas (marginal RR = 0.96, 95% CI 0.94-0.99) but not rural areas (marginal RR = 0.99, 95% CI 0.97-1.00); this heterogeneity was marginally significant (interaction p = 0.056). While we found an association between condomless sex and drought among urban participants (marginal RR = 1.02, 95% CI 1.01-1.03) and between drought and number of sexual partners among rural participants (IRR = 1.07, 95% CI 1.00-1.13), there was no evidence for effect heterogeneity by urban/rural for the relationship between drought and condomless sex and the number of sexual partners. Finally, we found evidence for effect heterogeneity by age (adolescent vs. adult) for the relationship between drought and recent HIV testing (interaction p < 0.001) such that drought was negatively associated with testing among adolescents (marginal RR = 0.90, 95% CI 0.86-0.93) but not adults (marginal RR = 0.99, 95% CI 0.97-1.01). We also found that drought was positively associated with condomless sex among adults (marginal RR = 1.01, 95% CI 1.00-1.02) and that drought was positively associated with the number of self-reported sexual partners among adolescents (IRR = 1.08; 95% CI 1.01-1.15); however, we did not find evidence for effect heterogeneity by age for the drought/condomless sex association and the drought/sexual partners association.

Table 2.

Adjusted marginal risk ratios for the associations between drought and HIV prevention outcomes among participants aged 15+ stratified by sex, rural/urban, and age category.

	Stratified by sex			Stratified by rural/urban			Stratified by age category
Outcome	Men	Women	Interaction P	Rural	Urban	Interaction P	Adolescent	Adult	Interaction P
Tested for HIV in the past 12 months	0.92*** (0.89, 0.95)	1.00 (0.97, 1.02)	<0.001	0.99 (0.97, 1.00)	0.96** (0.94, 0.99)	0.056	0.90*** (0.86, 0.93)	0.99 (0.97, 1.01)	<0.001
Condomless sex during last sexual encounter	1.02* (1.01, 1.04)	1.00 (0.99, 1.01)	0.49	1.01 (0.99, 1.01)	1.02* (1.01, 1.03)	0.17	1.02 (0.99, 1.04)	1.01* (1.00, 1.02)	0.72
Number of sexual partners in past 12 months	1.02 (0.95, 1.11)	1.08 (1.00, 1.16)	0.25	1.07* (1.00, 1.13)	1.02 (0.92, 1.13)	0.37	1.08* (1.01, 1.15)	1.04 (0.97, 1.12)	0.45

All models control for sex, marital status, age, education (none, primary, secondary, and higher), wealth index, urban residence, household size, and survey month. Standard errors are clustered at the EA level.

Asterisks denote level of significance

^***p<0.001

^**p<0.01

^*p<0.05

Discussion

In this representative, population-based study of 10 countries with high HIV prevalence in sub-Saharan Africa, drought was associated with lower HIV testing over the past 12 months and higher probability of condomless sex at their last sexual encounter. The negative association between drought and HIV testing in the past 12 months was strongest among men, adolescents, and people living in urban areas. While the effect sizes estimated in these analyses are relatively small and do not demonstrate causality, these findings may have important implications given the growing exposure to precipitation extremes. Between 1980 and 2014, 363 million people in sub-Saharan Africa were affected by drought [28], and this figure is expected to increase by 426.6% by 2081-2100 [29].

Burke et al (2015) found that infection rates in HIV-endemic areas in sub-Saharan Africa increased by 11% for every recent drought. The same study also found that the association between drought and HIV infection was strongest in countries with high HIV prevalence [6]. We add to this literature by examining drought and HIV testing, a measure of health care access and utilization [21], finding similarly an association in countries with high HIV prevalence. The association between drought and HIV testing was strongest among men, adolescents, and people living in urban areas of countries with high HIV prevalence. There remain significant barriers to health care access and HIV testing for men, including fear of testing positive for HIV, stigma associated with HIV, access to HIV testing, and perception of HIV risk, which could all be exacerbated during times of drought [10,11]. For instance, men may have to work longer hours during times of drought and economic hardship and thus may be unavailable to get testing during clinic operating hours [10]. Several barriers to HIV testing among adolescents may also be exacerbated during times of drought at the individual, family, and health facility levels [7,30]. At the individual level, drought may limit adolescents’ transportation resources to get testing or limit adolescents’ HIV knowledge if they are unable to attend school [30]. At the family level, caregivers may have competing priorities during times of drought and may be less involved with their adolescent’s routine health services [30]. Similarly, health care facilities with limited resources during times of drought may prioritize other services over routine HIV testing [30]. During non-drought times, urban centers generally have more availability of HIV testing than rural areas [31], but droughts may exacerbate poverty, interpersonal violence, and migration to urban centers to seek employment, which may lead to barriers for people to access testing in urban centers [8].

