Ranking County-Level Health-Related Measures Based on Longitudinal and Joint Outcome Models

• Activity Number: 83
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Ronald Gangnon*+ and Jessica Athens and Bridget Booske Catlin and Patrick L Remington
• Companies: University of Wisconsin-Madison and University of Wisconsin-Madison and University of Wisconsin-Madison and University of Wisconsin-Madison
• Keywords: empirical Bayes ; ranking ; mixed effects ; Monte Carlo

Abstract

Ranking county-level measures of health is useful for comparing population health. Here, we consider models that can exploit longitudinal observations or multiple outcomes to improve the precision of these ranks. Using data from the 2010 County Health Rankings project, we considered joint models for (1) poor physical health days and poor mental health days, (2) percent of births with low birth weight and fair/poor health prevalence and (3) age-specific mortality rates for 9 age cohorts and (4) longitudinal models for age-specific mortality rates within the 9 age cohorts. Using mixed effects models, we obtained empirical Bayes posterior samples. In addition to standard credible intervals, we propose an alternative interval-the comparison interval-which better communicates the relevant uncertainty about ranks. Models (1) and (4) improved rank precision. In contrast, model (2) did not improve rank precision, and model (3) degraded rank precision, due to inflation of the variance in the oldest age cohort. In conclusion, we find that models for joint outcomes and longitudinal data can be used to improve the precision of population-based health rankings.