Correlates of HIV testing among women in Malawi: results from the 2006 Multiple Indicator Cluster Survey


Corresponding Author Adamson S. Muula, Department of Community Health, University of Malawi, College of Medicine, Private Bag 360, Chichiri, Blantyre 3, Malawi. E-mail:


We report a study of women 15–49 years aimed at assessing correlates of HIV testing and having received test results in a nationally representative survey of women in Malawi. A total of 26 259 women were recruited into the study, of whom 3712 (14.1%) had ever been tested for HIV infection and received their results. We found that age and education were not significantly associated with HIV testing but marital status, wealth, region were. Contrary to our expectations that women who had delivered a child were more likely to have been ever tested when accessing prenatal and intra-partum care, we found that women who had delivered a child in the 2 years before the survey were less likely to have ever been tested. We suggest that by 2006 when the survey was conducted, prenatal and intra-partum care were not important avenues for HIV testing in Malawi.


Nous rapportons ici une étude sur des femmes de 15 à 49 ans visant àévaluer les corrélations dans le dépistage du VIH et ayant reçu les résultats du test, dans une enquête nationale représentative auprès des femmes au Malawi. 26 259 femmes ont été recrutées dans l’étude, dont 3712 (14,1%) n’avaient jamais été testées pour l’infection par le VIH et reçu leurs résultats. Nous avons constaté que l’âge et l’éducation n’étaient pas significativement associés au dépistage du VIH, mais l’état matrimonial, la richesse et la région l’étaient. Contrairement à nos attentes que les femmes qui avaient accouchées étaient plus susceptibles d’avoir été testées lors de l’accès aux soins prénataux et intra-partum, nous avons trouvé que les femmes qui avaient accouché au cours des deux années précédant l’enquête étaient moins susceptibles d’avoir été testées. Nous suggérons que jusqu’en 2006, lorsque l’enquête a été menée, les soins prénataux et intra-partum n’étaient pas des voies importantes pour le dépistage du VIH au Malawi.


Reportamos un estudio entre mujeres de 15 a 49 años que buscaba lo correlativo a haberse realizado la prueba para VIH y haber recibido los resultados de la misma, en una estudio nacional representativo en Malawi. Se incluyeron 26,259 mujeres, de las cuales 3,712 (14.1%) habían sido previamente testadas para VIH y habían recibido los resultados. Se halló que la edad y educación no estaban significativamente relacionadas con el haberse realizado la prueba para VIH, pero si lo estaban el estado civil, la riqueza y la región. En contra de nuestras expectativas de que las mujeres que tenían hijos tendrían más probabilidad de haber tenido acceso a la prueba durante los cuidados prenatales o relacionados con el parto, hallamos que las mujeres que habían dado a luz a un niño en los dos años anteriores a la realización del estudio tenían una menor probabilidad de haberse realizado la prueba. Se sugiere que para el 2006,cuando se realizó el estudio, los cuidados prenatales y durante el parto no eran vías importantes para la prueba de VIH en Malawi.


Malawi has an adult HIV sero-prevalence of about 12% (National Statistical Office and ORC Macro 2005). There is a growing number of HIV infected persons receiving treatment as a result of international and domestic investment in treatment (Banda et al. 2008; Harries et al. 2008; Makombe et al. 2008). Much of the recent data on HIV and AIDS in Malawi have come from operations research programmes with emphasis on HIV treatment. Published data on the scope, depth and spread of prevention programmes, such as HIV counselling and testing, are however scarce.

It has been suggested that the counselling that is provided before and after HIV testing may be critical in facilitating positive behavioural change (WHO 2004). However such a finding has not been consistently verified. In a meta-analysis of published research between 1985 and 1997, Weinhardt et al. (1999) observed that while HIV positive participants and HIV serodiscordant couples reduced unprotected intercourse and increased condom use more than HIV negative and untested participants, HIV negative participants did not modify their behaviour more than untested participants. Similarly, Matovu et al. (2005) in Uganda concluded that voluntary HIV counselling and testing (VCT) did not have an impact on subsequent risk behaviours or HIV incidence. Recently, Sherr et al. (2007) found that in Zimbabwe respondents who tested negative were more likely to adopt more risky behaviours in terms of having multiple sexual partners, while those who tested positive reported increased consistent condom use.

