Comparison of HIV-1 prevalence and risk factors between pregnant,non-pregnant,all women and the general population in Tanzania: implications for second-generation surveillance

Abstract

The aim of this study was to compare HIV-1 prevalence between pregnant, non-pregnant, all women and the general population (men and women) and investigate how well risk behaviours of pregnant women reflect those of all women in the general population in rural Kilimanjaro, Tanzania. A cross-sectional survey involving all individuals aged 15–44 years was conducted in Oria village between March and May of 2005. All consenting individuals were interviewed and offered HIV-1 and syphilis testing. The response proportion among women was found to be 73.7% (914/1241). Age-adjusted HIV-1 prevalence among pregnant women (n = 92) was 32.5% lower (5.4% versus 8.0%, P, 0.001) than that of all women but only 8.5% (5.4% versus 5.9%, P = 0.639) lower compared with that of the general population. HIV-1 risk factors among pregnant women were comparable to that of all women in the general population. In conclusion, pregnant women in the general population underestimate the HIV-1 magnitude among women but closely approximate that of the general population. HIV-1 risk factors are similar between pregnant and all women in the general population. Second-generation surveillance could utilize pregnant women risk behaviours to approximate that of all women in the general population.

Keywords

pregnant women HIV prevalence risk behaviours second-generation surveillance Tanzania

Introduction

Antenatal clinic (ANC) sentinel seroprevalence surveys are the major empirically based source of HIV-1 prevalence data. Although popular, ANC data are subject to biases. ANC attendance is more likely to vary by age, marital status, locality, socioeconomic status and education level; factors associated with HIV-1 serostatus.^1,2 Only sexually active non-contraceptive users attend ANC and HIV-1 tends to reduce fertility for biological and social reasons. The effect of reduced fertility among HIV-positive women has been well studied.^3–9 Overall, HIV-1 prevalence from ANC data have been reported to be 10–30% lower than that of women in the general population especially for those over 20 years.^1,10 In the efforts to validate the use of ANC data as a sentinel population in HIV-1 population estimation, comparison of ANC and general population HIV-1 prevalence has been tried.^11–14 Some studies have even employed a mathematical model in order to adjust ANC data for the purpose of extrapolating general female population estimates.^3,15–18

Differential attendance to ANC has been held as one of the cause of biased estimates in HIV-1 prevalence. Direct comparison of HIV-1 magnitude between pregnant, non-pregnant and all women (overall) from the general population may help in avoiding the effect of selection bias among ANC attendees. This may shed light on the magnitude of selection bias and whether the ANC estimates reflect the reality in the general population. Moreover, the World Health Organization (WHO) is recommending the use of ‘second-generation’ surveillance where inclusion of risk behaviours in routine ANC surveillance is advocated.^19–21 Only one study has tried to explore the usefulness of behavioural data from ANC in general population behavioural monitoring.²² However, it is not known how well risky sexual behaviours estimates from pregnant women in the general population approximate those of non-pregnant women and those of all women of reproductive age.

In this study, we had a unique opportunity to examine the above-mentioned issues by comparing HIV-1 prevalence and risky sexual behaviours between pregnant, non-pregnant, all women (pregnant and non-pregnant) and the general population (men and women) from a total village population survey in rural Kilimanjaro, Tanzania.

Methods

Study area

The study was conducted in Oria village in Kahe ward, Kilimanjaro, Tanzania. Details of this study area have been described elsewhere.^23,24 In brief, Oria is a rural village at the foot of Mount Kilimanjaro about 30 km from Moshi town in the Kilimanjaro region. The majority of the people engage in small-scale rice and maize farming and some keep goats and/or cattle. The village has one private and one public dispensary.

Study design and population

This was a population-based cross-sectional study conducted between March and May of 2005. All individuals aged between 15 and 44 years (n = 2093; women = 1241, men = 852) and listed as having an address in the village were eligible for participation.

