New study reveals widespread diagnostic errors in seriously ill hospitalized adults

University of California - San FranciscoJan 8 2024

A study of seriously ill patients from academic medical centers across the country has found that nearly a quarter had a delayed or missed diagnosis.

All the patients had either been transferred to the intensive care unit (ICU) after being admitted or died in the hospital. The researchers concluded that three-quarters of these diagnostic errors contributed to temporary or permanent harm, and that diagnostic errors played a role in about one in 15 of the deaths.

The most common errors identified in the study involved delayed rather than missed diagnoses, for example because a specialist was consulted too late or an alternate diagnosis was not considered soon enough, or because of problems ordering the correct test and interpreting the results.

Using statistical methods, they estimated that eliminating these problems with assessment and testing would reduce the risk for diagnostic errors by approximately 40%.

The study represents the largest assessment of diagnostic errors in which physicians reviewed each medical record. It appears Jan. 8, 2024, in JAMA Internal Medicine.

Academic medical centers often see the most challenging cases, and the data can help them increase patient safety by coaching physicians, improving communication between healthcare teams and patients, and developing more accurate diagnostic tools and techniques.

"Our study is similar to studies from the '90s describing the prevalence and impact of common patient safety events, such as medication errors, studies which catalyzed the patient safety movement," the paper's first author, Andrew Auerbach, MD, MPH, a professor in the UCSF in the Division of Hospital Medicine, said in reference to the groundbreaking 1999 Institute of Medicine report, "To Err is Human." "We hope our work provides a similar call to action to academic medical centers, researchers and policymakers."

The data may also be useful in designing artificial intelligence (AI) that can summarize lengthy medical records, suggest alternative diagnoses when patients fail to improve and ensure that the correct tests are ordered.

A national collaboration to improve safety

The study involved the 29 academic medical centers that are participating in the Hospital Medicine ReEngineering Network, a quality improvement collaborative that includes Beth Israel Deaconess Medical Center, Brigham and Women's Hospital, Johns Hopkins Hospital, Massachusetts General Hospital, the Mayo Clinic, UCSF Medical Center, Yale New Haven Hospital and Zuckerberg San Francisco General Hospital and Trauma Center.

While the study centered on some of the most respected medical centers in the country, the authors cautioned that the results may not generalize to all acute care hospitals.

The research was drawn from a pool of more than 24,000 hospitalized adults who were transferred to the ICU on their second hospital day or died in the hospital between Jan. 1, 2019, and Dec. 31, 2019. Patients who had been transferred to the ICU from the emergency department were excluded to eliminate cases that had been misdiagnosed there.

The researchers randomly selected cases from this large pool, settling on a final group of 2,428. The patients were extremely ill, and three-quarters (1,863) died in the hospital. The physicians first examined every chart for the presence or absence of diagnostic errors, then evaluated whether the mistake had caused harm. Two physicians who had been trained to identify errors reviewed each record, and a third was on hand to settle any disagreements.

Of the reviewed cases, 550 patients, or 23%, experienced a diagnostic error. The errors caused temporary or permanent injury or death in 436 of those patients. The researchers concluded that diagnostic error was a contributing factor in 121 of the deaths.

We know diagnostic errors are dangerous, and hospitals are obviously interested in reducing their frequency, but it's much harder to do this when we don't know what's causing these errors or what their direct impact is on individual patients. We found that diagnostic errors can largely be attributed to either errors in testing, or errors in assessing patients, and this knowledge gives us new opportunities to solve these problems."

Jeffrey L. Schnipper, MD, MPH, senior author of the Brigham's Division of General Internal Medicine and Primary Care

How AI can help physicians

The researchers say the study highlights the need to improve clinician training, evaluate physician workloads and develop more accurate diagnostic tools and techniques. This could include using AI to evaluate patients, select the most appropriate tests and reduce delays, although care must be taken to ensure the models are performing correctly without introducing errors or widening health disparities.

"In the end, helping physicians become better diagnosticians means coaching and training physicians, and helping physicians clearly explain diagnoses to patients," Auerbach said. "I suspect AI will help with many tasks, but we still have work to improve communication between patients and healthcare team members to fully advance the field."

This study was supported by the U.S. Department of Health and Human Services' Agency for Healthcare Research and Quality (AHRQ). Since 2019, AHRQ has received dedicated funding from Congress to support diagnostic excellence. This includes 10 Diagnostic Safety Centers of Excellence funded in 2022, one of which was awarded to UCSF.