P-BN60 Predicting post-operative length of stay following laparoscopic cholecystectomy: derivation and validation of a pre-operative risk score

Abstract

Abstract Background To predict the post-operative length of stay following laparoscopic cholecystectomy (LC) will help guide the utilisation of resources, particularly the allocation of day surgery beds. Prolonged post-operative stay (PPS) is associated with morbidity and the acknowledgement of pre-operative factors predisposing to morbidity and PPS will influence surgical decision-making. The aim of the present study was to determine pre-operative factors associated with PPS and both derive and validate a risk score to predict the risk of PPS following LC. Methods Patients who underwent emergency and elective LC between January 2015 and December 2019 across three surgical centres were included. Pre-operative, operative and post-operative data were collected retrospectively from multiple databases using a deterministic records-linkage methodology. The cohort was randomly divided into a derivation and validation cohort, by a 3:1 ratio, respectively. The derivation cohort was used to create a risk score for PPS (≥3 days post-operatively) using multivariate logistic regression. The risk score was then applied to the smaller set or ‘validation cohort’ for internal validation purposes and the predictive accuracy was assessed using a ROC curve. Results The rate of PPS was 10.6% (294/2768). PPS was associated with intra-operative complication (RR-17.0;p<0.001), bail-out procedures (RR-47.1;p<0.001) post-operative complications (RR-11.9;p<0.001), re-admission (RR-2.4;p<0.001) and post-operative imaging/intervention (RR6.5;p<0.001). Variables associated with PPS included ag ≥ 60 (OR-1.56;p=0.011), male sex (OR-1.47;p=0.022), ASA 2 (OR-1.63;p=0.019, ASA≥3 (OR-3.27;p<0.001), 2 hospital admissions (OR-1.56;p=0.046), ≥3 hospital admissions (OR-2.11;p=0.024), cholecystitis (OR-3.19;p<0.001), pre-operative ERCP (OR-3.49;p<0.001) and cholecystostomy (OR-2.77;p=0.025) and emergency LC (OR-5.20;p<0.001). The AUC (area under curve) for the derivation and validation cohorts were 0.83 and 0.81 respectively. The risks of PPS in low-risk (0-5), medium-risk (5.5-10) and high-risk (>10) groups were 3.3%, 14.9% and 45.1%, respectively. Conclusions The rate of PPS following LC is significant and strongly associated with peri- and post-operative morbidity. Our model concludes that multiple pre-operative patient factors can predict the likelihood of prolonged post-operative stay. A patient’s risk score could be used to assess the risk of morbidity. Pragmatic patient selection in accordance with the above model could help surgical centres improve the allocation of beds between tertiary centres and day-surgery units.