High HIV prevalence in a southern semi-rural area of Mozambique: a community-based survey


  • R González,

    Corresponding author
    1. Manhiça Health Research Center (CISM), Maputo, Mozambique
    2. CIBER Epidemiology and Public Health (CIBERESP), Barcelona, Spain
    • Barcelona Centre for International Heath Research (CRESIB, Hospital Clínic, University of Barcelona, Barcelona, Spain
    Search for more papers by this author
  • K Munguambe,

    1. Manhiça Health Research Center (CISM), Maputo, Mozambique
    Search for more papers by this author
  • JJ Aponte,

    1. Barcelona Centre for International Heath Research (CRESIB, Hospital Clínic, University of Barcelona, Barcelona, Spain
    2. Manhiça Health Research Center (CISM), Maputo, Mozambique
    Search for more papers by this author
  • C Bavo,

    1. Manhiça Health Research Center (CISM), Maputo, Mozambique
    Search for more papers by this author
  • D Nhalungo,

    1. Manhiça Health Research Center (CISM), Maputo, Mozambique
    Search for more papers by this author
  • E Macete,

    1. Manhiça Health Research Center (CISM), Maputo, Mozambique
    2. Health National Direction (DNS), Ministry of Health, Maputo, Mozambique
    Search for more papers by this author
  • PL Alonso,

    1. Barcelona Centre for International Heath Research (CRESIB, Hospital Clínic, University of Barcelona, Barcelona, Spain
    2. Manhiça Health Research Center (CISM), Maputo, Mozambique
    Search for more papers by this author
  • C Menéndez,

    1. Barcelona Centre for International Heath Research (CRESIB, Hospital Clínic, University of Barcelona, Barcelona, Spain
    2. Manhiça Health Research Center (CISM), Maputo, Mozambique
    3. CIBER Epidemiology and Public Health (CIBERESP), Barcelona, Spain
    Search for more papers by this author
  • D Naniche

    1. Barcelona Centre for International Heath Research (CRESIB, Hospital Clínic, University of Barcelona, Barcelona, Spain
    2. Manhiça Health Research Center (CISM), Maputo, Mozambique
    3. CIBER Epidemiology and Public Health (CIBERESP), Barcelona, Spain
    Search for more papers by this author

Correspondence: Dr Raquel González, Barcelona Centre for International Heath Research (CRESIB), Hospital Clínic, Universitat de Barcelona, Rosselló 132, 4-2, E-08036 Barcelona, Spain. Tel: +34 932275400, ext. 4144; fax: +34 932271850; e-mail: raquel.gonzalez@cresib.cat



Southern African countries have borne the brunt of the HIV/AIDS pandemic. Monitoring epidemiological dynamics is critical to identify the populations at greatest risk of infection and to guide control strategies.


A cross-sectional community-based study to determine age- and sex-specific HIV prevalence among individuals aged 18–47 years was carried out in Manhiça, southern Mozambique. Participants were randomly selected from the demographic surveillance system in place in the area and voluntary HIV counselling and testing were offered at home. In addition, HIV prevalence estimates from the antenatal clinic (ANC) were collected prospectively.


A total of 839 individuals were invited to participate in the study. Of these, 722 were recruited (50.7% women). The overall HIV prevalence in the community was 39.9% [95% confidence interval (CI) 35.9–43.8%]. By age, the prevalence was 23.2% (95% CI 17.9–28.6%) in individuals aged 18–27 years, 41.2% (95% CI 35.6–48.3%) in those aged 28–37 years and 44.8% (95% CI 38.4–51.2%) in those aged 38–47 years. HIV prevalence was higher among women than men in all age groups. The overall HIV prevalence estimate for women in the community (43.1%; 95% CI 37.6–48.5%) was 1.4 times higher than that for those attending the ANC (29.4%; 95% CI 26.7–32.0%).


