Restoration-Associated Active Coronal Caries in NHANES Adults: Reproducibility Package for Interpretable Nonlinear Modeling

Restoration history records previous treatment but does not establish current disease control. We characterized restoration-associated active coronal caries (RAACC), quantified its burden among adults with recorded restorations, and compared interpretable pre-examination and oral-chart-augmented predictive structure. Adults aged 20 years or older with Complete or Partial NHANES dentition examinations formed a 2015–2016 derivation cohort and a 2017–March 2020 later-cycle evaluation cohort. In the later cycle, survey-weighted RAACC prevalence was 5.25% overall and 6.58% (95% CI 5.13–8.40) among adults with a recorded restoration. Fair or poor self-rated oral health and unmet dental need were associated with RAACC. Explainable Boosting Machine Model A achieved an AUC of 0.713 using pre-examination information, whereas Model B achieved an AUC of 0.842 after adding examination-derived oral-chart features. The chart-augmented gain depended substantially on interaction structure. RAACC represents a treated-but-active same-tooth survey phenotype and should not be interpreted as secondary caries, restoration failure, causal evidence, or a clinically validated diagnostic system. This record contains the accompanying reproducibility package, including analysis code, aggregate results, model and interpretability artifacts, figures, documentation, environment records, and integrity manifests. It does not contain raw NHANES files, participant-level records, SEQN identifiers, or participant-level model matrices.