Jan 9, 2024 - Health

Patients are hurt by diagnostic errors more often than you think

Illustration of an EKG wave resembling blue check marks followed by a red "x."

Illustration: Annelise Capossela/Axios

Hospitalized patients who died or were transferred to the ICU during their stay experienced a diagnostic error nearly a quarter of the time — and in most cases the error caused harm, according to a new study that's prompting calls to rethink how health systems keep patients safe.

Why it matters: Hospitals can be risky places for patients, and the new study sheds light on how commonly human error in medicine harms and even kills patients.

  • It comes nearly 25 years after the landmark report "To Err Is Human" found that errors kill tens of thousands of patients annually — largely due to failures of systems rather than those of providers — and sparked a revolution in how health care sought to reduce potentially dangerous mistakes.

"Delayed diagnoses and misdiagnoses are more common than we would like to think, which is incredibly humbling," said Andrew Auerbach, lead author of the JAMA Internal Medicine study and professor at the UCSF School of Medicine.

  • The research was patterned on other studies of adverse events in hospitals dating back to the early 1990s, which helped lay the groundwork for how health care monitors and addresses other patient safety concerns, such as harmful drug effects.

Details: Researchers examined a random sample of nearly 2,500 patient records from 29 academic medical centers for adults hospitalized with general medical conditions and who were transferred to an ICU or died.

  • 550 of those patients, or approximately 23%, experienced a diagnostic error.
  • 18%, or 436 patients, experienced temporary or permanent harm as a result.
  • Of the 1,863 patients who died, a diagnostic error was deemed a contributing factor about 7% of the time.

Between the lines: Further study is needed to understand whether certain patients or certain conditions may be prone to missed or incorrect diagnoses.

  • It may be that physicians with higher workloads or certain types of patients are more likely to make an error, Auerbach told Axios.

The other side: In an accompanying article, JAMA editors called the study results "striking" as they also pointed out it looked only at especially sick patients, some of whom "may have had poor outcomes regardless of the errors."

  • Future research could compare hospitalized patients with similar diagnoses and severity of illness to understand the impact of diagnostic errors, they suggested.
  • "As the complexity of medical practice grows, we have a responsibility to patients to examine our role in contributing to patient harm through diagnostic error and invest in research and quality improvement initiatives to strengthen the diagnostic process in medical education and clinical care," wrote Grace Zhang, a JAMA editorial fellow and UCSF internal medicine resident, and Cary Gross, a JAMA associate editor and Yale University professor of medicine.

In thinking about how to prevent medical errors, critical lessons can be drawn from the experience of trying to mount an effective response to COVID-19, wrote UCSF medicine chair Bob Wachter in another perspective.

  • "Motivating ongoing prevention-oriented behaviors and policies requires continued dissemination of both data and stories, strong and consistent leadership, and a combination of carrots and sticks," he wrote.

The intrigue: The study comes as new AI tools aimed at improving the accuracy of diagnoses are exploding.

  • To be sure, the technology has limitations. For instance, a recent study found ChatGPT incorrectly diagnosed over 8 in 10 selected pediatric case studies.
  • There are opportunities for tools like generative AI to provide a helpful prompt at the point of care to assist with diagnosis, Auerbach said.
  • "But I don't know if AI is going to solve the time-pressure issue ... like how many patients you have on your service and how many handoffs you have going on," he said.

Our thought bubble: It may actually be decidedly low-tech measures that can help nudge providers toward reducing mistakes.

  • For example, one of the solutions being studied, Auerbach said, is the value of the "diagnostic pause."
  • That's just what it sounds like: A sort of "timeout" for a clinician to touch base with colleagues and review a checklist to make sure they didn't miss ordering any tests or overlook any clues from a patient's medical history.
  • The most important part of the exercise, he said: "It involves giving yourself time to think."
Go deeper