Related Experiment Videos
Predicting postoperative infections: prediction models and their evaluation: A systematic review
Nikki de Mul1, Katja M Scheffer-Wesdorp, Diede Verlaan
1From the Department of Intensive Care, University Medical Center Utrecht, Utrecht University, Utrecht, The Netherlands (NdM, KMSW, DV, LPGD, OLC, LMV), Department of Anaesthesiology, University Medical Center Utrecht, Utrecht University, Utrecht The Netherlands (NdM, WJMS, LMV), Department of Anaesthesiology, Cantonal Hospital Aarau, Switzerland (WJMS), Julius Center for Health Sciences and Primary Care, UMC Utrecht, Utrecht University, Utrecht, the Netherlands (MJMB, LPGD), European Clinical Research Alliance on Infectious Diseases (Ecraid), Utrecht, The Netherlands (MJMB, LPGD), Department of Anesthesiologie, Intensive Care and Pain Medicine, St Antonius Hospital, Nieuwegein, The Netherlands (GM, LMV).
This review found 267 models to predict postoperative infections, but none are implemented. The most validated models, ACS SRC and SURPAS, show limited clinical utility due to poor performance and study quality.
Area of Science:
- Medical Informatics
- Surgical Risk Assessment
- Predictive Modeling in Healthcare
Background:
- Pre-operative assessment is crucial for identifying patients at risk of postoperative infections.
- Personalized patient management relies on accurate infection risk prediction.
- Existing prediction models require evaluation for validation and implementation status.
Purpose of the Study:
- To systematically review and evaluate existing multivariable prediction models for postoperative infections.
- To assess the validation status and implementation of these predictive models.
- To identify the most common predictors and models used in postoperative infection risk assessment.
Main Methods:
- Systematic search of PubMed, Embase, and Cochrane Library databases.
- Inclusion of studies on the development, validation, and implementation of models using pre-operative predictors.
- Focus on elective, major noncardiac, non-intracranial surgery with 30-day infection risk.
Main Results:
- 267 distinct models were developed across 88 studies; 88 studies assessed model validity through 314 analyses.
- No studies reported model implementation.
- Surgical site infections (SSI), pneumonia, and general infections were predominantly predicted. Age, sex, and ASA score were common predictors.
- The American College of Surgeons Surgical Risk Calculator (ACS SRC) and SUrgical Risk Pre-operative Assessment System (SURPAS) were most frequently validated.
- Reported c-statistics for ACS SRC and SURPAS indicated variable and often suboptimal predictive performance.
- Most validation studies exhibited a high risk of bias.
Conclusions:
- While ACS SRC and SURPAS are the most validated models for predicting postoperative infections, their clinical utility is constrained.
- Poor and inconsistent predictive performance, coupled with low methodological quality in validation studies, limits the real-world applicability of these models.
- Further research is needed to develop and validate robust, implementable models for accurate postoperative infection risk prediction.