A systematic review evaluating HIV prevalence among conflict-affected populations, 2005–2020

Abstract

Historically, there has been concern that conflict may exacerbate the HIV epidemic. We conducted a systematic review to examine HIV prevalence in conflict-affected populations compared to district-level or countrywide HIV prevalence. Following PRISMA guidelines, studies presenting original HIV prevalence data published between 2005 and 2020 were drawn from PubMed, Scopus, and Embase. Data extracted included HIV prevalence, methods, dates, location, and population type. Studies were assessed for bias. Ten met criteria for data extraction; all focused on populations in sub-Saharan African. Most of the studies reported on mixed population settings while one was in a refugee camp. Six reported HIV prevalence higher than district- or country-level prevalence, while four reported lower HIV prevalence. Seven demonstrated moderate-to-high likelihood of bias in sampling, and five used methods limiting their comparability with local HIV prevalence. The relationship between armed conflict and HIV prevalence remains difficult to evaluate and likely varies by socioeconomic indicators.

Introduction

Armed conflict is often considered to increase HIV prevalence among refugees and internally displaced persons (IDPs)^1–4. Consequences of conflict including movement of armed groups and displaced persons^2–6, social disruptions, and widespread rape^7–9 create conditions conducive to the spread of the virus. Family separation, exacerbated poverty, and crippled health-care systems further render populations vulnerable to infection^{1,2,6,10–12}. However, empirical studies found no clear rise in HIV prevalence in association with armed conflict^10,13–16. HIV incidence between 1990 and 2012 actually decreased in 36 sub-Saharan African countries during years of conflict¹³, and intensity of armed-conflict was found to have an inverse relationship with HIV incidence and prevalence^13,16. In 2007, Spiegel et al. published a study comparing HIV prevalence in conflict-affected populations to nearby controls and found no clear difference in prevalence estimates¹⁵.

The observed decrease in HIV prevalence in contexts of armed conflict may be due to prolonged conflict’s tendency to isolate communities, limit mobility, and disrupt sexual activity^11,14,16. Many refugees migrate from populations with the lower HIV prevalence than host communities, and incoming humanitarian aid may further depress the spread of HIV¹⁵. Building on the findings of Spiegel et al.¹⁵, we examine how HIV prevalence among conflict- affected populations compares to district or countrywide prevalence.

Methods

This study includes a systematic review of articles examining HIV prevalence in conflict-affected populations carried out according to the PRISMA checklist¹⁷. The prevalence estimates were extracted from studies that met criteria for systematic review and then compared to district-level HIV prevalence in corresponding years. Countrywide prevalence was used when districtlevel prevalence was unavailable.

Search strategy and selection criteria

The databases PubMed, Scopus, and Embase were searched on December 13, 2017, for studies published from January 1, 2005, to the present. Studies published before 2005 were excluded because they were included in the 2007 systematic review by Spiegel et al.¹⁵. Databases were searched using a combination of the terms “HIV infection,” “human immunodeficiency virus infection,” “prevalence,” “sentinel surveillance,” “refugees,” “displaced persons,” “forced migrants,” “asylum seekers,” “conflict,” “post-conflict,” and “armed conflict and warfare.” Results were restricted to the English language.

Studies were included if they: (1) concerned populations in an area of ongoing conflict or displaced by conflict; (2) were peer reviewed; and (3) presented original HIV prevalence data. Studies were excluded if they focused on special populations not necessarily reflective of overall population trends, such as sex workers, military personnel, or tuberculosis patients. Studies were only included if the data came from low- or middle-income countries (LMIC). Excluded studies included those concerning refugees or asylum seekers in high income countries, where displaced populations and host populations were less comparable due to socio-economic and development-related factors.

Titles and abstracts were screened independently by two reviewers (DM and KO). Disagreements among reviewers were settled by discussion and tie-breaking vote by the third reviewer (PS). Abstracts that at least two reviewers agreed to include underwent full text review. This screening process was repeated during full text review.

Data extraction and quality review

Studies selected after full text review underwent quality assessment using the British Medical Journal’s guidelines for evaluating prevalence studies¹⁸ and data extraction. Due to data scarcity, quality ratings were not used to exclude studies, but informed our interpretation of findings. Likelihood of bias in each study due to sampling strategy, measurement, and analysis were assessed independently by DM and KO as either high or low. If reviewers disagreed on the likelihood of bias, then the likelihood of bias was recorded as moderate.

