We put our lives in our doctors, but the sad and scary fact is that doctors can get things wrong. About 100,000 Americans die each year from medical errors, and a recent study found that 10-15% of all clinical decisions about diagnosing and treating patients are wrong.
A research team led by Damon Sentra, professor and director of the Network Dynamics Group at the University of Pennsylvania Annenberg School for Communication, has discovered a simple and effective way to reduce errors in diagnosing and treating patients. Connect clinicians with other clinicians using structured networks.
In a study published today in the journal Proceedings of the National Academy of Sciences At (PNAS), researchers shared the results of a multi-year study involving nearly 3,000 physicians across the United States.
They found that when presented with case studies and asked to provide diagnostic and treatment recommendations to patients, clinicians who were anonymously shown the diagnostic decisions of their peers were, on average, twice as accurate in their recommendations than clinicians who made their own decisions.
Simply put, doctors make fewer mistakes when they have a support network.
A big risk of such information-sharing networks is that while some doctors may improve, the effects of averaging can lead better doctors to make worse decisions. But that doesn’t happen. There is consistent improvement rather than regression to the average. The worst clinicians get better, but the best clinicians don’t get worse. “
Damon Sentra, Prof. Elihu Katz, Communications, Sociology, Engineering
“There is a growing recognition that clinical decision-making should be viewed as a team effort involving multiple clinicians and patients,” said study co-author Elaine Kuhn of the University of California, San Francisco and San Francisco General Hospital and Trauma Center. “This study highlights that having other clinicians available at the time of decision-making improves clinical care.”
Not just the wisdom of people in clinical practice
Over the course of several months, the researchers tested clinicians’ treatment and diagnosis decisions through an app built specifically for this purpose and distributed on Apple’s App Store.
After signing up for the trial and downloading the app, physicians were asked to evaluate clinical cases. Based on actual documented patient cases – over 3 rounds. At the beginning of each round, the clinician reads the case study, after which she is given two minutes to answer two questions.
The first question asked physicians to estimate a patient’s diagnostic risk (for example, how likely is it that a patient with chest pain will have a heart attack in the next 30 days?) on a scale of 1 to 100. A second question prompted the physician to recommend the appropriate treatment among several options (eg, send home, administer aspirin, or refer for observation).
All clinicians were randomly assigned to one of two groups. One is a control group where members answer all questions individually, and the other is an experimental group where participants can connect with other anonymous clinicians on their social network and see their responses.
In rounds 2 and 3, participants in the control group had the same experience as in round 1 and answered the questions individually. However, network state participants can see the average risk estimates made by their social peers in her network during the previous round.
All participants were given the opportunity to correct their responses in each round, regardless of whether they were participating in the social network.
Centola’s team studied seven different clinical cases using the same experimental design. Each case was from a medical field known to have high diagnostic or therapeutic errors.
The researchers found that the overall accuracy of clinician decisions increased 2-fold in the network compared to the control group. Furthermore, among the initially worst performing clinicians, the network increased the proportion of clinicians who ultimately made correct recommendations by 15% over controls.
“Using a network of physicians can improve physician performance,” says Sentra. “We’ve known for a long time that doctors talk to each other. The real discovery here is that we can build an information-sharing network among doctors to greatly improve their clinical intelligence.”
level the playing field
Face-to-face consultation networks in healthcare are usually hierarchically structured, with senior doctors at the top and junior doctors at the bottom. “Young doctors with different cultural and personal perspectives are entering the medical community and are being influenced by these top-down networks,” said Sentra. “This is how persistent prejudices creep into the medical community.”
The researchers sought to recruit clinicians of various ages, specialties, expertise, and geographic locations for the experiment.
The researchers found that anonymized egalitarian networks removed status and seniority barriers that limit many aspects of learning in medical networks. “Egalitarian online networks have increased the diversity of opinions that influence clinical decision-making,” says Centola.
in the doctor’s office
“We don’t have to reinvent the wheel to implement these discoveries,” says Centola. “Some hospitals, especially in resource-poor areas, are using e-consulting technology where clinicians message external experts for advice. It usually takes 24-72 hours to get a response. Why not send this inquiry to a network of experts instead of just one?”
Centola points out that each experiment took less than 20 minutes. Plus, the network doesn’t have to be huge, he says. In fact, he ideally has 40 members.
“Having a network of 40 people greatly improves the collective intelligence of clinicians,” says Centola. “Profit increases beyond that, say he’s from 40 to 4,000, are minimal.”
Researchers are now working to bring network technology into the clinic. The University of Pennsylvania Hospital has already funded a pilot of this program, which is expected to begin later this year.
sauce:
Reference magazine:
Sentra, D. other. (2023) Experimental Evidence for Structured Information Sharing Networks to Reduce Medical Malpractice. PNAS. doi.org/10.1073/pnas.2108290120.