Title: Inferring Rates from Instantaneous State Variables: Application to HIV Incidence Estimation
Speaker: Alex Welte, University of Witwatersrand, South Africa
Date: Wednesday, January 20, 2010 12:00 - 1:00 pm
Location: DIMACS Center, CoRE Bldg, Room 431, Rutgers University, Busch Campus, Piscataway, NJ
For reasons of efficiency, it is attractive to be able to estimate rates, such as disease incidence, from information about the population state at a single time point, such as obtained in cross sectional surveys. Population dynamic equilibrium is the only well known condition where a simple estimator can be obtained, namely that incidence is the prevalence divided by the mean duration of the infected state. For the important problem of HIV, several complications arise which make this simple analysis inapplicable.
We present an analysis of this problem, in which we show how a test for 'recent infection' can be used to generate cross-sectional data which allows the construction of well defined consistent estimators in the case of epidemiological transients, even if the test's performance characteristics have considerable inter-subject variability.
The presentation considers some of the foundational problems of defining rates in the real world, introduces an intrinsic weighting scheme for measurements performed on non-equilibrium population states, and considers, for both a base model and a number of variations, the practical consequences of the assumptions and calibrations which are required.