Machine Learning-based Mortality Prediction Models for Cancer Patients Using Electronic Health Record Data: A Systematic Review and Critical Appraisal (Preprint)

Abstract

In the U.S., national guidelines suggest that aggressive cancer care should be avoided in the final months of life. However, guideline compliance currently requires clinicians to make judgments based on their experience as to when a patient is nearing the end of their life. Machine learning (ML) algorithms may facilitate improved end-of-life care provision for cancer patients by identifying patients at risk of short-term mortality.

This study sought to summarize the evidence in applying ML in 1-year or shorter cancer mortality prediction for assisting with the transition to end-of-life care for cancer patients.

We searched MEDLINE, Embase, Scopus, Web of Science, and IEEE to identify relevant articles. We included the studies describing ML algorithms predicting 1-year or shorter mortality for oncology patients. We used the prediction model risk of bias assessment tool to assess quality of included studies.

We included 15 articles involving 110,058 patients in the final synthesis. Twelve studies have a high or unclear risk of bias. Model performance was good: area under the receiver operating characteristic curve ranged from 0.72 - 0.92. We identified common issues leading to biased models, including using a single performance metric, incomplete report of or inappropriate modeling practice, and small sample size.

We found signs of encouraging ML performance in predicting short-term cancer mortality. Nevertheless, no included ML algorithms are suitable for clinical practice at the current stage due to the high risk of bias and uncertainty regarding real-world performance. Further research is needed to develop ML models using modern standards of algorithm development and reporting.