IADR Abstract Archives
Diagnostic accuracy of self-reported periodontitis using a predicted nomogram
Objectives: This study aimed to develop a nomogram using self-reported data to predict periodontitis among Danish adults.
Methods: Individuals aged >18 years were clinically evaluated for periodontitis and classified according to the case definition proposed by the Centers for Disease Control and Prevention in collaboration with the American Academy of Periodontology—CDC/AAP. The sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were used to compare the accuracy of self-reported data (sociodemographic information and eight questions proposed by the CDC/AAP to identify periodontitis) to predict cases of moderate/severe periodontitis and any form of periodontitis. A multivariable model was developed, including variables with p-value<0.2 identified by univariable regressions. Nomograms were generated based on the multivariable regression coefficients of the Akaike Information Criterion of the selected model. Decision curve analyses (DCA) were performed to evaluate the potential clinical utility of the nomogram by assessing the clinical net benefit at different thresholds.
Results: Of the 197 participants, 50 were classified as having moderate/severe periodontitis and 68 as having any periodontitis. In the multivariable model, the best-fit model to predict moderate/severe periodontitis included information on smoking, dental status, previous periodontal treatment, and tooth loss. For predicting any periodontitis, the model included age, education level, dental status, and previous periodontal treatment. The nomogram demonstrated strong discriminatory performance [moderate/severe cases: AUC 0.82 (95% CI 0.74, 0.90); any periodontitis: AUC 0.81 (95% CI 0.74, 0.88)], adequate calibration (moderate/severe cases: intercept=-0.062; any periodontitis: intercept=-0.129), and an insignificant overestimation of high risk and an underestimating of low risk (moderate/severe cases: slope=0.932; any periodontitis: slope=0.891). Across a wide range of thresholds, DCA demonstrated consistent clinical net improvement in both periodontitis case definitions.
Conclusions: Our nomogram demonstrated an excellent predictive capability to identify individuals having periodontitis using self-reported information, providing a feasible instrument for self-report-based surveillance of periodontitis.
Methods: Individuals aged >18 years were clinically evaluated for periodontitis and classified according to the case definition proposed by the Centers for Disease Control and Prevention in collaboration with the American Academy of Periodontology—CDC/AAP. The sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were used to compare the accuracy of self-reported data (sociodemographic information and eight questions proposed by the CDC/AAP to identify periodontitis) to predict cases of moderate/severe periodontitis and any form of periodontitis. A multivariable model was developed, including variables with p-value<0.2 identified by univariable regressions. Nomograms were generated based on the multivariable regression coefficients of the Akaike Information Criterion of the selected model. Decision curve analyses (DCA) were performed to evaluate the potential clinical utility of the nomogram by assessing the clinical net benefit at different thresholds.
Results: Of the 197 participants, 50 were classified as having moderate/severe periodontitis and 68 as having any periodontitis. In the multivariable model, the best-fit model to predict moderate/severe periodontitis included information on smoking, dental status, previous periodontal treatment, and tooth loss. For predicting any periodontitis, the model included age, education level, dental status, and previous periodontal treatment. The nomogram demonstrated strong discriminatory performance [moderate/severe cases: AUC 0.82 (95% CI 0.74, 0.90); any periodontitis: AUC 0.81 (95% CI 0.74, 0.88)], adequate calibration (moderate/severe cases: intercept=-0.062; any periodontitis: intercept=-0.129), and an insignificant overestimation of high risk and an underestimating of low risk (moderate/severe cases: slope=0.932; any periodontitis: slope=0.891). Across a wide range of thresholds, DCA demonstrated consistent clinical net improvement in both periodontitis case definitions.
Conclusions: Our nomogram demonstrated an excellent predictive capability to identify individuals having periodontitis using self-reported information, providing a feasible instrument for self-report-based surveillance of periodontitis.