Supervised Learning Algorithm Comparison in Discharge Status Prediction of Trauma Patients: Empirical Evaluation.

Area of Science:

• Medical Informatics
• Data Science in Healthcare
• Trauma Care Research

Background:

• Trauma center data analysis identifies key performance indicators affecting hospital budgets.
• Improved performance indicators can lead to reduced mortality rates and enhanced patient health outcomes.
• Identifying growth opportunities within trauma care is crucial for hospital financial health.

Purpose of the Study:

• To identify the optimal supervised machine learning algorithm for predicting trauma patient discharge status.
• To evaluate the efficacy of various algorithms in a clinical setting.

Main Methods:

• Retrospective study utilizing data from the Kashan Trauma Registry (March 2018 - February 2019).
• Evaluation of supervised algorithms: Naive Bayes, Logistic Regression, Support Vector Machine, Random Forest, and K-Nearest Neighbors.
• Performance assessment using accuracy, precision, recall, and F-measure metrics with a hold-out validation technique.

Main Results:

• The Random Forest algorithm demonstrated superior performance compared to other evaluated algorithms.
• Random Forest achieved the highest accuracy (84.6%), precision (79.6%), recall (76.8%), and F-measure (76.20%) using information gain.
• Specific performance metrics varied slightly based on the chosen metric (Gini index vs. information gain).

Conclusions:

• Supervised algorithms, when appropriately configured, can effectively aid in diagnosing trauma patient discharge status.
• Data balancing techniques show potential for improving algorithm performance in this context.
• Generalizability of findings is limited and depends on algorithm choice and parameter tuning.