World Population Profile: 1996

Population Projections Incorporating AIDS

Background

Although it has been clear for a number of years that mortality estimates and projections for many countries would have to be revised due to AIDS mortality, the lack of accurate empirical data on AIDS deaths, the paucity of data on HIV infection among the general population, and the absence of tools to project the impact of AIDS epidemics into the future have all hampered these efforts. Although the accuracy of data on AIDS deaths has not substantially improved, knowledge of HIV infection has expanded and modeling tools have become available to project current epidemics into the future.

The methodology used to project AIDS mortality for this report generally follows the method adopted for World Population Profile: 1994, with several modifications. The method consists of the following steps:

1. Establish criteria for selecting countries for which AIDS mortality will be incorporated into the projections.

2. For each selected country, determine the empirical epidemic trend and a point estimate of national HIV prevalence.

3. Model the spread of HIV infection and the development of AIDS in the population, generating alternative epidemic scenarios, and produce the seroprevalence rates and AIDS-related age-specific mortality rates which correspond to each epidemic scenario.

4. Use the empirical levels and trends (from step 2) to establish a factor representing each country's position on a continuum between high and low epidemics (from step 3). Use the derived factor to generate a unique interpolated epidemic.

5. Use weighted country total adult seroprevalence to determine an appropriate location on the total country epidemic curve implied by the interpolation factor. This projects adult HIV seroprevalence for the total country.

6. Interpolate AIDS-related mortality rates, by age and sex, associated with the estimated speed and level of HIV from epidemic results for the period 1990 to 2010.

In the sections that follow, each of these steps is described, and the method is illustrated.

Country Selection Criteria

The International Programs Center (Population Division, Bureau of the Census) maintains an HIV/AIDS Surveillance Data Base. This data base is a compilation of aggregate data from HIV seroprevalence studies in developing countries. Currently, it contains over 25,000 data items drawn from nearly 3,200 publications and presentations. As a part of the updating of the data base, new data are reviewed for inclusion into a summary table which, for each country, lists the most recent and best study of seroprevalence levels for high- and low-risk populations in urban and rural areas. (Note: High risk includes samples of prostitutes and their clients, sexually-transmitted disease patients, or other persons with known risk factors. Low risk includes samples of pregnant women, volunteer blood donors, or others with no known risk factors. For a more complete description of the selection criteria, see U.S. Bureau of the Census (1995).)

A review of the data in the summary table suggests that a reasonable cut-off point for selection would be countries that have reached 5 percent HIV prevalence among their low-risk urban populations or, based on recent trends, appear to be likely to reach this level in the near future.

A total of 21 countries now meet these criteria for the incorporation of AIDS mortality in the projections. All but two of these countries are in Africa. The countries are:

Botswana
Burkina Faso
Burundi
Cameroon
Central African Republic
Congo
Côte d'Ivoire
Ethiopia
Guyana
Haiti
Kenya
Lesotho
Malawi
Nigeria
Rwanda
South Africa
Tanzania
Uganda
Zaire
Zambia
Zimbabwe

AIDS mortality was incorporated into projections for two other countries, Brazil and Thailand, because some country-specific modeling work had already been completed. The description of the simplified approach taken in these special cases follows that of the more general procedure.

Empirical Epidemic Trends

For each of the 21 countries meeting the selection criteria, we reviewed the HIV seroprevalence information available in the HIV/AIDS Surveillance Data Base to establish urban seroprevalence trends over time (table B-1, cols.1-4) and to identify available rural data points (table B-1, cols. 5-6). The two data points judged to be most representative for the urban low-risk population were identified and used to calculate the annual change between the dates of the two studies. Rural data were used in conjunction with the urban data to establish a total-country seroprevalence estimate (table B-1, col. 7).

Alternative Scenarios

To project the impact in the selected countries, three alternative epidemic scenarios were developed, corresponding to low, medium, and high-impact AIDS epidemics. These scenarios were developed using iwgAIDS, which is a complex deterministic model of the spread of HIV infection and the development of AIDS in a population. It was developed under the sponsorship of the Interagency Working Group (iwg) on AIDS Models and Methods of the U.S. Department of State (Stanley et al. 1991).

All three of these epidemic scenarios incorporate increasing levels of behavior change in the form of increased condom use. This assumption corresponds to actual changes in behavior that are now beginning to occur in some countries.

Interpolation of a Unique Epidemic

The empirical urban trend from each country was used to interpolate among the three epidemic scenarios to derive an epidemic trend line matching the observed HIV seroprevalence increase between two data points. Thus, both the level and the rate of increase of the urban epidemic were matched through this procedure, resulting in an interpolation factor used in subsequent steps.

Projected Total Seroprevalence

At this point in the estimation procedure, no direct linkage has been made to the total-country prevalence or to a particular calendar year in this country's epidemic. The next step accomplishes these tasks. The total-country adult prevalence estimate (table B-1, col. 7) was matched with the one implied using the interpolation factor. From this comparison, an "offset" figure was calculated, corresponding to the number of years of difference between the start of the epidemics in the three scenarios and the empirical epidemic at the reference date.

AIDS-Related Mortality Rates

Based on the "interpolation factor" and the "offset" described above, AIDS-related age-sex-specific mortality rates ( nmx values) at 5-year intervals from 1990 to 2010 were interpolated and added to non-AIDS nmx values for the same period (non-AIDS nmx values were derived by making standard assumptions concerning the improvement in mortality conditions as described earlier in this appendix). Population projections were prepared with the combined nmx values as input, using the Rural-Urban Projection Program (RUP) of the Bureau of the Census.

The future course of the AIDS pandemic is uncertain, but making projections for affected countries requires that some assumptions be made about AIDS mortality as well as about non-AIDS mortality. For the projections underlying this report, it was assumed that the epidemics in each of the 23 affected countries would peak in 2010, with no further growth in HIV infection after that year. AIDS mortality was assumed to decline from the level reached in 2010 to nil by 2050, thus implying a return to "normal" mortality levels in the latter year. To implement the projection process, life tables for 2050 that assume no AIDS mortality were used.

The Special Cases of Brazil and Thailand

Modeling activities were also undertaken for Brazil and Thailand with the support of the Interagency Working Group. AIDS epidemics in these two countries have substantial homosexual and intravenous drug use components, while those in Africa do not (WHO/GPA 1993). For Brazil, AIDS-related age-sex-specific mortality rates were estimated from the iwgAIDS model and added directly to the non-AIDS mortality rates previously prepared for the projection program. For Thailand, AIDS-related mortality rates from recent epidemiological and demographic projections (TNESDB 1994) were added to the non-AIDS nmx values for the 1990 to 2010 period.

Caveats and Limitations

In developing the methodology for these projections, the International Programs Center has attempted to maximize the use of both the empirical data and the modeling tools available. However, there is much that is unknown about the dynamics of AIDS epidemics in countries around the world, and the methodology is necessarily imprecise. As the AIDS pandemic grows, future behavior changes and interventions being implemented in countries around the world may alter the projected course.

What if AIDS epidemics do not peak early in the next century as projected? Will entire populations become infected with HIV and eventually die from AIDS? The simulations used for this report suggest that this will not happen in any population, although population declines are possible with a sustained widespread epidemic. Variations in sexual behavior help to ensure that the majority of the population in countries around the world are not at high risk of HIV infection. With substantial proportions of the population at lower risk of infection, each of the epidemic scenarios displays a definite plateau in HIV seroprevalence after the initial rapid rise.

Source: U.S. Bureau of the Census, World Population Profile: 1996, pp. B-6 to B-10.