Uptake of HIV voluntary counselling and testing services in rural Tanzania: implications for effective HIV prevention and equitable access to treatment

Authors


Corresponding Author Alison Wringe, Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. E-mail: Alison.Wringe@lshtm.ac.uk

Summary

Objective  To describe the associations between socio-demographic, behavioural and clinical characteristics and the use of HIV voluntary counselling and testing (VCT) services among residents in a rural ward in Tanzania.

Methods  Eight thousand nine hundred and seventy participants from a community-based cohort were interviewed, provided blood for research HIV testing, and were offered VCT. Univariate and multivariate logistic regression was used to identify socio-demographic, clinical and behavioural factors associated with VCT use.

Results  Although 31% (1246/3980) of men and 24% (1195/4990) of women expressed an interest in the service, only 12% of men and 7% of women subsequently completed VCT. Socio-demographic factors, such as marital status, area of residence, religion and ethnicity influenced VCT completion among males and females in different ways, while self-perceived risk of HIV, prior knowledge of VCT, and sex with a high-risk partner emerged as important predictors of VCT completion among both sexes. Among males only, those infected with HIV for 5 years or less tended to self-select for VCT compared to HIV-negatives (adjusted odds ratio = 1.43; 95% CI: 0.99–2.14). This contributed to a higher proportion of HIV-positive males knowing their status compared to HIV-positive females.

Conclusions  In this setting, a disproportionate number of HIV-positive women are failing to learn their status, which has implications for equitable access to onward referral for care and treatment services. Evidence that some high-risk behaviours may prompt VCT use is encouraging, although further interventions are required to improve knowledge about HIV risk and the benefits of VCT. Targeted interventions are also needed to promote VCT uptake among married women and rural residents.

Abstract

Objectif  Décrire les associations entre les caractéristiques sociodémographiques, comportementaux et cliniques et l’utilisation des services pour conseil et dépistage volontaire (CDV) du VIH chez les résidents d’une zone rurale en Tanzanie.

Méthodes  8970 participants provenant d’une cohorte basée sur la communauté ont été interrogés, un échantillon de sang a été prélevé pour le dépistage du VIH et un CDV leur ont été offert. Des régressions logistiques univariée et multivariée ont été utilisées pour identifier les facteurs sociodémographiques, cliniques et comportementaux associés à l’utilisation du CDV.

Résultats  Bien que 31% (1246/3980) des hommes et 24% (1195/4990) des femmes aient exprimé un intérêt pour le service, seuls 12% des hommes et 7% des femmes ont ensuite complété un CDV. Des facteurs sociodémographique tels que l’état matrimonial, la zone de résidence, la religion et l’appartenance ethnique influençaient l’accomplissement du CDV par les hommes et les femmes de différentes façons. Tandis que le sentiment de risque du VIH, la connaissance préalable du CDV et des relations sexuelles avec un partenaire à haut risque sont apparus comme des indications importantes pour prédire l’accomplissement du CDVT pour les deux sexes. Parmi les hommes, ceux infectés par le VIH depuis 5 ans ou moins avaient tendance àêtre auto-sélectionnés pour le CDV que ceux de VIH négatifs (rapport de cotes ‘Odds ratio’ ajustés = 1,43; IC95%: 0,99 – 2,14). Cette situation a contribuéà une proportion plus élevée d’hommes séropositifs connaissant leur statut par rapport aux femmes séropositives.

Conclusions  Dans ce cadre d’étude, un nombre disproportionné de femmes séropositives ne sont pas au courant de leur statut, ce qui a des implications sur l’accès équitable et pour l’aiguillage vers les soins et les services de traitement. La preuve que certains comportements à haut risque peuvent inciter à l’utilisation du CDV est encourageante, même si des interventions supplémentaires sont nécessaires pour améliorer la connaissance sur les risques du VIH et les avantages du CDV. Des interventions ciblées sont également nécessaires pour favoriser l’utilisation du CDV par les femmes mariées et les habitants des régions rurales.

Abstract

Objetivo  Describir las asociaciones entre características socio-demográficas, de comportamiento y clínicas, y la toma voluntaria de los servicios de aconsejamiento y prueba para VIH (APV) entre los residentes de una zona rural en Tanzania.