Data extracted from each study included: (1) HIV prevalence; (2) methods of data collection; (3) dates of data collection; (4) country of study; (5) population type, such as refugee, IDP, or civilians in conflict-affected areas; and (6) population size. Studies reporting HIV prevalence among conflict-affected populations in addition to other diseases or populations were included only if they separated data according to disease or population, in which case only the relevant section of data was extracted.

Country level comparisons

Estimated HIV prevalence in selected studies was compared to country- and district-level prevalence as reported by UNAIDS, the Demographic and Health Surveys (DHS) database, or the country’s Ministry of Health (MoH) for the years closest to the data collection period of the corresponding study. UNAIDS’ estimates were based on sentinel surveillance, DHS estimates were based on population-based surveys, and MoH methods varied.

Results

Figure 1 outlines the study selection process. Ten studies underwent quality assessment and data extraction. Four were from northern Uganda, and six pertained to conflict-affected populations in Somalia, Sudan, Chad, the Democratic Republic of Congo (DRC), Guinea- Bissau, and Tanzania. One study examined HIV prevalence in a refugee camp¹⁹, while the others focused on IDPs in camps and communities, as well as non-displaced nationals. Key characteristics of selected studies are shown in table 1. Sources of country- and district-level HIV prevalence are listed in table 2.

Figure 1. Process of study selection.

Table 1.

Key Characteristics of ten Studies

Study	Country	Type of population	Years of data collection	Methods	Sample size	Estimated HIV Prevalence
Bannink-Mbazzi et al.20	Uganda (Acholi sub-region)	IDPs and non- displaced persons	2002–2011	Voluntary testing at ANC sites	132,783	6.2%
Fabiani et al.22	Uganda (Acholi sub-region)	IDPs and non- displaced persons	2000–2003	Sentinel surveillance at ANC site	4459	11.3%
Malamba et al.23	Uganda (Acholi sub-region)	IDPs and non- displaced persons	2011–2012	Stratified random sampling of communities + household census within communities	2388	12.2%
Patel et al.24	Uganda (Acholi sub-region)	IDPs and non- displaced persons	2010	Cross-sectional survey among purposively selected sample	384	12.8%
Abdalla et al.27	Somalia (Somaliland)	IDPs and non- displaced persons	2004	Sentinel surveillance at ANC sites	1561	1.4%
Charpentier et al.28	Chad	IDPs and non- displaced persons	2007	Cluster sampling	548	11.1%
Kaiser et al.29	Sudan	IDPs and non- displaced persons	2002–2003	Household cluster sampling + sequential ANC samples	2020	Household clusters: 0.4–4% ANC samples: 0.8% (Rumbek) 3% (Yei)
Kim et al.30	DRC	IDPs	2005	2-Stage random household survey	1284	7.6%
Månsson et al.31	Guinea-Bissau (Bissau city)	IDPs and non- displaced persons	1987–2004	Sentinel surveillance at ANC site	20,422	0–4.8%
Rutta et al.19	Tanzania	Refugees	2002–2004	Voluntary Counseling and Testing at ANC site	9346	3.2%

Table 2.

Sources of district and country level prevalence data

Country	Source of HIV Prevalence	Year
Uganda	Ministry of Health – North-central region – Sentinel Surveillance at ANC sites	2005
	AIDS Indicator Survey Mid-Northern region Population-based Made available by DHS	2011
Somalia	UNAIDS data – Sentinel Surveillance at ANC sites	2005
Chad	UNAIDS data – Sentinel Surveillance at ANC sites	2003
Sudan	UNAIDS data – Sentinel Surveillance at ANC sites	2003
DRC	UNAIDS data – Sentinel Surveillance at ANC sites	2007
Guinea-Bissau	UNAIDS data – Sentinel Surveillance at ANC sites	1990
	UNAIDS data Sentinel Surveillance at ANC sites	2009
Tanzania	AIDS Indicator Survey – Population-based – Made available by DHS	2004

Uganda

Bannink-Mbazzi et al.²⁰ reported findings from a prevention of mother-to-child HIV transmission (PMTCT) program in 24 antenatal care (ANC) clinics in the Acholi sub-region, which included IDPs and non-displaced persons. HIV testing was voluntary, and approximately 5% of women opted out. Among the 94.4% of women tested, 6.2% were seropositive²⁰. Seroprevalence remained relatively constant during the data collection period. Reliance on voluntary HIV testing may have introduced sampling bias²¹, but given the high percentage of women who agreed to test, the risk of sampling bias was deemed low. Risks of bias in measurement and analysis of data were also rated as low.

