Machine Learning for Predicting Financial Toxicity in Patients Receiving Complex Gastrointestinal Surgery
Year: 2025
Type: article
Study design: Observational Study
Abstract: Financial toxicity (FT), defined as the economic burden associated with medical care, has emerged as a critical factor affecting patient outcomes. However, this topic remains understudied in the context of complex gastrointestinal surgery for cancer patients. This study aimed to identify predictors of FT and develop a machine learning model capable of preoperatively stratifying patient risk. A dataset of 188 patients who underwent gastrointestinal surgery at Western Michigan Cancer Center during 2016-2023 was analyzed using a range of machine learning techniques. The Functional Assessment of Chronic Illness Therapy Comprehensive Score for Financial Toxicity instrument was used to assess FT, with a score of ≤26 indicating significant financial distress. Multivariable logistic regression was used for baseline analysis, while several classification models including Random Forest, Gradient Boosted Trees, Support Vector Machine, Naïve Bayes, K-Nearest Neighbors, and Neural Networks were evaluated to predict FT. The dataset was then split into training and testing sets using an 80/20 stratified sampling strategy, which ensured class distribution was preserved across both sets. Among all models tested, the Random Forest classifier demonstrated superior performance, achieving the highest accuracy and area under the curve. In addition, feature importance analysis revealed that social well-being, emotional well-being, and age were the most influential predictors of FT. These findings suggest that financial toxicity is not only prevalent but also predictable using readily available clinical and sociodemographic data. Integrating such predictive tools into preoperative workflows may support early intervention strategies aimed at mitigating the adverse effects of financial hardship on patient quality of life.
Institution: Western Michigan University
Language: English
Subfield: Economics and Econometrics
Domain: Social Sciences
Open Access status: hybrid
Estimated APC paid:
OpenAlex ID: W4414350276