Skip Navigation LinksHome > January 14, 2013 - Volume 27 - Issue 2 > HIV prevalence measurement in household surveys: is awarene...
AIDS:
doi: 10.1097/QAD.0b013e32835816ce
Editorial Comment

HIV prevalence measurement in household surveys: is awareness of HIV status complicating the gold standard?

Korenromp, Eline L.a; Gouws, Eleanorb; Barrere, Bernardc

Free Access
Article Outline
Collapse Box

Author Information

aDepartment of Public Health, Erasmus MC, University Medical Center Rotterdam, Netherlands

bUNAIDS Regional Support Team, Eastern and Southern Africa, Johannesburg, South Africa

cInternational Health and Development Division, Measure DHS, ICF International, Calverton, Maryland, USA.

Correspondence to Eline Korenromp, PhD, The Global Fund to fight AIDS, Tuberculosis and Malaria, Vernier, Geneva, Switzerland. Tel: +41 58 791 1732; e-mail: ekorenromp@orange.fr

Received 4 July, 2012

Accepted 12 July, 2012

Since the introduction of HIV serological testing in Demographic and Health Surveys (DHS) in 2001, national serosurveys have become the gold standard of HIV prevalence estimation in countries with generalized epidemics [1]. By June 2012, 48 DHS or AIDS Indicator Surveys (AIS) in 33 countries (of which 28 are in Africa, 26 with generalized HIV epidemics) have included HIV serotesting of adult men and women, which is anonymous, informed and voluntary [2,3]. These nationally representative surveys use standardized sampling, questionnaires and testing protocols to produce data that are comparable over time and among countries.

In an analysis of repeat household survey data from rural Malawi, Floyd et al.[4] show how HIV prevalence is increasingly underestimated as more and more people become aware of their HIV status and refuse to retest once they know that they are infected. Exploring several methods to adjust for selective nonparticipation, the study shows a wide margin of uncertainty in resulting prevalence estimates. Across three surveys, the proportion of participants who knew their HIV status on the basis of previous testing rose from 54 to 89% between 2007 and 2010. HIV-positive women and men were two-fold to four-fold more likely to refuse HIV testing, so that crude measurements underestimated true prevalence by 1.1-fold to 1.4-fold in 2007/2008; by 1.2-fold to 2.0-fold in 2008/2009 and by 1.2-fold to 2.3-fold in 2009/2010. As a result of the increasing bias, the unadjusted survey series misleadingly suggested a recent prevalence decline in men, whereas best-adjusted estimates indicated a slight rise.

These results, however, are not generalizable to DHS and AIS, for several reasons. In Karonga, testing acceptance rates decreased over subsequent surveys, from 62 to 56% among men and from 70 to 64% among women between 2007/2008–2009/2010. In contrast, DHS and AIS have generally achieved high and increasing testing coverage rates over the last decade. In all of nine African countries that included HIV testing in two subsequent DHSs, coverage increased slightly from the first to the second survey (except for women in Senegal), reaching a median of 84% of men and 91% of women tested in the second survey (Table 1). Across all DHS and AIS that included serotesting to date, median test coverage increased from 79% over 2001–2006 to 83% over 2007–2011 for men and from 87 to 88% for women.

Table 1
Table 1
Image Tools

Despite recent increases in the proportion of respondents who know their HIV status, prior test exposure in national DHS to date remains much lower than in the Karonga study. Across the nine countries referred to above, any prior testing exposure was on average 39% among men and 57% among women in the second DHS, compared to 87 and 92% in men and women in Karonga by 2009/2010.

In addition, the survey settings differ substantially; whereas participants in the Karonga surveillance site were exposed to three repeat surveys within 3 years, DHS or AIS are usually conducted at 4–5 year intervals, and they only sample a fraction of the national population, with successive surveys drawing independent, fresh samples of geographical clusters and households.

Differences in testing rates can also be explained by different testing protocols: in Karonga, surveys used a rapid diagnostic test and shared results with all participants during the interview. DHS, in contrast, do not currently provide test results during the survey, but refer participants to a nearby testing and counselling centre [2].