Data collection procedures

The research team completed a house-to-house registration of all eligible individuals. Eligibility was based on being at the given age range and having an address in the village. The village had nine hamlets (small administrative units) and data collection was done from hamlet to hamlet. Appointments for data collection were set by the research team and the hamlets leaders. At the set dates, the research team went from house to house. Participants received information regarding the study aims and procedures, and those who agreed to participate provided a written informed consent. A structured questionnaire was administered in a private place to ensure confidentiality. During the interviews, information on the socio demographic factors (age, sex, education, current occupation and marital status), HIV-1 risk factors (number of sexual partners in the past four weeks and 12 months, exchange of money/goods for sex, past month treatment for genital ulcer disease, genital discharge and past five years history of treatment for sexually transmitted infection) was collected. Data on pregnancy status were also collected. These included current pregnancy status, parity and time since last delivery. The interviews were followed by pretest counselling and subsequent blood collection for HIV-1 and syphilis testing. Appointments were made with each participant for post-test counselling within two weeks from the date of sample collection. Efforts were made to provide results to all who wanted to know their status as this would serve as an entry point for care and prevention. Individuals who indicated that they did not want to know their test results were also re-visited and re-counselled. All missed participants during recruitment were visited three times at home or at their nearby work place before being considered as non-respondents.

Ethical issues

Ethical approval for the study was provided by the Ethical Committee of the Ministry of Health in Tanzania and the Norwegian Committee for Medical Research Ethics in Norway. The village government of Oria granted permission for the study. All participants gave written informed consent for the interview and blood sampling separately. HIV-1 test results were issued to respondents in person after pretest and post-test counselling. Further medical follow-up was offered for all HIV-1-positive respondents. All residents of the village had access to free treatment for common medical conditions from a research mobile clinic throughout the survey duration.

Laboratory methods

Two independent enzyme-linked immunoassay (ELISA) systems (Voronostica Uniform II plus O; Organon, Boxtel, the Netherlands and Enzygnost Anti HIV 1/2 Plus; Dade Behring, South Africa) were used in the detection of HIV-1 antibodies. In case of discordant or weakly concordant ELISA results, Western blot (Bio-Rad Laboratories Ltd, Dartford, UK) was used for confirmation. Past infection with syphilis was tested using rapid plasma reagin test (Immutrep RPR, Omega Diagnostics, Scotland, UK). All specimens were tested at the clinical laboratory of the Kilimanjaro Christian Medical Centre, Moshi, Tanzania.

Statistical analysis

Frequencies and cross tabulation of various demographic characteristics was done for descriptive purposes. The χ² test for differences in proportions was applied for categorical variables; continuous normally distributed variables were presented as means with their respective standard deviations (SDs). To test for differences between means, Student's t-test and analysis of variance (ANOVA) were employed. Age-specific HIV-1 prevalences were calculated using the district (Moshi rural) population as standard population. Strength of association between HIV-1 infection and various pregnancy statuses was estimated by calculating the odds ratios with 95% confidence intervals (CIs). Multivariable logistic regression with HIV-1 seropositivity as a dependent variable was used to control for potential confounders (age, marital status and parity). Adjusted odds ratios (AOR) with 95% CI were then presented. The Statistical Package for Social Sciences (SPSS) for windows version 15.0 (SPSS Inc., Chicago, IL, USA) was used for the data analysis.

Results

A total of 1528 individuals (men and women) participated in the study. The overall response proportion was 73.0% (1528/2093). Of all participants, 914 (59.8%) were women (response proportion 73.7%) and 614 were men (response proportion 72.1%). The mean age did not differ significantly between women and men (mean age 27.9 [SD ±8.6] years for women versus 28.7 [SD±8.8] for men, P = 0.056). The overall age-adjusted HIV-1 prevalence in the village was 5.9% (8.0% for women and 3.2% for men). During the time of the survey, 92 (11.2%) of the women were pregnant and 822 were not pregnant. A total of 708 (77.5%) reported to have ever been pregnant and the majority of them, 433 (61.2%), reported to have delivered in the past five years preceding the survey. Half of those who have ever been pregnant 354 (50.0%), had had 2–4 pregnancies.

Table 1 presents background characteristics of pregnant women, non-pregnant women and all women overall. A significant proportion of pregnant women were younger (15–34 years), while majority of non-pregnant women were older (35–44 years). Pregnant women therefore had lower mean age (26.6, SD±6.3) compared with non-pregnant women (30.1, SD±7.5), P < 0.001. The mean age of pregnant women was more comparable to that of all women (27.9, SD±8.8). Compared with non-pregnant women, a significant proportion of pregnant women were married or cohabiting. Overall, pregnant women had comparable demographic characteristics to those of all women in the general population than did non-pregnant women.

