Is the Performance of a Periodontal Prediction Model for Identification of Diabetes affected by Participants’ Characteristics?

Abstract

Objective: To evaluate whether the diagnostic accuracy of a novel periodontal prediction model (PPM) for identification of adults with diabetes varies according to participants’ characteristics. Basic Research Design: The study was carried out among 250 adults attending primary care clinics in Riyadh (Saudi Arabia). The study adopted a case-control approach, where diabetes status was first ascertained, and data collection carried out afterwards using questionnaires and periodontal examinations. Variations in the performance of the PPM by demographic (sex and age), socioeconomic (education) and behavioural factors (smoking status and last dental visit) were evaluated using receiver-operating characteristic (ROC) regression. Results: The PPM including 3 periodontal parameters (missing teeth, percentage of sites with pocket depth ≥6mm and mean pocket depth) had an area under the ROC curve (AUC) of 0.69 (95% Confidence Interval: 0.61-0.78), which dropped to 0.64 (95% CI: 0.53-0.75) after adjustment for covariates. Larger variations in performance were found by participants’ sex, age and education, but not by smoking status or last dental visit. The PPM performed better among male (adjusted AUC: 0.76; 95% CI: 0.53 to 0.99), younger (0.67; 95% CI: 0.50 to 0.84) and less educated participants (0.76; 95% CI: 0.60, 0.92). Conclusions: The diagnostic accuracy of a novel periodontal prediction model to identify individuals with diabetes varied according to participants’ characteristics. This study highlights the importance of adjusting for covariates on studies of diagnostic accuracy