Abstract
Background/Aims
Electronic medical records (EMR) and claims data offer two potential data sources for researchers to examine healthcare utilization patterns and cost of care. In particular, combining the clinical and epidemiological variables typically available in EMR with cost information available in the claims data is not only intuitively sensible but also increasingly more feasible with growing standardization of EMR across healthcare delivery systems.
Methods
In this study, we compare EMR and claims data within a cohort of rheumatoid arthritis patients who received care from Geisinger Health System and had concurrent Geisinger Health Plan (GHP) coverage. We also develop a cost “imputation” method to obtain GHP claims-based cost estimates within EMR even for those who did not have GHP coverage.
Results
EMR-based estimated means of total cost of care and utilizations tend to substantially underestimate the total cost of care. In particular, EMR substantially understates emergency department (ED) visits [4% (125 of 3,131) in EMR vs. 11.2% (352 of 3,131) in claims], X-rays [4% (125 of 3,131) vs. 22% (689 of 3,131)], and CT scans [5.1% (160 of 3,131) vs. 7.3% (229 of 3,131)]. Use of biologic agents appear to be slightly higher in EMR than in claims [7.2% (226 of 3,131) vs. 6.7% (210 of 3,131)], although the difference is not statistically significant. The correlation between log-transformed EMR-based cost of care and log-transformed claims-based cost of care was modest (R2 = 0.81).
Conclusions
The findings confirm that there is significant disagreement between EMR and claims data and suggest that each represent a different set of “reality.” The main source of such discrepancies between EMR and claims appears to be missing utilization of certain types of care in EMR. In particular, discrepancies seem greater for the types of services for which patients have more alternative choices in the area. Claims data reflects health plan’s coverage decisions and utilization management, while EMR reflects clinicians’ decisions and practice patterns. Thus, researchers should carefully consider which “reality” they are interested in capturing in their analysis. Lastly, the fact that both EMR and claims are collected for clinical and administrative purposes, not for research purposes, must be emphasized.
Keywords: Cost of care, Electronic medical records