The high HIV prevalence found in this region suggests that the epidemic is in a mature stable phase. The lower rates in the ANC than in the community suggest that ANC evaluations may underestimate community HIV prevalence. Resources to monitor HIV infection dynamics are needed to guide targeted control strategies in countries in which the epidemic exacts the greatest toll.


Despite recent advances in the development of prevention strategies and the global scale-up of HIV antiretroviral drugs, the control of the HIV/AIDS epidemic continues to be challenging, especially in sub-Saharan Africa, where approximately 22.5 million (68.5%) of the 32.8 million people infected with HIV world-wide live [1]. Although the number of new infections slightly decreased in 2008, recent estimates from sub-Saharan countries indicate a modest increase in the HIV prevalence, which can probably be attributed to improved access to antiretroviral treatments and consequent increased survival [2].

Accurate community-based HIV prevalence estimates are needed to assess the evolution of the epidemic in specific settings to allow adequate monitoring and evaluation of control strategies. HIV prevalence data derived from antenatal clinics (ANC) have traditionally been used to monitor HIV epidemic trends in many countries, as the prevalence in pregnant women is assumed to correlate well with HIV prevalence in other adults aged 15–49 years in the general population [3]. However, since 1998 the Joint United Nations Programme on HIV/AIDS (UNAIDS) has also recommended that population-based surveys should be conducted to enable the population to be more widely represented and to compensate for potential biases in the ANC estimates, such as their poor general representativeness [3-6].

Monitoring basic epidemiological HIV infection data is especially important in southern African countries, as they bear the brunt of the HIV/AIDS pandemic. Mozambique is one of the 10 countries with the highest HIV prevalence in the world, with 1.4 million [95% confidence interval (CI) 1.2–1.5 million] people living with HIV according to UNAIDS estimates [1, 7]. Since 1988, a national surveillance system has been monitoring HIV prevalence through ANC sentinel sites [4]. Currently, there are 36 ANC sentinel sites contributing to the surveillance system in the country. The Mozambican ministry of health estimated from the ANC surveillance system data that the national HIV prevalence in adults aged 15–49 years was 13% in 2001 and increased to 16.2% in 2004 [8]. Recent published results from the first population-based survey, conducted in 2009, showed that the national HIV prevalence was 11.5% (95% CI 10.3–12.6%) in individuals aged 15–49 years [4]. The country has rapidly scaled up the use of antiretroviral (ARV) drugs and it has been estimated that approximately 170 000 people had initiated ARV therapy by the end of 2009, which represents a coverage of 38% [9].

In order to tailor prevention programmes to distinct populations, local epidemiological data are necessary. The primary objective of this study was to determine the age- and sex-specific community HIV prevalence in adults aged 18–47 years old in an area of southern Mozambique. In addition, the results from the community-based survey were compared with HIV prevalence estimates derived from the ANC surveillance data of the local district hospital.


Study area and population

The study was carried out in Manhiça District, a semi-rural area in Maputo Province, in southern Mozambique. Since 1996, the Centro de Investigação em Saúde de Manhiça (CISM) has been running a continuous demographic surveillance system (DSS) for vital events and migrations. In 2007, there were 160 000 inhabitants in the district [10]. Currently, the DSS covers nearly 90 000 inhabitants, 36 000 of whom live within a 10-km distance of the centre of Manhiça town, which constitutes the main study area of the CISM. The trends of the demographic indices in Manhiça District have been described in detail elsewhere [11]. The majority of the population is Changana, with a small proportion of the Ronga ethnic group. The main occupations are farming, petty trading and employment on the two large sugar cane estates in Maragra and Xinavane.

Since 2003, the CISM has collaborated with the Mozambican HIV/AIDS control programme through the establishment and continuous support of voluntary counselling and testing centres at health facilities, the provision of ARV drugs and diagnostic tests, and contributions to the clinical management of patients, among other activities. In addition, several clinical studies have been carried out at the hospital [12, 13]. Estimates from the ANC of Manhiça District Hospital (MDH) showed an HIV prevalence of 23.6% in a study performed in 2003–2004 [14, 15]. However, basic epidemiological data on HIV/AIDS trends at the community level were lacking. The study protocol and informed consent form were reviewed and approved by the National Committee on Health Bioethics of Mozambique and the Hospital Clínic of Barcelona Ethics Committee (Spain).

