Objective To establish and validate a risk prediction model for mechanical ventilator-associated pneumonia (VAP) in older patients with lung cancer.

Methods  A total of 548 older patients with lung cancer were enrolled between January 2022 and May 2025. Subsequently, the patients were divided into a training set (n = 384) and a validation set (n = 164) at a ratio of 7∶3. According to whether the patients developed VAP, the training set was further divided into a non-VAP subgroup (n = 231) and a VAP subgroup (n = 153), and the validation set was divided into a non-VAP subgroup (n = 99) and a VAP subgroup (n = 65). Multivariate stepwise logistic regression was conducted to identify the predictors of VAP in older patients with lung cancer, and a nomogram model was established accordingly. The model was validated using the receiver operating characteristic (ROC) curve and the calibration curve.

Results  A total of 153 participants (39.84%) in the training set and 65 participants (39.63%) in the validation set developed VAP. Univariate analysis revealed that such factors as age, lung cancer stage, mechanical ventilation duration, Acute Physiologic Assessment and Chronic Health Evaluation (APACHE) Ⅱ score, procalcitonin (PCT), pentraxin 3 (PTX3), retinol-binding protein (RBP), soluble triggering receptor expressed on myeloid cells-1 (STREM-1), and human cartilage glycoprotein-39 (YKL-40) were associated with VAP in older lung cancer patients (P < 0.05), and there was no multicollinearity. Multivariate logistic stepwise regression revealed that age (odds ratio OR = 1.182, 95% CI: 1.014-1.377), mechanical ventilation duration (OR = 3.929, 95% CI: 1.374-11.233), APACHE Ⅱ (OR = 1.770, 95% CI: 1.296-2.416), PTX3 (OR = 1.019, 95% CI: 1.007-1.030), RBP (OR = 1.150, 95% CI: 1.083-1.260), STREM-1 (OR = 1.168, 95% CI: 1.083-1.260) were predictors of VAP in older lung cancer patients (P < 0.05). A nomogram model was constructed with the following equation: Prob = 1/(1 + e^－Y), Y = －43.147 + 0.167 × age + 1.368× mechanical ventilation duration + 0.571 × APACHE Ⅱ score + 0.019 × PTX3 + 0.140 × RBP + 0.155 × STREM-1. The area under the curve (AUC) of the ROC in the training set was 0.909 (95% CI: 0.878-0.940), and the AUC of the validation set was 0.843 (95% CI: 0.776-0.910). The average absolute error tested by the Bootstrap method was 0.03, indicating that the model showed a certain level of consistency during both the training and validation processes.

Conclusion  Age, mechanical ventilation duration, APACHE Ⅱ score, and serum levels of PTX3, RBP, and STREM-1 are all predictors for the occurrence of VAP in older lung cancer patients. The nomogram model constructed based on these factors demonstrates acceptable predictive performance.