The objectives for VCT services in Malawi are (Stanley & Matinga 2004): to provide an opportunity for health education, to provide information for reducing the risk of HIV infection, and to refer clients to support services (including treatment). With regard to entry into care, an individual who knows that he or she is HIV infected is more likely to access HIV care and support services than someone who has not been tested (WHO 2003). Hence, VCT maximizes the number of people exposed to the prevention messages of testing, and ensures that all eligible persons access treatment and care for HIV/AIDS. The introduction of rapid HIV test kits in 2002 may have led to the increase in the number of persons who were tested for HIV infection in public health institutions between 2002 and 2006 from 149 540 to 661 400, respectively (Barnaba et al. 2004). Further increase in the number of clients tested for HIV in the later years may have been due to the national roll out of the free ART programme in June 2004 (Kamoto et al. 2008). The number of persons who have been counselled, tested and received results is bound to increase further with the introduction of mobile HIV counselling and testing (HCT) sites nationwide during HCT week, thus providing access to VCT services to rural areas ( We are unaware of data that have been reported on the socio-demographic correlates of HIV testing and receiving results of the test among women in Malawi. We therefore conducted this study to estimate the lifetime prevalence of having ever had HIV testing and known the results of the test among women 15–49 years in Malawi. The data used in our study was collected in 2006, 2 years after the Malawi public health services initiated an HIV treatment programme where drugs are provided free in both public and private health facilities. Although data was collected from both males and females, these were sampled differently such that they could not be analysed as one data set. In this paper we only describe females who used VCT for HIV.


Study design

We conducted a secondary analysis of the 2006 Malawi Multiple Indicator Cluster Survey (MICS) data, which we obtained from ORC Macro, Calverton MD USA. The MICS is a household (HH) survey initiative developed by UNICEF to assist countries in filling data gaps for monitoring the socio-economic situation of children and women. Its design enables the estimation of nationally representative estimates and allows cross-national comparisons of indicators due to its standard method. The MICS was developed in response to the World Summit for Children to measure progress towards an internationally agreed upon set of mid-decade goals in child and maternal health. A full description of the MICS technique and the Malawi MICS is presented elsewhere (National Statistical Office and UNICEF 2006).

Sampling, data collection and processing

The Malawi MICS used a two-stage sampling method to select the 1200 HHs per district. Within each district of the country, 40 census enumeration areas (clusters) were selected with probability proportional to population size. A HH list was drawn from each cluster and a systematic sample of 30 HHs was eventually identified.

A total of 31 200 HHs (26 districts multiplied by 1200 HHs) were selected in 1040 clusters (26 districts multiplied by 40 clusters) under MICS. All 1040 selected clusters were covered during the fieldwork period between mid-July and mid-November 2006. Data entry overlapped with data collection from August 2006 to end of December 2006. MICS is thus one of the largest HH surveys undertaken in Malawi.

Response rates

Of the 31 200 HHs selected for the sample, all of them were found to be occupied. This is because the house listing operation and the canvassing of HHs took place at the same time. Of these, 30 553 were successfully interviewed for a HH response rate of 97.9%. In the interviewed HHs, 27 073 women (age 15–49) were identified. Of these, 26 259 were successfully interviewed, yielding a response rate of 97.0%.

Outcome variable

We categorized the outcome variable into two groups: (1) respondents who had been tested and received the results were put in one group, and (2) respondents who had not been tested at all, or those who had been tested but did not receive the results into another group. Group 1 was coded as 1 (yes) while group two were coded 0 (No).

