Risk factors for postoperative pulmonary infection in patients with non-small cell lung cancer: analysis based on regression models and construction of a nomogram prediction model.

OBJECTIVE

This study was designed to analyze risk factors for postoperative pulmonary infection (PPI) in patients with non-small cell lung cancer (NSCLC) based on regression models and to construct a corresponding nomogram prediction model.

METHODS

A total of 244 patients with NSCLC who received surgical treatment from June 2015 to January 2017 were retrospectively analyzed. According to the PPI, they were assigned to a pulmonary infection group (n=27) or non-pulmonary infection group (n=217). The independent risk factors for PPI in NSCLC patients were screened through least absolute shrinkage and selection operator (LASSO) and logistic regression analysis, and a corresponding nomogram prediction model was constructed.

RESULTS

A total of 244 NSCLC patients were included, including 27 with PPI (11.06%). According to LASSO regression-based screening, age, diabetes mellitus (DM), tumor node metastasis (TNM) staging, chemotherapy regimen, chemotherapy cycle, post-chemotherapy albumin (g/L), pre-chemotherapy KPS and operation time were all significant and found to be the influencing factors for PPI. The risk model constructed based on LASSO was 0.0035770333 + (0.0020227686* age) + (0.057554487* DM) + (0.016365428* TNM staging) + (0.048514458* chemotherapy regimen) + (0.00871801* chemotherapy cycle) + (-0.002096683* post-chemotherapy albumin (g/L) + (-0.00090206* pre-chemotherapy Karnofsky performance score (KPS)) + (0.000296876* operation time). The pulmonary infection group got significantly higher risk scores than the non-pulmonary infection group (P<0.0001). According to receiver operating characteristic (ROC) curve-based analysis, the area under the curve (AUC) of risk score in predicting pulmonary infection was 0.894. Based on 4 independent predictors, a risk-prediction nomogram model was constructed to predict pulmonary infection in NSCLC patients after surgery. The internal verification C-index was 0.900 (95% CI: 0.839-0.961), and the calibration curves were well fitted with the ideal ones.

CONCLUSION

The prediction model based on a regression model for PPI in NSCLC patients demonstrates good prediction efficiency, which is conducive to early screening of high-risk patients and further improvement of treatment regimen.