How can the real scale of the HIV/AIDS pandemic be measured?

Since 1997, The Joint United Nations Programme on AIDS (UNAIDS) and the World Health Organisation (WHO) have provided estimates for the number of people living with HIV in different countries every two years.  As new methods are developed for calculating existing and future trends in the disease, figures can be startlingly different.  How can accurate decisions be made on health spending if policy-makers are unclear about how many people are HIV positive?

Researchers review theprocess, methods and procedures that have been used in the past and currentround of estimates of HIV/AIDS burden. There are several major differencesbetween previous estimates and those made in 2004. First, in the past, a singlepercentage figure, such as 16.5 percent, was given as an estimate of the numberof people globally or nationally with HIV. This presented a false sense ofaccuracy.  However, now a range is given,for example, between 13.5 percent and 20.1 percent of a population is HIVpositive to represent the uncertainty about the estimates.

Second, individual countrieshave become more involved in preparing the estimates and the use of localknowledge has meant that the figures have become more accurate.  The countries no longer just comment on figuresproduced international organisations. They receive training in estimationmethods and are supplied with specialist computer software for the work. Ratherthan just plotting a single disease trend for a whole country, governmentagencies now often calculate the trends for different regions within a country.

Third, since the 2002 roundof estimates, new data from nationally representative household surveys havebecome in available in many countries. These data have been used to refine theassumptions used in making estimates and have been the basis of calibratingdata from sentinel surveillance sites.

The study found that thefollowing methods can reduce levels of uncertainty when calculating how many peopleare HIV positive in a country:

• Other sources can be used to verify the data, forexample, national censuses could be used to look at the changes inmortality over time.
• Data sources are being checked.  In Zambia several urban clinics werecategorised as rural clinics by mistake. HIV is often higher in cities and the estimate for the whole ofZambia was brought down once the mistake was discovered.
• It has become acceptable to undertake HIV testingduring household surveys of other kinds of data.  This gives a more accurate reading ofthe whole population than the usual method of testing pregnant women whoattend antenatal clinic.

In Kenya, the nationalestimate was adjusted considerably following a household survey because itemerged that twice as many women as men were infected with the disease.  Improving ways of collecting data andanalysing future trends means that HIV figures can change quite dramaticallyfrom one year to the next.  Such a changecan be confusing.  It is not that theepidemic has taken a down-turn but that the processing of information has beenrefined.  Medical bodies and governmentagencies have to be open to new methods to make their predictions as accurateas possible.  The study suggests that:

• It would help to include several years using thenew method to give people a better understanding of the informationprovided. 
• The use of a range of possibilities rather than asingle figure is a more appropriate way of providing the information. 
• The information is accurate enough to make soundpolicy decisions in health spending, particularly if estimates are madefor different social groups, such as sex worker, and for different partsof a country.

All of these suggestions wereimplemented in the 2004 round of estimates released by UNAIDS and WHO.