Objective: To assess the coding accuracy of primary and secondary discharge diagnoses in the Quebec hospital discharge database for elderly persons with myocardial infarction (MI).
Design: Retrospective chart review in a convenience sample of six Montreal hospitals. The diagnoses listed in the medical chart were compared with those listed in the hospital discharge database. For each subject, the Charlson comorbidity index was calculated twice, once based on the medical chart and again based on the hospital discharge database.
Patients: Subjects aged 65 years and over who had an MI coded as the primary discharge diagnosis in the hospital discharge database and who were discharged alive.
Main results: For 234 MI survivors, the positive predictive value (ie, probability that a patient with MI reported in the hospital discharge database had an MI diagnosed by the discharging physician) for coding MI was 0.96 (95% CI 0.94, 0.98). Comorbid medical conditions and complications of the MI were under-reported in the hospital discharge database, which meant that the Charlson index based on the hospital discharge database was an average of 0.71 units lower than the Charlson index based on the medical chart.
Conclusions: When studying survivors of MI by using hospital discharge databases, the advantages must be weighed against potential drawbacks in the quality of the information. Hospital discharge databases are almost as reliable as medical charts for identifying MI patients, but there is substantial under-reporting of comorbid medical conditions.