A semi-parametric Bayesian model of inter- and intra-examiner agreement for periodontal probing depth

TL;DR

This paper introduces a Bayesian hierarchical model to analyze inter- and intra-examiner agreement in periodontal probing depth, accounting for examiner bias, site-specific effects, and measurement correlation, with applications to simulated and real calibration data.

Contribution

It presents a novel semi-parametric Bayesian approach using Dirichlet process mixtures to identify site-specific measurement biases in periodontal probing.

Findings

Model accurately recovers examiner bias and heterogeneity in simulated data

Provides cluster-adjusted agreement estimates for periodontal measurements

Effectively analyzes calibration training data

Abstract

Periodontal probing depth is a measure of periodontitis severity. We develop a Bayesian hierarchical model linking true pocket depth to both observed and recorded values of periodontal probing depth, while permitting correlation among measures obtained from the same mouth and between duplicate examiners' measures obtained at the same periodontal site. Periodontal site-specific examiner effects are modeled as arising from a Dirichlet process mixture, facilitating identification of classes of sites that are measured with similar bias. Using simulated data, we demonstrate the model's ability to recover examiner site-specific bias and variance heterogeneity and to provide cluster-adjusted point and interval agreement estimates. We conclude with an analysis of data from a probing depth calibration training exercise.