Ann Surg. 2023 May 4. doi: 10.1097/SLA.0000000000005897. Online ahead of print.
ABSTRACT
OBJECTIVES: To develop and validate a predictive model to predict the risk of postoperative mortality after emergency laparotomy taking into account the following variables: age, age ≥ 80, ASA status, clinical frailty score, sarcopenia, Hajibandeh Index (HI), bowel resection, and intraperitoneal contamination.
SUMMARY BACKGROUND DATA: The discriminative powers of the currently available predictive tools range between adequate and strong; none has demonstrated excellent discrimination yet.
METHODS: The TRIPOD and STROCSS statement standards were followed to protocol and conduct a retrospective cohort study of adult patients who underwent emergency laparotomy due to non-traumatic acute abdominal pathology between 2017 and 2022. Multivariable binary logistic regression analysis was used to develop and validate the model via two protocols (Protocol A and B). The model performance was evaluated in terms of discrimination (ROC curve analysis), calibration (calibration diagram and Hosmer-Lemeshow test), and classification (classification table).
RESULTS: One thousand forty three patients were included (statistical power = 94%). Multivariable analysis kept HI (Protocol-A: P=0.0004; Protocol-B: P=0.0017), ASA status (Protocol-A: P=0.0068; Protocol-B: P=0.0007), and sarcopenia (Protocol-A: P<0.0001; Protocol-B: P<0.0001) as final predictors of 30-day postoperative mortality in both protocols; hence the model was called HAS (HI, ASA status, sarcopenia). The HAS demonstrated excellent discrimination (AUC: 0.96, P<0.0001), excellent calibration (P<0.0001), and excellent classification (95%) via both protocols.
CONCLUSIONS: The HAS is the first model demonstrating excellent discrimination, calibration, and classification in predicting the risk of 30-day mortality following emergency laparotomy. The HAS model seems promising and is worth attention for external validation using the calculator provided.
PMID:37139796 | DOI:10.1097/SLA.0000000000005897