Study design and procedures

A cross-sectional community-based study was designed to determine the age- and sex-specific HIV prevalence in adults. The lower age limit was thus established at 18 years. The sample size was calculated for three age categories assuming a 20% HIV prevalence and a 95% CI with a precision of 5% using pass software (NCSS, Kaysville, Utah, USA). It was deemed that 232 subjects per age group would be needed. The width of the defined age groups was designed to be equal to 10 years for each of the three groups (18–27, 28–37 and 38–47 years of age) in order to facilitate future determination of incidence in a second cross-sectional survey [16].

In addition, HIV screening data from the routine ANC of the MDH were prospectively collected, stratified by the predefined age groups and compared with the respective population-based estimates of the same year. At the time the study was conducted, local guidelines for prevention of mother-to-child transmission of HIV were based on ARV monotherapy administration to the mother from 28 weeks of gestation, plus combined ARVs during labour, and administration of ARVs to the infant for up to 4 weeks. HIV counselling, testing and treatment are available and provided free of charge at the health services in Manhiça Hospital.

For the cross-sectional study, Microsoft® Visual FoxPro 5.0 software (Microsoft Corporation, Redmond, WA) was used to generate random lists from the DSS of adults living in the study area stratified by age group and sex, and organized by neighbourhood. The study inclusion criteria were: age 18–47 years, being resident in the main study area, and being willing to participate in the study after signing an informed consent form. Study candidates were recruited regardless of their previously known HIV status. Prior to the study initiation, community sensitization activities were carried out, consisting of informative meetings about the study and its objectives with the neighbourhood leaders.

The selected individuals were visited at home by a study field worker who explained briefly the objectives of the study. If the candidate agreed, he/she was given an appointment card and another home visit was made by a mobile team to provide more information about the study. Voluntary HIV counselling and testing were offered in the households [17]. If the subject was absent, he/she was revisited one more time. In the case of repeated absence or migration, the next listed candidate was visited.

At the scheduled date and time, a trained HIV counsellor visited the household. In order to ensure privacy and confidentiality, the counsellor identified an adequate area to perform HIV testing and informed consent. Again, in the case of absence at the time of the visit, candidates were visited only one more time to offer participation in the study. Recruitment was stopped once the minimum sample size for each age and sex group was reached.

Basic sociodemographic information was recorded on the study case report form (CRF). Rapid HIV testing was performed by fingerprick following national recommendations using two rapid tests: the Determine HIV 1/2 test (Abbott Laboratories, North Chicago, IL; sensitivity 100%; specificity 99.6%) and the UniGold HIV test for confirmation of an HIV-positive result (Trinity Biotech, Bray, Ireland; sensitivity 100%; specificity 99.7%). HIV positivity was defined as positive results for both tests. The test result was recorded on the study CRF. Participants with a positive result were offered medical follow-up at the Manhiça out-patient clinic, which included CD4 cell counts, clinical management and provision of ARV treatment if needed, following national guidelines.

In addition to the population-based study, data from the routine HIV screening of pregnant women attending the ANC of the MDH were collected prospectively from March to September 2010.

Data management and statistical methods

Data from the study CRFs were double-entered at the CISM using the OpenClinica software for clinical data management (www.openclinica.org). The statistical analysis was performed using stata software version 11 (Stata Corp., College Station, TX). One-way and two-way contingency tables were generated for description of the categorical variables and calculation of proportions and P-values. The probability of sampling was taken into account to extrapolate the data from the survey to the community by weighting the sample groups (defined by sex and age) and using DSS data [18].