Métodos  Se entrevistaron 8970 participantes de una cohorte basada en la comunidad, se les tomó sangre para realizar la prueba para VIH y se les ofreció AVP. Se utilizó una regresión logística univariada y multivariada para identificar los factores socio-demográficos, clínicos y de comportamiento asociados con el uso de AVP.

Resultados  Aunque un 31% (1246/3980) de los hombres y un 24% (1195/4990) de las mujeres expresaron interés en el servicio, solo un 12% de los hombres y un 7% de las mujeres subsecuentemente completaron el AVP. Factores socio-demográficos tales como el estatus marital, el área de residencia, la religión o la etnia tienen influencia en la compleción del AVP de forma diferente entre hombres y mujeres, mientras que la auto-percepción de riesgo de VIH, el conocimiento anterior del AVP y el mantener sexo con un compañero de alto riesgo surgieron como predictores importantes a la hora de completar AVP entre ambos sexos. Entre los hombres, aquellos infectados con VIH durante 5 años o menos, tendían a auto elegirse para AVP comparado con los VIH negativos (odds ratio ajustado = 1.43; 95%CI:0.99–2.14). Esto contribuyó a una proporción más alta de hombres VIH positivos que conocían su estatus comparado con mujeres VIH-positivas.

Conclusiones  En este lugar, un número desproporcionado de mujeres VIH-positivas están dejando de conocer su estatus, lo cual tiene implicaciones para el acceso equitativo a centros de atención especializada y tratamiento. La evidencia de que algunos comportamientos de alto riesgo podrían provocar el AVP es alentador, aunque se requieren más intervenciones para mejorar el conocimiento sobre el riesgo de VIH y los beneficios de AVP. Las intervenciones dirigidas también son necesarias para promover la toma de AVP entre mujeres casadas y residentes rurales.

Introduction

HIV voluntary counselling and testing (VCT) services are a key component of national AIDS programmes in sub-Saharan Africa. In terms of HIV prevention, post-test counselling sessions are an opportunity to provide education, and to promote strategies for reducing the risk of HIV transmission or acquisition among sexually active adults (WHO 2004). Moreover, with the current expansion of antiretroviral therapy (ART) programmes, VCT sites are now acting as a gateway for HIV-infected persons to receive medical care beyond treatment for opportunistic infections (WHO 2003). Indeed, maximizing VCT use among HIV-infected adults will be the first step to ensuring that they have the best chances of starting ART in a timely fashion, and that treatment coverage is as fair as possible.

Data on VCT uptake from a rural cohort study in Rakai, Uganda in 1999 showed socio-economic differentials in terms of VCT uptake, lower uptake among HIV-positive adults compared to those who were HIV negative, and little evidence to suggest higher uptake among those reporting sexual risk behaviours (Matovu et al. 2005). In a more recent study from Manicaland, in rural Zimbabwe, VCT uptake was not associated with sexual risk, but was positively associated with educational attainment, and among women, with being HIV positive (Sherr et al. 2007). However, these studies were conducted prior to the availability of ART, and the extent to which access to these drugs will act as a motivator to use VCT services among HIV-positive or high-risk groups is still uncertain.

This paper describes VCT uptake in relation to socio-demographic characteristics, sexual behaviour, HIV status and spouses’ VCT use and HIV status, after VCT provision during a serological survey conducted in rural Tanzania between September 2003 and March 2004, shortly before the introduction of free ART through the national treatment programme.

Methods

Study setting

The study was conducted in Kisesa ward in Mwanza Region in North-west Tanzania, described in detail by Mwaluko et al. (2003). The ward has a population of 30 000 people living in six villages approximately 20-km east of the regional capital Mwanza, along the main road to Kenya. Economic activities revolve around farming, and per capita income in 2004 was estimated to be <120 USD/year (Mwaluko et al. 2003).

Between 1994 and 2004, four serological surveys of the adult population were conducted, accompanied by height and weight measurements and questionnaire surveys on sexual behaviour and health. Over the same period, 19 rounds of demographic surveillance collecting information on residence and survival status of all household members, on pregnancy among women of reproductive age and on births and migration were completed. In the fourth serological surveillance round (SERO4), conducted between September 2003 and March 2004, consenting study participants provided dry blood spots for HIV testing without result disclosure for research purposes, and were additionally offered a separate service of VCT for HIV in temporary, village-based facilities.

Antiretroviral therapy was not available through the public sector during SERO4, although the Tanzanian government had announced its intention to start providing free ART through major hospitals – including two in Mwanza City – by the end of 2004. Radio programmes and local newspapers provided the main source of information about the national ART programme during this period.

