Use of ICD-9-CM codes in the estimation of alcohol-involved injury: search for a surrogate II

This study presents a follow-up analysis of 22,427 injury cases drawn from the California Regional Trauma Registry. Whereas the earlier analysis developed a surrogate for alcohol-involved injuries using E-codes as reflective of injury type, this analysis explores the possibility of using ICD-9-CM diagnosis codes that have a longer history and are available in more jurisdictions. Findings reported herein indicate that the original demographic patterns predicting testing and alcohol involvement patterns were maintained in ICD-9-CM diagnostic code-based models. Moreover, although variables representing demographic background, time of day, and day of week clearly were the most powerful model predictors, ICD-9-CM diagnostic codes reflecting physician assessment that the patient had a chronic alcohol problem did improve the fit of models, and thus provide additional information concerning testing and blood alcohol concentration patterns. In contrast, ICD-9-CM diagnostic codes reflecting the specific nature of the injury, although statistically significant, explained little additional variability in alcohol involvement. Nevertheless, the overall model did accurately classify approximately 75% of those in the sample for whom blood alcohol concentration status information was available, suggesting its appropriateness for surrogate development.