Prediction of risk factors for pulmonary infection complication in esophageal cancer and chemoradiotherapy patients based on random forest algorithm

OBJECTIVE To explore the risk factors for pulmonary infection (PI) complication in patients with esophageal cancer (EC) undergoing chemoradiotherapy and construct a random forest prediction model on which the prevention and treatment strategies are formulated. METHODS The clinical data were retrospectively collected from 227 EC and chemoradiotherapy patients who were treated in the First Affiliated Hospital of Nanchang University from May 2020 to Dec. 2023. The patients were divided into the infection group with 46 cases and the non-infection group with 181 cases according to status of PI complication. The risk factors were screened out and analyzed by means of multivariate logistic regression model, and the risk prediction model was constructed based on the risk factors by using R software (4.3. 2 version). RESULTS The result of multivariate logistic regression analysis showed that smoking, age, malnutrition and tumor location were the risk factors for the PI complication in the EC and chemoradiotherapy patients(P＜0.05). The result of construction of the random forest model indicated that the factors affecting the PI complication, ranked in descending order of importance, were as follows: tumor location, smoking, age and malnutrition. The AUC of the random forest prediction model was 0.774(95%CI:0.714 to 0.827), while the AUC of the logistic regression model was 0.743(95%CI:0.681 to 798); as compared with the AUC between the two models, the result showed that Z=1.981,P=0.0476. CONCLUSIONS The tumor location, smoking, advanced age and malnutrition are the independent risk factors for the PI complication in the EC and chemoradiotherapy patients. The random forest prediction model has high accuracy and can provide theoretical bases for formulating targeted prevention and treatment strategies so as to reduce the incidence of PI.