RES E AR C H A R T I C L E Open Access
What explains gender inequalities in
HIV/AIDS prevalence in sub-Saharan Africa?
Evidence from the demographic and
health surveys
Drissa Sia
1*
, Yentéma Onadja
2
, Mohammad Hajizadeh
3
, S. Jody Heymann
4
, Timothy F. Brewer
5
and Arijit Nandi
6
Abstract
Background: Women are disproportionally affected by human immunodeficiency virus (HIV)/acquired immunodeficiency
syndrome (AIDS) in sub-Saharan Africa (SSA). The determinants of gender inequality in HIV/AIDS may vary across countries
and require country-specific interventions to address them. This study aimed to identify the socio-demographic and
behavioral characteristics underlying gender inequalities in HIV/AIDS in 21 SSA countries.
Methods: We applied an extension of the Blinder-Oaxaca decomposition approach to data from Demographic and
Health Surveys and AIDS Indicator Surveys to quantify the differences in HIV/AIDS prevalence between women and
men attributable to socio-demographic factors, sexual behaviours, and awareness of HIV/AIDS. We decomposed
gender inequalities into two components: the percentage attributable to different levels of the risk factors between
women and men (the “composition effect”) and the percentage attributable to risk factors having differential effects on
HIV/AIDS prevalence in women and men (the “response effect”).
Results: Descriptive analyses showed that the difference between women and men in HIV/AIDS prevalence varied
from a low of 0.68 % (P =0.008)inLiberiatoahighof11.5%(P < 0.001) in Swaziland. The decomposition analysis
showed that 84 % (P < 0.001) and 92 % (P < 0.001) of the higher prevalence of HIV/AIDS among women in Uganda
and Ghana, respectively, was explained by the different distributions of HIV/AIDS risk factors, particularly age at first sex
between women and men. In the majority of countries, however, observed gender inequalities in HIV/AIDS were
chiefly explained by differences in the responses to risk factors; the differential effects of age, marital status and
occupation on prevalence of HIV/AIDS for women and men were among the significant contributors to this
component. In Cameroon, Guinea, Malawi and Swaziland, a combination of the composition and response effects
explained gender inequalities in HIV/AIDS prevalence.
Conclusions: The facto rs that explain gender inequality in HIV/AIDS in SSA vary by country, suggesting that
country-specific interventions are needed. Unmeasured factors also contributed substantially to the difference in
HIV/AIDS prevalence between women and men, highlighting the need for further study.
Keywords: Gender inequality, HIV/AIDS, Blinder-Oaxaca decomposition, Sub-Saharan Africa
* Correspondence: drissa.sia@uqo.ca
1
Département des sciences infirmières, Campus de Saint-Jérôme, Université
du Québec en Outaouais, 5, rue Saint-Joseph, bureau J-3226, Saint Jérôme,
Québec J7Z 0B7, Canada
Full list of author information is available at the end of the article
© The Author(s). 2016 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Sia et al. BMC Public Health (2016) 16:1136
DOI 10.1186/s12889-016-3783-5
Background
Countries in sub-Saharan Africa (SSA) remain the most
severely affected by the human immunodeficiency virus
(HIV)/acquired immunodeficiency syndrome (AIDS)
pandemic, accounting for 68 % of all persons living with
HIV/AIDS worldwide [1, 2]. Compared to men, women
living in SSA are disproportionally affected by HIV/
AIDS, accounting for 59 % of all infections in this region
[1–3]. The gender disparity starts at a young age, with 15
to 24 year old women in SSA being more than twice as
likely as men to become newly infected with HIV [1, 2, 4].
There is an increasing recognition that prevention and
treatment programs must address gender inequalities in
HIV/AIDS [5]. Nevertheless, the mechanisms that give
rise to these inequalities are poorly understood.