Fabiani et al.²² reported HIV-1 prevalence from a single ANC site in the Acholi sub-region’s Gulu district. Total HIV prevalence among women tested was 11.3%. Authors reported that age-standardized HIV prevalence remained stable during the 4 years of data collection, ranging 6.0–6.8% for those < 20 years of age, 9.6–13.1% for those 20–24 years of age, and 14.4–16.1% for those ≥ 25 years of age. Age-specific multivariate analysis showed that among women younger than 25, HIV-1 infection was higher for those who had lived ≤ 2 years at their current residence (Adjusted odds ratio [AOR] 1.42, 95% confidence interval [CI]: 1.01–2.01). Of the 419 women who reported ≤ 2 years at their current residence, the prevalence was higher among those from other districts (31/157; 19.7%, 95% CI: 13.8–26.8%) or from Gulu Municipality (13/82; 15.9%, 95% CI: 8.7–25.6%) than for those coming from rural areas (15/180; 8.3%, 95% CI: 4.7–13.4%) (p = 0.010). Other predictors of HIV infection included age ≥ 25 years (AOR 2.56, 95% CI 1.91–3.44), partner’s age ≥ 35 years (AOR 2.68, 95% CI 1.72–4.16), urban residence in Gulu Municipality (AOR 1.76, 95% CI 1.41–2.18), and being unmarried (AOR 1.60, 95% CI 1.27–2.01). Likelihood of sampling bias for this study was deemed high due to inclusion of only one ANC site and oversampling of women aged 15–24, which authors explained as an effort to better assess changes in prevalence. However, UNAIDS and WHO guidelines recommend increasing overall sample size to ensure sufficient representation of women 15–24 years of age, rather than oversampling this specific group. Likelihood of measurement and analysis bias were rated as low.

Malamba et al.²³ carried out an open cohort study in the Acholi sub-region’s Gulu, Nwoya, and Amuru districts from 2011 to 2012, using stratified random sampling to select eight of 32 possible permanent, transient, or internally displaced study communities. A house-to-house census in selected communities enumerated the population, and proportional and non-proportional quota sampling methods were used to select study participants aged 13–49 years. Overall HIV prevalence was 12.2% (95% CI: 7.6–18.8), 14.6% (95% CI: 9.3–22.2) among females, and 8.5% (95% CI: 5.6–12.7) among males. HIV seropositivity was higher among the 8% of participants who had experienced rape or sexual abuse. Seropositivity was associated with major depressive disorder, post-traumatic stress disorder, and traumatic events. All three parameters of bias were rated as low for this study.

Patel et al.²⁴ randomly selected two sub-counties (Awach and Ongako) hosting IDPs in Gulu District, within which they purposefully selected 12 transit camps for IDPs. Proportional and non-proportional quota sampling methods were used to recruit 15- to 29-year-olds to participate in a survey that included HIV testing. Overall HIV prevalence was 12.8% (CI: 9.6–16.5); it was higher among females 15.6% (CI: 10.8–21.6) and non-abductees 13% (CI: 9.3–17.5). History of abduction was found not to be associated with serostatus, but participants in Awach as opposed to Ongako were more likely to be seropositive. Residents of the remote Awach region reported greater mobility than those in Ongako. This study demonstrated high likelihood of sampling bias because it was unclear whether purposely selected transit camps were representative of IDPs in the Gulu District. Furthermore, study participants were mobilized by hired community members and may be more representative of community members’ personal networks than the general IDP population. Likelihood of bias in measurement and analysis was rated low.

The HIV prevalence estimated by Bannink-Mbazzi et al.²⁰ using sentinel surveillance from 2002 to 2011 is lower than the district level HIV prevalence of 8.2% estimated by the MoH in 2005²⁵ and similar to the 6.3% estimated by the AIDS indicator survey in 2011²⁶ (Table 3).

Table 3.