Serological survey

Detailed questionnaires on socio-demographic characteristics, sexual behaviour, health status and previous VCT use were completed for all subjects during interviews conducted by same-sex interviewers. Finger-prick blood was collected from all consenting adults, and tested at the National Institute of Medical Research (NIMR) in Mwanza, with serologic diagnosis of HIV based on two ELISAs (Uniform 2, bioMerieux bv, Boxtel, The Netherlands and Enzygnost HIV1/HIV2, Dade Behring Marburg GmbH, Marburg, Germany).

Voluntary counselling and testing service

During the interview, individuals were asked whether they would like to undergo VCT. Individuals expressing desire for VCT were directed to a separate purpose-constructed hut after the interview, for pre-test counselling with a trained counsellor. Venous blood was collected and routinely tested at NIMR-Mwanza using the same testing protocol as for serological survey samples. Clients were asked to return to the same location for their test results and post-test counselling 1 week later.

Data

Data from the first three serological surveillance rounds were linked to the SERO4 data set using numerical identifiers that could not be linked to named individuals. Marital change variables and household structure variables were constructed using data collected during the demographic surveillance rounds. Spouses were identified from the demographic surveillance data, after which the HIV status of spouses was linked in from the serological data.

The percentage change in body mass index (BMI) was calculated using height and weight data collected during SERO3 (2001) and SERO4. To estimate duration of HIV infection, seroconversion dates were allocated for individuals whose HIV status was known to have changed, by choosing a random date between last negative and first positive test (Wambura et al. 2007). For prevalent cases, infection dates were then estimated using maximum likelihood methods, based on age at first positive test, sex-specific incidence patterns and survival post-infection.

Symptoms consistent with sexually transmitted infections included abnormal discharges or blood from genital areas and genital ulcers. A ‘high-risk’ sexual partner was defined as a non-regular sexual partner working as a driver or bar-maid.

Statistical methods

The main outcomes of interest were defined as (i) expressing desire for VCT; (ii) attending VCT and (iii) completing VCT by returning for the test results and post-test counselling. The proportion achieving each outcome was described using the participants of the serological survey as the denominator, as well as by using the numerator of the previous analysis when considering the second and third outcomes. For each comparison, crude analyses of the associations between the various risk factors and the outcome were performed for both sexes together and stratified by sex. Cross tabulations and chi-square tests were conducted to identify possible confounders, and Mantel-Haenszel stratum-specific odds ratios were examined to check for possible interaction. Logistic regression was used to build multivariate models for identifying the factors that were independently associated with each outcome. HIV status was included in the model as an a priori risk factor of central interest. Interaction terms were added to the model at the end, and were retained where statistically significant.

Ethical approval

Ethical approval for the study was granted by the Tanzanian Medical Research Coordinating Committee (MRCC).

Results

Population characteristics

Among 8970 eligible participants who consented to interview and provided a blood sample for the serological survey, 44% were male, with a median age of 29 years, compared to 31 years for females. HIV prevalence was slightly lower among males (7.5%), than among females (8.2%) and a slightly lower proportion of men than women had been infected for >5 years.

The proportions reporting a high self-perceived risk of HIV (26%), and sex with a high-risk partner in the past year (4%) were similar among men and women. Men were more likely to report multiple sexual partners in the past year (46%) than women (13%), and to have previously heard of VCT (65%vs. 43%). Two percent of men and 1% of women reported previous VCT use.

Desire for VCT

Overall, 31% (1246/3980) of men and 24% (1195/4990) of women expressed desire for VCT (Table 1). VCT desire was highest for both sexes among those with the most years of education, those from non-Sukuma tribes, Muslims and among males only, VCT desire was high among those with recent marital status change (40%). Desire for VCT was lowest among men and women aged 45 years and above who were without education, widowed or following traditional beliefs.

Table 1.   Distribution of voluntary counselling and testing (VCT) desire and VCT uptake, with regard to socio-demographic characteristics, by sex
VariableLevelMalesFemales
DesireTestedDesireTested
Totaln%*n%†%*Totaln%*n%†%*
  1. *As a % of all SERO4 participants.

  2. †As a % of all those who expressed a desire for VCT.

  3. ‡Secondary education or higher.

