Prediction of Periodontal Disease From Multiple Self-Reported Items in a German Practice-Based Sample

Ascertainment of periodontal disease using self-reported measures would be useful for large epidemiologic studies. This study evaluates whether a combination of self-reported items with established risk factors in a predictive model can assess periodontal disease accurately.

Responses of 246 subjects to a detailed questionnaire were compared to their periodontal disease history as assessed from radiographs. Multiple regression modeling was used to construct predictive models using self-reported items and established risk factors.

Depending on the definition of gold-standard periodontal disease, two or three self-reported items were selected for the predictive models, in addition to age, gender, and smoking. Self-reported tooth mobility was associated strongly with periodontal disease independent of other risk factors and was selected in all models. For dichotomous definitions of periodontal disease, discrimination of predictive logistic regression models was good with areas under the receiver operating characteristic curve >0.80. Assessment of periodontal disease history based on extreme quantiles of model-predicted values yielded high sensitivity and specificity.

The combination of several self-reported items may be useful for ascertainment of periodontal disease in epidemiologic studies.