HIV prevalence in conflicted-affected groups versus Countrywide or District level HIV prevalence

Country	Year*	Study population	Prevalence in conflict-affected population (95% CI)	Countrywide/district level prevalence
	2003	IDPs and non- displaced persons	11.322†	--
	2005		--	8.225†‡
Uganda	2010		12.8 (9.6–16.5)24	--
	2011		6.220†	6.326‡
	2012		12.2 (7.6–18.8)23	--
Somalia	2004	IDPs and non- displaced persons	1.4 (0.9–2.1)27†	0.938†
	2005	IDPs and non- displaced persons	--	3.339†
Chad	2007		11.1 (4.5–16.5)28	--
Sudan	2003	IDPs and non-displaced Persons	Yei: 3.0 (0.8–5.2)†	2.340†
			Rumbek: 0.8 (0–1.6)29†
	2005	IDPs	7.6 (4.1–11)30	--
DRC	2007		--	1.941‡
	1990	IDPs and non- displaced persons	--	0.342†
	1997		2.5 (1.7–3.3)31†	--
	1999		5.2 (4.1–6.3)31†	--
Guinea-Bissau	2009		--	2.542†
Tanzania	2004	Refugees	3.219†	7.032

^*The listed year is the last year for which data was collected. If data were collected for a range for years, the prevalence estimate for that range is presented in this table is association with the last year of that range.

^†Based on sentinel surveillance data.

^‡Based on district level prevalence data. Otherwise, data are countrywide.

Assessing significance of the difference is difficult since authors did not report CI. Fabiani et al.’s,²² sentinel surveillance estimate of 11.3 is higher than the district level prevalence in both 2005 and 2011. However, Fabiani et al. only reported findings concerning HIV-1, while UNAIDS data included both HIV-1 and HIV-2. Like Bannink-Mbazzi et al.²⁰ and Fabiani et al.²², the MoH estimate relies on sentinel surveillance while AIS uses population-based methods. Population-based methods used by Malamba²³ and Patel²⁴ produced prevalence estimates higher both district-level estimates, though Malamba’s²³ 95% CI of (7.6–18.8) encompasses the 2005 district-level estimate.

Somalia

In the self-declared autonomous region of Somaliland, Abdalla et al.²⁷ tested 1561 consecutive blood samples at four sentinel ANC sites serving IDP and non-displaced persons, selected for their routine collection of blood, accessibility, and willingness to participate. HIV prevalence at different ANC sites ranged from 0.6% to 2.3%, with a median of 1.4% (95% CI: 0.9–2.1). Women living in urban areas had a higher HIV prevalence than women in rural areas (3.1% vs. 1.3%).

This difference was not statistically significant, possibly because only 6.3% of participants reported rural residence, as all four ANC clinics were in urban centers. Age groups with the highest prevalence were 25–29 years (2.4%) and 15–19 years (1.8%). Likelihood of sampling bias was rated as moderate. Convenience sampling of ANC sites may not be representative of ANC patients across Somaliland. Likelihood of bias in analysis was also moderate, as CIs were only calculated across the 4 ANC sites, obscuring differences between sites. Likelihood of bias in measurement was low. The estimated countrywide HIV prevalence of 0.9 was within the 95% confidence HIV prevalence in reported by Abdalla et al.²⁷, suggesting no significant difference between the two populations.

Chad

Charpentier et al.²⁸ reported the results of a population-based serosurvey conducted by Chad’s National AIDS program in 2007. The survey covered 15 of Chad’s 18 regions, excluding Borkou-Ennedi-Tibesti, Guera, and Salamat due to accessibility issues. Non-displaced persons, IDPs, and refugees were included. Population areas were randomly selected using cluster sampling. The number of subjects in each population area was determined in a way that would be representative of the country’s overall demographics based on prior census data, and included 10,587 adults. Five percent of participants (n = 548) were randomly selected for inclusion in the serosurvey, with a median of 28 individuals per region. Overall HIV-1 seroprevalence was 11.1% (95% CI: 4.5–16.5). The prevalence in different regions ranged from 0 to 38%. Ouaddai, which, along with Wadi-Fira hosted many IDPs and refugees, had the highest prevalence of HIV of all included regions at 38% (95% CI: 22.7–54.2). Wadi-Fira had a prevalence of 13%, which was unremarkable given the range of 0–22% in all included regions other than Ouaddai. Likelihood of sampling bias was high, as sample sizes of only 10–53 individuals per region are likely insufficient to estimate HIV prevalence in respective areas.

Although the original sample of 10,587 was representative of nationwide demographics, it is unclear whether the subset randomly selected for serosurveys remained representative.