Table 1

Comparison of background characteristics between pregnant, non-pregnant and all women in the general population in a rural village, Kilimanjaro, Tanzania

		Percent of
Variable	Category	Pregnant (n = 92)	Non-pregnant (n = 822)	All women (n = 914)	P value
Age groups	15–24	45.7	31.3	38.5
	25–34	40.1	32.7	36.4
	35–44	14.2	36.0	25.1	<0.001
Marital status	Single	23.3	25.9	24.6
	Married/cohabiting	71.4	54.8	63.1
	Divorced/separated	5.3	19.3	12.3	0.013
Education	None	12.5	14.9	13.7
	Primary education	78.2	79.0	78.6
	Secondary+	9.3	6.1	7.7	0.985
Occupation	Farmers	85.5	89.3	87.4
	Employed	2.9	2.5	2.7
	Petty business	8.2	11.6	9.9	0.144
Religion	Catholics	24.2	25.4	24.8
	Protestant	29.5	27.1	28.3
	Muslim	46.3	47.5	46.9	0.460

Significant associations highlighted in bold

Figure 1 compares HIV-1 prevalence between pregnant, non-pregnant women and all women in the general population by age groups, while Figure 2 compares the HIV-1 prevalence between pregnant women, non-pregnant women and the general population by age groups. Under the age of 24 years, pregnant women had significantly higher HIV-1 prevalence as compared with non-pregnant women, all women and the general population (Figures 1 and 2). Above the age of 24 years, pregnant women had a significantly lower HIV-1 prevalence as compared with non-pregnant women, all women and the whole general population. This difference was much larger when pregnant women were compared with all women than when compared with the general population (Figures 1 and 2). The overall age-adjusted HIV-1 prevalence among pregnant women was 32.5% lower than that of all women in the general population (5.4% versus 8.0%, P < 0.001). However, this prevalence was only 8.5% (5.4% versus 5.9%, P = 0.639) lower when compared with that of the whole general population.

Figure 1

Comparison of HIV-1 prevalence between pregnant, non-pregnant women and all women in the general population in rural Kilimanjaro, Tanzania. *No HIV-1 seroconversion among the non-pregnant women in this age group, ^†Age-adjusted HIV-1 prevalence

Figure 2

Comparison of HIV-1 prevalence between pregnant, non-pregnant women and the general population by age groups in rural Kilimanjaro, Tanzania. *No HIV-1 seroconversion among the non-pregnant women in this age group, ^†Age-adjusted HIV-1 prevalence. General population include both men and women

Multivariable analysis of pregnancy status and HIV-1 infection prevalence are presented in Table 2. Women who were currently pregnant had 30% less HIV-1 infection compared with those who were not pregnant. HIV-1 prevalence was found to increase with increased duration since last delivery. Women with last delivery >10 years preceding the survey had 1.6 (95% CI: 1.1–9.5) times higher HIV-1 prevalence compared with those with 0–5 years. Moreover, HIV-1 prevalence was observed to increase with parity but this was not significant after controlling for potential confounders in the multivariable analysis (Table 2).

Table 2

Multivariable analysis of pregnancy status and HIV-1 infection among women in a rural village Kilimanjaro, Tanzania

				Odds ratio (95% CI)
Variable	Category	No	%HIV+	Crude	Adjusted^*
Currently pregnant	No	822	8.2	1	1
	Yes	92	5.4	0.7 (0.2–0.8)	0.7 (0.2–0.9)
Ever been pregnant	No	298	9.6	1	1
	Yes	616	7.3	0.8 (0.4–3.8)	0.9 (0.3–4.3)
Years since last delivery	0–5	497	5.1	1	1
	6–10	141	7.8	1.2 (0.2–4.6)	0.9 (0.3–1.2)
	>10	60	12.1	2.1 (0.9–9.0)	1.6 (1.1–9.5)
Parity (number of deliveries)	1	151	7.0	1	1
	2–4	311	8.3	1.2 (0.4–3.9)	1.1 (0.2–5.1)
	>4	154	9.6	1.4 (0.7–4.2)	1.2 (0.3–6.6)

CI-confidence interval; OR = odds ratio,

Odds ratio-adjusted for age, marital status and parity; Significant associations highlighted in bold

