In:
British Journal of Surgery, Oxford University Press (OUP), Vol. 111, No. Supplement_6 ( 2024-07-03)
Abstract:
To develop a novel prognostic score to predict 30-day surgical-site infection (SSI) following gastrointestinal surgery, and externally validate in comparison to existing prognostic models. Method This was a secondary analysis of 3 independent prospective international cohort studies conducted on a global basis. This included adults undergoing gastrointestinal surgery. Model development was performed in the GlobalSurg-2 dataset (January-July 2016). The primary outcome was 30-day SSI, with two predictive techniques explored: penalised regression (LASSO) and machine learning (XGBOOST). Final model selection based on prognostic accuracy (Area Under the Curve [AUC], 95% confidence interval) and clinical utility. Novel and previous scores were externally validated in GlobalSurg-1 (July-November 2014), and GlobalSurg-3 (April-October 2018). Results 30,029 patients were eligible: 14,019 (SSI=12.3%) in the derivation cohort, and 8,464 (SSI=11.4%) and 7,546 (SSI=15.7%) in the validation cohorts (GlobalSurg-1 and GlobalSurg-3 respectively). The LASSO model was selected due to similar discrimination to XGBoost in the GlobalSurg-2 dataset (AUC: 0.738, 0.725-0.750 versus 0.737, 0.709-0.765), but greater explainability. The final GloSSI score included six variables: country income, ASA, diabetes, and operative contamination, approach, and duration. Discrimination remained good on external validation in GlobalSurg-1 (AUC: 0.730, 0.715-0.744), but moderate discrimination in GlobalSurg-3 (AUC: 0.606, 95% CI 0.588-0.623). Nonetheless, this demonstrated superior performance to external validation of all previous models evaluated within GlobalSurg or prior datasets. Conclusions The GloSSI score allowed accurate prediction of SSI with 6 simple variables routinely available at surgery across income settings. This can inform use of intraoperative and postoperative interventions to modify the risk of SSI and minimise associated harm.
Type of Medium:
Online Resource
ISSN:
0007-1323
,
1365-2168
DOI:
10.1093/bjs/znae163.082
Language:
English
Publisher:
Oxford University Press (OUP)
Publication Date:
2024
detail.hit.zdb_id:
2006309-X
