Utility of automated data in identifying femoral shaft and subtrochanteric (diaphyseal) fractures

Summary In women age 45 years and older, enrolled in an integrated group practice in 2007, use of ICD9 diagnostic codes, including the “not otherwise specified” code (821.00) resulted in a high false-positive rate for identifying femoral diaphyseal fractures. Restriction to more specific site-codes missed 36% of these rare fractures. Introduction The aim of this study was to assess the utility of automated data in identifying the occurrence of femoral diaphyseal fractures. Methods We identified all women age 45 years and older enrolled in a Pacific Northwest integrated group practice during 2007. Using the computerized database we selected all ICD9 codes that could be related to a femur fracture occurring in the diaphyseal region. We then quantified the percent of codes confirmed by medical record review to have occurred in the correct anatomic location during the year of interest (positive predictive value). Results Of the 95,765 eligible women, 161 (0.17%) had an ICD9 diagnostic code potentially related to a femoral diaphyseal fracture in 2007; of these 58 (36%) had a fracture of the femoral diaphysis, and 38 (24%) of the fractures occurred in 2007. The most frequent code was 821.00, described as “femur fracture not otherwise specified”, applied to 107 women; 21 of the 58 diaphyseal fractures had this code. Conclusion In this study, use of ICD9 codes that included the “not otherwise specified” code (821.00) resulted in a high false-positive rate for identifying diaphyseal fractures. However, restriction to more specific site codes would have missed at least 36% of the diaphyseal fractures. Furthermore, the codes did not provide any information about the characteristics of the fracture. Our findings support validating cases selected using ICD codes before they are used as a surrogate for the occurrence of femoral diaphyseal fractures.