Mean age at sexual debut were 18.0, 17.7 and 17.6 years among pregnant, non-pregnant and all women, respectively. The corresponding mean ages at first marriage were 19.9, 19.2 and 19.3 years, respectively. From ANOVA, the mean age at sexual debut and marriage did not differ significantly (P = 0.967 and 0.965, respectively) between the three groups. Table 3 depicts comparison of HIV-1 risk factors between pregnant and non-pregnant women and all women (pregnant and non-pregnant) in the study village. A number of variables were used to assess the difference in risk behaviours between the three groups. A significant proportion of non-pregnant (74.5%) women reported no use of condom during last casual sex compared with pregnant (61.7%) and all women (68.1%) (P = 0.002). When these proportions were compared between pregnant and all women, no significant difference was observed (61.7% versus 68.1%, respectively, P = 0.234). Although a large proportion of non-pregnant women reported more risk behaviours compared with pregnant women and all women, the three groups did not differ significantly with regards to reported risk behaviours (Table 3).

Table 3

Comparison of HIV-1 risk factors between pregnant, non-pregnant and all women in the general population in a rural village Kilimanjaro, Tanzania

		Percent of
Variable	Category/test	Pregnant (n = 92)	Non-pregnant (n = 822)	All women (n = 914)	P value^*
First sex†	≤ 17 years	38.4	40.6	39.5	0.828
First married‡	≤ 17 years	17.1	15.7	16.4	0.870
Marital status	Polygamous	11.5	7.3	9.4	0.145
Ever condom use†	Ever use	31.6	28.4	30.0	0.613
Condom use last casual sex§	Not used	61.7	74.5	68.1	0.002
Exchanged money/goods for sex at last sex†	Yes	8.4	10.0	9.2	0.708
Partners past four weeks	≤2 partners	3.8	5.6	4.7	0.507
Partners past 12 months	≤2 partners	10.4	13.4	11.9	0.466
Genital ulcer	Had past month	1.1	1.7	1.4	0.931
Genital discharge	Had past month	3.1	4.5	3.8	0.436
STI	Past five years	12.9	18.5	15.7	0.175
Ever had syphilis	RPR positive	1.3	2.5	1.9	0.895

RPR = rapid plasma regain; STI = sexually transmitted infection; Significant associations highlighted in bold

P value from χ² for difference in proportions

†

From those who ever had sex

‡

From those reported to have been married

From those reported to have casual sex

Discussion

The results from this study showed that pregnant women in the general population were significantly younger and more frequently married/cohabited compared with non-pregnant women. HIV-1 prevalence was significantly lower (32.5% lower) among pregnant women as compared with non-pregnant women in this rural population but closely similar to that of the general population (men and women combined). Young (<24 years) pregnant women overestimated HIV-1 prevalence among both all women and the whole general population while older women (>24 years) underestimated the same estimates. Pregnant, non-pregnant women and all women had similar age at sexual debut and age at marriage. Reported risk behaviours among pregnant women were comparable to that of all women in the general population. Increased duration since last delivery was significant associated with HIV-1 infection.

This study shows that although HIV-1 testing among pregnant women tends to significantly underestimate the actual HIV-1 magnitude in the female population, they approximate closely the overall general population estimates (women and men combined). This observation could be explained by the fact that women in Africa have disproportionally high HIV-1 prevalence compared with men (8.0% versus 3.2% in this population)²⁴ and involving men estimates adjust downward the overall general population estimates close to that of pregnant women. Similar findings have been recently published from Zambia using women attending ANC.²⁵ This study therefore underpins these findings using pregnant women from the general population by comparing them with non-pregnant, all female population (pregnant and non-pregnant) and the whole village population (men and women). Therefore, in areas where ANC utilization is high, ANC HIV-1 estimates reflect the reality in the general population and selection bias might have very little effect. The importance of these findings is the usefulness of pregnant women in directly extrapolating general population HIV-1 estimates.

Differences in HIV-1 risk factors between pregnant, non-pregnant women and overall women in the general population have yet to be described. A study in Mwanza Tanzania tried to examine the usefulness of ANC collected behavioural data in reflecting the general population behavioural data.²² However, this study involved a selected sample of pregnant women attending ANC and did not involve non-pregnant women. This study had a unique opportunity to compare unbiased sample (from the general population versus ANC sample) of pregnant, non-pregnant women and all women (overall) in terms of various HIV-1 risk factors. It was shown in this study that although pregnant women were younger than non-pregnant women, these two groups had similar age at sexual debut and marriage. Moreover, pregnant women's reported risk behaviours were comparable to that of all women of reproductive age in the general population. This indicate that in places where ANC utilization is high in such a way that women attending ANC represent women of reproductive age in the general population, estimation of HIV-1 risk factors among them could be held to approximate that of all women in the general population. These findings support the WHO ‘second-generation’ surveillance recommendation by showing that collection of various HIV risk factors from ANC attendee may not only explain their HIV estimates but also risk factors of all women in the general population.^19–21 The increased HIV-1 infection with time since last delivery has been observed in many studies.^4,6–8 The impact of HIV on fertility and social explanation like widowhood among HIV-positive women has been held to explain this observation. This is supported by the finding in this population where HIV-1 prevalence was high among women who reported to be widowed/separated.²⁴

