The person you might feel most comfortable identifying isn’t even really a person, it is artificial intelligence (AI) providing access to reliable, accurate, and up-to-date knowledge.

Not to worry physicians, like me, and advanced practice providers will still be in demand. We will be the necessary interface between a patient’s subject needs and the objective-proven best practices—diagnosis, therapies, and prognoses. Medical professionals will be able to spend more time with patients educating, communicating, coaching, consoling, and generally being supportive. We will also be an important second check on the “digital” recommendations.

 

Why does the health profession need to embrace new technology? Two reasons:

• Just four in ten Americans trust physicians and hospitals according to a JAMA article last month. In 2020, pre-COVID, 71% of Americans trusted healthcare.
• Surprisingly the third leading causes of death in the United States are medical errors. What’s worse, is this upsetting fact made national headlines in 1999 with the Institute of Medicine’s report, “To Err is Human,” and since then the facts may not have improved.       Recognizing a problem is the first step to finding a solution. We can now look back 25 years and sadly realize we are unchanged or perhaps worse with an estimated 200,000 number of unnecessary deaths.

Time for a change; we don’t want to revisit the stories of Kodak vs. digital cameras or the Swiss watch manufacturers' disregard for battery-powered watches. Using digital technology responsibly and safely is going to be better than avoiding the future. Swimming with the current is always smarter and sometimes life-saving.  

 

Let’s start with the basics. Passing the US Medical License Exam requires three sets of medical boards, one after two years of medical school, one upon graduation, and a final after one year of graduation from medical school. Large-language models, that are a form of CHAT GPT, posted a 67.6% accuracy on practice tests. The passing rate varies year-to-year but usually is in the 60s.

 

After becoming licensed, physicians take specialty boards. The University of Toronto let loose GPT-4 for radiology boards which have a passing rate of 70%. GPT-4 answered 81% correctly. Certain types of questions and content had up to 90% correct.

 

What about diagnosis? “Better than you think,” is part of the title of a Modern Healthcare article. “There needs to be more research into AI’s use,” is a quote from National Library of Medicine senior investigator which everyone—caregivers and care-receivers should appreciate.

 

The Modern Healthcare article shares a few examples. Dementia and Alzheimer’s Disease are difficult to diagnose early. Treatments are maturing to decrease the severity and progression when started before too much loss of function. Clinicians using AI are 26% more accurate in early recognition of the disease.

 

Cancer diagnoses are more accurate and have outperformed traditional approaches for breast cancer, skin cancer, and other maladies. AI can also review medical records for missing information, predict surgically curable pancreatic cancer, and suggest additional genetic testing. According to a quote by Dr. Harlan Levine, President of health innovation and policy at City of Hope from Modern Healthcare in May 2024:

• About 50% of metastatic lung cancer patients don’t get the right genomic test
• Another 15% get the test but don’t get the right targeted therapy

AI outperformed radiologists at New York-Presbyterian and Columbia University in detecting structural heart abnormalities in patients with congestive heart failure.

 

Other examples abound but obstacles exist as well outlined in a National Library of Medicine comprehensive paper entitled, “Artificial intelligence in disease diagnosis: a systematic literature review, synthesizing framework and future research agenda.”

 

Adoption of new knowledge, processes, and procedures always goes through growth spurts, short retreats, some distrust, and finally acceptance. Keeping the time between positive benefits and acceptance reasonable will help everyone in the long run—patients, physicians, advance practice practitioners, health care systems, big pharma, device manufacturers, and medical insurers. Unnecessary suffering can be avoided while quality improves, and costs decrease.