Preoperative risk stratification of patient mortality following elective craniotomy; a comparative analysis of prediction algorithms

Abstract

Factors associated with mortality following craniotomy for non-traumatic etiologies are not well studied. We performed a retrospective case-control study to investigate the utility of the Gupta, RCRI, KPS, ASA, and NSQIP risk calculators in predicting complication and mortality rates among a neurosurgical patient population undergoing elective craniotomy and used ROC curves to determine their relative predictive accuracy. We identified 17 patients who died following elective craniotomy and 17 control patients who survived, matched for sex, age, comorbid conditions, and intervention. The deceased cohort experienced a greater degree of surgical complications, with concordantly higher preoperative risk scores as assessed by Gupta, KPS, ASA, and NSQIP scales. Cause of death was secondary to a surgical complication in 11 (65%) cases and nonsurgical in 6 (35%) cases. When comparing our deceased and survival cohorts for specific complication profiles, higher RCRI score was significantly associated with postoperative cardiac death. Poor preoperative ASA and KPS score were both associated with mortality. NSQIP risk was significantly associated with mortality and nonsurgical and cardiopulmonary complications. ROC analyses comparing predictive scales found no clearly superior prediction algorithm for postoperative complications or mortality; NSQIP and RCRI were highly predictive of cardiac death, KPS and NSQIP with nonsurgical complications, and Gupta, ASA, KPS, and NSQIP with postoperative mortality. NSQIP scores correlated most broadly with nonsurgical complications and mortality. Future efforts at quality improvement in healthcare value will be rooted in the development of more accurate and specific risk assessment strategies specific to neurosurgery.

Keywords: Complication; Craniotomy; Morbidity; Mortality; Quality improvement; Risk; Safety.