Findings in this study have the following implications: firstly, estimated HIV-1 risk behaviours among pregnant women could be used as a proxy measure of risk among all women of reproductive age in the general population. This may be crucial in the implementation of the WHO ‘second-generation’ surveillance. Secondly, HIV-1 estimates from pregnant women could continue to be used as a proxy measure of the general populations estimates and lastly, ANC estimates reflect the actual estimates of respective group (pregnant women) in the general population where ANC utilization is high and selection bias may not have a significant effect.

In this study, a response proportion of 73.7% was achieved among women. As a response proportion above 70.0% have been recommended for population surveys,²⁶ it may be stated that a representative sample of women were included giving the results good external validity. The approach of involving the whole village population in this study may also help to eliminate potential sampling biases. Differential non-response among men may affect general population estimates. In this study, no significant difference in participation was observed between men and women. Comparison of sociodemographic characteristics between participants and non-participants did not show any significant difference indicating that participants were a representative sample of the village population.²⁴ Estimation of reported risky sexual behaviours is always associated with biases that could be due to desirability reporting or the fact that the behaviours of partners are not known.^19,27,28 Moreover, women tend to under-report their HIV risk behaviours in Africa as part of cultural femininity ideologies.²⁷ But these biases may have affected both pregnant and non-pregnant women non-differentially hence introducing less effect on the findings presented.

Conclusions

Pregnant women from the general population significantly underestimate HIV-1 estimates of all women of reproductive age but closely approximate that of the general population (women and men). Reported HIV-1 risk factors among pregnant women are comparable to those of all women of reproductive age in the general population. Long period since last delivery is associated with HIV-1 infection probably due to infertility and widowhood among HIV-1-infected women. A representative sample of pregnant women attending ANC could be used to estimate general population HIV prevalence. Risky sexual behaviours among these women could also be used to explain HIV estimates of all women of reproductive age in the general population as recommended in the ‘second-generation surveillance’.

Funding: This work was in part funded by a grant from The Norwegian Programme for Development, Research and Higher Education (NUFU).

Footnotes

Acknowledgements

The authors are grateful to the participants for their time. This work was facilitated by the collaborating institutions: Muhimbili University of Health Sciences, Kilimanjaro Christian Medical College, Centre for Education in Health Development Arusha in Tanzania, and the Universities of Oslo and Bergen, Norway.

References

Glynn

, Buve

, Carael

Factors influencing the difference in HIV prevalence between antenatal clinic and general population in sub-Saharan Africa.

AIDS 2001; 15: 1717–25

Gregson

, Zhuwau

, Anderson

, Chimbadzwa

, Chiwandiwa

SK.

Age and religion selection biases in HIV-1 prevalence data from Antenatal clinics in Manicaland, Zimbabwe.

Cent Afr J Med 1995; 41: 339–46

Fabiani

, Nattabi

, Ayella

, Ogwang

, Declich

Differences in fertility by HIV serostatus and adjusted HIV prevalence data from an antenatal clinic in northern Uganda.

Trop Med Int Health 2006; 11: 182–7

Glynn

, Buve

, Carael

Decreased fertility among HIV-1 infected women attending antenatal clinics in three African cities.

J Acquir Immune Defic Syndr 2000; 25: 345–52

Zaba

, Gregson

Measuring the impact of HIV on fertility in Africa.

AIDS 1998; 12 (Suppl. 1): S41–50

Zaba

, Wawer

, Sewardda

Population-based study of fertility in women with HIV-1 infection in Uganda.

Lancet 1998; 351: 98–103

Zaba

, Terceira

, Mason

, Gregson

The contribution of HIV to fertility decline inrural Zimbabwe, 1985-2000.

