TY - JOUR
T1 - Development and external validation of the 'Global Surgical-Site Infection' (GloSSI) predictive model in adult patients undergoing gastrointestinal surgery
AU - NIHR Global Research Health Unit on Global Surgery and GlobalSurg Collaborative
AU - Picciochi, M.
AU - Pinkney, T.
AU - Li, E.
AU - Morton, D.
AU - Harrison, E. M.
AU - Lilford, R.
AU - Bhangu, A.
AU - Nepogodiev, D.
AU - Belli, A.
AU - Kamarajah, S. K.
N1 - Publisher Copyright:
© The Author(s) 2024. Published by Oxford University Press on behalf of BJS Foundation Ltd.
PY - 2024/6/21
Y1 - 2024/6/21
N2 - Background: Identification of patients at high risk of surgical-site infections may allow surgeons to minimize associated morbidity. However, there are significant concerns regarding the methodological quality and transportability of models previously developed. The aim of this study was to develop a novel score to predict 30-day surgical-site infection risk after gastrointestinal surgery across a global context and externally validate against existing models. Methods: This was a secondary analysis of two prospective international cohort studies: GlobalSurg-1 (July-November 2014) and GlobalSurg-2 (January-July 2016). Consecutive adults undergoing gastrointestinal surgery were eligible. Model development was performed using GlobalSurg-2 data, with novel and previous scores externally validated using GlobalSurg-1 data. The primary outcome was 30-day surgical-site infections, with two predictive techniques explored: penalized regression (least absolute shrinkage and selection operator ('LASSO')) and machine learning (extreme gradient boosting ('XGBoost')). Final model selection was based on prognostic accuracy and clinical utility. Results: There were 14 019 patients (surgical-site infections = 12.3%) for derivation and 8464 patients (surgical-site infections = 11.4%) for external validation. The LASSO model was selected due to similar discrimination to extreme gradient boosting (AUC 0.738 (95% c.i. 0.725 to 0.750) versus 0.737 (95% c.i. 0.709 to 0.765)), but greater explainability. The final score included six variables: country income, ASA grade, diabetes, and operative contamination, approach, and duration. Model performance remained good on external validation (AUC 0.730 (95% c.i. 0.715 to 0.744); calibration intercept -0.098 and slope 1.008) and demonstrated superior performance to the external validation of all previous models. Conclusion: The 'Global Surgical-Site Infection' score allows accurate prediction of the risk of surgical-site infections with six simple variables that are routinely available at the time of surgery across global settings. This can inform the use of intraoperative and postoperative interventions to modify the risk of surgical-site infections and minimize associated harm.
AB - Background: Identification of patients at high risk of surgical-site infections may allow surgeons to minimize associated morbidity. However, there are significant concerns regarding the methodological quality and transportability of models previously developed. The aim of this study was to develop a novel score to predict 30-day surgical-site infection risk after gastrointestinal surgery across a global context and externally validate against existing models. Methods: This was a secondary analysis of two prospective international cohort studies: GlobalSurg-1 (July-November 2014) and GlobalSurg-2 (January-July 2016). Consecutive adults undergoing gastrointestinal surgery were eligible. Model development was performed using GlobalSurg-2 data, with novel and previous scores externally validated using GlobalSurg-1 data. The primary outcome was 30-day surgical-site infections, with two predictive techniques explored: penalized regression (least absolute shrinkage and selection operator ('LASSO')) and machine learning (extreme gradient boosting ('XGBoost')). Final model selection was based on prognostic accuracy and clinical utility. Results: There were 14 019 patients (surgical-site infections = 12.3%) for derivation and 8464 patients (surgical-site infections = 11.4%) for external validation. The LASSO model was selected due to similar discrimination to extreme gradient boosting (AUC 0.738 (95% c.i. 0.725 to 0.750) versus 0.737 (95% c.i. 0.709 to 0.765)), but greater explainability. The final score included six variables: country income, ASA grade, diabetes, and operative contamination, approach, and duration. Model performance remained good on external validation (AUC 0.730 (95% c.i. 0.715 to 0.744); calibration intercept -0.098 and slope 1.008) and demonstrated superior performance to the external validation of all previous models. Conclusion: The 'Global Surgical-Site Infection' score allows accurate prediction of the risk of surgical-site infections with six simple variables that are routinely available at the time of surgery across global settings. This can inform the use of intraoperative and postoperative interventions to modify the risk of surgical-site infections and minimize associated harm.
UR - http://www.scopus.com/inward/record.url?scp=85200890886&partnerID=8YFLogxK
U2 - 10.1093/bjs/znae129
DO - 10.1093/bjs/znae129
M3 - Article
AN - SCOPUS:85200890886
SN - 0007-1323
VL - 111
JO - British Journal of Surgery
JF - British Journal of Surgery
IS - 6
M1 - znae129
ER -