External validation of a model using health administrative data to predict acetabular fracture probability: Brief report

Abstract

Analyses using population-based health administrative data can return erroneous results if case identification is inaccurate (“misclassification bias”). An acetabular fracture (AF) prediction model using administrative data decreased misclassification bias compared to identifying AFs using diagnostic codes. This study measured the accuracy of this AF prediction model in another hospital. We calculated AF probability in all hospitalizations in the validation hospital between 2015 and 2020. A random sample of 1000 patients stratified by expected AF probability was selected. Patient imaging studies were reviewed to determine true AF status. The validation population included 1000 people. The AF prediction model was very discriminative (c-statistic 0.90, 95% CI: 0.87–0.92) and very well calibrated (integrated calibration index 0.056, 95% CI: 0.039–0.074). AF probability can be accurately determined using routinely collected health administrative data. This observation supports using the AF prediction model to minimize misclassification bias when studying AF using health administrative data.