Estimated diagnostic errors and serious misdiagnosis-related harms by clinical setting in the USA annually*
| Clinical setting | Visits (n)† | Diagnostic errors (n)‡ | Serious harms (n)‡ |
| Inpatient | 34 million | ~2.4 million (95% CI 1.6M to 3.3M)9 | ~376 000 (95% CI 185K to 566K)9§ |
| Emergency department | 140 million | ~7.2 million (95% CI 5.0M to 11.2M)11 | ~433 000 (95% CI 266K to 1062K)11 |
| Primary care clinics | 521 million | ~32.8 million (95% CI 31.9M to 33.7M)13 | ~231 000 (speculative)2 |
| Specialty care clinics | 515 million | ~10–30 million (speculative) | ~114 000 (speculative)2 |
| All visits combined | 1.2 billion | ~50–100+ million (speculative)¶ | ~909 000 (PR 684K to 1170K)** |
*The estimates in this table are based on best available evidence. However, there are limitations in making estimates or extrapolations based on studies that use different methods, data sources and measures. In cases where there are extremely high levels of uncertainty or limited data, such estimates are denoted as ‘speculative’.
†Annual US visits are from data published by the American Hospital Association (inpatient) and Centers for Disease Control and Prevention National Center for Health Statistics (emergency department, primary care, specialty care) from the most recent years available: inpatient (2022), emergency department (2021), ambulatory clinics (2019).
‡When prior estimates were based on visit data from earlier years, diagnostic errors and serious misdiagnosis-related harms have been scaled from source studies to the total visits from the most recent years available. For Dalal et al9 and Singh et al13 values are derived as the mathematical product between published error rates and number of visits. For those listed as ‘speculative’, a rationale is provided in the text, footnotes or cited sources.
§The proportion of serious harms attributable to the inpatient clinical setting is ~41% (n=~376, 000/~909, 000), with a plausible range between ~20% (~185, 000/~909, 000) and ~62% (~566, 000/~909, 000) (using the numerator’s upper and lower confidence bound divided by the point estimate for the denominator).
¶Rates of diagnostic error reported across frontline clinical settings in various studies are in the range of 5.1% to 7.2%.9 11 13 If we assume specialty clinic error rates (on which there are no aggregate data) are the same as those seen in primary care, then the visit-volume-based weighted, mean error rate across all clinical settings would be 6.2% (PR 5.8 to 6.7). With 1.2 billion healthcare visits annually in the US, the total number of errors would then be ~75 million (PR 70 to 81). If the error rate in specialty care were lower (eg, roughly one-third of what is seen in primary care), the total would be ~50 million. However, most of these studies systematically underestimate diagnostic errors with minor harms either unintentionally or by design (see text). Experts in the field have speculated an overall error rate in the 10% to 15% range, which would translate to 121–181 million errors. Therefore, we offer a speculative range of 50–100+ million US diagnostic errors annually across all clinical settings.
**The combined number of serious harms for all visits (~909 000) should be less than the sum across clinical settings (~1 154 000), mainly because it is expected that some of the serious harm counts are partly overlapping (eg, from a diagnostic delay that begins in the emergency department but carries forward through to the inpatient setting). The proportion of deaths (46.7%)2 corresponds to ~424 000 deaths each year due to diagnostic error.
K, thousand; M, million; PR, plausible range.