Harmful diagnostic errors common among hospitalized patients

Key takeaways:

• Diagnostic errors contributed to 6.6% of deaths among hospitalized patients.
• Solutions for diagnostic errors may involve coaching and cognitive interventions.

Diagnostic errors, including problems with patient assessments and test interpretations, were common and harmful among hospitalized patients, sometimes leading to death, a study in JAMA Internal Medicine found.

According to Andrew D. Auerbach, MD, MPH, a professor of medicine at the University of California San Francisco School of Medicine, and colleagues, diagnostic errors are thought to contribute to events such as death, various harms and ICU transfers, but “few past studies used structured approaches to detect diagnostic errors.”

“For example, a recent study of inpatient adverse events did not screen specifically for diagnostic processes and detected diagnostic error in only 10 of nearly 1,000 adverse events reviewed,” they explained.

So, the researchers conducted a retrospective multicohort study with an adjudicative process to assess diagnostic error types, harms and frequencies among hospitalized adult patients who were transferred to the ICU or died from Jan. 1 to Dec. 31 of 2019. The study sample included the records of 2,428 patients (mean age, 63 years; 45.6% women) across 29 academic research centers.

Overall, 23% (95% CI, 20.9-25.3) of patients experienced a diagnostic error.

Auerbach and colleagues found that the errors contributed to either temporary harm, permanent harm or death in 17.8% (95% CI, 15.9-19.8) of patients.

Of the 1,863 patients who died, diagnostic error contributed to 6.6% (95% CI, 5.3-8.2) of deaths.

“Within the group of patients who died and had a diagnostic error, the error contributed to the death in 29.4% (95% CI, 24.0%-35.3%),” the researchers noted.

Meanwhile, the diagnostic process faults most highly associated with diagnostic errors included:

• patient assessment problems (adjusted RR = 2.89; 95% CI, 2.23-3.73); and
• problems with test ordering and interpretation (aRR = 2.85; 95% CI, 2.16-3.76).

Both faults “appear to be the most important targets for safety improvement programs.” Auerbach and colleagues wrote.

They added that solutions to testing problems could fall on informatics tools like alerts or predictive models, whereas clinical assessments may need cognitive interventions, evaluation of physician workloads and coaching.

The study was limited by its inability to determine what cognitive process was linked to what type of diagnostic error, the researchers explained, and the overall severity and prevalence of diagnostic errors was not fully captured by the specific sample size.

They concluded that the findings provide an impetus for further research and interventions to reduce diagnostic errors “by targeting gaps in test selection and interpretation and physicians’ ability to debias and rethink diagnoses as high-priority areas.”