Do current comorbidity indices accurately predict adverse events after operative fixation of hip fractures? A retrospective database review

Abstract

Background: The incidence of hip fractures continues to grow due in part to an aging population and increasing activity levels. With the potential implementation of bundled-care payment models for the operative fixation of hip fractures, a metric that accurately predicts postoperative length of stay and adverse events would allow for adjustments in bundled payments for “high-risk” patients. Methods: A retrospective review of the National Inpatient Sample was performed to identify patients with a femoral neck fracture or intertrochanteric femoral fracture between 2002-2014. Multivariable logistic regression models, with either the Elixhauser Comorbidity Measure (ECM) or Charlson Comorbidity Index (CCI) were created to predict inpatient mortality and adverse events. A base model that just contained patient demographic characteristics also was evaluated. Last, a combined model that used each index, along with the base model was created. The predictive discrimination of each model was evaluated using the C-statistic. Results: A total of 477,648 hip fractures were identified. The mean age of our cohort was 82.3±7.3 yr, with an inpatient mortality rate of 2.2%. The model incorporating the base demographic variables and ECM provided the best predictive models, with a C-statistics of 0.767 for inpatient mortality, 0.713 for cardiac complications, 0.818 for pulmonary complications, 0.818 for renal complications, and 0.615 for thromboembolic complications. Conclusions: A combined model that includes basic demographic variables and the ECM outperforms either the CCI or ECM in isolation for predicting inpatient mortality and adverse events after hip fractures. Level of Evidence: Level IV.