Secular Trends in Hip Fracture Mortality and Predictors of Mortality from the NSQIP Database

Abstract Context Hip fractures incur high morbidity and mortality. Data on secular trends in mortality from hip fractures and risk predictive models are scarce. Objective We aim to describe secular trends in 30-day mortality post-hip fracture surgery from the 2011-2017 National Surgical Quality Improvement Program (NSQIP) database, identify preoperative and on-discharge predictors of 30-day mortality and develop risk calculators. Methods We calculated yearly proportions of deaths and examined survival using Kaplan Meier curves. We implemented logistic regressions models, using SPSS and created calculators using Excel. Results In 84,824 cases of hip fracture surgery, the overall 30-day mortality was 6.8%. It decreased from 8.1% to 6.5% between 2011 and 2017 (p<0.001). Significant preoperative predictors of 30-day mortality on admission were male gender, age, lower BMI, white race, poorer functional health status, higher creatinine, lower hematocrit, >10% weight loss in the past 6 months, congestive heart failure within 30 days before surgery, and chronic obstructive pulmonary disease. Predictors on discharge included preoperative predictors with the exception of white race, hematocrit and > 10% weight loss in the past 6 months, and the addition of unplanned intubation, cerebrovascular accident, myocardial infarction, and pneumonia. The parsimonious preoperative risk calculator for mortality had 10 variables, an area under the curve (AUC) of 0.739, and a model fit R2 of 0.9716. The on-discharge calculator had 11 variables, an AUC of 0.800, and an R2 of 0.9924. Conclusion 30-day mortality post-hip fracture surgery decreased significantly from 2011 to 2017. Readily available clinical risk factors predict mortality, preoperatively and on discharge. While most predictors are non-modifiable, the calculators may better inform clinical decision-making.