Machine Learning-Based Prediction of 1-Year Survival Using Subjective and Objective Parameters in Patients With...

Area of Science:

• Oncology
• Palliative Care
• Machine Learning

Background:

• Palliative care is crucial for cancer patients with limited life expectancy.
• Machine learning (ML) can enhance survival outcome prediction in oncology.
• Identifying patients who benefit most from palliative care is essential.

Purpose of the Study:

• To evaluate the predictive power of objective and subjective self-reported variables for 1-year cancer mortality.
• To explore the importance of variables from electronic health records and patient self-assessments.
• To compare the efficacy of different machine learning models in predicting mortality.

Main Methods:

• Utilized data from 265 advanced cancer patients (April 2020-March 2021).
• Collected objective clinical data and subjective patient-reported outcomes.
• Employed logistic regression, decision trees, and random forests with 20-fold cross-validation.
• Analyzed performance using ROC-AUC and PR-AUC metrics.

Main Results:

• Objective clinical variables demonstrated superior predictive performance (LR: 0.81 ROC-AUC, 0.72 F1 score).
• Subjective patient-reported variables showed lower predictive accuracy (LR: 0.55 ROC-AUC, 0.52 F1 score).
• Machine learning models highlighted the greater importance of objective data.

Conclusions:

• Objective variables are significantly more predictive of 1-year mortality than subjective patient-reported variables.
• Subjective burden, as measured in this study, is not a reliable predictor of cancer patient survival.
• Further research is warranted to refine ML models for mortality prediction using patient-reported data.