Swift Development of a No-Code AI Model for Predicting Hospitalization Duration in Burn Patients Using Musculoskeletal System Data

Epidemiological studies show that burn victims often have extended hospital stays (LOS), affecting individuals and society significantly. Accurate LOS prediction is crucial for optimizing hospital management and reducing costs. This study introduces a machine learning approach to predict LOS for burn patients using diagnostic records, employing five algorithms: Logistic Regression, LightGBM, XGBoost, CatBoost, and random forest. The model predicts LOS across various timeframes and evaluates performance across different demographics. Logistic Regression achieved an F1-score of 0.72 and an AUROC of 0.84 for predicting stays over 14 days. Other predictions yielded an F1-score of 0.61 and RMSE of 0.93 and 6.11. A no-code training platform reduced model training time by 95%. The study demonstrates the effectiveness of machine learning and no-code AI platforms in predicting LOS, improving treatment outcomes, and optimizing hospital resources.