10
Jul

Accurate and credible UNAIDS data on the HIV epidemic: the cornerstone of the AIDS response

Once a year, UNAIDS releases its estimates on the state of the worldwide HIV epidemic. Since the data can literally affect life and death decisions on access to services for treatment and prevention—and are used to decide how to spend billions of dollars a year—they need to be as accurate as possible and be regarded as credible by everyone who uses the information.

How we collect and interpret HIV data has huge consequences—a pregnant woman visiting an antenatal clinic can help calculate the size of the country’s HIV epidemic, can help shape national policies for the response to HIV and can influence the size of grants to respond to HIV from the Global Fund to Fight AIDS, Tuberculosis and Malaria, the United States President’s Emergency Plan for AIDS Relief and others.

So, how do we do it?

Collecting data on the ground

We don’t count people. We can’t—many people who are living with HIV don’t know that they are, so can’t be counted. And everyone in a country can’t be tested every year to calculate the number of people living with HIV. Instead, we make estimates.

The data that are published in our reports, quoted in speeches and used by governments around the world to plan and implement their AIDS responses originate on the ground, in a clinic, in a hospital or anywhere else that people living with HIV access, or need, HIV services.

Take the example of a pregnant women visiting an antenatal clinic as part of her routine antenatal care. She will be offered an HIV test, which will show her to be either HIV-positive or HIV-negative. An HIV-positive result will, of course, open up for the individual mother the range of prevention of mother-to-child transmission of HIV services available to keep her well and her baby HIV-free, but the result, whether negative or positive, will also be used to determine the wider impact of services and the success of country programmes.

Some countries operate a so-called sentinel survey system, in which a network of reporting sites collect data. If the clinic is one of these, a sample of blood will be anonymized and collected with results from the other sentinel sites, resulting in a large set of data to estimate trends from the sentinel sites over time.

In other countries, data from all women who are being tested routinely at all antenatal care sites are used to estimate HIV prevalence. Her test result will be recorded and passed on to the country’s national-level HIV reporting agency.

Data from antenatal clinics, when combined with information from broader, but less frequently collected, population-based surveys, are the basis for HIV data collection in countries where HIV has spread to the general population.

For countries that have HIV epidemics mainly confined to key populations, data from HIV prevalence studies among those key populations are most often used. These prevalence studies are combined with the estimated number of people in those key populations—an estimate that is difficult to make, given that behaviours of key populations are outlawed in many countries.

In countries in which doctors are required to report cases of HIV, and if those data are reliable, those direct counts are used to estimate the epidemic. An increasing number of countries are setting up systems that use reported cases of HIV diagnoses.


Survey types

Population-based survey: a survey that is conducted in a random selection of households in a country. The survey is designed to be representative of all people in the country.

HIV prevalence study: a study of a specific population that collects blood samples from the population to determine how many people in that population are living with HIV. Typically, the results of that test are provided to the survey respondent.


Number crunching

Once a year, the country’s reporting agency will, helped by UNAIDS and partners, make estimates of the number of people living with HIV, the number of people on HIV treatment, the number of new HIV infections, etc., using software called Spectrum, which uses sophisticated calculations to model the estimates.

For estimates relating to children, a whole range of information, such as fertility rates, age distributions of fertility and the number of women in the country aged 15–49, is taken into account when computing the final numbers.

Estimates for different populations and age groups are calculated by Spectrum, taking into account different types of demographic and other data, building up a comprehensive picture of the country’s HIV epidemic.

The Spectrum estimates are sent to UNAIDS at the same time as the collection of the annual Global AIDS Monitoring reporting on the response to the HIV epidemic in the country. UNAIDS compiles and validates all the Spectrum files and uses the country-level data to make global estimates of the HIV epidemic and response.

UNAIDS publishes estimates for all countries with populations greater than a quarter of a million people. For the few countries of that size that do not develop Spectrum estimates, UNAIDS develops its own data, based on the best available information.

Ranges are important

In 2015, there were 36.7 million [34.0 million–39.8 million] people living with HIV in the world. The numbers in the brackets are ranges—that is, we are confident that the number of people living with HIV is somewhere within the range, but can’t say for sure what the definite number is.

All UNAIDS data have such ranges, but why can’t we be more accurate? UNAIDS data are estimates, which vary in their accuracy, depending on several factors. The size of sample taken for the estimate affects the range—a large sample means a small estimate range, and vice versa; if a population-based survey is conducted in a country, the estimate range will be smaller; and the number of assumptions made for an estimate has an impact on how narrow the range will be.

If it’s found to be wrong, it’s fixed

UNAIDS’ models are regularly updated in response to new information. For example, this year’s data will show a slight rise in the reported number of children becoming infected with HIV. This isn’t a real rise in young children acquiring HIV, but an adjustment in our knowledge of how infections occur in real life—in fact, once we apply this updated knowledge to previous years, we see that the number of new HIV infections among infants was higher then too.

Our new knowledge shows us that, after childbirth, higher numbers of women who are breastfeeding are becoming infected with HIV and hence passing the virus on to their children. Models had not fully captured the length of time for which women breastfed and were therefore at risk of passing on the virus through their milk if they became infected with HIV. With the model adapted to take into account women breastfeeding for longer than one year, the number of infants contracting HIV increased slightly for all years since the start of the epidemic.

Because of such finetuning, estimates from one year can’t be compared with estimates from a previous year. When UNAIDS publishes its yearly data, we revise all previous years’ estimates, taking into account the revised methodology. For example, the estimate published in 2006 for the worldwide number of people living with HIV in 2005 was 38.6 million—this was before we had incorporated national household surveys into estimates. By 2016, with the additional information from surveys, the number for 2005 had been revised to 31.8 million. Likewise, the estimate for AIDS-related deaths in 2005 was 2.8 million, which, by 2016, had been revised down to 2.0 million in 2005.

This finetuning has steadily improved the accuracy of our estimates, with the result that recent revisions are becoming smaller—the estimated number of people living with HIV in 2013 made in 2014 was 35.0 million, not far off the current estimates of 35.2 million.

Source: UNAIDS