Discordance between physicians and coders in assignment of diagnoses

Objective. To measure concordance between physicians and medical record cod ers in their assignment of diagnoses. Design. Prospective cohort series. Setting. Five hundred and fifty-bed, tertiary-care, university teaching hos pital. Study participants. In-patients who were discharged from either the Cardiac Sciences Program (n=125), the Renal Program (n=43), or the HIV-AIDS Progra m (n=25) during the period May 18-July 1, 1995. Interventions. None. Main outcome measures. Physicians and coders assigned diagnoses for individ ual in-patients based on their independent interpretations of the patient c hart and discharge summary sheet. AU assigned diagnoses were coded using th e ICD-9-CM classification system. Concordance was measured for the most res ponsible diagnosis and for all assigned diagnoses. Difference in calculated resource intensity weights based on physicians' and coders' assignment of diagnoses was also calculated. Results. Concordance rates for the most responsible diagnosis in each progr am were: Cardiac Sciences [27%; 95% confidence interval (CT)= 20-36%], Rena l Program (35%; 95% CI = 21-53%), and HIV-AIDS Program (20%; 95% CI, 6-41%) . Concordance rates for all. diagnoses per chart were similar: Cardiac Scie nces (20%; 95% CI, 14-25%), Renal Program (25%; 95% CI, 20-33%), and HIV-AI DS Program (29%; 95% CI, 25-44%). Resource intensity weights assigned by co ders for the Cardiac Sciences and HIV-AIDS Program were significantly highe r than those assigned by the physicians.