The Pathologic Fracture Mortality Index: A Novel Externally Validated Tool for Predicting 30-day Postoperative Mortality

Abstract

Current mortality predictive tools, in the setting of completed or impending pathologic fractures, are nonspecific. Clinical decision making and mortality prediction in research would benefit from creation of a high-fidelity scoring system for calculating the risk of 30-day postoperative mortality. The purpose of this study is to develop a validated research and clinical tool that is superior to existing methods in estimating postoperative mortality risk after fixation of pathologic fractures. One thousand two hundred nineteen patients who underwent fixation for either completed or impending pathologic fractures in the National Surgical Quality Improvement Program (2012 to 2018) database were analyzed. Multivariable logistic regression with diagnostics was used to develop a predictive model in a derivation cohort and then validated in a validation cohort. Area under the curve (AUC) from receiver operator curve analysis was used to assess accuracy. A score was derived and compared with the American Society of Anesthesiologists classification and modified five-component frailty index (mF-I5). The score was validated in an exclusive cohort of patients who underwent fixation for pathologic fractures at a tertiary care center. Of 1,219, a total of 177 (15%) patients did not survive beyond 30 days postoperatively. AUC for our predictive model was 0.76 in the derivation and 0.75 in the validation National Surgical Quality Improvement Program cohorts. The derived Pathologic Fracture Morbidity Index included seven data points: anemia, alkaline phosphatase > 150 U/L, albumin < 3.5 mg/dL, pulmonary disease, recent weight loss, functional dependence, and white blood cell count >12,000. The PFMI (AUC = 0.75) was more accurate than ASA (AUC = 0.60) or mF-5 (AUC = 0.58) (P < 0.01). The AUC for PFMI in predicting 30-day mortality in the exclusive cohort (N = 39) was 0.74. The PFMI is a validated tool that may be used for predicting postoperative 30-day mortality after fixation of pathologic fractures, with higher level of accuracy compared with ASA or mF-I5.