Objective:

To evaluate various machine learning (ML) and deep learning (DL) algorithms for predicting one-year mortality among elderly hip fracture patients, and to investigate the effectiveness of an automated ML platform in enhancing model development and evaluation.

Key Findings:

• Manually optimised Extreme Gradient Boosting (XGB) algorithm showed superior predictive performance (AUC = 0.846), indicating strong predictive capability.
• Automated ML model via LLM and TPOT frameworks demonstrated comparable performance (AUC = 0.844) with higher recall but lower precision, suggesting a trade-off in model characteristics.
• Key predictors of one-year mortality included baseline serum albumin and urea levels, patient age, intraoperative hypothermia, and number of chronic diseases.

Interpretation:

ML-based models, particularly the XGB algorithm, significantly enhance predictive accuracy for one-year mortality among elderly hip fracture patients, while automated ML frameworks offer a practical alternative for clinicians with limited technical expertise.

Limitations:

• Retrospective design may introduce biases that could affect the reliability of the findings.
• Single-center study limits generalizability, suggesting the need for multi-center validation.

Conclusion:

The study highlights the potential of ML and automated ML approaches in improving mortality risk prediction for elderly hip fracture patients.