Concordance Between Claims-Based and Electronic Health Record-Based Comorbidity Measures: An Application Using Linked Data.

PURPOSE: The National Cancer Institute Comorbidity Index (NCI-CI) captures patient-level comorbidity burden using diagnosis codes found in administrative claims or electronic health record (EHR) data. However, it is less well understood how claims- and EHR-based comorbidity measures compare. We determined the concordance between the claims-based and EHR-based NCI-CI using a linked data source.

MATERIALS AND

METHODS

This study used a linkage of tumor registry, EHR, and Medicare claims data. Eligible patients were 65 years and older and were diagnosed and/or received their first course of treatment in 2019-2020. We formed comparison pairs determined by the data source, period relative to the diagnosis date, length of interval for calculating the index, and type of comorbidity index. The primary pair was the 12-month prediagnosis NCI-CI calculated using claims (NCI-CIclaims) and EHR (NCI-CIEHR) data. We compared pairs using descriptive statistics, correlation coefficients, and histograms. We also identified patient-level factors associated with claims/EHR comorbidity discordance using models of categorical outcomes.

RESULTS

The final sample included 1,616 individuals. The median values for NCI-CIclaims and NCI-CIEHR were 0.6 (range: 0-4.11) and 0.29 (range: 0-3.06), respectively. We identified moderate correlation (0.49-0.52) between the claims-based and EHR-based comorbidity measures across all comparison pairs. Among individuals with claims-based comorbidities, 32 percent did not report any of those comorbidities based on the EHR data. Older age (eg, age 80+ years) was positively associated with the odds of having a discordant pair.

CONCLUSION

In this study, we found moderate correlations between claims-based and EHR-based approaches to NCI comorbidity measures. Mean comorbidity index values were consistently higher using claims data potentially because of more complete capture of health care utilization and diagnoses outside of the hospital system using claims data.