Characteristics of study participants

A total of 1124 adults were approached to determine their availability to participate in the study and were given an appointment card for a later mobile team visit. Of those who made an appointment, 839 adults (74.6%) met with the mobile team, received the study information and were invited to participate in the study. Reasons for not receiving the study information were refusal (3.7%), absent twice at the second household visit (8.1%), not eligible (10.3%), and unknown (3.3%). Of the 839 adults invited to participate, 722 agreed to participate and were recruited (acceptance rate 86.1%). Almost 60% (68 of 117) of the individuals who did not agree to participate in the study were men. This sex difference in the acceptance rate was statistically significant only in the 28–37-year-old group (P = 0.016). Table 1 shows the acceptance rate by sex and age group. Twenty-seven out of 117 individuals (23%) who did not participate in the study claimed that they already knew their HIV status. Almost half of the participants (45.1%) were unemployed. Their sociodemographic characteristics are shown in Table 2.

Table 1. Acceptance rate by sex and age group
Age groupInvited (n)Recruited (n)Acceptance (%) (recruited/invited) P-valuea
  1. aTest for assessing dependence between sex and test acceptance.
18–27 years    
28–37 years    
Female14112286.5 0.016
38–47 years    
Table 2. Sociodemographic characteristics of study participants (n = 722)
  n %
Sex (n = 722)  
Age group (n = 722)  
18–27 years23432.4
28–37 years24233.5
38–47 years24634.1
Education (n = 716)  
No schooling10314.4
Primary school42258.9
Secondary school17224.0
Higher education192.7
Employment (n = 700)  

Population-based age and sex-specific HIV prevalence estimates

The overall HIV prevalence was 39.9% (95% CI 35.9–43.8%). Four (0.6%) out of 722 tested individuals had an indeterminate HIV test result. Young adults (18–27 years) had the lowest HIV prevalence rates (23.2%; 95% CI 17.9–28.6%). The HIV prevalence in older adults (28–47 years) was found to be significantly higher than in younger individuals (P < 0.0001). The overall proportion of HIV-infected individuals tended to be higher among women (43.1%; 95% CI 37.6–48.5%) than men (37.6%; 95% CI 33.0–43.2%) but this difference between sexes was not statistically significant (P = 0.33) (Table 3).

Table 3. Population-based HIV test results in the Manhiça district by age group and sex
  n Prevalencea (%)95% CI
  1. CI, confidence interval.
  2. aTaking into account the survey sampling design.
Age group   
18–27 years (n = 234)   
28–37 years (n = 242)   
38–47 years (n = 246)   
Male (n = 356)   
Female (n = 366)   

Figure 1 shows the age- and sex-specific HIV prevalence determined in the community-based survey. The estimated HIV prevalence in women aged 18–27 years was 30.8% (95% CI 22.3–39.2%) and in men of the same age it was 17.1% (95% CI 10.0–24.0%). In the 28–37-year age group, the proportion of individuals with HIV infection rose to 45.9% (95% CI 37.0–54.8%) in women and to 39.2% (95% CI 30.4–48.0%) in men. Finally, in adults aged 38–47 years the HIV prevalence was 46.5% (95% CI 37.7–55.2%) in women and 43.7% (95% CI 34.7–52.7%) in men. Although the HIV prevalence was consistently higher in women than in men in all age groups, the only statistically significant difference between men and women was found in the youngest age group (P = 0.014).

Figure 1.

Population-based sex- and age-specific HIV prevalence. Brackets show 95% CI. CI, confidence interval.

Comparison between the community survey and the ANC data

The community-based estimates were compared with the HIV surveillance data from the ANC of the MDH, stratifying by the predefined age groups. The proportion of women at the ANC who were infected with HIV was 23.5% (155 of 660; 95% CI 20.2–26.7%) in the 18–27-year age group, 42.7% (108 of 253; 95% CI 36.6–48.8%) in those aged 28–37 years, and 35.9% (14 of 39; 95% CI 20.6–51.1%) in those aged 38–47 years (Fig. 2). HIV prevalence estimates from the ANC tended to be lower than those for women tested in the community in the three age groups. Globally, HIV prevalence was 1.4 times higher in women tested in the community (43.1%; 95% CI 37.6–48.5%) than in pregnant women attending the ANC (29.4%; 95% CI 26.7–32.0%; P < 0.0001). However, after stratifying by age group, there were no significant differences in HIV prevalence between women at the ANC and the community. The overall HIV community prevalence (men and women) tended also to be higher than the ANC estimates.

