Prediction of periodontal disease: modelling and validation in different general German populations

AIM: To develop models for periodontitis using self-reported questions and to validate them externally.

METHODS: The Study of Health in Pomerania (SHIP-0) was used for model development. Periodontitis was defined according to the definitions of the Center for Disease Control and Prevention-American Academy of Periodontology, the 5th European Workshop in Periodontology, and Dietrich et al. (≥2 teeth with inter-proximal clinical attachment loss of ≥4 mm and 6 mm as moderate and severe periodontitis) respectively. These models were validated in SHIP-Trend and the Fourth German Oral Health Study (DMS IV).

RESULTS: Final models included age, gender, education, smoking, bleeding on brushing and self-reported presence of mobile teeth. Concordance-statistics (C-statistics) of the final models from SHIP-0 were 0.84, 0.82 and 0.85 for the three definitions respectively. Validation in SHIP-Trend revealed C-statistics of 0.82, 0.81 and 0.82 respectively. As bleeding on brushing and presence of mobile teeth were unavailable in DMS IV, reduced models were developed. C-statistics of reduced models were 0.82, 0.81 and 0.83 respectively. Validation in DMS IV revealed C-statistics of 0.72, 0.78 and 0.72 for the three definitions respectively. All p values of the goodness-of-fit tests were >0.05.

CONCLUSIONS: The models yielded a moderate usefulness for prediction of periodontitis.