J Nurs Manag. 2026;2026(1):e1006360. doi: 10.1155/jonm/1006360.
ABSTRACT
AIM: To develop and validate a machine learning-based associational prediction model for nurse-led identification of financial toxicity (FT) among stroke patients.
BACKGROUND: FT is increasingly recognized among patients with chronic conditions, yet evidence in stroke patients remains limited. Early identification may help nurses provide timely financial, psychosocial, and discharge-related support.
METHODS: A total of 575 stroke patients were recruited. Based on Health Ecology Theory, factors were grouped into five levels. The dataset was randomly split 7:3 into training and test sets. Least Absolute Shrinkage and Selection Operator (LASSO) was applied for feature selection, and five machine learning models were trained and evaluated on the independent internal test set. The optimal model was interpreted using SHapley Additive exPlanations (SHAP) and evaluated in an external cross-sectional validation dataset of 207 patients from another hospital. A web-based calculator was developed for individualized FT assessment.
RESULTS: The prevalence of FT was 62.3%. The eXtreme Gradient Boosting (XGBoost) model demonstrated the best model performance, achieving an AUC of 0.823 in the internal test set and 0.865 in the external cross-sectional validation set. SHAP analysis based on the XGBoost model identified age, fear of progression, complications, primary caregiver, and out-of-pocket cost as the most important associated factors of FT. A web-based calculator based on the XGBoost model was further developed to support individualized FT assessment in nursing practice.
CONCLUSIONS: The XGBoost model showed good internal and external performance in identifying FT among stroke patients. The web-based XGBoost tool may support nurse-led contemporaneous FT assessment in routine practice. Longitudinal validation is warranted to establish its clinical utility.
IMPLICATIONS FOR NURSING MANAGEMENT: The tool may help nurse leaders embed structured FT assessment into admission and discharge workflows, enabling nurses to identify financially vulnerable stroke patients and initiate timely financial navigation, psychosocial support, and multidisciplinary referral.
PMID:42478054 | DOI:10.1155/jonm/1006360