All 3980124631605491549901195244954110
Age years15–2416254232619646121700447261874211
25–349043844219150211291422331824314
35–4460125542126492189221824934310
45+8501842292501111071081033313
ResidenceRural2305636283084813261646018183407
Roadside8032713412747161108316291404413
Semi-urban8723393917050191266419331724114
Adults in house1+ other3701115431552481544781048234394210
No other251873551592047914330553811
Marital statusMarried1942687353284817262665925249389
Polygamous141503531622253014928634212
Never married159640926194471279618123884911
Sep/divorced10335341954183539326464913
Widowed36514360827226106232
Marital changeNo29278442940248142793656232664110
Yes53421440113532183422827893911
Unknown3571283660471795022424974310
EducationNone62697154041616631831155303
Primary 1–47911762289511178817422764410
Primary 5–722188824039345182342735313224414
Secondary‡344135398361241967940425321
Ethnic groupSukuma377511593155448154602105023418409
non-Sukuma205874251592538814537775320
ReligionChristian2960103035515501743151075254424110
Muslim10543411740161546643375624
Traditional9141731973428521541016303

Among both sexes, VCT desire was highest among those reporting sex with a high-risk partner or inconsistent use of condoms, those with a high self-perceived risk of HIV, and those who had previously heard of or used VCT (Table 2). Among women, the proportion who wanted VCT rose with the number of reported sexual partners (P < 0.001; test for trend). Both men and women were more likely to express an interest in VCT if they were HIV positive, with this being most apparent for males who had been infected for <5 years (42%), and among women who had been infected for >5 years (34%). Finally, among men and women with an HIV-positive partner, 13% (10/80) of the men’s spouses and 20% (19/94) of women’s spouses underwent VCT. Ninety per cent of men and 37% of women with an HIV-positive partner who used the VCT service also expressed desire for VCT themselves.

Table 2.   Distribution of voluntary counselling and testing (VCT) desire and VCT uptake, with regard to clinical and behavioural characteristics, by sex
VariableLevelMalesFemales
DesireTestedDesireTested
Totaln%*n%†%*Totaln%*n%†%*
  1. *As a % of all SERO4 participants.

  2. †As a % of all those who expressed a desire for VCT.

All 3980124631605491549901195244954110
HIV statusNegative3680112230542481545821073234454110
Postive <5 years22695424749212948328344112
Postive ≥5 years7429391655221143934164114
Spouse VCTNo spouse17354492621648121421300211404710
Use & HIV statusHIV−ve, no VCT10223193113241139211601749315
HIV+ve, no VCT7024341667237514193214
HIV−ve, VCT9557604477461796335467326
HIV+ve, VCT10990778701973757126
Spouse, no HIV449141317251169272412687369
Spouse no ID455193429147201068329311243812
Body mass index lossNo75624633126511790219622874410
<5%427127306047145101032041408
≥5%281107384744175711252245368
No BMI data25167663037249153007771263224211
STI symptoms in last 12 monthsNo2314715313374715363580822332419
Yes16665313226850161355387291634212
Risk partner in past 12 monthsNo3435113633557491644801068244354110
Yes154875641472722911048544924
No partners in past 12 months0216452120449604601027454
117255133024548143513916263644010
2+165466340333502059320334994917
Condom use in last 12 monthsNever use24347693236848153331859263353910
Inconsistent68532347168522535218352925026
Consistent1274939295923572137157126
No sex in past year60167112740488476932424
Perceived risk of HIVNone147129520138479199727614109395
Low831325391544719973312321203812
High10634804524250231244472382034316
DK615146247149127761351763478
Heard of VCTNo13812141580376286537713125334
Yes259910324052551202125818383704517
Previous VCTNo3895118730564481449281158234774110
Yes855969416948623760184929

VCT use

Forty-nine per cent of the men and 41% of the women who wished for VCT subsequently attended pre-test counselling and underwent a HIV test (Tables 1 and 2). Having expressed a desire for the service, VCT attendance was highest among males who were the sole adult in their household (59%), those in polygamous marriages (62%), those with secondary education (60%) and those from non-Sukuma tribes (59%). In terms of sexual behaviour, VCT attendance was highest among men reporting consistent condom use (59%), and lowest among those reporting zero partners during the past year (44%). Having expressed a desire to test, VCT use was also high among men who had previously undergone VCT (69%), and among men whose wives had undergone VCT (78% for those with a HIV-positive wife and 77% for those with a HIV-negative wife). Among men who expressed a desire to test, VCT use was lowest among those who had not previously heard of the service (37%).