Explanatory variables

Explanatory variables assessed were: age, education, marital status, whether respondent had given birth or not in the previous two years to the survey, wealth index, region, and ownership of television. We had hypothesized that with the scaling-up of prevention of mother to child transmission (PMTCT) prevention programmes, women who had given birth would be more likely to have been tested and known their test results than those who had not delivered a baby.

Although ownership of television was part of the wealth index, we hypothesized that ownership of television would be associated with utilization of VCT through watching VCT clips on television.

Malawi is divided into three regions: Northern, Central and Southern. The Southern region is the most populous, and has about half (47%) of Malawi’s poor people living in it, 40% living in Central region, and 11% in the Northern region (2% of no fixed abode) (Benson et al. 2004).

Wealth was defined based on HH assets (such as radio, bicycle, car, television, type of roofing, floor etc) reported by the survey participant. Each asset was assigned a weighting value, using principal component analysis as described by the World Bank and ORC Macro. A HH was assigned a standardized score for each owned asset. For each HH, these scores were summed and HHs ranked into five wealth quintiles (Filmer & Pritchett 1998, 1999; Rutstein & Johnson 2004).

Statistical analysis

Data were analysed using spss version 14.0 (SPSS, Chicago, IL, USA). Frequencies were calculated for the main outcome variable and explanatory variables (Table 1). A weighted logistic regression analysis was conducted to estimate the unadjusted odds ratios for the main outcome and each of the selected explanatory variables (Table 2). A backward stepwise approach of variable selection was used. Variables found to be statistically significant at the 0.05 level in bivariate logistic regression analysis were entered into a multivariate model. The results for the adjusted odds ratios are shown in Table 3. Medians (Q, Q3) were used to summarize the distributions of continuous variables that were not normally distributed.

Table 1.   Frequency of socio-demographic and economic variables among study participants
FactorMedian (Q1, Q3)
Age26 (20, 34)
Education (Completed years in school)4 (2, 6)
Frequency (%) 
Marital/union status
 Currently married/in union18 684 (71.2)
 Formerly married/in union3317 (12.6)
 Never married/in union4258 (16.2)
Wealth index (quintile)
 First (most poor)5166 (19.7)
 Second 5167 (19.7)
 Third5212 (19.8)
 Fourth5114 (19.5)
 Fifth (least poor)5600 (21.3
Gave live birth in the previous 2 years
 No16 049 (61.1)
 Yes10 210 (38.9)
 Northern2857 (10.9)
 Central11 685 (44.5)
 Southern11 716 (44.6)
Owned television
 Yes1792 (6.8)
 No24 461 (93.2)
Table 2.   Associations in bivariate analysis between selected factors and ever been tested for HIV and received the results among females in Malawi
FactorEver been tested and received the resultsNever been testedCrude Odds Ratio OR (95% CI)
Median (Q1, Q3)Median (Q1, Q3)
  1. Percentages are calculated for each row.

