Risk prediction models for postoperative pneumonia in lung cancer patients: a systematic review and meta-analysis

This is the authors' abstract. We don't add key points for this paper.

Background Postoperative pneumonia (POP) is a common complication after lung cancer surgery. Numerous risk prediction models for POP have been developed. However, the quality and predictive performance of these models remain unclear. Methods A literature search was carried out in CNKI, Wanfang Database, VIP Database, SinoMed, PubMed, Embase, the Cochrane Library, CINAHL, and Web of Science. The search covered studies on risk prediction models for postoperative pneumonia in lung cancer patients from inception to July 20, 2026. The research question was formulated using the PICOTS framework. Data extraction was performed following the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS), and risk of bias and applicability were assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST). Meta-analysis was conducted using STATA 17.0 software. Results A total of 26 prediction models from 20 studies were included. Models were developed predominantly with logistic regression, with machine-learning approaches used in two studies. Common predictors included age, operation time, smoking, diabetes mellitus, preoperative pulmonary function, chronic obstructive pulmonary disease (COPD), and TNM stage. All studies exhibited a high risk of bias, mainly due to inadequate data sources and suboptimal reporting in the analysis domain; 16 studies also raised applicability concerns, mainly due to participant selection limited to specific lung cancer subgroups. The pooled AUC from the meta-analysis was 0.89 (95% CI: 0.87-0.91). Conclusion Although existing postoperative pneumonia prediction models for lung cancer patients show good discriminative performance, their clinical applicability is limited by a high risk of bias, which may lead to overestimation of discrimination. Future studies should develop more robust models with larger sample, rigorous designs, and multicenter external validation. Systematic review registration https://www.crd.york.ac.uk/prospero/ , identifier CRD42024554224.