A variation in this pattern was observed among females: having expressed a desire for VCT, attendance was highest among those who had never married (49%), those who were separated or divorced (49%), those with secondary education (53%), non-Sukuma (53%) and Muslims (56%). VCT attendance also tended to be high among those reporting risky sexual behaviours, such as sex with a high-risk partner (49%), two or more partners in the past year (49%), inconsistent condom use (50%), as well as among those reporting consistent condom use (71%) and those with a spouse who had undergone VCT. Among women who expressed a desire to test, VCT use was lowest among the oldest women, those with traditional beliefs (30%) and those with no education (29%).

The characteristics of VCT users were also described as a proportion of the whole population, as shown in the final columns for each sex in Tables 1 and 2, and suggest some contradictory patterns in terms of desire for, and subsequent use of VCT services. For example, among men, desire for VCT was high among those in monogamous marriages (35%), but having expressed a desire to test, their VCT use was low (48%). Among women, those who were the sole adult in their household were more likely to desire VCT, but were less likely to undergo VCT having expressed an interest, resulting in similar testing uptake among women regardless of the composition of their household (approximately 10%).

Completion of VCT

Figure 1 shows the percentage of adults by sex and HIV status completing VCT, relative to their desire for, and subsequent attendance at VCT. Although there were more HIV-positive females than males in the survey, a higher proportion of HIV-positive males than females completed VCT (17%vs. 10%; P < 0.01), resulting in more men than women being aware of their positive HIV status. Desire for VCT was less likely to translate into use of VCT among HIV-negative males and females compared to their HIV-positive counterparts. Overall, desire for VCT was lowest among HIV-negative females (1051/4582, 23%), while completion of VCT, having attended, was highest among HIV-positive females (42/48, 88%).

Figure 1.

 Uptake of VCT services by HIV status and sex.

Table 3 shows the adjusted odds ratios for socio-demographic, clinical and behavioural factors that were independently associated with completing VCT, having attended SERO4, stratified by sex. Overall, 12% (486/3980) of males and 7% (368/4990) of females who attended SERO4 completed VCT. Among males, the odds of VCT completion were higher among those who had recently changed their marital status (aOR = 1.48; 95% CI: 1.11–1.98) and among those who were from non-Sukuma tribes (aOR = 1.77; 95% CI: 1.21–2.58). Males who had been HIV-positive for up to 5 years were also more likely to complete VCT (aOR = 1.43; 95% CI = 0.99–2.14), as were those whose wives had attended VCT, whether the wife was HIV-positive (aOR = 11.35; 95% CI: 2.00–22.10) or HIV-negative (aOR = 3.85; 95% CI: 2.42–6.12). Males who reported sex with a high-risk partner had higher odds of VCT completion (aOR = 1.53; 95% CI: 0.98–2.39), as did those with a high perceived risk of HIV (aOR = 1.80; 95% CI: 1.37–2.37). VCT completion was negatively associated with having no education (aOR = 0.54; 95% CI: 0.35–0.84), with following traditional beliefs (aOR = 0.54; 95% CI: 0.38–0.76), and with having no prior knowledge of VCT (aOR = 0.44; 95% CI: 0.32–0.60). There was no evidence that age, area of residence, household structure, marital status, recent STI symptoms, recent BMI change, number of sexual partners or condom use were associated with VCT completion among males.

Table 3.   Adjusted odds ratios* for factors associated with completing VCT among all SERO4 attendees, by sex
VariablesLevelMalesFemales
aOR95% CIaOR95% CI
  1. *Adjusted for all other variables shown.

