AEGiS-15IAC: Modeling HIV testing behavior and its impact on incidence estimation.

15th International AIDS Conference

Bangkok, Thailand - July 11-16, 2004

Modeling HIV testing behavior and its impact on incidence estimation.

Int Conf AIDS 2004 Jul 11-16; 15:(abstract no. ThOrC1425)

Song R
Centers for disease control and prevention, Atlanta, United States

BACKGROUND: Two estimation methods are widely used in estimating HIV incidence density. One is the cohort-based Kittayaporn method that apportions seroconverters and person-time across calendar years. The other is the well-known STARHS method that uses a less-sensitive test to identify a recent seroconversion. Both methods are unbiased if testing for HIV is independent of both the hazard of being infected in general and the event of seroconversion in particular. Unfortunately, this may not be true in studies where HIV tests are voluntary and motivated by risk behavior.

METHODS: Testing behaviors are characterized by two testing intensity parameters. One is related to risk level and the other reflects a change in testing behavior after seroconversion. Statistical procedures are developed to test (1) whether the testing frequency is independent of risk level and (2) whether seroconversion has any impact on testing frequency. Simple examples and complex simulations are used to evaluate the performance of the two incidence estimation methods.

RESULTS: Both the Kittayaporn and STARHS estimators are biased when the risk level and the testing frequency vary over time within the population considered. A change in testing frequency following seroconversion has significant impact on the STARHS estimator, but not much on the Kittayaporn estimator. Incidence density may be overestimated if persons at higher risk test more frequently or test sooner than usual after seroconversion. Underestimation may occur with both estimators when persons at lower risk test more frequently than persons at higher risk.

CONCLUSIONS: Voluntary testing can cause bias for both the Kittayaporn and STARHS estimators. Tests on the independence between risk and testing behavior using the proposed procedures are recommended. Estimation results must be interpreted with caution when the independence assumption is violated.

Keywords: AEGIS, Incidence, HIV Seropositivity, Acquired Immunodeficiency Syndrome, HIV, HIV Infections, Risk-Taking, HIV Seroprevalence, HIV Antibodies, Attitude, HIV Core Protein p24, epidemiology, immunology, methods

040711
ThOrC1425