Estimating Survival Probabilities Based on Complex Survey Data with Mortality Follow-Up

• Activity Number: 648
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Barry Graubard and Victoria Landsman*+
• Companies: National Cancer Institute and Ontario Institute for Cancer Research
• Keywords: Survival analysis ; Complex survey data ; Smoking related mortality ; Variance estimation

Abstract

This paper is motivated by collaborative work on the estimation of smoking-related mortality in the US. An estimator for the survival probabilities for different smoking groups is proposed. Careful attention is paid to the complications of analyzing survey data with complex sample designs, and, in particular, to estimating standard errors of the proposed estimator in this context. The leaving-one-out jackknife estimator that accounts for the complex sample design is described, but its implementation can be time consuming. Hence, an analytic formula for variance estimation is highly desirable. The method of Taylor deviates to obtain an analytic expression for variance estimator is outlined. The methods are illustrated with the data from the US National Health Interview Survey with linked mortality files. Extensions using multiple data sources to combine estimates across countries are discussed.