Gender inequalities in HIV/AIDS might be attributable
to the differential distribution of risk factors for women
and men. For example, the lower socioeconomic pos-
ition of women in SSA may place them at greater behav-
ioural risk for HIV infection [6–8]. Women are more
likely to be uneducated, unemployed, and impoverished
than men, which predisposes them to transactional sex-
ual exchanges [9]. These sexual exchanges are often
made with casual sex partners and without protection
[10]. Thus, economic inequality between women and
men may increase vulnerability to HIV among sexually
active women [11]. Besides economic differences, un-
equal power relationships and the subordinate position
of wom en relative to men also place women at higher
risk for contracting HIV [12–18]. Wom en in SSA gener-
ally have less power to negotia te safe sex, including con-
dom use [19]. Additionally, cultural factors encouraging
older men—who are more likely to be HIV-infected—to
have younger female partners (i.e., intergenerational sex)
[20] limit women’s ability to negotiate safe sex and in-
crease the risk of HIV infection for women relative to
men potentially exacerbating gender inequalities in HIV
infection [21]. Moreover, social norms permitting vio-
lence against women, including domestic violence,
spousal abuse, and rape might increase the probability of
infection among women. This violence has many impli-
cations for the spread of HIV/AIDS. For example, it is
associated with lack of condom use as well as traumatic
injury amon g women in SSA [19], which increases risk
of HIV infection [16, 22–25].
The differential responses of women and men to HIV/
AIDS risk factors may also contribute to observed gender
inequalities in HIV/AIDS. For example, a recent study by
Magadi [3] using pooled data from 20 SSA countries
showed that conditioning on HIV risk factors, including
sexual behaviors, did not explain the increased odds of
HIV/AIDS among women relative to men [3], suggesting
that traditional HIV risk factors may have differential and
more detrimental effects for women compared to men.
Few studies [26, 27], however, have assessed whether risk
factors have different effects on the probability of HIV/
AIDS for men and women. One study showed that un-
married women have twice the risk of HIV compared to
unmarried men [28], suggesting that the differential effects
of marital status may contribute to gender inequalities in
HIV/AIDS. Additionally, although men and women may
have similar distributions of household wealth, women
have less control over household decision-making and fi-
nancial resources and thereby may lack power to negotiate
safe sexual practices (for example, condom use) with their
partners [9, 29, 30], which puts them at higher risk for
HIV/AIDS. Unmeasured biological factors may also be
important [31]. For example, male-to-female transmission
of HIV is more biologically efficient than female-to-male
transmission [32–35]. However, gender inequalities in
HIV/AIDS vary substantially across world regions and are
unlikely to be explained by biological differences alone.
Political, organizational and legislative [5], social [12–18],
and other cultural factors not already mentioned may also
play important roles.
Gender inequalities in HIV/AIDS prevalence vary across
countries [36]. Clarifying the determinants of gender in-
equalities in the SSA region, including whether they are ex-
plained by the differential distributions (a “composition
effect”) or effects (a “response effect”) of HIV/AIDS risk
factors for women and men, may help to inform country-
specific interventions for mitigating them [37, 38]. How-
ever, the characteristics explaining gender inequalities in
HIV/AIDS prevalence in this region have not been system-
atically evaluated. Using data from the Demographic and
Health Surveys (DHS) and AIDS indicator surveys (AIS),
we recently elucidated the factors explaining gender in-
equalities in HIV/AIDS prevalence in Kenya, Lesotho and
Tanzania [39]. This study showed that composition effects
mainly explained gender inequalities in HIV/AIDS in
Tanzania, whereas in Kenya and Lesotho they were partly
explained by differences in the effects of measured HIV/
AIDS risk factors for men and women, including socio-
demographic characteristics (i.e., age and marital status)
and sexual behaviours (i.e., age at first sex). In the current
study, we extended our previous work by: 1) measuring the
magnitude of the gender inequality in HIV/AIDS preva-
lence across 21 SSA countries using available DHS; 2)
quantifying the extent to which gender inequalities in
HIV/AIDS were attributable to composition or response
effects using a decomposition analysis; and 3) estimating
the contribution of each risk factor to gender inequalities
in HIV/AIDS prevalence across SSA countries.
