Three city data sources (CDSs)—police reports, ambulance reports, and medical examiner (ME) logs—were evaluated for their usefulness in epidemiologic studies of trauma. The CDSs were employed to identify all cases of penetrating injury to the chest and/or abdomen severe enough to require care in a medical institution during 1979 and 1980 in Baltimore city. The percent of cases identified by source was: police, 66.8%; ambulance, 47.9%; ME, 16.6%; police plus ambulance, 89.4%; police plus ME, 82.9%; and ambulance plus ME, 50.1%. Hospital admissions to six study hospitals due to chest and/or abdomen penetrating injury were located and matched to the CDS reports: 89.2% of the hospitalized cases were reported in one or more CDS, and 34.7% of the cases identified by one or more CDS could not be located in the hospital records. Using hospital records as the standard, each source was determined to have the following completeness of case reporting: police, 66.2%; ambulance, 72.9%; and ME, 92.2%. The authors conclude that existing CDSs should be used with caution, and that the usefulness of data from multiple sources far outweighs that from any single source.