Abstract: PURPOSE: To explore the risk factors for postoperative pulmonary infection in patients with oral cancer and construct a risk prediction model based on the POSSUM score. METHODS: A retrospective cohort study was conducted, enrolling 201 patients with oral cancer who underwent surgical treatment at the First Affiliated Hospital of Bengbu Medical University from January 2021 to December 2023. The patients were divided into infection group and non-infection group according to the occurrence of postoperative pulmonary infection. The POSSUM scoring system was used to evaluate the perioperative risks of the patients. Potential risk factors were screened out by univariate analysis, and independent risk factors were identified by multivariate logistic regression analysis. The nomogram prediction model was established using the rms package in R language. The performance of the model was verified by the receiver operating characteristic (ROC) curve, calibration curve and Hosmer-Lemeshow test. RESULTS: Among the 201 patients, totally 36 cases developed postoperative pulmonary infection and 165 cases had no infection, with the incidence rate of postoperative pulmonary infection being 17.9%. Gram-negative bacteria were the main pathogenic bacteria. Statistically significant differences were observed between the infection group and non-information group in terms of tracheotomy status, postoperative antibiotic application, length of hospital stay, albumin level, operation time and POSSUM score (P<0.05). Binary multivariate logistic regression analysis showed that tracheotomy status, length of hospital stay and POSSUM score were independent risk factors for postoperative pulmonary infection in patients with oral cancer (P<0.05). The area under the ROC curve (AUC) of the risk prediction model constructed based on the above risk factors was 0.878, which was higher than the AUC values of tracheotomy status, length of hospital stay and POSSUM score alone. Meanwhile, the model had high credibility in the calibration verification curve. The Hosmer-Lemeshow goodness-of-fit test indicated that the nomogram model had good consistency in predicting postoperative pulmonary infection in oral cancer patients (χ²=4.064, P=0.851; df=8). CONCLUSIONS: The nomogram model constructed based on tracheotomy status, length of hospital stay and POSSUM score has good predictive performance for postoperative pulmonary infection in oral cancer patients. It can provide a quantitative tool for clinical early identification of high-risk patients and formulation of individualized prevention and control strategies.

Key words: Oral cancer, Pulmonary infection, POSSUM score, Nomogram, Risk prediction