Analysis Strategies for Planned Missing Data in an Oral Health Study

• Activity Number: 423
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Lauren Harrell*+
• Companies: University of California Los Angeles
• Keywords: item response theory ; planned missingness ; multiple imputation ; periodontal ; missing data ; hierarchical model

Abstract

In a study of the dental health outcomes among methamphetamine users, subjects were randomized to receive either the full mouth or a half-mouth periodontal exam, where only two of the four quadrants of the mouth are observed. Because the half-mouth periodontal examination has been shown to underestimate the prevalence of periodontal disease in a given population, alternative approaches to analyzing the periodontal exam measurements are investigated. A hierarchical normal model, in which different covariance structures and potential spatial effects are explored, can be used as a basis for multiple imputations of site-level measurements of bleeding, pocket depth, and attachment loss. Alternatively, an item response theory model is calibrated to describe the behavior of the measurements at each location in relation to the latent progression of periodontal disease.