Effect of partial recording protocols on severity estimates of periodontal disease

Abstract

The study aim was to assess bias magnitudes of periodontal disease severity estimates for specific partial recording protocols (PRPs) in epidemiological studies. Estimates of mean clinical attachment loss (MCAL) and mean probing pocket depth (MPPD) were derived for 20 different PRPs using full-mouth periodontal data from 1437 dentate Brazilian subjects 14-103 years old having at least four teeth. Biases, relative biases and intra-class correlations for all PRPs were evaluated. Graphical methods were used to assess how well the PRP-based estimates agreed with full-mouth scores across levels of disease. Slightly higher levels of disease were evidenced on lingual than on buccal sites. Seven multi-site PRPs and the Ramfjörd PRP produced small biases in MPPD (-0.17 to 0.04 mm) and MCAL with relative biases under 8% and 4% in absolute value for MPPD and MCAL, respectively. Biases for full- and random half-mouth-based PRPs were similar. The three-site random half-mouth MB-B-DL and the Ramfjörd PRPs produced the smallest biases, with relative biases <3% in absolute value for MPPD and MCAL. Bias for MPPD or MCAL estimates varies by site type, number of sites per tooth and number of quadrants included in the PRP.