Abstract: Objective: To analyze the risk factors of Hospital-Acquired Pneumonia in patients with Traumatic Brain Injury, and to construct a risk prediction model. Methods: The clinical data of 596 patients with Traumatic Brain Injury hospitalized in a tertiary grade A hospital in Guizhou province from January 1st, 2019 to July 31st, 2021 were analyzed retrospectively. Five machine learning algorithms were used to construct risk prediction models, including Logistics Regression, Naive Bayes, Support Vector Machine, Kashin Nearest Neighborg and Multi-layer Representation. The accuracy, recall, F1 value, AUC value were performed to evaluate and compare the model. Results: The incidence of Hospital-Acquired Pneumonia was 34.90%. Among the five models, Multi-layer Representation model had higher accuracy, recall, F1 value and AUC value. Conclusion: Multi-layer Representation model is effective and suitable for the early prediction of Hospital-Acquired Pneumonia in patients with Traumatic Brain Injury. It is expected to provide reference for diagnosis, treatment and prevention strategies for patients with Traumatic Brain Injury.

Key words: Traumatic Brain Injury; Hospital-Acquired Pneumonia; influencing factor; machine learning; predictive model