Mapping sequential patterns of periodontal breakdown using longitudinal electronic health records data

To identify probabilistic, sequential patterns of periodontal breakdown using sequential pattern mining (SPM) on a large longitudinal cohort.

Electronic health records from 19,428 patients with periodontitis (2011-2022) were analyzed. A custom SPM algorithm was developed to identify frequent temporal subsequences of tooth mobility and furcation involvement, enforcing strict temporal precedence. 'High-risk' states (eg, generalized clinical attachment loss [CAL] > 3.0 mm) were evaluated as precursors to localized failure. Association rules were ranked by confidence (conditional probability).

The analysis revealed distinct biomechanical cascading patterns. In the mandibular anterior sextant, mobility in the central incisors was a strong probabilistic precursor to subsequent failure in the lateral incisors and canines (confidence 7.3%). A significant 'antagonistic cascade' was identified, where mandibular incisor mobility preceded maxillary incisor instability (confidence 8.1%). For molars, furcation involvement acted as a specific temporal sentinel for future mobility (confidence ~5% to 6%). Generalized high CAL was the strongest predictor of localized failure (confidence 18.9%), significantly outperforming active inflammation (bleeding on probing) as a driver of mechanical instability.

Progressive biomechanical compromise (such as tooth mobility and advanced furcation involvement) follows predictable, probabilistic sequential patterns. The identification of specific 'sentinel events,' such as mandibular incisor mobility and molar furcation involvement, allows for a shift from reactive site-by-site management to predictive intervention.

These findings underscore the critical role of generalized attachment loss in driving localized mechanical failure cascades.