Likelihood of bias in measurement and analysis was assessed as low. The 2005 UNAIDS estimate of countrywide HIV prevalence at 3.3 was below the range of the 95% CI of 4.5–16.5% calculated by Charpentier et al.²⁸ in 2007. However, while UNAIDS data are based on ANC data, Charpentier et al.²⁸ used a population-based sample. Charpentier et al.²⁸ also only collected HIV-1 data, whereas the UNAIDS HIV prevalence includes both HIV-1 and HIV-2.

Sudan

Kaiser et al.²⁹ reported HIV prevalence in the towns of Yei, Western Equatoria, and Rumbek, Bar-elGhazal, in conflict-affected southern Sudan, based on data collected in 2002–2003. A two-stage household cluster was carried out in Yei town, the surrounding 20 km, and Rumbek town. Of 2444 eligible IDPs and non-displaced persons, 2020 (74%) consented to serosurveys. In addition, 625 sequential samples were taken from ANC clinics in Yei and Rumbek. Among respondents of the household cluster survey, HIV prevalence was found to be 0.4% in Rumbek (95% CI: 0–0.8), 0.7% in areas surrounding Yei (95% CI: 0–1.8), and 4.4 in Yei town (95% CI: 2.9–5.9). Among women attending the ANC clinics, HIV prevalence was 0.8% in Rumbek (95% CI: 0–1.6) and 3.0% in Yei (95% CI: 0.8–5.2). Although HIV prevalence was higher in Yei, which had a greater concentration of IDPs and returnees than Rumbek, no significant association was found between HIV prevalence and displacement status. Likelihood of bias in sampling was assessed as moderate, as the selection of ANC sites may not be representative of ANC patients in respective regions. Likelihood of bias in measurement and analysis was low.

The UNAIDS prevalence estimate of 2.3% fell below the 95% CI of ANC data from Rumbek, but within the CI calculated for ANC data from Yei. The UNAIDS’ estimate was also above the 95% CI for the population-based prevalence estimates from both Rumbek and Yei towns.

DRC

Kim et al.³⁰ examined HIV prevalence among reproductive-aged women in an IDP camp and two surrounding host communities along the Congo river in DRC in 2005. The selection process for communities was not described. A two-stage random sampling strategy was used to select households for inclusion, and then randomly select one woman aged 15–49 from each household.

The sample was weighted to account for differences in selection probabilities. A total of 1059 women from host communities and 225 from the IDP camp provided blood samples (n = 1284). In host communities, 0.4% (95% CI: 0–0.8) of respondents were refugees from another country, and 4% (95% CI: 2.8–5.1) reported being IDPs. In the camp, 24.4% (95% CI: 18.8–30.1) reported being refugees, and 75.6% (95% CI: 70.0–81.2) were IDPs. HIV prevalence was 3.1% (95% CI: 2.1–4.1) among the host communities and 7.6% (95% CI: 4.1–11.0) in the camp. The prevalence of HIV among refugees in the camp (n = 55) was not significantly different from the overall prevalence in the camp. History of sexual violence during the conflict was significantly associated with HIV infection in the camp. No economic variables were reported. Likelihood of sampling bias was moderate due to narrow geographic focus, which may not be representative of IDPs or host communities in the region overall. Likelihood of bias in measurement and analysis was low. The DHS district-level HIV prevalence of 1.9% was lower than the prevalence and 95% CIs reported by Kim et al.³⁰.

Guinea-Bissau

Månsson et al.³¹ report HIV-1 and HIV-2 prevalence from a sentinel ANC site in Guinea-Bissau between 1987 and 2004. Anonymous HIV testing of women giving birth took place every 1–2 years between August and December at Simao Mendes National Hospital (SMNH) in Bissau city, testing 20,422 women irrespective of displacement status over the 18-year period.

Approximately 1500 women were tested each year. The prevalence of HIV-1 increased from 0 to 4.8% throughout the data collection period. Between 1997 and 1999, HIV-1 prevalence doubled from 2.5% (95% CI: 1.7–3.3) to 5.2% (95% CI: 4.1–6.3), and was stable around 5% after 1999. The prevalence increased for all age groups until 1999, but decreased from 5.1% to 1.8% among women 15 to 19 years old after 1999. HIV-2 prevalence decreased from 8.3% to 2.5% between 1987 and 2004. Likelihood of sampling bias was high due to reliance on a single ANC site. This sample may be representative of Bissau if most women in the city gave birth there, but we do not know if this was the case. Authors state that the number of women surveyed each year stayed relatively constant since 1992, but it is unclear whether the proportion of women in Bissau giving birth at SMNH stayed constant. Likelihood of bias in measurement and analysis was low. The countrywide prevalence of 0.3% in 1990 and 2.5% in 2009 estimated by UNAIDS are lower than the prevalence and 95% CIs reported by Månsson et al.³¹.