HIV control programs increasingly provide testing and counseling services, through provider-initiated testing, community outreach and campaigns [5,6]. The question as to whether the Karonga scenario could be expected in future national household surveys depends much on future testing coverage. If increasing test exposure raises refusal ratios of known HIV positive to HIV negative individuals in the future, the resulting bias may remain limited provided overall test coverage can be maintained at above 90%. Encouragingly, the 2010 DHS in Malawi found fewer missing tests among men and women who self-reported a previous HIV-positive test (7 and 4%, respectively), compared to men and women self-reporting a previous HIV-negative test (8 and 6%) or no previous test (10 and 8%).

Nevertheless, the Karonga study provides a timely reminder that survey-based prevalence estimates may be more uncertain than often assumed. The sophisticated adjustments on the Karonga data were possible only by the study's unique longitudinal design, while not addressing the need for an adjustment method for single cross-sectional surveys.

The currently accepted method used to assess the potential effect of nonresponse on prevalence estimates is multiple imputation of missing results, based on household and individual characteristics of refusers and absentees as far as these have been measured and correlated with HIV status among those tested [2,7,8]. However, the demographic and behavioural variables available for imputation explain only about 20% of total variation in HIV prevalence in most countries [8]. In Karonga, imputation based on observed associations that ignore the effect of prior HIV status knowledge resolved only 1–21% of the refusal bias. A larger bias indeed becomes apparent when adjusting DHS data for nonrandom test refusals, that is for correlation between refusal and (unknown) HIV status. Across 20 DHS conducted between 2001 and 2006, extrapolation of relative HIV risk among nontesters from earlier longitudinal studies found that crude measurements underestimated true prevalence by up to 1.34-fold [1,9]. The application of Heckman-type selection models to control for selection bias in national HIV prevalence estimates [10], while still being investigated, also suggests that the uncertainty in survey-based prevalence measures may be larger than is apparent from standard imputation.

The uncertainty in survey-based prevalence estimates support recommendations to monitor HIV prevalence in generalized epidemics by triangulating antenatal clinic (ANC) surveillance and national serosurveys [11–13]. UNAIDS recommends that trends in national prevalence be derived from ANC surveillance data, after calibrating prevalence levels and the male-to-female ratio on the basis of national household survey measurements [14]. The calibration adjusts for known biases in ANC data, including differential fertility and HIV exposure of ANC clients– the effects of which vary with age, with stage of the epidemic and with antiretroviral (ART) coverage– and geographically unrepresentative ANC sites [15].

As HIV testing and ART are being rolled out, the use and interpretation of prevalence data from both national surveys and sentinel surveillance will need to evolve. In addition to potential test refusal bias, improved survival for people on ART complicates the interpretation of adult prevalence. Infection incidence is, therefore, becoming increasingly important for monitoring epidemics and for assessing prevention success, and options to measure or estimate incidence deserve exploration: testing for recent infection in surveys [16], epidemiological modeling based on age-specific prevalence measurements [17] or frequent monitoring of prevalence trends among young adults [18].

Even if future national serosurveys maintain the current high rates of test acceptance rates, refusal bias due to knowing one's HIV status may become more important as prior test exposure increases. It will be important for national surveys to continue to monitor test refusals and assess the extent to which test refusal is correlated with awareness of HIV status. Robust quantitative methods are needed to adequately adjust for nonresponse bias and to estimate the associated level of uncertainty. Control programs and their target populations would be disserviced by premature conclusions about success in preventing new infections, if epidemic trends inadvertently get confounded by success in improving awareness of people's HIV status.

Back to Top | Article Outline

Acknowledgements

Conflicts of interest

Our views do not necessarily represent the decisions, policy or views of UNAIDS, ICF International, or USAID. We declare that we have no conflicts of interest.

Back to Top | Article Outline

References

1. Garcia-Calleja JM, Gouws E, Ghys PD. National population based HIV prevalence surveys in sub-Saharan Africa: results and implications for HIV and AIDS estimates. Sex Transm Infect 2006; 82 (Suppl 3):iii64–iii70.

2. ICF Macro. HIV prevalence estimates from the Demographic And Health Surveys. In. Calverton, Maryland: ICF Macro; 2010. http://measuredhs.com/pubs/pdf/OD61/OD61.pdf [Accessed 30 November 2012].

