Design: Retrospective cohort, linked national VA and Medicare data
Population: 1.41 million male veterans aged ≥65 enrolled in both VA and fee-for-service Medicare
Primary outcome: PSA testing and prostate biopsy rates per 1000 person-years
Effect: Age ≥80: PSA IRR 0.70 (0.69–0.71), biopsy IRR 0.67 (0.60–0.75), VA vs Medicare

This retrospective cohort used linked Veterans Affairs and Medicare data from 2004 to 2023, taking a 30% sample of male veterans aged 65 or over enrolled in both systems — 1.41 million men and 8.67 million person-years. Each man's main primary care setting was assigned over a rolling three-year window.

From 2018 to 2023, among men aged 80 or over, adjusted PSA testing was 187.9 per 1000 person-years in VA primary care against 267.9 in fee-for-service Medicare (IRR 0.70, 0.69–0.71), and biopsy rates were 3.38 against 5.05 (IRR 0.67). In men with under five years' predicted life expectancy, VA testing was also lower (IRR 0.84). In men aged 65 to 69, VA testing was slightly higher (IRR 1.10).

The pattern is what good screening should look like: more testing where benefit is plausible, less where it is not. Even so, about one in five men over 80 in VA care was still tested each year, so low-value testing persists in both systems. The comparison is observational and specific to US care structures.

Before ordering a PSA in a man over 75, estimate his life expectancy: benefit takes about a decade to appear.
Routine PSA screening in men with under 10 years' life expectancy is low-value care.
A raised PSA in a frail older man leads to biopsy, anxiety and treatment with little survival gain.
Clinical decision support that links screening to age and life expectancy is one way to reduce this testing.

Why it matters

The same patients were tested less where the system tied testing to likelihood of benefit.

The statistics, in plain English

An incidence rate ratio of 0.70 means about 30% fewer PSA tests per year of follow-up in VA care. The very narrow intervals reflect the huge sample, not necessarily that the systems are otherwise comparable.