Factors associated with respite care use among veteran caregivers: a machine learning analysis

Respite care provides temporary relief to family caregivers yet remains underused, and the factors shaping its utilization among Veteran caregivers are not well understood. This evaluation examined caregiver‑ and Veteran‑specific characteristics associated with respite care use within the Department of Veterans Affairs (VA) Caregiver Support Program's Program of General Caregiver Support Services (PGCSS).

We analyzed survey and administrative data from 1,727 caregivers of Veterans enrolled in PGCSS who completed baseline surveys between 2018 and 2021. Caregivers were predominantly female (96%) with a mean age of 62 years; Veterans averaged 70 years. Respite use within 2 years of survey completion was identified through linked VA data. Guided by Andersen's Healthcare Utilization Model, 34 caregiver and Veteran variables were evaluated with random forest models to identify characteristics that most strongly differentiated respite users from non‑users.

Respite care was used by 23.5% of caregivers. Use was more common among older caregivers and Veterans (predisposing factors), among caregivers reporting greater burden, depression, or financial strain and Veterans with higher frailty, functional limitation, or dementia (need factors), and among caregivers perceiving stronger communication and collaboration with the clinical team (enabling factors). Model performance was strong (testing accuracy = 0.79 with all variables; 0.77 with the top 15), and results remained consistent in a sensitivity analysis limited to caregivers of Veterans who survived the 2‑year follow‑up period.

Both caregiving intensity and care‑recipient complexity characterize respite use even within a system of broad service availability. Findings provide a foundation for future hypothesis‑driven studies and inform efforts to align respite programs more closely with caregiver-Veteran needs.