Bootstrap Estimates of Synthetic Cohort Incidence Estimates

• Activity Number: 554
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Bryan Sayer*
• Companies: Social & Scientific Systems
• Keywords: incidence ; prevalence ; HIV ; bootstrap

Abstract

We estimate the district level incidence of HIV for females in Uganda for the time period 2008-2013 using cross-sectional program data from the Prevention of Mother To Child Transmission (PMTCT) program. We then construct bootstrap confidence intervals of the incidence rates. The estimates and confidence intervals are compared to similar estimates obtained using survey data. The survey data (AIDS Indicator Surveys for Uganda, 2004/5 and 2011) estimates allow us to calculate bootstrap confidence intervals two ways: one using bootstrap sampling weights and the second using the given sampling weights but resampling n cases using with replacement sampling. Incidence is estimated using pairs of cross-sectional data points and known mortality estimates from the Masaka district cohort study, plus population estimates from the Uganda Bureau of Statistics. Incidence cases are the change in prevalence plus the number of deaths occurring in the time interval. The denominator is the mid-point estimate of the number of HIV negative women. Confidence intervals obtained from the bootstrap procedure indicate a wide range of incidence rates across districts.