Methods
Data
We used available data from the international DHS and
the AIS to analyse the sources of gender inequality in
Sia et al. BMC Public Health (2016) 16:1136 Page 2 of 18
HIV/AIDS prevalence across 21 SSA countries surveyed
between 2003 and 2012 (Table 1). Each DHS is a cross-
sectional survey that collects and disseminates nationally
representative hous ehold data, including comparable in-
formation on socio-d emographic, behavioral, nutritional,
health and other characteristics over time [22, 40, 41].
The DHS uses a multistage stratified design with prob-
abilistic sampling that gives a defined probability of
selection to each elementary unit [42]. Each DHS survey
was stratified by urban and rural status and also by
country-spe cific geographic or administrative regions
[43]. To ensur e comparability across countries and time,
the DHS uses standardized measurement tools and tech-
niques and an identical core questionnaire that is pre-
tested and then administered by trained interviewers
[44]. Further details concerning the DHS survey meth-
odology are available elsewhere [45].
The AIS has been fielded in selected low- and middle-
income countries since 2001 [3, 46]. Unlike sentinel surveil-
lance, the AIS is a population-based survey that provides
nationally-representative HIV prevalence data based on
anonymous and voluntary testing of men and women aged
15–49 who were interviewed in the DHS, although some
countries have also tested older adults [47, 48]. Due to the
anonymous nature of the sur vey, respondents cannot be
provided with their results. However, all respondents are
offered referrals for free voluntary counselling and testing
(VCT) and AIDS educational materials. In some coun-
tries, mobile VCT teams follow up after interviewers to
counsel and test respondents who agree to be tested. The
comparative nature of the DHS and the possibility to link
HIV status from the AIS to the full DHS survey data,
while conserving anonymity, provide a unique opportunity
to examine factors contributing to gender inequalities in
HIV/AIDS in different contexts in Africa. Data from the
standard DHS linked to HIV prevalence data from the
AIS were available for 313,207 respondents across 21
countries, with seven countries surveyed twice between
2003 and 2012.
We used secondar y data collected by the international
Demographic and Health Survey (DHS) program after
obtaining participants’ consent. Due to the anonymous
nature of our data, our study was exempted from ethical
committee review.
Measures
Our outcome of interest was HIV serostatus, determined
by a confirmatory HIV-positive antibody blood serum
result. Sex of the respondent (male vs. female, as defined
in the DHS and AIS), used as a proxy for gender, was
the key explanatory variable. Other cov ariates included
socio-demographic characteristics, sexual behaviours,
and HIV/AIDS awareness. Socio-demographic character-
istics included urban/rural residence, the sex of the
household head, the respondent’s age at the time of sur-
vey, educational attainment (none, primary, or secondary
and above), marital status (married, never married, or sep-
arated/divorced/widowed), and occupational type (agricul-
tural, unemployed, domestic, trade, manual, office/service,
or professional/manager). Applying a relative approach to
poverty [49–51], household wealth was measured by a
composite index created by principal component analysis
(PCA) using information on household assets (ownership
of radio, television, refrigerator, bicycle, motorcycle/
scooter, car/truck, and telephone), housing quality, and
environmental conditions (electricity, source of drinking
water, type of toilet facility, floor material). The wealth
index was split into country-specific quintiles. Sexual be-
haviors included age at first marriage, age at first sex, pre-
marital sex, sexual behavior risk (i.e., if a condom was not
used at last sexual intercourse or having intercourse with
a partner other than a spouse), and having multiple sex
partners in the past year. Following the approach of
Magadi [3], the PCA technique was employed to create a
country-specific index of HIV/AIDS awareness using nine
questions on knowledge of the modes of HIV transmis-
sion and ways to avoid infection.
Statistical analysis
We calculated the prevalence of HIV/AIDS for women
and men across countries. The Chi-square test was used
to estimate gender inequalities in HIV/AI DS as the dif-
ference in prevalence comparing women to men. Then,
in countries where gender was significantly associated
with HIV/AIDS prevalence, we explored the source s of
gender inequalities in HIV/AIDS pre valence using an
extension of the Blinder-Oaxaca (BO) decomposition
[52, 53]. This involved decomposing the observed
women-men gaps in the prevalence of HIV/AIDS into
two components: composition and response effects. Com-
position effects represent the contribution to gender
inequalities in HIV/AIDS prevalence due to gender differ-
ences in the distributions of observable HIV/AIDS risk
factors between women and men (i.e., socio-demographic
characteristics, sexual behaviors, and HIV/AIDS aware-
ness). Response effects reflect the contribution to gender
inequalities in HIV/AIDS due to gender differences in the
effects of measured HIV/AIDS risk factors, as well as
unmeasured factors not included in the model [52–54].
