Home-based primary care for veterans with dementia was associated with less emergency department use in an instrumental-variable analysis

A cohort analysis using an instrumental-variable approach examined 133,846 veterans aged 65 and over with dementia in the Veterans Health Administration between 2021 and 2023, of whom 17.36% (23,240) received home-based primary care Primary Study (Miller et al., PMID 42853183). The outcome was emergency department or urgent care use. Home-based primary care was defined as starting in the first month a veteran had two or more claims within 91 days.

Enrolees differed from non-enrolees: they were older, more likely to be female, Hispanic or Black or African American, to live in a metropolitan area, to have post-traumatic stress disorder or traumatic brain injury, and to have more comorbid conditions. Their raw emergency department use was higher (7.1% against 5.9%). To address selection into the programme the authors used lagged programme penetration as the instrument (F-statistic above 100), and on that basis enrolees were 3.60 percentage points less likely to use emergency or urgent care (P<0.001).

The headline estimate therefore depends on the instrument and not on the raw comparison, which ran in the opposite direction. The outcome is service use, not days at home, patient experience or carer burden, and the Veterans Health Administration is an integrated payer-provider system, so the estimate cannot be carried to the NHS without an NHS comparison.

It matters for the standing record because an outcome did move, in a real system, in a large dementia population. That qualifies the published pattern that process measures move while patient-outcome measures are flat, untested or unmeasured: here a service-use outcome fell, though it remains a measure of use and not of how the person’s time was spent.