What real-world EHR studies can and cannot answer
11 September 2026 · Cohira team
A federated EHR network queries de-identified records across hundreds of healthcare organisations at once. No recruitment, no waiting for follow-up: the events have already happened.
What it does well
Comparative effectiveness and safety questions between treatments used in routine care. Propensity-score matching builds two comparable cohorts in hours, and outcomes arrive with real-world denominators behind them.
What it cannot do
It cannot randomise. Confounding by indication never fully disappears, and coding quality varies by site. A good real-world study says so, tests its assumptions with sensitivity analyses, and frames its conclusion as association where that is what the design supports.
What reviewers look for
A pre-specified protocol, a matching diagnostic table, clearly reported attrition, and honest limitations. Reported to STROBE or RECORD, that combination survives peer review.