1) Sample-size bias: The sample sizes of studies are generally very small. Small samples of non-random population may tend to overestimate the seroprevalence.
2) Nonrepresentative samples: Many surveys are taken of populations convenient for the medical team drawing blood for testing. Many are taken in clinics or hospitals where the available sample of people may be sicker than those who do not attend the clinic.
3) Geographic bias: Samples may also be taken in more-accessible rather than less-accessible geographic areas. This would tend to bias upward the estimate of HIV seroprevalence, if taken to represent the country as a whole.
4) Testing bias: The ELISA test has been the predominant test used to determine seropositivity but it gives a number of false positives. Test results must be confirmed by a second test. Not all studies report confirmatory testing.
5) HIV-1 and HIV-2 overlap: In countries where both HIV-1 and HIV-2 are present, tests are done for both viruses and often find people who test positive for both. Care must be used with this data to avoid either double counting or omitting various categories of infected population.
In order to minimize the biases using current estimates, CIR has developed several criteria to select sample estimates: larger samples are generally favored over smaller samples, more recent estimates are selected over older estimates, and better documented data are usually selected over poorly documented data. These criteria only attempt to minimize the biases in the data, not eliminate them. All seroprevalence data must be used with caution.