Risk prediction models for postoperative intra-abdominal infection in patients with digestive system tumors: a...

Postoperative intra-abdominal infection (IAI) is a severe complication in digestive system tumor patients, increasing hospital stays, costs, and mortality. Accurate prediction enables early intervention and better prognosis. However, existing prediction models lack comprehensive evaluation due to diverse study designs, data sources, and assessment methods. A systematic review is needed to develop a structured prediction model for postoperative IAI in patients with digestive system tumors, and to provide references for the optimization or development of such prediction models in the future.

A computerized search was conducted for relevant studies in PubMed, Web of Science, Embase, Cochrane Library, CINAHL, CNKI, CBM, WanFang Data and VIP databases, with the search time restricted to the establishment of the database to 6 February 2025. Literature screening was performed independently by two researchers and data information was extracted, and the risk of bias and applicability of the model were evaluated using PROBAST.

A total of 22 studies with 9,127 patients were included in the literature. The area under the operating characteristic curve (AUC) of the subjects included in the model ranged from 0.702 to 0.987, and the predictive performance of the model was good in all cases (AUC >0.700). Three cases were internally validated, two cases were externally validated, and two cases were evaluated using a combination of internal and external validation for the model. The most common predictors included length of surgery, comorbid diabetes mellitus, serum albumin level, length of drain retention, and age.

Currently, the prediction model for the risk of postoperative IAI in patients with digestive system tumors is still in the research and development stage. Based on the PROBAST assessment, all studies were considered to have a high risk of bias. Subsequent studies should refer to the reporting guidelines of the PROBAST. Additionally, they should focus on large sample sizes and rigorously designed multicenter external validation to further evaluate the efficacy and feasibility of the models in clinical practice.