Patient-specific predictors of failure to rescue after pancreaticoduodenectomy

Abstract

Background: Failure to rescue (FTR) is a recently described outcome metric for quality of care. However, predictors of FTR have not been adequately investigated, particularly after pancreaticoduodenectomy. We aim to identify predictors of FTR after pancreaticoduodenectomy.

Methods: We reviewed all patients who developed serious morbidity after pancreaticoduodenectomy from 2005 to 2012 in the ACS-NSQIP database. Logistic regression was used to identify preoperative and postoperative risks for 30-day mortality within a development cohort (randomly selected 80%). A score was created using weighted beta coefficients. Predictive accuracy was assessed on the validation cohort (remaining 20%) using a receiver operator characteristic curve and calculating the area under the curve (AUC).

Results: The FTR rate was 7.2% after pancreaticoduodenectomy (n = 5,027). We identified 5 independent risk factors: age ≥65 and albumin ≤3.5 g/dL, preoperatively; and development of shock, renal failure, and reintubation, postoperatively. The generated score had an AUC = 0.83 (95% CI, 0.77-0.89) in the validation cohort. Using the score: 1*Albumin ≤3.5 g/dL + 2*Age ≥ 65 + 2*Shock + 5*Renal failure + 5*Reintubation, FTR rates increased with increasing score (p < 0.001).

Conclusion: FTR rates have previously been shown to be associated with hospital factors. We show that FTR is also associated with preoperative and postoperative patient-specific factors.