Figure 2.

Antenatal clinic (ANC) and community estimates of HIV prevalence in women from Manhiça district by age group. Brackets show 95% CI.


This is the first study to assess sex- and age-specific HIV prevalence in a Mozambican community through individualized random sampling. Mozambique is one of the countries with the greatest burden of HIV infection in the world, and the high HIV prevalence found in this study confirms the magnitude of the epidemic in the southern region of the country. The current results are consistent with recent local hospital-based estimates from previous studies which showed an HIV seropositivity of 37.8% in adults attending the out-patient clinic with reported fever [19] and an HIV prevalence of 49% in women at delivery [20].

An important factor when analysing population HIV prevalence estimates is the level of nonresponse, as this can result in substantial biases in the population estimate [6, 21]. In this study the refusal rate excluding participants contacted but not invited was lower (13.9%) than that found in South Africa, which reached up to 50% [21, 22]. As observed in other settings, the refusal rate among men was higher than that in women [23]. This gender pattern is likely to be explained by cultural and behavioural factors. It has been suggested that, in cases of a high refusal rate, the HIV estimates should be corrected for selection on unobserved variables [24]. It was shown in Zambia that nonparticipation bias could affect the HIV prevalence estimate in men, which increased from 12% to 21% after adjusting for nonparticipation [24]. However, it did not affect HIV prevalence estimates in women. In addition, the use of mortality rates to adjust survey HIV prevalence estimates in rural South Africa increased the overall prevalence by around 7% [21]. In situations of high nonparticipation rates in surveys conducted in low-income settings, it has also been suggested that the data collected should be carefully verified and the interviewers should be closely monitored to ensure validity of the results [25]. The current survey did not capture all subjects who were absent from the household at the time of the invitation and at the time of the mobile team visit. Consequently, although the actual rate of refusal to participate in the study was relatively low, the number of absences could have introduced a bias. For instance, it could be hypothesized that sick individuals tend more frequently to stay at home than healthy individuals, and thus the HIV prevalence estimates may be biased towards a higher proportion of infected people.

As reported in most sub-Saharan countries [1, 6, 22, 26, 27], a gender disparity in the prevalence of HIV infection was also found in this study in all age groups, although the only statistically significant difference in HIV prevalence between women (30.8%) and men (17.1%) was observed in the youngest age group (aged 18–27 years). This difference may be attributable to the previously demonstrated increased vulnerability of women to HIV infection [28-30]. Biological, social and behavioural risk factors (such as age differences between sexual partners) have been suggested to contribute to the difference in HIV prevalence between the sexes in other African countries [30, 31]. In particular, in this area male partners are on average 5 years older than their female counterparts [32]. In addition, the observed gender difference in the youngest age group may be linked to the high migration rate of men in the Manhiça area (on average 100 per 1000 person- years) which peaks in 25-year-old men [11]. This migration pattern may indeed have contributed to a reduction in the number of young men present in Manhiça at the time of the survey. In addition, as previously mentioned, nonparticipation of men could also lead to a lower apparent HIV prevalence in men than in women [24].

