Hi Brijesh.

Estimation of prevalence of HIV is based on type of population. Basically there are three types of population: general, bridge (clients of female sex workers) and high risk groups.

The different sources of data available for general population (with logical approach of 'something is better than nothing') are from PPTCT (data from among pregnant women attending Ante Natal Care (ANC) in the district), Health Sentinel Surveillance (HSS)-ANC (data from among pregnant women attending ANC at designated sites), and blood banks (data from among the blood donors).

The different sources of data available for bridge population (with logical approach of 'something is better than nothing') are from HSS - male STD patients and ICTC walk in males.

High risk groups (HRG) include female sex workers (FSWs), Men who have sex with men and transgenders (MSM-T), Injection Drug Users (IDU), truckers and migrants. However, there is no readily available universal data regarding truckers and migrants except from those sites which have a TI project running for the same). Hence, HRG population data is limited to FSW, MSM-T and IDUs. Data for the prevalence for these groups are drawn from the Targetted Intervention (TI) projects implemented for the same.

Sources of data regarding HRGs are predominantly proxy for the HRGs. They are HSS-MSM-T and ICTC walk in females.

Each of these data sources (irrespective of the population it represents) have its own strengths and limitations, considering the proportion of these groups within the district's population, and the absolute risk of HIV among the HRGs (dependant on the existing prevalence and chances of getting infected due to their high risk behaviour). That is, the best available source data for one district may not hold good for another district.

For example, strengths of PPTCT data depends on coverage of ANC, and number of pregnant women out of total who are getting tested for HIV in the above coverage. If large numbers of pregnant women are covered and from among them, most of them are tested, it would give the shortest confidence intervals. If the PPTCT centres increase in numbers across the past years, it would not be helpful for trend analysis, as the continuous expansion of PPTCT centres to the smaller towns and villages over the past few years probably has changed the profile of those who undergo HIV testing.

HSS-ANC data has also got its own strengths and limitations. If there are consistent sites across the years, it would enable a trend analysis. But with interpretation has to be with caution, as they cover a small sample of 400 per site, and there are variations in testing and sampling procedures over the years. There is also a chance that more-risky pregnant women would be covered and tested over the years, thereby over-estimating the prevalence. Thus it would provide large confidence intervals around the annual estimates.

If there are larger number of units of blood collection involved, blood bank data could be reliable. This in turn depends on the health care seeking behavior of population, and utilization and effectiveness of the existing health system. Disadvantages are that the risk profile and the sexual activity status of the donor is not known, especially in the camp approach for voluntary blood donation.

Regarding the HSS data on male STD patients, trend analysis could be possible if the sites are consistent. But, sample size is very small (less than or equal to 150 in some districts), thereby it may result in large confidence intervals around the estimate. As risk profiles of the individuals tested are not available, and as there could be variations in testing and sampling procedures over the years, HSS data may tend to overestimate the prevalence among the clients of sex workers.

For data on ICTC walk in males, it is assumed that a substantial proportion of the data is from bridge population. Sometimes, and often, this could be only data for males in the district. This is in turn limited by lack of information on the risk profile, which in turn, limits comparability across districts and time periods.

HSS data on sexual minorities (which is assumed to be MSM-T) is available late in the history, as far as NACO is concerned. So a logical trend analysis may not be fruitful. Sample sizes less than 250 per site would provide large confidence intervals around the estimate.

It is assumed that the substantial proportion of ICTC walk in females would be HRGs. (Please understand that the average proportion of HRGs with respect to total population in an Indian district would be 1-2%). With the effect of guidelines for testing of FSWs once in six months (as per NACO TI projects), this would better be a HRG indicator among the women. However, other women would also be there. Thus, there is a possibility that the risk profile of the ICTC walk in females may be not known. Also, comparability across the districts may be limited in this case.

Hence, the best available figures are often taken to arrive at an estimate of the prevalence from among these populations, district wise. Thus, prevalence data of HIV in districts are not absolute but indicative (please understand, something is better than nothing, here). Hence the figures of estimates (please note, estimates) have to be addressed with caution, as there is no compulsion for every citizen to undergo the test, routinely, every six months.

However, the data is adequate enough to categorize districts and plan interventions for prevention, care and support programs. That is, to say logically, they are functional indicators rather than absolute values.

What seems to be more appropriate is a population-proportion based model for collating data from different sources, to arrive at a relative more-reliable prevalence data, which is hardly attempted.

Thank you so much sir for excellent information... I am trying to learn more about this topic only way is the scientists like you peoples suggestions and views.