Synergy of diagnosis coding between administrative claims and electronic health records of large patient populations across multiple healthcare organizations.

Objective: This study evaluates the completeness of diagnostic information captured in electronic health records (EHRs) compared to administrative claims data across multiple U.S. healthcare organizations between 2010 and 2022. Materials and Methods: Deidentified claims and EHR data of ∼17 million patients across 56 health care organizations were linked. For each Clinical Classification Software (CCS) diagnostic group, proportions of patient-years were computed for diagnoses recorded exclusively in administrative claims, in EHR, or in both sources. Agreement metrics were calculated for high-level CCS diagnostic groups. Trends in the EHR-claims diagnosis coding gap were estimated using linear regression. Results: The completeness of diagnosis data captured in the EHR data, as compared to administrative claims, improved ∼10% from 2010 to 2019. Nonetheless, ∼45% of person-diagnosis data were only captured in claims, and were missing in EHRs, across all years. The missingness of diagnosis data in EHRs, as compared to claims, varied across CCS diagnostic groups with some categories being missed more often than others. The EHR-claims gap of diagnostic codes also affected comorbidity measures such as the Charlson Comorbidity Index, which narrowed significantly from 2010 to 2019, but did not meaningfully change from 2020 to 2022. Discussion: Capturing comprehensive diagnosis data is essential for accurate risk adjustment, yet EHR data systematically under-document multimorbidity compared with claims. Improving EHR interoperability can enhance completeness of EHR-derived diagnosis data and potentially narrow the EHR-claims gap of diagnostic codes. Conclusions: EHRs capture significant amounts of diagnostic data, but increased interoperability of EHRs and integration of claims feeds are essential to achieving comprehensive risk stratification capability.