Age (years)27 (22, 34)26 (20, 33)1.02 (1.01, 1.02)
Education (completed years in school)4 (2, 6)4 (2, 6)1.01 (1.00, 1.03)
 n (%)n (%) 
Marital/union status
 Currently married/in union2660 (14.2)16 024 (85.8)1.02 (0.96, 1.07)
 Formerly married/in union591 (17.8)2727 (82.2)1.32 (1.24, 1.42)
 Never married/in union462 (10.9)3796 (89.1)1
Wealth index (quintile)
 First (most poor)597 (11.6)4569 (88.4)0.83 (0.77, 0.89)
 Second607 (11.7)4560 (88.3)0.84 (0.78, 0.91)
 Third583 (11.2)4629 (88.8)0.80 (0.74, 0.86)
 Fourth744 (14.5)4370 (85.5)1.08 (1.00, 1.15)
 Fifth (least poor)1181 (21.1)4419 (78.9)1
Gave live birth in the previous 2 years
 No3052 (19.0)12 997 (81.0)1.84 (1.76, 1.93)
 Yes661 (6.5)9549 (93.5)1
 Northern552 (19.3)2306 (80.7)1.33 (1.24, 1.42)
 Central1529 (13.1)10 157 (86.9)0.84 (0.80, 0.88)
 Southern1632 (13.9)10 084 (86.1)1
Owned television
 Yes446 (24.9)1346 (75.1)1.47 (1.39, 1.55)
 No3266 (13.4)21 196 (86.6)1
Table 3.   Associations of selected factors and ever been for HIV and received the results in a multivariate analysis among females in Malawi
FactorAdjusted odds ratios [AOR (95% CI)]
Age (years)
Education (completed years in school)
Marital/union status
 Currently married/in union1.30 (1.22, 1.37)
 Formerly married/in union1.52 (1.41, 1.65)
 Never married/in union1
Wealth index (quintile)
 First (most poor)0.87 (0.79, 0.95)
 Second0.87 (0.80, 0.95)
 Third0.82 (0.75, 0.90)
 Fourth1.08 (1.00, 1.17)
 Fifth (least poor)1
Gave live birth in the previous 2 years
 No1.98 (1.88, 2.08)
 Northern1.25 (1.17, 1.34)
 Central0.90 (0.85, 0.96)
Owned television
 Yes1.16 (1.08, 1.25)


Socio-demographic characteristics of the study participants

We recruited 26 259 women aged 15–49 years into the study, of whom 3712 (14.1%) had ever been tested for HIV and received the results. As shown in Table 1, the median age of respondents was 26 (Q1 = 20, Q3 = 34) years. Participants had completed a median 4 years of school (Q1 = 2, Q3 = 6) years. The majority of the respondents were currently married or were in union (71.2%), had no television (93.2%), and were either from the central (44.5%) or southern (44.6%) regions of Malawi. Only about a quarter (21.3%) of the respondents were in the fifth quintile of wealth index.

Factors associated with HIV testing and getting test results

At bivariate analyses, only education was not significantly associated with ever having been tested for HIV and received the results (Table 2). However in the multivariate analysis, age in addition to education was also not significantly associated with the outcome (Table 3). Compared to respondents who were never married/in union, currently or formerly married/in union were more likely to have ever tested and received the results for HIV infection [adjusted odds ratio (AOR) = 1.30; 95% CI 1.22, 1.37), and AOR = 1.52; 95% CI 1.41, 1.65), respectively]. Respondents in wealth quintile 1 to 3 were less likely to have ever been tested and received the results compared to respondents who were in the fifth wealth quintile. Furthermore, respondents who had a live birth in the previous 2 years to the survey were 98% (95% CI 1.88, 2.08) more likely to have ever been tested and received results compared to respondents who had not given birth to a live baby. Participants who were from the northern and central regions were 25% [95% CI (1.17, 1.34)] more likely, and 10% [AOR = 0.90, 95% CI (0.85, 0.96)] less likely to have ever been tested and received the results compare to participants from the Southern region. Having access to a television was associated with increased odds of ever been tested and received the results for HIV infection [AOR = 1.16, 95% CI (1.08, 1.25)].


In this study in which a nationally representative sample of women 15–49 years were included, we found a VCT uptake rate of 14.1%, and being currently or formerly married or in union, high wealth index, recent child birth, region, and ownership of television were associated with having ever been tested for HIV and received test results. Women who had given birth in the previous two years to the survey were less likely to have been tested and received results. Age and education level were not significantly associated with previous testing and receiving results of the test in multivariate analysis.

Our finding of 14.1% VCT uptake compares with the rate that was obtained in the MDHS-2004 (National Statistical Office and ORC Macro 2005) of 12.9% of women in the 15–49 years age group. With such a low rate of VCT uptake, most HIV infected women would unknowingly transmit the virus to their sex partners as they may not see the need to have protected sex. VCT uptake could be increased through strengthening of VCT services in the prenatal care services that would need its attendance increased.

