Combining Self-Reported Information with Radiographic Bone Loss to Screen Periodontitis: A Performance Study

Abstract

Background/Objectives: The objective of this study is to evaluate the diagnostic performance of a combined screening approach using self-reported periodontal information and radiographic periodontal bone loss (R-PBL) in identifying individuals with periodontitis. Methods: An exploratory cross-sectional study was conducted including adult participants with available panoramic radiographs and responses to a validated self-reported periodontal screening questionnaire. R-PBL was assessed on interproximal sites and classified according to established thresholds. Self-reported information followed a validated strategy based on the Center for Diseases Control tool. The performance of individual and combined indicators was analyzed against the 2018 case definition for periodontitis, calculating sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUC). Results: A total of 150 participants were included, equally divided between periodontitis cases and controls, with a mean age of 46.5 years. The R-PBL model demonstrated the best predictive performance for both periodontitis (AUC: 0.833) and severe periodontitis (AUC: 0.796), with the highest precision and net benefit across thresholds. The Either model showed similar performance, particularly in sensitivity, while SR and Both models underperformed. Decision curve analysis confirmed the superior clinical utility of 'R-PBL' and 'Either' models in guiding decision-making. Conclusions: Combining self-reported information with radiographic bone loss showed adequate screening performance for periodontitis. This dual approach may provide a feasible strategy for identifying high-risk individuals in settings where full clinical examination is not possible.

Keywords: oral health; periodontal disease; periodontitis; prediction model; radiographic periodontal bone loss; self-report.

Figure 1

ROC curve with 95% confidence interval. ROC curves for the expected outcome of periodontitis for 4 different models (Either, SR, R-PBL and Both) and the dashed gray line is the reference line.

Figure 2

Decision curve analysis (DCA) plot for each model predicting periodontitis. The blue line, labeled as the “Expected Outcome Model”, illustrates the net benefit of employing DDS across various probability thresholds. In contrast, the dashed line, marked “Treat All”, indicates the net benefit if everyone is deemed high-risk, while the dotted line, labeled “Treat None”, shows the net benefit if no one is considered high-risk.

Figure 3

ROC curve with 95% confidence interval. ROC curves for the expected outcome of severe periodontitis for 4 different models (Either, SR, R-PBL and Both) and the dashed gray line is the reference line.

Figure 4

Decision curve analysis (DCA) plots for each model predicting severe periodontitis. The blue line, labeled as the “Expected Outcome Model”, illustrates the net benefit of employing DDS across various probability thresholds. In contrast, the dashed line, marked “Treat All”, indicates the net benefit if everyone is deemed high-risk, while the dotted line, labeled “Treat None”, shows the net benefit if no one is considered high-risk.