Predicting postoperative pulmonary infection in elderly patients undergoing major surgery: a study based on logistic...

Area of Science:

• Medical Informatics
• Surgical Outcomes
• Geriatric Medicine

Background:

• Postoperative pulmonary infection (POI) significantly impacts prognosis, especially in elderly patients undergoing major surgery.
• The predictive capabilities of logistic regression (LR) and machine learning (ML) for POI in high-risk populations require further investigation.

Purpose of the Study:

• To evaluate the predictive performance of LR and ML algorithms for POI in elderly surgical patients.
• To develop a clinical tool for identifying high-risk patients and guiding perioperative management.

Main Methods:

• A retrospective cohort study included 9481 elderly patients undergoing major surgery.
• Features selected by least absolute shrinkage and selection operator regression were used for LR and ML models.
• Random forest analysis was employed for ML model interpretability.

Main Results:

• LR achieved the highest AUC (0.80), outperforming ML models like decision trees (AUC 0.75).
• The LR model demonstrated superior accuracy (88.22%), specificity (90.29%), precision (44.42%), and F1 score (54.25%).
• A web-based nomogram based on LR stratified patients into distinct risk intervals for POI.

Conclusions:

• Logistic regression is more effective than popular ML algorithms for predicting POI in elderly surgical patients.
• The developed nomogram facilitates the identification of high-risk elderly patients and supports perioperative management planning.