Popul Stud 2003; 57: 149–64

Carpenter

, Nakiyingi

, Ruberantwari

Estimates of the impact of HIV infections on fertility in a rural Ugandan population.

Health Trans Rev 1997; 7 (Suppl. 2): 113–26

Nguyen

, Gange

, Wabwire-Mangen

, Sewankambo

, Sewardda

Reduced fertility among HIV-infected women associated with viral load in Rakai district, Uganda.

Int J STD AIDS 2006; 17: 842–6

10.

Zaba

, Boerma

, White

Monitoring the AIDS epidemic using prevalence data among young women attending antenatal clinics: problems and prospects.

AIDS 2000; 14: 1633–45

11.

Changaluch

, Grosskurth

, Mwita

Comparison of HIV prevalences in community-based and antenatl clinic surveys in rural Mwanza, Tanzania.

AIDS 2002; 16: 661–5

12.

Kwesigabo

, Killewo

JZJ

, Sandstrom

Sentinel surveilance and cross-sectional survey on HIV infection prevalence: a comparative study.

E Afr Med J 1996; 73: 298–302

13.

Kwesigabo

, Killewo

JZJ

, Urassa

Monitoring of HIV-1 infection prevalence and trends in the general population using pregnant women as sentinel population: 9 years experience from the Kagera region of Tanzania.

J Acquir Immune Defic Syndr 2000; 23: 410–7

14.

Kigadye

, Klokke

, Nicoll

Sentinel surveillance for HIV-1 among pregnant women in developing country: 3 years experience and comparison with a population serosurvey.

AIDS 1993; 7: 849–55

15.

Fabiani

, Fylkness

, Nattabi

, Ayella

, Declich

Evaluating two adjustment methods to extrapolate HIV prevalence from pregnant women to the general female population in sub-Saharan Africa.

AIDS 2001; 17: 399–405

16.

Glynn

, Buve

, Carael

, Zaba

Adjustment of antenatal clinic HIV surveillance data for HIV-1 associated differences in fertility.

AIDS 1999; 13: 1598

17.

Nicoll

, Stephenson

, Griffioen

The relationship of HIV prevalence in pregnant women to that in women of reproductive age: a validated method of adjustment.

AIDS 1998; 12: 1861–7

18.

Somi

, Matee

, Swai

RO.

Estimating and projecting HIV prevalence and AIDS deaths in Tanzania using antenatal surveilance data.

BMC Public Health 2006; 6: 120 (doi:10.1186/1471-2458-6-120)

19.

Garnett

, Garcia-Calleja

, Rehle

, Gregson

Behavioural data as adjunct to HIV surveilance data.

Sex Transm Infect 2006; 82 (Suppl. 1): i57–62

20.

World Health Organization. United Nations Joint Programme on AIDS. Second Generation Surveillance for HIV: The Next Decade. Geneva: UNAIDS/WHO, 2006

21.

Rehle

, Lazzari

, Dallabetta

Second generation HIV surveillance: better data for decision making.

Bull World Health Organ 2004; 82: 121–7

22.

Urassa

, Kumogola

, Isingo

HIV prevalence and sexual behaviour changes measured in an antenatal clinic setting in northern Tanzania.

Sex Transm Infect 2006; 82: 301–6

23.

Klouman

, Masenga

, Klepp

KI.

HIV and reproductive tract infections in a total village population in rural Kilimanjaro, Tanzania: women at increased risk.

J Acquir Immune Defic Syndr Hum Retrovirol 1997; 14: 163–8

24.

Mmbaga

, Hussain

, Leyna

Prevalence and risk factors for HIV-1 infection in rural Kilimanjaro region of Tanzania: implications for prevention and treatment.

BMC Public Health 2007; 7: 58

25.

Dzekedzeke

, Fylkness

Reducing uncertanities in global HIV prevalence estimates: the case of Zambia.

BMC Public Health 2006; 6: 83 (doi:10.1186/1471-2458-6-83)

26.

Mishra

, Vaessen

, Boerma

JT.

HIV testing in national population-based surveys: experience from the Demographic and Health Surveys.

Bull World Health Organ 2006; 84: 537–45

27.

Dare

, Cleland

JG.

Reliability and validity of survey data on sexual behaviour.

Health Trans Rev 1994; 4: 93–110

28.

Fenton

, Johnson

, Mcmanus

Measuring sexual behaviour: methodological challenges in survey research.

Sex Transm Infect 2001; 77: 84–92