Making Populationlevel HIV prevalence estimates in South Africa

Making Population-level, HIV prevalence estimates in South Africa Should we use the NM/HSRC data or ASSA projections?

Antenatal HIV prevalence data n n n ANC data is widely used in developing countries to monitor the HIV epidemic Wealthier countries use more sophisticated surveillance tools ANC data useful for monitoring trends over time – not so useful for making population level estimates

Int J Epidemiol. 2002 Apr; 31(2) : 449 -55. How well do antenatal clinic (ANC) attendees represent the general population? A comparison of HIV prevalence from ANC sentinel surveillance sites with a population-based survey of women aged 1549 in Cambodia. Saphonn V, et al n Although ANC data can be used to estimate trends over time, it should be realized that ANC data may overestimate the actual prevalence in the younger age group in rural areas in Cambodia.

AIDS. 1998 Jul 9; 12(10): 1227 -34. Studying dynamics of the HIV epidemic: populationbased data compared with sentinel surveillance in Zambia. Fylkesnes K n ANC-based data might draw a rather distorted picture of current dynamics of the HIV epidemic.

AIDS. 2002 Mar 8; 16(4): 661 -5. Comparison of HIV prevalences in community-based antenatal clinic surveys inrural Mwanza, Tanzania. Changalucha J n In this rural population, the HIV prevalence in ANC attenders underestimated the prevalence among women in the general population.

East Afr Med J. 1996 May; 73(5): 298 -302. Sentinel surveillance and cross sectional survey on HIV infection prevalence: acomparative study. Kwesigabo G, n Seroprevalence among antenatal clinic attenders was significantly lower than that of females from the general population sample (p =0. 016).

AIDS. 2003 Feb 14; 17(3): 399 -405. Evaluating two adjustment methods to extrapolate HIV prevalence from pregnantwomen to the general female population in sub-Saharan Africa. Fabiani M, n The non-adjusted HIV prevalence among ANC attendees underestimates the prevalence among the general female population by 8. 0% in Chelston in 1998 and by between 20. 7% and 31. 9% in all other cases.

AIDS. 1993 Apr; 7(4): 567 -72. Sentinel surveillance for HIV-1 infection: how representative are blood donors, outpatients with fever, anaemia, or sexually transmitted diseases, and antenatalclinic attenders in Mwanza Region, Tanzania? Borgdorff M, n Patients with fever and antenatal clinic attenders may reflect trends, but data from patients with fever markedly overestimate, and data from antenatal clinic attenders underestimate, population HIV-1 prevalence.

Population used for the ANC HIV survey

Antenatal data much less useful for making population level estimates about the prevalence of HIV because of the serial extrapolations based on assumptions, often backed by sparse data, make the final estimates subject to great potential error.

Data source for ANC surveys n Prevalence measured is among pregnant women attending a small proportion of public sector antenatal clinics for 6 to 8 weeks a year.

From the sample to all pregnant women n Extrapolated to all pregnant women n Assumptions are made about: • women not using ANC services, • prevalence among private sector users.

From pregnant women to all women n Extrapolated to all women assumptions made about HIV prevalence among: • • • contraceptive users, infertile, non-sexually active, elderly and young women

From pregnant women to all women n HIV/AIDS reduces fertility and so the HIV prevalence among the over 25 year old pregnant women is probably an underestimate of the true HIV prevalence. With increased access to HIV VCT, more HIV infected women know their status and hence are likely to tend to prevent pregnancy which results in an underestimate of the true HIV prevalence. Pregnant women under 20 are not truly representative of this age group because not all women in this age group are sexually active. This probably results in the antenatal survey overestimating HIV prevalence

From women to men & children n Extrapolated to men and children assumptions about: • the male to female prevalence ratios and • adult to children ratios

Specific weaknesses in using South African antenatal data • • • In some clinics the nurses obtain consent from the patients which may result in biased results if a substantial proportion refuse. No data on the refusal rate. Certain provinces do not follow the protocol. For example in KZN the clinics were hand selected instead of being randomly selected using the “probability proportional to size methods”. This has resulted in all 35 clinics being on main roads There is limited quality control or validation of the South African survey and the data has never been released to the public making it impossible to double check the analysis