The percentage of gender inequality in HIV/AIDS ex-
plained by a given component for each risk factor is de-
fined by the amount of the difference in HIV/AIDS
prevalence explained by the component divided by the
total difference in HIV/AIDS prevalence between women
and men multiplied by 100. The BO method allowed us to
assess which factors were associated with each source of
inequality. Initially limited to continuous dependent vari-
ables, the BO decomposition approach has been extended
Sia et al. BMC Public Health (2016) 16:1136 Page 3 of 18
to the case of non-linear dependent variables [55–59].
Estimates were obtained using the statistical routine
designed for non-linear outcomes described by Powers,
Yoshioka and Yun [54]. This approach overcomes poten-
tial problems related to path dependence and identifica-
tion [54]. All analyses, both descriptive and multivariate,
were weighted using the available DHS sampling weights
and accounted for clustering at the household level. We
used STATA version 12 software for all analyses.
Results
Gender inequalities in HIV/AIDS
Table 1 reports response rates, samples size and preva-
lence of HIV/AIDS by gender, country and survey year.
Women had a significantly higher prevalence of HIV/
AIDS than men in all countries and years sampled, apart
from Burkina Faso in 2003 and 2010, Mali in 2006,
Niger in 2006, Sao Tome & Principe in 2008/09, Senegal
in 2011 and Sierra Leone in 2011. The absolute difference
Table 1 Response rates (%), samples size and prevalence of HIV/AIDS (%) by gender, country and survey year
Survey
year
Age range
(years)
Response
rates (%)
a
Women Men Gender inequality
in HIV/AIDS
prevalence
Countries Women Men n
b
HIV +
c
Prevalence1
(%)
d
n
b
HIV +
c
Prevalence
(%)
d
Women-Men p-value
e
Burkina Faso (BF) 2003 15–49 15–59 89 4189 84 1.82 3341 59 1.95 −0.13 0.713
2010 15–49 15–59 95 8346 100 1.18 7034 60 0.84 0.34 0.086
Cameroon (CM) 2004 15–49 15–59 91.34 5154 349 6.63 5041 203 3.92 2.71 <0.001
2011 15–49 15–59 93 7253 434 5.57 6945 215 2.89 2.69 <0.001
Congo Brazzaville (CG) 2009 15–49 15–49 97.6 6349 240 4.13 5760 134 2.06 2.07 <0.001
Côte d’Ivoire (CI) 2005 15–49 15–49 78 4547 247 6.21 3917 110 3.11 3.1 <0.001
Ethiopia (ET) 2005 15–49 15–59 79 5942 142 1.86 5107 70 0.91 0.95 0.003
2011 15–49 15–59 85.75 15505 358 1.86 12998 182 0.98 0.88 <0.001
Ghana (GH) 2003 15–49 15–59 85 5289 138 2.71 4265 68 1.63 1.08 <0.001
Guinea (GN) 2005 15–49 15–59 91 3842 68 1.89 2925 35 1.1 0.79 0.010
Liberia (LR) 2005 15–49 15–49 84 6482 147 1.91 5206 62 1.23 0.68 0.008
Malawi (MW) 2004 15–49 15–54 67 2864 421 13.32 2404 243 10.23 3.09 0.002
2010 15–49 15–54 87 7396 890 12.88 6509 530 8.39 4.5 <0.001
Mali (ML) 2006 15–49 15–59 88 4743 69 1.54 3886 38 1.11 0.44 0.109
Mozambique (MZ) 2009 15–64 15–64 91 5901 875 12.67 4404 442 9.04 3.63 <0.001
Niger (NE) 2006 15–49 15–59 88 4441 39 0.71 3232 33 0.71 0 0.974
D.R. Congo (CD) 2007 15–49 15–59 88 4632 81 1.62 4304 43 0.92 0.7 0.027
Rwanda (RW) 2005 15–49 15–59 96.5 5663 222 3.61 4728 115 2.2 1.41 <0.001
2010 15–49 15–59 98 6952 266 3.71 6296 154 2.41 1.3 <0.001
Sao Tome &
Principe (ST)
2008/2009 15–49
15–59 – 2550 37 1.29 2160 39 1.8 −0.5 0.215
