Effect of partial recording protocols on estimates of prevalence of periodontal disease

Background: The aim of this study was to assess the degree of underreporting in the estimates of prevalence of periodontal attachment loss due to different partial recording protocols (PRP) in epidemiological studies, and to derive a correction factor to adjust for this bias. Methods: The study sample included 1,460 dentate persons 14 to 103 years old who were examined clinically to assess the clinical attachment loss at six sites per tooth. Seven PRP based on full-mouth or half-mouth designs were assessed, and the bias and sensitivity in the assessment of attachment loss prevalence for these protocols were assessed. Results: All partial protocols underestimated the prevalence of attachment loss. Bias estimates for any full-mouth PRP were smaller than those for the corresponding site-combination PRP for the half-mouth design. The PRP using the mesio-buccal (MB), mid-buccal (B), and disto-lingual (DL) sites of teeth in all four quadrants showed the smallest bias and highest sensitivity of prevalence estimates among the seven PRP evaluated, uniformly across the range of attachment loss severity level. The three site PRP incorporating the DL site produced less bias than the three site PRP including the disto-buccal (DB) site. There was a 3% to 12% gain in sensitivity for 2 to 5 mm attachment loss thresholds for the three site half-mouth PRP compared with the two site MB, B half-mouth PRP. Conclusions: The bias in the assessment of attachment loss is influenced by the partial recording design and the type and number of sites assessed, and is also influenced by the severity of attachment loss in the study population. These factors should be considered when selecting a partial recording method in large surveys. Furthermore, inflation factors designed to adjust for the bias due to the use of partial systems should be calculated and reported so that comparisons of results with other surveys are more meaningful.