AEGiS-14IAC: Proposals for estimating HIV incidence in the United States from HIV reporting data and the detuned assay.

14th International AIDS Conference

Barcelona, Spain - July 7-12, 2002

Proposals for estimating HIV incidence in the United States from HIV reporting data and the detuned assay.

Int Conf AIDS 2002 Jul 7-12; 14:(abstract no. WeOrC1308)

Karon J, Kaplan EH, Brookmeyer R, Song R
Division of HIV/AIDS, Surveillance &Epidemiology, Centers for Disease Control and Prevention, Atlanta, Georgia, United States

BACKGROUND: Development of the detuned assay has led the US Centers for Disease Control and Prevention to consider new approaches for estimating HIV incidence in the United States from new HIV diagnoses reported to HIV surveillance.

METHODS: Two statistical models have been developed to account for the fact that individuals choose whether (and how frequently) to seek HIV testing, that test results are only reported if HIV+ on the sensitive antibody test, and that additional information on prior (HIV-negative) testing dates can be obtained via surveys. Both models account for heterogeneity in testing frequency and assign a weight to each detected recently infected person. Model II uses detuned assay results; Model I also uses time since last negative test for some seroconverters, regardless of detuned assay results. The performance of these models was assessed by simulation.

RESULTS: Assuming that persons test twice per year on average, 100 simulations gave the following results for 1-year incidence estimates and standard deviations of the estimates: [table: see text] Simulations presuming that prior test dates would not be available for 25% of the infected persons revealed that while Model I remained accurate, Model II became badly biased.

CONCLUSIONS: These analyses show that it may be feasible to estimate HIV incidence from statistical models based on data from HIV surveillance and detuned assay results. Further refinements that address issues such as non-reporting bias (due to those at risk who do not seek HIV testing), samples that are not tested using the detuned assay, and changes in HIV test-seeking behavior leading to testing must be investigated before a final system for estimating HIV incidence can be designed and implemented.

Keywords: AEGIS, HIV, Incidence, HIV Infections, HIV Seropositivity, United States, Models, Statistical, HIV Core Protein p24, HIV Antibodies, HIV Antigens, Bias (Epidemiology), epidemiology, immunology, organization & administration

020707
WeOrC1308