The impacts of drought are multifaceted and occur on both micro and macro systemic levels, all of which may explain the association with lower likelihood of HIV testing. For example, drought can lead to food insecurity through reducing food availability due to crop loss and animal death, reduced food access through economic disruptions, reduced food utilization through worsened diet quality, and compromised stability of the food system due to market volatility and increased food prices [1,12]. Food insecurity, in turn, is associated with several negative outcomes along the cascade of care among people living with HIV, including compromised ART adherence [32–36], incomplete viral suppression [36], opportunistic infections [32], and increased HIV-related morbidity and mortality [35,37,38]. Our study suggests that drought-induced food insecurity may also negatively impact HIV testing, an important factor along the cascade of care that is necessary in ensuring patients receive adequate care. To further compound this, drought can worsen mental health through emotional stress, depression, anxiety, and increased alcohol consumption [13–15] (as can subsequent financial, food, and water insecurity resulting from drought [39,40]). Drought is also associated with increases in intimate partner violence, which may further impact mental health [16]. These financial and mental health challenges may impact an individual’s ability to access and afford adequate health care [41–43] or pay for transportation to travel to clinics for HIV testing [44,45]. Drought may also place families in a last-resort situation of migration [46,47]. In this context, the change in location may reduce access to health care given the new geographical location [48]. Lastly, extreme weather events (e.g., droughts) can adversely impact the health care system or infrastructure. This may occur through increased heat waves leading to wildfires [46,49], as well as impacts to electricity production [50], which may disrupt power supplies to needed heating and cooling systems that store HIV testing materials [46], or through storms and floods blocking roads and affecting electricity. Future studies could incorporate longitudinal research designs to better determine the mechanisms by which drought may impact HIV testing.

Low et al. (2019) found that in Lesotho, drought was associated with early sexual debut and transactional sex among adolescent girls and young women ages 15-59 years [8]. We similarly found that drought was associated with HIV transmission risk behaviors including lower probability that a participant’s last sexual encounter was with a condom, particularly among men, people living in urban areas, and adults. In addition, drought was associated with a greater number of sexual partners over the past 12 months in adolescents and those living in rural areas. Drought could lead to food insecurity, and adolescents in particular may rely on transactional sex to survive periods of drought [8]. The association between drought and number of sexual partners was stronger in rural areas, where food insecurity and income shocks may be most pronounced due to limited diversification of economic activity.

This study includes several limitations. Data on HIV transmission risk behaviors were based on self-report which may be subject to recall and social desirability biases. These data are from a period (2011-2016) prior to and at the start of key changes in HIV prevention, including the roll-out of at-home self-testing and PrEP. It is possible that the more widespread availability of self-testing would reduce barriers to testing faced by those in drought conditions; however, further study is necessary. The DHSs lacked consisted PrEP utilization data in the time period covered, but it is possible that drought may increase barriers to accessing PrEP through many of the same pathways that negatively impact testing, including economic disruptions and food insecurity. This should be an area of future research. Heterogeneity in findings across countries may be due to inconsistencies in data collection across countries and across years, although DHSs attempt to achieve standardization in data collection procedures across locations and time points, or could reflect other differences across countries. Uganda, the only country where drought was associated with higher HIV testing, typically experiences heavy rains, so drought (based on our definition) in this context may have been less disruptive than heavy rains. The distribution of satellite data and ground stations used by CHIRPS precipitation dataset may not be consistent and some countries may have less accurate precipitation data than others, leading to the potential for misclassification of drought in some settings. By classifying drought as percentile of annual precipitation over the past 30 years, rather than absolute rainfall estimates, we reduced the potential for misclassification.

Strengths of the study include an analysis of 10 high-HIV prevalence countries with nationally representative data, representing varying agricultural systems, environments, and sociodemographic characteristics. Data collection spanned over 6 years representing a range of drought conditions.

Conclusion

This study has potentially important public health and clinical implications. First, it contributes to the small but growing literature on climate change and HIV, identifying for the first time the association between drought and lower HIV testing rates. It is particularly important to note that drought is associated with both higher HIV transmission risk behaviors but lower testing rates, a dangerous combination, in countries with high HIV prevalence. HIV testing programs should consider increasing their outreach after periods of drought. Climate change causes higher temperatures and lower levels of precipitation, leading to increased frequency and intensity of drought. This is expected to worsen in the coming years. As drought becomes more common and widespread, this in turn may further potentiate HIV-transmission risk behaviors and lead to further disruptions to HIV testing in vulnerable populations. It is essential that future work focuses on the mechanisms linking drought to poor health outcomes, and on identifying groups that are particularly vulnerable to drought, to adequately design and implement interventions. Furthermore, in addition to interventions that help people adapt to the negative health impacts of drought, there is a need for more research testing health co-benefits of climate solutions as a way to help mitigate impacts. Finally, at a societal level, measures to curb emissions and address the root causes of climate change are crucial to lessen the impacts of drought and other extreme weather events on global health and wellbeing.

Availability of data and material:

Demographic and Health Survey data are publicly available at https://dhsprogram.com. CHIRPS precipitation data are publicly available at https://chc.ucsb.edu/data/chirps.