Background: The use of the French version of the DRG model is focused on cost allocation, based on the case-mix system and the use a weight called ISA (Synthetic Index of Activity) for each DRG. However, this administrative database is becoming more and more used by both researchers and health policy makers for health planning and benchmarking. In France, data abstraction and coding of medical records is done by physicians. The objective of this study was to determine the accuracy of a database of the discharge summaries used for DRGs and to compare consequences of inappropriate coding on budget estimation and risk adjustment.
Methods: Samples of discharge summaries from six cardiology units were recoded by trained physicians in data abstracting and coding. Comparison between initial and recoded diagnoses (errors on main diagnosis or on comorbidities) used by the DRG system algorithm, and the original and final case-mix were performed. The before and after abstracted data were stratified and compared by principal diagnosis (myocardial infarction or congestive heart failure) and discharge status (dead or alive).
Main results: Comorbidities were underreported by physicians of cardiology units compared to reabstracted data (mean number of secondary diagnoses per summary: 2.1 vs. 3.6, p<0.001), especially those which had a minimal impact on the DRG classification. In spite of a 15% rate of wrong DRGs, there was no significant difference in the total amount of ISA after data reviewing. Underreporting of comorbidities is more important for medical records of dead patients at discharge but, without significant effect on rate of change in DRG and amount of ISA.
Conclusion: Discharge summaries used in the French DRGs system consistently underestimate the presence of comorbid conditions, which has direct implications for policy-makers comparing performance between hospital units. Both clinical practitioners and policy makers should be aware of this bias when assessing patient's quality of care or performing health planning through discharge summaries.