Multivariate Small-Area Forecasting: Jointly Modeling Cause-Specific Mortality in U.S. Counties

• Activity Number: 351
• Type: Topic Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Government Statistics Section
• Author(s): Kyle Foreman*
• Companies: Institute for Health Metrics and Evaluation
• Keywords: Bayesian statistics ; small area estimation ; epidemiology ; forecasting

Abstract

Mortality forecasts are typically limited in that they pertain only to national rates, predict only all-cause mortality, or do not capture the patterns of correlation between similar diseases. We present a novel hierarchical Bayesian model that jointly forecasts cause-specific mortality rates for small areas. We examine its effectiveness by applying it to United States vital statistics data for 1979 to 2011 and produce estimates to 2025. We also discuss computational issues that arise from fitting large Bayesian models with many parameters and techniques we have experimented with to improve model convergence and computational efficiency.