Integrating Machine Learning into Care Coordination Pathways: A Framework for Long-Term Care Facilities

Background:
Long-term care facilities (LTCFs) face persistent care coordination challenges as nursing, recreational therapy, social work, and rehabilitation teams work with shared resident populations using fragmented, discipline-specific workflows. While prior studies have explored isolated machine learning (ML)
applications in long-term care, there is currently no systematic, workflow-integrated conceptual framework that explains how ML can be leveraged to support interdisciplinary care coordination.

Objective:
This paper proposes a conceptual, design-oriented framework for integrating ML-based decision support into interdisciplinary care coordination pathways in LTCFs.

Framework:
The proposed framework is non-empirical and comprises five interconnected components: (1) a data integration layer designed to accommodate routinely available clinical, functional, and psychosocial information across disciplines; (2) an ML analytics layer that outlines appropriate classes of models rather than prescribing specific algorithms, selected to align with interpretability and accountability requirements in long-term care; (3) role-specific interfaces translating analytical outputs into interpretable, actionable decision support for care teams; (4) communication mechanisms enabling coordinated interdisciplinary responses; and (5) governance and feedback processes that conceptually address data quality oversight, fairness considerations, and ethical deployment. The framework emphasizes human–AI collaboration.

Application:

An illustrative (non-empirical) use case demonstrates the framework's application to the identification of social isolation risk. The data integration layer consolidates routinely available resident information. The ML analytics layer applies explainable classification models to identify at-risk residents based on declining participation patterns and limited social engagement. Role-specific interfaces present recreational therapists with prioritized resident lists and interpretable risk factors. Communication mechanisms enable coordinated case review, targeted social programming, and proactive intervention.

Implications:
By explicitly linking ML outputs to interdisciplinary decision-making processes, this framework provides LTCFs with a systematic roadmap for leveraging ML to shift care coordination from reactive to proactive. The anticipated benefits include improved decision timeliness, enhanced coordination across disciplines, and more efficient alignment of therapeutic resources while preserving human oversight and ethical accountability.