Newman-Toker · BMJ quality & safety 2024 · cross-sectional epidemiological modeling study · n=21.5 million hospital discharges and registry data

Burden of serious harms from diagnostic error in the USA.

Cited 211 times in the scientific literature.

Level 4 - case-series / case-control

Cross-sectional epidemiological modeling study combining observational discharge/registry data with literature-derived rates

PubMed 37460118 · doi:10.1136/bmjqs-2021-014130 · record verified 2026-08-29

What was done

This cross-sectional modeling study estimated the annual US burden of serious misdiagnosis-related harms (permanent morbidity and mortality). Annual incident vascular events and infections were derived from 21.5 million sampled US hospital discharges (2012–2014), and incident cancers were extracted from 2014 US-based cancer registries. Disease-specific incidences for 15 major conditions ('Big Three': vascular, infections, cancer) were multiplied by literature-derived diagnostic error and harm rates, with uncertainty ranges evaluated using Monte Carlo simulations and sensitivity analyses.

What was found

Annual US incidence was estimated at 6.0 million vascular events, 6.2 million infections, and 1.5 million cancers. Across these categories, weighted mean diagnostic error and serious harm rates were 11.1% and 4.4%, respectively. Extrapolated across all conditions, total US serious harms were estimated at 795,000 annually (plausible range: 598,000–1,023,000; conservative sensitivity analysis: 549,000). Fifteen dangerous diseases accounted for 50.7% of all serious harms, and the top five (stroke, sepsis, pneumonia, venous thromboembolism, and lung cancer) accounted for 38.7%.

Why it matters

This study provides a rigorous, comprehensive national estimate of preventable death and permanent disability caused by misdiagnosis. Because approximately half of these serious harms are concentrated in just 15 conditions, quality improvement efforts have specific, tractable targets.

Limits

The findings depend on mathematical modeling and extrapolation from historical literature-based error rates rather than direct prospective clinical audits. The primary datasets reflect 2012–2014 administrative and registry data, which may not capture recent changes in diagnostic technologies or clinical workflows.

Cited by