A non-clinical risk model for periodontitis derived from a national health examination survey: development and...

Periodontitis is a prevalent non-communicable disease (NCD), but most risk stratification tools require periodontal examination, limiting use in public health settings. This study developed and externally validated a non-clinical model for identifying prevalent periodontitis using routinely collected NCD surveillance variables. Because survey cycles were cross-sectional, the model was diagnostic rather than prognostic.

Data were obtained from the Korea National Health and Nutrition Examination Survey, a nationally representative survey with a complex sampling design. A survey-weighted multivariable logistic regression model was developed using cycle VII data (2016-2018; n=12,144) and externally validated without re-estimation using non-overlapping cycle VI data (2013-2015; n=12,523). Prevalent periodontitis was defined as a Community Periodontal Index score ≥3 in any sextant. Twelve pre-specified non-clinical predictors spanning demographic, socioeconomic, metabolic, and behavioural domains were entered without data-driven selection. Performance evaluation incorporated discrimination, calibration, reclassification, and decision curve analysis. Reporting followed the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis statement.

Discrimination improved from an area under the receiver operating characteristic curve (AUC) of 0.732 for the demographic-only model to 0.755 for the full model. Performance was retained during external validation (AUC, 0.728; calibration slope, 0.984; intercept, -0.106). Decision curve analysis demonstrated a positive net benefit across clinically plausible thresholds in both cohorts.

Routinely collected NCD surveillance data can identify adults at high risk of periodontitis with externally reproducible performance. Non-clinical models utilising these data may support periodontal case-finding within NCD screening programmes, particularly in primary care and community settings.