Nutritional management adherence via an ePRO platform in patients with cancer: a machine learning model studyResearch in context
Summary: Background: Electronic patient-reported outcome (ePRO) systems have significant potential for providing individualized and continuous nutritional management for patients with cancer. However, adherence to ePRO-guided nutritional interventions varies significantly, and the key factors assoc...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-07-01
|
| Series: | EClinicalMedicine |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2589537025002627 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Summary: Background: Electronic patient-reported outcome (ePRO) systems have significant potential for providing individualized and continuous nutritional management for patients with cancer. However, adherence to ePRO-guided nutritional interventions varies significantly, and the key factors associated with adherence remain poorly understood. This study aimed to assess adherence to nutritional targets via ePRO platforms and identify predictors influencing adherence to optimize management strategies. Methods: This multicenter, prospective longitudinal cohort study included 8268 patients with cancer from March 2021 to April 2024 (registration number: ChiCTR2100047535). Adherence was defined as the ratio of actual to prescribed intake for both total energy (TEI) and total protein (TPI). Adherence to TEI and TPI targets was monitored throughout the study period based on prescriptions generated by ePRO-guided nutritional management. The proportion of actual/prescribed intake <60% was set as low adherence. Explainable machine learning models were used to identify predictive features, with SHapley Additive exPlanation (SHAP) analysis ranking variable importance. Two-way fixed-effect logistic regression models were applied to further assess longitudinal predictors of adherence. Findings: Among 8268 patients (median age, 61 years; 61.9% male, 38.1% female), 2727 (33.0%) and 3332 (40.3%) failed to meet TEI and TPI targets, respectively. The LightGBM model achieved superior predictive performance (area under the receiver operating characteristic curve: TEI = 0.861, TPI = 0.821). Key predictors of lower adherence to both TEI and TPI included advanced TNM stage (TEI: odds ratio [OR] = 1.18, 95% CI: 1.11–1.26; TPI: OR = 1.39, 95% CI: 1.27–1.53), poorer Eastern Cooperative Oncology Group performance status (TEI: OR = 1.18, 95% CI: 1.11–1.26; TPI: OR = 1.39, 95% CI: 1.27–1.53), higher Patient-Generated Subjective Global Assessment scores (TEI: OR = 1.08, 95% CI: 1.08–1.09; TPI: OR = 1.08, 95% CI: 1.07–1.10), elevated platelet counts (TEI: OR = 1.01, 95% CI: 1.00–1.01; TPI: OR = 1.01, 95% CI: 1.00–1.01), walking time <60 min/day (TEI: OR = 2.42, 95% CI: 2.18–2.69; TPI: OR = 2.59, 95% CI: 2.19–3.06), sleep duration <8 h/day (TEI: OR = 1.48, 95% CI: 1.25–1.76; TPI: OR = 1.41, 95% CI: 1.29–1.52), and nausea (TEI: OR = 1.32, 95% CI: 1.23–1.41; TPI: OR = 1.44, 95% CI: 1.37–1.51). Conversely, factors associated with higher adherence included female sex and higher levels of serum albumin, alanine transaminase, and glucose. Interpretation: In this large multicenter study, over one-third of patients failed to meet ePRO-guided nutritional targets, highlighting substantial challenges in adherence. Further, we identified key predictors associated with low adherence to ePRO-guided nutritional management in patients with cancer, which might help identify at-risk patients and guide future research. Funding: Wu Jieping Medical Foundation (320.6750.2021-02-22). |
|---|---|
| ISSN: | 2589-5370 |