Risk prediction model for surgical site infection in patients with gastrointestinal cancer: a systematic review and meta-analysis.

backgroundCurrently, various risk prediction models for surgical site infection (SSI) in patients with gastrointestinal tumors have been developed, but comprehensive comparisons regarding the model construction process, performance, and data sample bias are lacking. This study conducts a systematic review of relevant research to evaluate the risk bias and clinical applicability of these models. MATERIALS AND

methodsThe Web of Science, PubMed, Cochrane Library, Embase, CINAHL, CBM, CNKI, Wanfang, and VIP databases were searched for studies related to SSI prediction models in gastrointestinal cancer patients published up to August 19, 2024. Two researchers independently screened the literature, extracted the data, and evaluated the quality. A meta-analysis was conducted on the common predictive factors included in the model, using odds ratio (OR) values and 95% confidence interval (CI) as effect statistics. The Q test and heterogeneity index I

resultsA total of 28 articles were included, and 39 models were constructed. The area under the receiver operating characteristic curve (AUC) for the models ranged from 0.660 to 0.950, indicating good predictive performance. Eight studies conducted internal validation, eight studies conducted external validation, and two studies used a combination of internal and external validation for model evaluation. The overall risk of bias in the literature was high, but the applicability was good. The results of the meta-analysis revealed that factors such as underlying diseases, surgical factors, demographic factors, and laboratory-related indicators are the main predictors of surgical site infections in patients with gastrointestinal tumors.

conclusionsCurrently, risk prediction models for surgical site infections in patients with gastrointestinal cancer remain in the developmental phase, and there is a high risk of bias in the areas of study subjects, outcomes, and analysis. Researchers need to enhance research methodologies, conduct large-scale prospective studies, and refer to the reporting standards of the bias risk assessment tool for predictive models to construct predictive models with low bias risk and high applicability.