How AI Could Improve the Physician Experience

A great joy of my job is the opportunity to interact with each new generation of medical students. These future physicians choose our profession for reasons personal and universal, from a fundamental desire to help others to a passion for problem-solving and scientific discovery. They look forward to a career that will challenge and inspire — and at its best, this continues to be the case for people in medicine. But increasingly and concerningly, administrative responsibilities and other stressors are detracting from their experience in clinical care and biomedical research. A recent survey found that more than 90% of physicians report feeling burnt out on a regular basis — and sadly this isn’t surprising.

That same survey also offers a ray of hope, however. More than 80% of those doctors said they believe artificial intelligence could help improve their daily professional lives — and there is ample evidence to support this conviction. By easing the burden of clerical tasks, AI has the potential to free more time for physicians to focus on patient care and high-impact research, while also enhancing the care that they can provide their patients.

Helping Us Get Away from the Computer

Physicians want to spend as much time as possible with their patients. Many also are conducting research that informs and advances standards of care. But the requirements of modern medicine often get in their way. Clinicians spend a staggering 15 hours a week on paperwork and administration, translating to an average of nearly 50% of their time doing administrative work and just 27% in direct face time with patients. Of course, it’s important to document patient visits and respond to their messages. But the medical field has long acknowledged that physicians’ time could be better spent, if those tasks could be effectively delegated. We are now getting closer to a solution.

As we’ve seen with the explosion of Chat-GPT and other generative AI models, one of the most basic ways we can use machine learning is to train models to create and compose based on existing information. In medicine, this means directing AI-powered tools to draft documents in support of our clinicians. We’re testing tools that do exactly this at Stanford Medicine. One uses OpenAI’s GPT-4.0 technology to generate draft responses to patient emails, and another uses ambient listening in an exam room, with patient consent, to generate a formatted medical note of the clinical interaction. In both cases, care providers edit and approve the document before it is finalized creating a human-in-the-loop interaction with the AI. We’re still fine-tuning these innovations, but early feedback indicates that clinicians’ feelings of burnout were appreciably improved.

Helping Us Predict Our Patients’ Health

Another area where AI has great potential is in helping physicians support their patients’ health. Clinicians are deeply invested in the wellbeing of the people in our care, and we want to do our very best to keep them as healthy as possible. We’d all like a crystal ball that could tell us what conditions are on the horizon for our patients or when their health is on the brink of deterioration. AI can enhance our ability to make these predictions. Researchers around the world are using artificial intelligence to analyze vast sets of patient data, with an eye toward identifying patterns from the past to inform the future. For example, an AI review of records for more than 40,000 patients undergoing routine cardiac computerized tomography (CT) scans helped scientists at the University of Oxford develop a tool to predict the risk of heart attack for patients who are not yet showing any signs of disease. Similar systems are under development for a wide range of conditions, ranging from type 2 diabetes to pancreatic cancer to Parkinson’s disease. AI can also help us predict a patient’s health in the shorter term. Hospitals are already using systems trained through extensive datasets to recognize early signs that a patient’s condition is likely to worsen, with clinical teams at the ready to intervene. In many cases, results are promising — including showing improved patient outcomes — as with warning systems designed to flag early signs of sepsis, a life-threatening reaction to an existing infection.

?

Helping Us Get Better at What We Do

These uses fall into a category that I call the “virtual army of assistants,” and they have vast potential to expand the care that physicians are able to provide for our patients. Another function of AI, however, is to help our teams get better at what we do —?and physicians always want to push the envelope of what we can achieve for the people that we serve. An area that benefitted first from AI innovations, and consequently has seen the most advancement, is imaging. Many tools are already in use to help radiologists analyze complex scans more quickly and accurately; and similar systems are under development to assist pathologists in their evaluation of tumors and other tissue samples. AI can also be used for training. Here at Stanford Medicine, researchers are working on technologies, trained on videos of actual surgeries, that can provide feedback on proper technique for future surgeons. A similar concept is deployed in some of our operating rooms. The OR Black Box captures millions of datapoints — from audiovisual data about team interactions to equipment readings on patient vital signs —?during a surgery. Teams can then use AI to sift through the information to identify opportunities for improving quality and safety practices.

Rejuvenating the Physician Experience

Taken together, these innovations offer a glimpse of a future where physicians can focus more on the health and healing of their patients — and to greater effect — leading to more satisfaction at work and in their careers as a whole. To be sure, all AI technologies used in a health care setting must be thoroughly and continuously vetted to ensure fairness, usefulness, and accuracy. But the systems that meet these high standards — and there will likely be many — promise to help us reimagine the physician experience to rejuvenate the fulfillment long associated with practicing medicine.

These resources provide more information about the potential uses and considerations of AI in health care. ?

RAISE Health Updates. This joint initiative from Stanford Medicine and the Stanford Institute for Human-Centered Artificial Intelligence (HAI) guides the responsible use of AI in biomedical research, education, and patient care. Stay connected, for regular updates about events, research, and other topics.

?

There’s One Hard Question My Fellow Doctors and I Will Need to Answer Soon (New York Times). In this guest essay, Daniela J. Lamas, MD, explores the potential of AI to aid diagnosis, interpret medical data, and predict patient outcomes. She also discusses how it could impact medical training and the doctor-patient relationship, and the importance of human expertise in health care.

?

ChatGPT may have better bedside manner than some doctors, but it lacks some expertise (CNN). This article discusses a study in which responses to medical questions from the public were answered by ChatGPT and physicians. The research found that responses from AI were rated higher in quality and empathy.

?

Doctors Receptive to AI Collaboration in Simulated Clinical Case without Introducing Bias?(Stanford HAI). A study that paired physicians with a prototype ChatGPT-like tool in a mock medical environment found that the doctors were willing to collaborate with the AI tool to improve patient outcomes.

Image by ipopba