We had hypothesized that women who had recently given birth to a child would be more likely to have been tested for HIV than if they had not. This was based on understanding that these women would receive testing when they present for prenatal care and delivery services. According to the Malawi Demographic and Health Survey 2004 (MDHA-2004), only 4.6% of women who had delivered a child in the 5 years preceding the survey reported never to have accessed prenatal care at all (National Statistical Office and ORC Macro 2005). Our study however found that having recently delivered a child (and possibly having had at least one prenatal care visit) was negatively associated with having ever been tested. This suggests that by 2006, prenatal care was not an important entry point for HIV testing in Malawi. In the MDHS-2004 about half (52.5%) of the respondents were counselled during antenatal visit, and only 3.1% of them were counselled, tested and received results.

Currently or formerly (current divorced, separated or widowed) married women may have perceived themselves at greater risk and were more likely to have been tested and received results for the following plausible reasons: women who have ever been married have higher HIV prevalence than those who have never been married before. The MDHS-2004 reported that HIV prevalence was 37.4% among widowed women, 25.5% and 12.5% among divorced/separated and currently married, respectively, compared to 5.3% among those never married at all (National Statistical Office and ORC Macro 2005). It is possible therefore that ever married women in the MICS study had realized the risk of HIV associated with marriage in Malawi. Some formerly married women may have been widowed due to HIV/AIDS, and thus used the VCT services to check on their HIV status. Our finding that women who were currently of formerly married were more likely to utilize VCT services concurs with that found by Matovu et al. (2005) who reported that VCT acceptance was higher among currently married and previously married women.

It is interesting also to note in our study that women in the highest wealth quintiles were more likely to have been tested than women in the lower wealth quintile. This finding is consistent with a result in the same study that women from the richer Northern region were more likely to undergo VCT and obtain results compared to women from poorer regions. Educational level however was not associated with testing. The MDHS-2004 reports that 14.9% of women in the highest wealth quintile were HIV infected in 2004 compared to 4.4% of women in the lowest quintile (National Statistical Office and ORC Macro 2005). Previous results have shown or suggested that among women in Africa, wealth may be associated with HIV risk through their spouses’ risk behaviour. Wealth may make men more desirable to non-marital partners, and may facilitate mobility, which exposes an individual to potential sexual partnerships outside one’s usual geographical residence or work environment (Gabrysch et al. 2008). In a study in Kenya, Hargreaves (2002) also reported higher HIV infection prevalence among high socio-economic groups.

These data were obtained via a cross sectional survey and so cannot therefore ascribe causality to any of the explanatory findings (Kaufman & Cooper 1999; Kaufman et al. 2003). Furthermore, data were collected through self-reports. To the extent that study participants miss-reported, both intentionally and inadvertently, our results may be biased. However, data collection in the 2006 MICS study was done in private between the interviewer and study participant in order to ensure truthful reporting and minimize intentional misreporting.

The 2006 MICS study did not obtain information on sexual behaviour, such as condom use. We can only speculate that inclusion of such behavioural factors in multivariate analysis may have confounded the relationships between some of the factors we considered in our study and the outcome. It has also been reported that condom use is associated with socio-economic status, as well as with use of VCT services (Bankole et al. 2007). Inclusion of condom use in our study may have confounded the relationship between the economic factors in our study (wealth index, region, and ownership of television) and VCT uptake. However we are unable to speculate on the direction and magnitude of bias in our study as a result of non-inclusion of sexual behavioural factors that were not collected in the 2006 MICS study in our multivariate model.


Using data from a nationally representative survey of women in Malawi, we found that wealth, having ever been married or in union, region, and ownership of television were positively associated with having ever been tested for HIV and received the results. Having given birth recently was negatively associated with uptake of VCT. We conclude that by 2006, attendance to prenatal care may not have provided benefit of HIV testing among women in Malawi, probably due to shortage of health personnel manning this service.


We thank ORC Macro for making these data available to us for analysis. Permission to analyse these data was also obtained from the Malawi National Statistical Office. However, neither ORC Macro nor the NSO influenced the analysis, design, or decision to publish these findings.