The diagnoses of 431 general internal medicine patients from an urban outpatient department (OPD) were analyzed using two methods of case mix description: 1) a visit-based method which captures a single diagnosis for each visit; 2) a patient-based method which captures multiple diagnoses for a patient over one year. Nine of the top ten diagnoses were the same using either method, but the prevalence of diagnoses was two- to twelvefold higher with the patient-based method. Next the OPD was compared by the visit-based method with a national survey of doctors' private offices. Although the visit-based case mix in the OPD appeared to be the same as that in doctors' private offices, the analysis suggested that differences may be hidden by the method of describing case mix. The authors conclude that a visit-based approach to case mix description makes urban OPDs resemble doctors' private offices because the visit-based method undercounts those patients with chronic diagnoses, co-morbid conditions, and psychosocial problems, so common in the urban OPD. These findings have major implications for ambulatory reimbursement schemes, most of which capture only one diagnosis for each visit.