Prioritizing Patient Selection in Clinical Trials: A Machine Learning Algorithm for Dynamic Prediction of In-Hospital...

Area of Science:

• Critical Care Medicine
• Machine Learning
• Biostatistics

Background:

• Intensive Care Unit (ICU) patient selection for clinical trials faces challenges.
• Existing scoring systems like SAPS have limitations in predicting mortality.
• Red blood cell Distribution Width (RDW) shows potential as a mortality predictor.

Purpose of the Study:

• To develop and evaluate a machine learning prognostic scoring system for ICU mortality.
• To assess the performance of the new algorithm against existing methods.
• To identify key predictors of ICU mortality for improved patient stratification.

Main Methods:

• Developed a Mixed-effects logistic Random Forest for binary data (MixRFb) algorithm.
• Integrated Random Forest (RF) classification with a mixed-effects model.
• Compared MixRFb performance against RF and SAPS-based scoring using receiver operating characteristic curves.

Main Results:

• MixRFb achieved a higher area under the curve (0.882) compared to SAPS-based scoring (0.814).
• Age and RDW were identified as the most significant predictors of ICU mortality.
• The algorithm demonstrated superior efficacy in predicting in-hospital mortality.

Conclusions:

• The MixRFb algorithm offers improved prediction of ICU mortality.
• Age and RDW are key factors in predicting mortality in ICU patients.
• This tool can enhance clinical trial patient selection, improving outcomes and ethics.