Estimation of Periodontal Disease Incidence in the Presence of Diagnostic Error

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Anthony J. Parker* and Elizabeth H. Slate+
• Companies: Medical University of South Carolina and Medical University of South Carolina
• Keywords: measurement error ; diagnostic error ; SIMEX

Abstract

The probed pocket depth (PPD) is a periodontal measurement that is subject to diagnostic error. By dichotomizing the PPD such that values of 5 mm or greater indicate periodontal disease, two statistical models, the Carlos-Senning (CS) and Lu models (see, e.g. Reed and McHugh, Biometrics, 1970), that have been used to obtain estimates of caries incidence adjusted for diagnostic error, were adapted for estimation of the incidence of periodontal disease. In a simulation study using both uncorrelated and more realistic within-subject correlated PPD data, we evaluated the performance of the adapted CS and Lu estimators and compared to a naïve estimator ignoring diagnostic error. A SIMEX-corrected estimator was also evaluated. We illustrate the methods using real PPD data.