Doctors, residents, and pre-med students forced to endure organic chemistry and medical school anatomy were floored when they learned OpenAI's ChatGPT successfully passed U.S. Medical Licensing Exams.
Can a bot really replace physicians?
If you believe the tech leaders (including OpenAI officials) who recently arguing AI could lead to human extinction, the answer is no -- and it would be incredibly dangerous to try.
Perhaps, though, we are asking the wrong question.
Let's try again: Can an AI bot replace some physicians? The answer is yes. How? Clinicians can, and should, use AI to improve their operational efficiency. This will allow them to spend more time with, and even see more, patients without sacrificing quality of care or burning out. And if doctors can see more patients, our current physician shortage can be lessened, if not fully erased.
And we get to keep doctors in the workforce.
Documentation Requirements Lead to Doctor Burnout
When determining operational staffing needs for emergency departments, inpatient services, and clinics, managers and operators review historical patient volumes and aim to closely match physician and nursing staffing capacities accordingly.
Factored into these staffing calculations is electronic documentation, a significant (and necessary!) patient encounter component. According to research in the , healthcare providers dedicate almost half of each day to documentation. They also spend an on documentation after scheduled office hours. (This time is often not considered in operational decisions, though it should be.)
Clinicians are required to document for liability and billing reasons, and for the next provider who will treat a patient. As an emergency medicine physician, I am intimately familiar with the daily burden of medical record documentation. This taxing system has serious implications not only for physicians, but patients too. Burnout . And, as any parent or exhausted teacher knows, burnout and exhaustion can lead to a curt efficiency that can quickly slide into distrust or, at the very least, relationship disengagement. In the absolute worst cases, burnout can decreased clinician job satisfaction, depression, and even suicide.
Documentation in Practice Today
Here is a practical example of what this looks like: A few weeks ago, as I entered data into an electronic health record for a patient I had just seen, a nurse alerted me to a new "code blue" patient. I sprinted to the patient, shocked her, and put a breathing tube and life-sustaining drips in place. I then left my critical patient's bedside to return to the screen -- and I knew the patient I had just shocked would also require painstaking documentation that would yank me away from my patients on the emergency department floor yet again.
Electronic medical records are helpful when reviewing medical histories or ordering tests, and serve as effective billing, research, and data collection tools. Documentation is a necessary part of patient care, but whenever I'm on the floor I worry about how this additional task is impacting my work, the clinicians around me, and the precious patients we are trying hard to care for. I don't know a single person who went into medicine to spend days (and nights) doing paperwork.
How AI Can Improve Operational Efficiency
There has to be a better way -- a tool that can help doctors accurately record patient encounters to preserve the continuity and quality of patient care while keeping us on the floor.
AI could be that tool.
AI software could, for example, draw from its data store to enhance patient records. (Tools such as ChatGPT can access digital data, affording physicians access to an expansive library of information to support their patient encounters.) , AI "acts as a scribe and advisor, nudging clinicians with documentation suggestions to make record keeping as thorough as possible." These nudges reduce clinician stress and facilitate accurate billing and reimbursement. AI also can help formulate differential diagnoses, allowing physicians to make more accurate assessments at a quicker pace.
There is more too. Stanford University researchers have speech recognition and natural language processing technology can support the note-creation process in real time by listening in on patient-physician conversations. AI can also collect, sort, and assemble clinical information from multiple sources, including laboratory results and pharmacy records, faster than humans while automating and optimizing coding and billing processes based on risk-adjustment factors.
In January, University of California (UC) Davis Health it will pursue AI functionality that interprets care team conversations in real time to reveal insights that may improve patient care, the clinician experience, and billing accuracy. UC Davis researchers have also received a $15 million, 5-year grant renewal from the National Cancer Institute to fund AI projects aimed at improving breast cancer screening and risk prediction and reducing health disparities. Likewise, practices such as , a federally qualified health center in New York, began using AI medical scribes nearly a decade(!) ago.
As we delve into the applications of AI in medicine, its role will continue to grow and evolve. However, crucial concerns surrounding data storage, regulations, patient privacy, and safety need to be resolved. AI, while not a complete substitute for physicians, can revolutionize the field by easing administrative burdens and enhancing efficiency. It's essential that we utilize the time saved from reduced documentation to foster stronger connections with our patients, ensuring a more compassionate and personalized healthcare experience.
is a practicing emergency medicine physician, founder of , and a professor of practice at the University of North Carolina's Kenan-Flagler Business School. Previously he served as president of emergency medicine and chief impact officer for one of the nation's largest national medical groups.