Prevalence of harmful diagnostic errors in hospitalised adults: a systematic review and meta-analysis
Abstract
Background: Diagnostic error is commonly defined as a missed, delayed or wrong diagnosis and has been described as among the most important patient safety hazards. Diagnostic errors also account for the largest category of medical malpractice high severity claims and total payouts. Despite a large literature on the incidence of inpatient adverse events, no systematic review has attempted to estimate the prevalence and nature of harmful diagnostic errors in hospitalised patients. Methods: A systematic literature search was conducted using Medline, Embase, Web of Science and the Cochrane library from database inception through 9 July 2019. We included all studies of hospitalised adult patients that used physician review of case series of admissions and reported the frequency of diagnostic adverse events. Two reviewers independently screened studies for inclusion, extracted study characteristics and assessed risk of bias. Harmful diagnostic error rates were pooled using random-effects meta-analysis. Results: Twenty-two studies including 80 026 patients and 760 harmful diagnostic errors from consecutive or randomly selected cohorts were pooled. The pooled rate was 0.7% (95% CI 0.5% to 1.1%). Of the 136 diagnostic errors that were described in detail, a wide range of diseases were missed, the most common being malignancy (n=15, 11%) and pulmonary embolism (n=13, 9.6%). In the USA, these estimates correspond to approximately 249 900 harmful diagnostic errors yearly. Conclusion: Based on physician review, at least 0.7% of adult admissions involve a harmful diagnostic error. A wide range of diseases are missed, including many common diseases. Fourteen diagnoses account for more than half of all diagnostic errors. The finding that a wide range of common diagnoses are missed implies that efforts to improve diagnosis must target the basic processes of diagnosis, including both cognitive and system-related factors. PROSPERO registration number: CRD42018115186.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Diagnostic errors are as high as 10-15%, impacting patient health and system costs.
“Estimates evaluate diagnostic errors as high as 10-15%, impacting patient health and system costs [1][2][3] .”
From Enhancing human-AI collaboration: The case of colonoscopy (2024) · cites this paper as [2]
Cites
Metaprop: a Stata command to perform meta-analysis of binomial data
Nyaga, V. N., Arbyn, M., & Aerts, M. (2014) · Archives of Public Health · doi:10.1186/2049-3258-72-39 · Cited by 2,139 · see citing works
Diagnostic error in internal medicine
Graber, M. L., Franklin, N., & Gordon, R. (2005) · Archives of Internal Medicine · doi:10.1001/archinte.165.13.1493 · Cited by 1,549 · see citing works
The Challenges in Defining and Measuring Diagnostic Error
Zwaan, L., & Singh, H. (2015) · Diagnosis · doi:10.1515/dx-2014-0069 · Cited by 173 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
Enhancing human-AI collaboration: The case of colonoscopy
Introzzi, L., Zonca, J., Cabitza, F., Cherubini, P., & Reverberi, C. (2024) · Digestive and Liver Disease · doi:10.1016/j.dld.2023.10.018 · Cited by 28 · see citing works
The economics of diagnostic safety
Slawomirski, L., Kelly, D., de Bienassis, K., Kallas, K.-A., & Klazinga, N. (2025) · OECD Health Working Papers · doi:10.1787/fc61057a-en · Cited by 20 · see citing works
Automation Bias in Large Language Model Assisted Diagnostic Reasoning Among AI-Trained Physicians
Qazi, I. A., Ali, A., Khawaja, A. U., Akhtar, M. J., Sheikh, A. Z., & Alizai, M. H. (2025) · medRxiv · doi:10.1101/2025.08.23.25334280 · Cited by 16 · see citing works
Garber, A., Garabedian, P., Wu, L., Lam, A., Malik, M., Fraser, H., Bersani, K., Piniella, N., Motta-Calderon, D., Rozenblum, R., et al. (2023) · JAMIA Open · doi:10.1093/jamiaopen/ooad031 · Cited by 10 · see citing works
- Authors
- Gunderson, C. G., Bilan, V. P., Holleck, J. L., Nickerson, P., Cherry, B. M., Chui, P., Bastian, L. A., Grimshaw, A. A., & Rodwin, B. A.
- Affiliations
- Yale University (US)
- Thomas Jefferson University (US)
- Corresponding author
- Gunderson, Craig G
- Venue
- BMJ Quality & Safety
- Publisher
- BMJ
- Published
- 2020-04-08
- Volume
- 29
- Issue
- 12
- Pages
- 1008-1018
- ISSN
- 1475-3898 (ISSN-L)
- 2044-5415
- 2044-5423
- OpenAlex
- W3015860778
- Cited by
- 130
- References
- 3
- Citations by year
- 2026: 19 · 2025: 30 · 2024: 20 · 2023: 26 · 2022: 16 · 2021: 14 · 2020: 5
- FWCI
- 3.14
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- BMJ Quality & Safety (publishedVersion)
- PubMed (publishedVersion)
- HighWire Press Open Archive (submittedVersion)
- Topics
- Health Sciences › Medicine › Family Practice › Clinical Reasoning and Diagnostic Skills
- Health Sciences › Health Professions › Emergency Medical Services › Patient Safety and Medication Errors
- Health Sciences › Medicine › Emergency Medicine › Hospital Admissions and Outcomes
- Keywords
- diagnostic errors, prevalence, hospitalized adults, patient safety, missed diagnosis, delayed diagnosis, misdiagnosis, medical malpractice, cancer, pulmonary embolism
- MeSH terms
- Adult, Diagnostic Errors, Hospitalization, Humans, Inpatients, Prevalence, Patient Safety
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review