Tanzania

Rutta et al.¹⁹ reported the results of a PMTCT program in the Greater Lukole refugee camp in Tanzania’s Kagera region from 2002 to 2004 that included voluntary counseling and testing (VCT). Of the 101,129 women utilizing ANC services, 9346 (92.3%) agreed to HIV testing. HIV prevalence was 3.2%. Only 14% of male partners agreed to undergo HIV testing, and 2.3% of the 1,454 tested couples were seropositive. Likelihood of bias in sampling, measurement, and analysis was low in this study. The prevalence of 3.2% reported by Rutta et al.¹⁹ is lower than the countrywide prevalence of 7.0% in the same year³².

Discussion

In comparing HIV prevalence in conflict-affected populations to countrywide or district-level prevalence, we found insufficient evidence to conclude that armed conflict is associated with an increase in prevalence. Rather, armed conflict appears to have mixed effects on HIV prevalence. While some studies found HIV prevalence in conflict-affected populations to be lower than^19,29 or similar to^20,22,27 the district or country overall^19,29, others found it to be higher^{23,24,28,30,31}.

Mixed results may be explained by limited comparability between prevalence estimates, a major limitation of this study. Of the ten studies included, five used population-based methods while district or countrywide estimates were based on ANC data or vice versa^{19,23,24,28,30}. Other studies only examined HIV-1 prevalence^22,28, while corresponding countrywide prevalence included HIV-1 and HIV-2. Furthermore, sampling bias was common among the selected studies, possibly due to challenges in accessing conflict-affected populations¹⁴. Studies relying on VCT for data collection^19,20 underestimate HIV prevalence, as individuals with known HIV status are more likely to refuse testing²¹. The use of ANC data may also underestimate HIV prevalence, as seropositive women are less likely to give birth³³. Limited availability of district-level HIV prevalence further complicated comparison between conflict-affected and control populations.

Although the search strategy did not have geographic restrictions, all studies reporting original prevalence data came from seven sub-Saharan African countries. However, at the time of this writing, the top two countries of origin for refugees globally were Syria and Afghanistan³⁴, which have different patterns of violence and HIV spread than countries in sub-Saharan Africa. In Afghanistan, primary modes of HIV transmission are injection drug use and unsafe paid sex³⁵. Although two studies on displaced Afghans were included in full-text review³⁶, neither met criteria for data extraction. Given the geographically limited scope of studies reporting HIV prevalence among conflict-affected populations, findings may not be applicable to some of the largest groups of displaced persons. Another major limitation is that many of the studies meeting criteria for data extraction used older data. The fact that none of the ten examined studies collected data after 2012 suggests that the most recent HIV prevalence trends among conflict-affected populations may not be represented.

Theoretical arguments regarding the impact of armed conflict on HIV prevalence highlight changes in affected communities’ social and institutional environments, such as increased mobility, poverty, disrupted social norms, damaged public health infrastructure, and incoming humanitarian aid^1,5,6,10,14. However, studies reviewed in this article did not examine these effects of conflict in relation to HIV prevalence, focusing instead on general demographic characteristics such as age³¹ and marital status²². Effects of conflict that was studied were individualistic, such as displacement status²⁹, history of rape²³ or abduction²⁴, and did not shed light on how conflict changed participants’ social and institutional environments. Assessments of respondents’ environments were limited to urban versus rural^22,27, or temporary versus permanent settlements^23,30. Data on healthcare access or changes in socioeconomic status were not included. The limited nuance in empirical studies relative to the theoretical arguments regarding armed-conflict and HIV prevalence may explain the mixed results observed in this review.

Conclusion

As we witness the unprecedented forced displacement of over 70.8 million people globally³⁷, understanding how armed conflict affects HIV prevalence becomes increasingly important. Future studies that examine the specific consequences of how armed conflict and forced displacement change affected communities’ social and institutional environments are necessary to allow the development of more specific and nuanced interventions to prevent transmission of HIV and reduce its consequences. These interventions may consist of outreach programs in conjunction with VCT to accommodate the growing proportion of refugees dispersed in urban settings instead of concentrated in camps, or counseling of serodiscordant couples from countries like Afghanistan, where HIV is primarily spread through IV drug use.