Age (years)15–24  1 
25–34  1.150.85–1.77
35–44  1.230.51–1.77
45+  0.640.37–1.11
Area of residenceRural  1 
Peri-urban  1.501.12–2.02
Urban  1.080.80–1.47
Marital statusMarried (monogamous)  1 
Married (polygamous)  1.310.90–1.90
Never married  4.981.13–21.82
Separated/divorced  6.201.40–27.41
Widowed  2.490.45–13.85
Marital changeNo 1   
Yes1.481.11–1.98  
Unknown1.040.72–1.51  
EducationNone0.540.35–0.840.470.31–0.70
Primary 1–40.870.64–1.170.880.63–1.24
Primary 5–71 1 
Secondary or higher1.170.84–1.621.360.85–2.17
EthnicSukuma1   
Non-Sukuma1.771.21–2.58  
ReligionChristian1 1 
Muslim1.100.61–1.982.141.33–3.46
Traditional0.540.38–0.760.530.26–1.07
HIV statusNegative1 1 
Positive <5 years1.430.99–2.141.200.77–1.88
Positive ≥5 years1.180.62–2.281.530.78–2.97
Spouse HIV status & VCT useNo spouse1 1 
HIV−ve, no VCT0.730.55–0.972.220.50–9.76
HIV+ve, no VCT1.350.70–2.62  
HIV−ve, VCT3.852.42–6.1212.712.87–56.22
HIV+ve, VCT11.352.84–45.3711.711.96–70.05
Spouse, no HIV0.890.63–1.283.050.71–13.17
Spouse unidentified0.960.68–1.353.540.83–15.22
Risk partner in past 12 monthsNo1 1 
Yes1.530.98–2.391.621.06–2.47
Perceived risk of HIVNone1 1 
Low1.361.01–1.841.010.72–1.43
High1.801.37–2.371.320.97–1.81
Don’t know1.110.77–1.610.780.52–1.18
Heard of VCTNo0.440.32–0.600.370.28–0.48
Yes1 1 
Previous VCTNo1   
Yes2.991.85–4.84  

In contrast, among females, the odds of completing VCT were higher among those who were resident in roadside villages (aOR = 1.50; 95% CI: 1.12–2.02), Muslims (aOR = 2.14; 95% CI: 1.33–3.46), and those who had never married (aOR = 4.98; 95% CI: 1.13–21.82) or who were separated or divorced (aOR = 6.20: 95% CI: 1.40–27.41). There was also evidence to suggest that the odds of VCT completion increased with each additional level of completed education (P < 0.01; test for trend). Women whose husband underwent VCT had higher odds of completing VCT themselves, regardless of the HIV status of their husband (aOR = 12.71; 95% CI: 2.87–56.22 for those with HIV negative husbands, and aOR = 11.71; 95% CI: 1.96–70.05 for those with HIV-positive husbands). The odds of completing VCT were also higher among women who reported sex with a high-risk partner (aOR = 1.62, 95% CI: 1.06–2.47), and lower among those who reported no prior knowledge of VCT (aOR = 0.37; 95% CI: 0.28–0.48). There was no evidence to suggest that household structure, marital change, recent BMI change, ethnicity, reported STI symptoms, number of reported sexual partners, or condom use were independently associated with VCT completion among females.

Discussion

We investigated socio-demographic, clinical and behavioural factors in relation to VCT desire, attendance and uptake. Overall, completion of VCT in Kisesa ward during the serological survey was low, despite a relatively high initial interest in testing, and the fact that access to ART was on the horizon. The fact that desire for HIV testing does not necessarily translate into use of VCT, particularly among HIV-positive individuals, suggests that fear of HIV-related stigma and discrimination represent important barriers to HIV testing in Kisesa ward, as in other African settings (Maman et al. 2001; Nuwaha et al. 2002). Moreover, caution needs to be exercised when translating socio-demographic correlates of desire for VCT with potential demand for testing services (Mbago 2004).

Several gender differences were observed in terms of VCT use in this rural setting, and these may have important equity implications in terms of access to HIV prevention messages and referrals to HIV treatment and care services. In particular, there was evidence that males who had been infected for 5 years or less were more likely to use VCT compared to those who were HIV-negative (aOR = 1.43; 95% CI: 0.99–2.14), while a similar pattern was not seen among females, suggesting that a disproportionate number of women who may be in need of ART are failing to access VCT and appropriate onward referral services.

Women in monogamous marriages emerged as a particularly disadvantaged group in terms of VCT use in Kisesa, representing a particular cause for concern in a population where 57% of HIV-infected women are in monogamous marriages, and since marriage is a known risk factor for HIV infection (Glynn et al. 2001; Gregson et al. 2002). A possible explanation for these findings is that monogamously married women believe themselves to be at a low risk of HIV infection. Furthermore, in Kisesa, where HIV-related stigma remains rife (Mshana et al. 2006), it is likely that as in other areas of sub-Saharan Africa, partner violence, marriage dissolution, and a lack of autonomy and empowerment are probable disincentives for women to undergo HIV testing, even when it is provided free of charge (Maman et al. 2001; Pool et al. 2001).

