Implementation of Validated Caries Indices in an Electronic Health Record

Objectives: The objective of this study was to implement and validate caries indices in a dental institution that were previously developed and validated in an Electronic Health Record (EHR) at two time points.
Methods: Using algorithms previously developed and validated, 24 caries indices (DMFT, dmft, ds, dt, ect.) were implemented at UCSF within the EHR (axiUm, Exan Group, Coquitlam, British Columbia, Canada). Once implemented, a data sample was extracted from a set of 11,500 patient records who were seen for a dental exam in 2015-2016 (T0 exam date) and subsequently seen for a dental exam in 2017-2018 (T1 exam date). Purposive sampling was used to extract 200 dental records with 3 types of dentition (Primary, Mixed, Permanent) and 3 levels of decay experience (Low, Moderate, High). The computer derived caries indices were examined by manual review of the EHR by two calibrated abstractors. Discrepancies between the 2 abstractors were resolved by consensus and the consensus manual review was compared to the computer derived caries indices. Correlation statistics computed the automated EHR calculation by intra-class correlations (Lin’s concordance correlations (LCCs)).
Results: Abstractor consensus had high correlation with EHR caries indices at each time points: LCCs ≥ 0.99 for DMFT (T1, T0), dmft (T0), DMFSdmfs (T1, T0), DMFS (T1, T0), dmfs (T1, T0), MT (T1, T0), FT (T1, T0), mt (T1, T0), ft (T0) MS (T1, T0), FS (T1, T0), ms (T1, T0), fs (T1, T0), 0.99 > LCCs ≥ 0.98 for DMFTdmft (T1, T0), dmft (T1), ft (T1), ds (T1), and 0.98 > LCCs ≥ 0.92 for dt (T1, T0), DS (T1), ds (T0). Permanent decayed surfaces had slightly lower agreement at the initial time point (LCCs = 0.77), which improved at the latter time point (LCCs = 0.86).
Conclusions: We conclude that caries indices derived from an EHR, were successfully implemented and validated at a dental institution where they can be used for research and patient care purposes.