At the end of 2010, the Mozambican Ministry of Health published the final results of the first population-based national survey on HIV infection prevalence, carried out in 2009 [4]. This national survey found an overall HIV prevalence of 11.5% in individuals aged 15–49 years, and stratification by regions showed a prevalence of 19.8% for Maputo Province. The difference between the results of the current survey in Manhiça (overall prevalence of about 40%) and those of the national survey in the same province may be explained by various factors. Firstly, differences in the methodology used (cluster sampling in the national survey versus individualized sampling in the current survey) may partly explain the disparity. People from the same village tend to resemble each other more than people from different villages in terms of disease risk [33]. In addition, individuals in a household cluster are not usually ‘independent’ of each other. However, a significant design effect seems unlikely in a national survey including over 10 000 participants [34]. Another factor contributing to discrepancies between the local Manhiça and national surveys may lie in the age limits of the two surveys. Unlike the national survey, teenagers were not included in the current cross-sectional survey. HIV infection rates in teenagers are usually lower than in adults and including them in a survey could decrease the overall community HIV prevalence estimate. As this was the first HIV population-based survey in the Manhiça community, its acceptability was unknown and the survey was thus limited to adults. Future community studies in this and similar settings should include individuals younger than 18 and older than 47 years.

ANC prevalence estimates generally provide useful information for monitoring HIV epidemic trends over time and have traditionally been used to estimate national rates [6]. The current findings show that, in this area, data derived from the ANC surveillance underestimate the HIV prevalence rates of women in the community, in all age groups but especially in the youngest group (18–27 years). These results are in agreement with those of other studies [5, 35, 36]. The representativeness of participants and nonresponse bias have been suggested as explanations for discrepancies between ANC and community estimates [3]. A plausible reason for the underestimation of the number of women infected with HIV in Manhiça based on the data from the ANC is the association of HIV infection and subfertility [37, 38]. HIV-infected women are generally less likely to become pregnant and would therefore be underrepresented at the ANC services [37].

It has been hypothesized that ‘hotspots’ for HIV infection may exist in small southern African communities [39]. For instance, migration [11] is known to be important in Manhiça District and could play an important role in local HIV transmission patterns [20]. Its location on the north–south highway and railway corridor between Maputo and Beira may also contribute to the particularly high HIV prevalence estimate found in the Manhiça area.

In agreement with studies from Zambia and Cameroon [35, 40], HIV prevalence in the Manhiça community increased with age in both women and men. However, some population-based studies from South Africa have shown a decrease in HIV infection rates in the third decade of age [31, 41]. Age-pattern differences compared with surrounding countries such as South Africa and Zimbabwe may be caused by the above-mentioned methodological differences (nonparticipation and mortality rates adjustment) and higher migration rates in Manhiça district [21, 42].

In conclusion, these results show high (> 35%) HIV infection rates in adults in this southern area of Mozambique. Furthermore, in this area HIV prevalence estimates from routine ANC surveillance tended to underestimate the magnitude of the epidemic, especially in the youngest age group. The estimated HIV infection rates will help to identify populations at greatest risk for infection which need to be prioritized in prevention programmes and strategies [43]. Indeed, HIV/AIDS education programmes commonly target adolescents and younger adults, but our results suggest that prevention programmes should also be extended to older adults [44]. Improving the prevention tools already available is crucial, but the development and testing of innovative prevention strategies such as circumcision, prevention strategies that include HIV-infected individuals and test and treat may need to be tailored to different age and risk groups, especially in sub-Saharan countries such as Mozambique, where the epidemic continues to exact a severe toll.


This work was supported by the European and Developing Countries Clinical Trials Partnership (EDCTP) as part of the AfrEVacc consortium and co-funded by the Fondo de Investigaciones Sanitaria from the Spanish Ministry of Health. The CISM receives core funding from the Spanish Agency for International Cooperation (AECI) and the HIV VCT units and personnel from the Manhiça Health Centre are supported by the Agència de Cooperació Catalana. R.G. was supported by a grant from the Spanish Ministry of Health (Contrato post-Formación Sanitaria Especializada ‘Rio Hortega’, Fondo de Investigaciones Sanitarias, Instituto de Salud Carlos III, ref. CM07/0015). We are grateful to the study participants, field workers, HIV counsellors and all the staff from the Demography and Social Sciences Departments at the CISM, especially to Charfudin Sacoor, Elpidia Pedro, Carolina Mindu and Helena Boene.