3. MEASURE DHS+. Demographic and Health Surveys (DHS). Calverton, MD, USA: USAID. http://www.measuredhs.com/ [Accessed 30 November 2012].

4. Floyd S, Molesworth A, Dube A, Crampin AC, Houben R, Chihana M, et al.Underestimation of HIV prevalence in surveys when some people already know their status, and ways to reduce the bias. AIDS 2013; 27:233–242.

5. World Health Organization HIV/AIDS Programme, UNAIDS. Guidance on provider-initiated HIV testing and counselling in health facilities. Geneva: World Health Organization HIV/AIDS Programme; UNAIDS; 2007.

6. Lugada E, Millar D, Haskew J, Grabowsky M, Garg N, Vestergaard M, et al. Rapid implementation of an integrated large-scale HIV counseling and testing, malaria, and diarrhea prevention campaign in rural Kenya. PLoS One 2010; 5:e12435.

7. Mishra V, Vaessen M, Boerma JT, Arnold F, Way A, Barrere B, et al. HIV testing in national population-based surveys: experience from the Demographic and Health Surveys. Bull World Health Organ 2006; 84:537–545.

8. Mishra V, Barrere B, Hong R, Khan S. Evaluation of bias in HIV seroprevalence estimates from national household surveys. Sex Transm Infect 2008; 84 (Suppl 1):i63–i70.

9. Reniers G, Araya T, Berhane Y, Davey G, Sanders EJ. Implications of the HIV testing protocol for refusal bias in seroprevalence surveys. BMC Public Health 2009; 9:163.

10. Barnighausen T, Bor J, Wandira-Kazibwe S, Canning D. Correcting HIV prevalence estimates for survey nonparticipation using Heckman-type selection models. Epidemiology 2011; 22:27–35.

11. Garcia Calleja JM, Jacobson J, Garg R, Thuy N, Stengaard A, Alonso M, et al. Has the quality of serosurveillance in low- and middle-income countries improved since the last HIV estimates round in 2007? Status and trends through 2009. Sex Transm Infect 2010; 86 (Suppl 2):ii35–ii42.

12. Garcia-Calleja JM, Zaniewski E, Ghys PD, Stanecki K, Walker N. A global analysis of trends in the quality of HIV sero-surveillance. Sex Transm Infect 2004; 80 (Suppl 1):i25–30.

13. Lyerla R, Gouws E, Garcia-Calleja JM. The quality of sero-surveillance in low- and middle-income countries: status and trends through 2007. Sex Transm Infect 2008; 84 (Suppl 1):i85–i91.

14. Gouws E, Mishra V, Fowler TB. Comparison of adult HIV prevalence from national population-based surveys and antenatal clinic surveillance in countries with generalised epidemics: implications for calibrating surveillance data. Sex Transm Infect 2008; 84 (Suppl 1):i17–i23.

15. Montana LS, Mishra V, Hong R. Comparison of HIV prevalence estimates from antenatal care surveillance and population-based surveys in sub-Saharan Africa. Sex Transm Infect 2008; 84 (Suppl 1):i78–i84.

16. Mastro TD, Kim AA, Hallett T, Rehle T, Welte A, Laeyendecker O, et al. Estimating HIV Incidence in populations using tests for recent infection: issues, challenges and the way forward. J HIV AIDS Surveill Epidemiol 2010; 2:1–14.

17. Hallett TB, Stover J, Mishra V, Ghys PD, Gregson S, Boerma T. Estimates of HIV incidence from household-based prevalence surveys. AIDS 2010; 24:147–152.

18. The International Group on Analysis of Trends in HIV Prevalence and Behaviours in Young People in Countries most Affected by HIV. Trends in HIV prevalence and sexual behaviour among young people aged 15–24 years in countries most affected by HIV. Sex Transm Infect 2010; 86(Suppl 2):ii72–ii83.

Keywords:

HIV/AIDS; monitoring; evaluation; prevalence; survey; surveillance; impact evaluation; adult; Africa

© 2013 Lippincott Williams & Wilkins, Inc.

Login

Search for Similar Articles
You may search for similar articles that contain these same keywords or you may modify the keyword list to augment your search.