Machine Learning-based Prediction of Unplanned Acute Care in Outpatients Receiving S-1 Chemotherapy
Misaki Teramoto1, Tsubura Noda2, Kenji Kawasumi3
1Faculty of Pharmaceutical Science, Tokyo University of Science, Tokyo, Japan.
In Vivo (Athens, Greece)
|June 30, 2026
Summary
Machine learning accurately predicts unplanned acute care (UAC) in patients receiving S-1 chemotherapy. This approach improves early risk identification, enhancing safety for outpatient cancer treatment.
Area of Science:
- Oncology
- Medical Informatics
- Machine Learning in Healthcare
Background:
- Oral anticancer agents are increasingly used in outpatient settings.
- Treatment-related adverse events necessitate unplanned acute care (UAC).
- Early identification of high-risk patients is crucial for managing UAC.
Purpose of the Study:
- To develop a machine learning model for predicting UAC in S-1 chemotherapy outpatients.
- To compare the performance of machine learning models against traditional logistic regression.
- To improve the safety and management of outpatient chemotherapy.
Main Methods:
- Retrospective study of 579 outpatients receiving S-1 therapy.
- Development of predictive models using logistic regression and Support Vector Machine (SVM).
- Evaluation using recall-oriented metrics, with F2 score as the primary measure; SHAP for feature selection.
Main Results:
- The SVM model showed superior performance in predicting UAC compared to logistic regression.
- SVM achieved higher recall (0.769 vs. 0.615) and F2 score (0.368 vs. 0.325) on the test dataset.
- SHAP analysis aided in feature selection and model interpretation.
Conclusions:
- A Support Vector Machine (SVM)-based machine learning model effectively predicts UAC in S-1 chemotherapy outpatients.
- The model reduces false-negative predictions, aiding in early risk stratification.
- This approach enhances the safety of outpatient chemotherapy by identifying high-risk patients.