The correlated binomial model in the analysis of clinical diagnostic test parameters for recurrence of periodontitis

In the present study, recall data of 29 patients with advanced periodontitis were utilized to calculate diagnostic test parameters for bleeding scores or increased residual probing depth on increase of periodontal probing depth of > or = 2 mm after 2 years, comparing the simple binomial and correlated binomial model. 5.6% of sites were identified, where the periodontal probing depth increased 2 mm or more during 2 years of maintenance. Within-patient correlation coefficient of sites were found to be small, in general (p < 0.02), with the exceptions for predictive value of a positive test for frequent (> or = 3/4 times) bleeding on probing and residual probing depth of > or = 4 mm (p = 0.137 and p = 0.086, respectively), and specificity for no bleeding (p = 0.151). Point estimates of test parameters generally did not differ between simple binomial and correlated binomial model. The respective standard deviations were either over- or underestimated in the simple binomial model. It was concluded that the suggestion of a limited diagnostic value of bleeding scores or increased residual postoperative probing depth to predict recurrence of periodontitis cannot be rejected after application of a correlated binomial model.