This study evaluates the quality of injury-related coding in state hospital data and their usefulness to injury researchers. Using 1997 hospital discharge records from 19 states, hospitalized non-fatal injury-related cases were identified by first selecting all cases that met broad criteria for injury, and then dropping cases that appeared incorrectly coded as injuries and cases related to medical care. Based on our criteria, 1,129,980 non-fatal hospitalized cases were identified as probable acute injuries. Three-quarters were coded with a traditional injury diagnosis in the primary diagnosis field, and 90% had a traditional injury diagnosis somewhere in the first six diagnosis fields. Of cases with an injury diagnosis code in the first three diagnosis fields, 88.1% were E coded. E coding completeness varied by state, with some states reporting high rates of E coding by using non-specific E codes. Other challenges included E-coded cases where no injury diagnosis was reported and apparent miscoding of the E code. We conclude that it is possible to combine multiple states' data if researchers are aware of the challenges they may encounter. In order to capture all injury-related cases, it is important to scan secondary diagnosis fields.