Reliability of the minimum basic dataset for diagnoses of cerebrovascular disease.

INTRODUCTION: The minimum basic dataset is the largest available hospital care administrative database that is used in clinical studies and hospital management in association with diagnosis-related groups (DRGs). In 2011, the quality of the national MBDS in hospital discharges was audited, in order to assess its reliability. This paper presents a sub-analysis of the results from that analysis which are referred to cerebrovascular disease (CVD).
METHODS: Using all discharge reports from the Spanish MBDS in 2009, a representative sample was obtained by stratified sampling and 11 209 records were evaluated. Outcome indicators were obtained to measure any differences observed between the national MBDS being evaluated and the hospital's original MBDS. Analysis of codes for CVD as a primary diagnosis was performed for ICD-9-CM diagnostic categories 430 through 438. We evaluated error rates in the selection and classification of main diagnoses, as well as in DRG assignment.
RESULTS: There were 397 discharges of cases of CVD which included 21 different DRGs. Diagnostic coding showed a concordance rate of 81.87%; the selection error rate was 2.26% and the classification error rate was 15.87%. The error rate in the DRG was 16.12% and associated with the greatest impact on the mortality risk level.
CONCLUSIONS: While the errors we observed must be taken into account, data suggest that the quality of the MBDS for CVD is sufficient to ensure delivery of valid information. The hospital discharge registry serves as a valuable tool for use in studies of this disease.