Keywords: Spatial Statistics, Zero Inflated Models, Missing Data, Dental Data, Methamphetamine
Caries status, a key component of oral health, is collected as the count of decayed, missing, or filled surfaces (DMFS) on each tooth. Missing teeth, which can occur for a number of reasons, present challenges for modeling DMFS data, and approaches such as treating all surfaces as adding to the DMFS count give rise to misleading conclusions. This poster will review such challenges in the context of a prospective study of 571 methamphetamine users fitting models for DMFS data that account for and model the generative process of missing teeth. The models considered here include spatial random effects with conditionally autoregressive relations to account for within-mouth patterns of association among teeth incorporating contributions from adjacent teeth and from corresponding teeth on the other side of the person’s mouth. Alternative models will be run for comparison purposes that make no accommodation for missing teeth or make ignorability assumptions that are not easy to motivate. The presentation will also discuss additional sources of data that can inform models for DMFS outcomes in ways that account for uncertainty about underlying missingness mechanisms.