Public-oriented personalized long-term care service recommendation using machine learning
6 affiliations
- 1Department of Software Engineering and Management, National Kaohsiung Normal University, Kaohsiung, Taiwan
- 2Institute of Population Health Sciences, National Health Research Institutes, Miaoli, Taiwan
- 3National Center for Geriatrics and Welfare Research, National Health Research Institutes, Yunlin, Taiwan
- 4Graduate Institute of Biomedical Sciences, China Medical University, Taichung City, Taiwan
- 5Department of Risk Management and Insurance, Tamkang University, New Taipei City, Taiwan
- 6School of Public Health, College of Public Health, Taipei Medical University, Taipei, Taiwan
Abstract
Background: Long-term care prediction models often rely on large multidomain assessments, which can limit interpretability and reuse across multiple service outcomes.
Objective: To determine whether a common reduced predictor set could support prediction of 20 later recorded long-term care service-use outcomes while quantifying performance loss relative to the full set.
Methods: This retrospective cohort study linked initial Care Management Evaluation Form assessments to administrative service records for 9,297 adults from southern Taiwan. The development cohort included 8,367 participants, and the fixed-holdout re-evaluation cohort included 930. We trained logistic regression, decision tree, random forest, extreme gradient boosting, and a shallow artificial neural network. Development-only recursive feature elimination ranked 378 processed predictors and produced nested sets of 31 and 16 predictors. Performance was assessed using the area under the receiver operating characteristic curve, area under the precision-recall curve, Brier score, and log loss.
Results: With all 378 predictors, macro area under the receiver operating characteristic curve ranged from 0.711 to 0.772. The corresponding ranges were 0.706 to 0.745 with 31 predictors and 0.685 to 0.702 with 16 predictors. The 31-predictor set incurred smaller losses than the 16-predictor set. Performance varied by algorithm and service. The artificial neural network did not consistently outperform logistic regression, random forest, or extreme gradient boosting.
Conclusion: The shared 31-predictor set provided the strongest balance between parsimony and discrimination and emerged as the preferred reduced configuration for subsequent geographic, temporal, and prospective evaluation.
Related
- Machine learning classification techniques for smart agriculture: a review of disease detection and crop recommendation
- A recommendation model based on profile enhancement for personalized medical expert in online health communities
- Personalized Boolean network dynamic profiling and machine learning approaches to identify combinatorial treatments for prostate cancer
- Multi-service honeypot framework for high-fidelity cyber-attack data collection and machine learning-based threat detection in cloud environments
- Explainable machine learning classification of rural rescue service proximity coverage using OpenStreetMap data
BibTeX
@article{chien2026publicoriented,
title = {{Public-oriented personalized long-term care service recommendation using machine learning}},
author = {Shuo-Chen Chien and Yu-Hung Chang and Chia-Ming Yen and Ying-Erh Chen and Chia-Chun Liu and Yu-Ping Hsiao and I-Chien Wu and Chih-Cheng Hsu and Hung-Yi Chiou and Ren-Hua Chung},
journal = {Digital Health},
year = {2026},
month = sep,
doi = {10.1177/20552076261491128},
url = {https://doi.org/10.1177/20552076261491128},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - Public-oriented personalized long-term care service recommendation using machine learning
AU - Shuo-Chen Chien
AU - Yu-Hung Chang
AU - Chia-Ming Yen
AU - Ying-Erh Chen
AU - Chia-Chun Liu
AU - Yu-Ping Hsiao
AU - I-Chien Wu
AU - Chih-Cheng Hsu
AU - Hung-Yi Chiou
AU - Ren-Hua Chung
PY - 2026
DA - 2026/09/16
JO - Digital Health
DO - 10.1177/20552076261491128
UR - https://doi.org/10.1177/20552076261491128
AB - Background: Long-term care prediction models often rely on large multidomain assessments, which can limit interpretability and reuse across multiple service outcomes. Objective: To determine whether a common reduced predictor set could support prediction of 20 later recorded long-term care service-use outcomes while quantifying performance loss relative to the full set. Methods: This retrospective cohort study linked initial Care Management Evaluation Form assessments to administrative service records for 9,297 adults from southern Taiwan. The development cohort included 8,367 participants, and the fixed-holdout re-evaluation cohort included 930. We trained logistic regression, decision tree, random forest, extreme gradient boosting, and a shallow artificial neural network. Development-only recursive feature elimination ranked 378 processed predictors and produced nested sets of 31 and 16 predictors. Performance was assessed using the area under the receiver operating characteristic curve, area under the precision-recall curve, Brier score, and log loss. Results: With all 378 predictors, macro area under the receiver operating characteristic curve ranged from 0.711 to 0.772. The corresponding ranges were 0.706 to 0.745 with 31 predictors and 0.685 to 0.702 with 16 predictors. The 31-predictor set incurred smaller losses than the 16-predictor set. Performance varied by algorithm and service. The artificial neural network did not consistently outperform logistic regression, random forest, or extreme gradient boosting. Conclusion: The shared 31-predictor set provided the strongest balance between parsimony and discrimination and emerged as the preferred reduced configuration for subsequent geographic, temporal, and prospective evaluation.
SN - 2055-2076
ER -