The strongest predictor of VCT use among both sexes, but particularly among women, was VCT use by the person’s spouse. When compared to women with no spouse, women whose husbands also underwent VCT during the SERO survey had around 12 times the adjusted odds of completing VCT themselves, although the confidence intervals around these estimates are wide due to the small numbers of identified spouses who underwent VCT. Nevertheless, efforts to promote couple-counselling may help improve uptake of VCT among married women in this setting, as well as providing many other well-documented benefits of counselling partners together in terms of risk reduction and effective access to care (WHO 2004; Were et al. 2006).

In contrast to findings from rural Uganda (Matovu et al. 2005), but broadly consistent with those from rural Zimbabwe (Sherr et al. 2007), we found that VCT use increased with increasing levels of education. In this context, where only 9% of men and 4% of women in the ward have completed secondary school, these findings highlight the need to disseminate information on the risks of HIV, and on the availability and benefits of VCT and ART beyond schools, and in a format that does not disadvantage those without a formal education. As knowledge of VCT emerged as one of the strongest predictors of VCT use among both sexes, community-led interventions on the benefits and availability of free testing services are likely to have an important impact on the uptake of VCT.

There was strong evidence that significantly fewer followers of traditional religions completed VCT than Christians, among both sexes. This represents an important challenge for improving VCT uptake in Kisesa, where 23% of males and 10% of females report following traditional religions, particularly since these beliefs may translate into practices surrounding HIV prevention, identification and treatment that are inconsistent with biomedical approaches (Plummer et al. 2006). In other African settings, initiatives aimed at promoting referrals to testing sites by traditional healers have shown some success in increasing VCT uptake among individuals initially seeking care through traditional channels (Peltzer et al. 2006), and similar approaches may be useful in rural Tanzania.

In contrast to rural Uganda (Matovu et al. 2005), there was evidence to suggest that in Kisesa, a high perceived risk of HIV motivated both men and women to undergo VCT, and further analyses showed a perceived high risk of HIV to be associated with higher risk sexual behaviours. Furthermore, unlike the Manicaland study, where there was no evidence that risky sexual behaviour was associated with VCT use (Sherr et al. 2007), we found that individuals reporting sex with a high-risk partner in the previous 12 months were more likely to complete VCT. Further analyses showed that among males, there was no difference in terms of HIV prevalence among those reporting sex with a high-risk partner in the past year compared to those who do not, and this is consistent with previous reports who suggest that males may over-report their sexual activity in this setting (Nnko et al. 2004). However, among females, HIV prevalence was almost twice as high among those reporting sex with a high-risk partner in the previous 12 months compared those who did not (15%vs. 8%), providing some cause for optimism that VCT may be relatively accessible to a particular sub-group of high-risk women.

In contrast to risk behaviours, there was no evidence to suggest that HIV-like symptoms, such as STI symptoms in the past year, or a sharp decrease in BM, were independently associated with completion of VCT, for either males or females. Neither was there evidence to suggest that completion of VCT was higher among those infected with HIV for <5 years than among HIV-negative people. One explanation for these findings may be that recent risky behaviour is a stronger motivator to undergo testing than past risks or current illness, suggesting that community campaigns to promote VCT uptake need to focus on treatment-seeking behaviours among those with HIV-like symptoms, to facilitate efficient access to ART therapy.

The limitations of this study relate to the biases inherent in community-level surveys. First, participation rates in the serological survey were higher among females than males, and it is therefore possible that the observed gender differences in VCT uptake would be smaller with higher levels of participation among men. Those who attend serological surveys also might be in poorer health than those who do not attend, because free medical treatment is provided to participants during the survey. However, since there was no evidence that VCT uptake was higher among those with recent BMI loss, or recent STI symptoms, it is unlikely that this bias affected our results.

In conclusion, socio-demographic factors influenced VCT uptake among males and females in different ways in this setting where access to ART was on the verge of becoming available. Interventions are required to improve knowledge about VCT, with a particular focus placed on promoting access to VCT among married women, and those in remote rural areas. Future analyses of VCT uptake in this setting will be able to assess the impact of VCT on behaviour change and HIV incidence, as well as monitoring the extent to which the increasing availability of ART through the Tanzanian national programme attracts infected persons to learn their HIV status.

Ancillary