Senegal (SN) 2005 15–49 15–59 80 4466 48 0.88 3250 16 0.44 0.44 0.009
2011 15–49 15–59 80 5590 61 0.83 4327 32 0.51 0.32 0.071
Sierra Leone (SL) 2008 15–49 15–59 86 3466 64 1.73 3009 32 1.16 0.57 0.068
Swaziland (SZ) 2006/2007 15–49 15–49 85 4584 1438 31.15 3602 704 19.7 11.45 <0.001
Uganda (UG) 2011 15–59 15–59 96 11967 944 8.21 9399 551 6.11 2.1 <0.001
Zambia (ZM) 2007 15–49 15–59 75 5713 947 16.09 5161 649 12.29 3.8 <0.001
Zimbabwe (ZW) 2005/2006 15–49 15–54 70 7494 1553 21.12 5555 782 14.75 6.37 <0.001
2010/2011 15–49 15–54 75 7852 1463 17.71 6045 811 12.45 5.05 <0.001
Note: n sample size
a
More information on response rates is available at the following link: http://www.measuredhs.com/What-We-Do/survey-search.cfm?pgtype=main&SrvyTp=country
b
Numbers of women and men in the sample. These frequencies are unweighted numbers
c
Numbers of women and men with HIV positive test. These frequencies are unweighted numbers
d
A Weighted percentage of persons with HIV positive test among women and men using sampling weights provided by the DHS and AIS
e
p-value based on Chi-squared test for the difference in HIV/AIDS prevalence between women and men
Sia et al. BMC Public Health (2016) 16:1136 Page 4 of 18
between women and men in HIV/AIDS prevalence ranged
from a low of 0.68 % (P = 0.008) in Liberia (2005) to a high
of 11.5 % (P < 0.001) in Swaziland (2006–7). Fig. 1 maps
gender inequalities in HIV/AIDS prevalence in 21 SSA
countries (using the most recent survey for countries with
more than one available); inequalities were more pro-
nounced in the southeastern region of SSA relative to the
northwestern region.
Sample characteristics
Descriptive analyses (Additional file 1: Table S1) showed
that there were differences in the distributions of HIV/
AIDS risk factors between women and men. On average,
women were younger than men in all countries other
than Mozambi que, Swaziland, Liberia, Zimbabwe and
Malawi, likely due to differences in the samp ling frames
for women and men, which ranged from 15 to 49 years
for women and 15 to 64 years for men. In general,
compared to men, women wer e more likely to be mar-
ried (e.g., 76.7 % versus 63.8 % in Sierra Leone; 62.8 %
versus 50 % in Cameroon) and to be separated/divorced/
widowed (e.g., 9.1 % versus 5.4 % in Ghana; 18.4 % ver-
sus 5.5 % in Mozambique). However, there were some
exceptions. For example, in Malawi the percentages of
married women and married men were statistically simi-
lar. There was no difference between the percentages of
separated/divorced/widowed women an d men in Mali.
In general, women were more socioeconomically disad-
vantaged than men. For example, compared to men,
fewer women had secondary or above education (e.g.,
11.1 % versus 31.9 % in Guinea; 25.3 % versus 36.6 % in
Uganda). Additionally, women were more likely than
men to be unemployed or employed in trading, whereas
men were more likely to be employed in professional/
managerial occupations. The descriptive statistics results
also showed that a higher percentage of women reported
Fig. 1 Gender inequalities in HIV/AIDS prevalence in SSA countries
Sia et al. BMC Public Health (2016) 16:1136 Page 5 of 18