Mike Fruscione

Routine versus difficult cholecystectomy: using predictive analytics to assess patient outcomes

BACKGROUNDnThe American College of Surgeons National Surgical Quality Improvement Program® (NSQIP) Surgical Risk. Calculator (SRC) estimates postoperative outcomes. The aim of this study was to develop and validate a specific predictive outcomes model for cholecystectomy procedures.nnnMETHODSnPatients who underwent cholecystectomy between 2008 and 2016 and were deemed too high risk for acute care general surgery (GS) and had surgery performed by the Division of Hepatopancreatobiliary Surgery (HPB) were identified. Outcomes of the HPB cholecystectomies were matched against cholecystectomies performed by GS. New predictive models for postoperative outcomes were constructed. Area under the curve was used to assess predictive accuracy for both models and internal validation was performed using bootstrap logistic regression.nnnRESULTSnA total of 169/934 (18%) cholecystectomies were identified as too high risk for GS. These 169 patients were matched with 126 patients who had cholecystectomy performed by GS. For GS and HPB cholecystectomies, the proposed model demonstrated better discriminative ability compared to the SRC based on ROC curves (proposed model: 0.589-0.982; SRC: 0.570-0.836) for each of the predicted outcomes.nnnCONCLUSIONnFor patients undergoing cholecystectomy, customized models are superior for predicting individual perioperative risk and allow more accurate, patient-specific delivery of care.

Developing and validating a center-specific preoperative prediction calculator for risk of outcomes following major hepatectomy procedures

BACKGROUNDnThe American College of Surgeons NSQIP® Surgical Risk Calculator (SRC) was developed to estimate postoperative outcomes. Our goal was to develop and validate an institution-specific risk calculator for patients undergoing major hepatectomy at Carolinas Medical Center (CMC).nnnMETHODSnOutcomes generated by the SRC were recorded for 139 major hepatectomies performed at CMC (2008-2016). Novel predictive models for seven postoperative outcomes were constructed and probabilities calculated. Brier score and area under the curve (AUC) were employed to assess accuracy. Internal validation was performed using bootstrap logistic regression. Logistic regression models were constructed using bivariate and multivariate analyses.nnnRESULTSnBrier scores showed no significant difference in the predictive ability of the SRC and CMC model. Significant differences in the discriminative ability of the models were identified at the individual level. Both models closely predicted 30-day mortality (SRC AUC: 0.867; CMC AUC: 0.815). The CMC model was a stronger predictor of individual postoperative risk for six of seven outcomes (SRC AUC: 0.531-0.867; CMC AUC: 0.753-0.970).nnnCONCLUSIONnInstitution-specific models provide superior outcome predictions of perioperative risk for patients undergoing major hepatectomy. If properly developed and validated, institution-specific models can be used to deliver more accurate, patient-specific care.