What the Ability of AI to Make Clinical Decisions Could Mean for the Future of Patient Care

Last December, news outlets reported that ChatGPT, a new artificial intelligence (AI)-powered algorithm had performed at or near the passing threshold for all three US medical licensing exams without prior specialized training or reinforcement. This could represent a major milestone in the way medicine is practiced in the future and for healthcare in general. Since the boom of computer technology in the late 1990s, technology has increasingly played an important part in our everyday lives. For most people, gone are the days when one needed to go to the store to buy groceries or to the bank to make a financial transaction. Today, one can have groceries delivered at home, apply for a loan, pay a credit card statement, transfer money to a loved one, or make a purchase simply by hitting a few keystrokes on a laptop computer or smart device while seating in the comforts of his or her own home. These technological advances have significantly changed the landscape of human interactions for the better or worse depending on how one looks at it. Technology has shaped the way society and people behave and interact with one another. It helps us optimize the way we do things by making us more efficient, saving us time, reducing human errors, and eliminating bias in some cases. Technology is not only important in making it easy to perform our basic human daily tasks but also has some wide applications in more complex fields such as medicine, pharmacy, dentistry, engineering, biochemistry, genetics, etc. AI is one branch of technology that is gaining more traction within these complex fields as AI seeks to simulate human intelligence processes using computer systems. These computer systems are powered by algorithms such as natural language processing, speech recognition, and machine learning that behave almost like humans. A study published in 2021 by two Canadian researchers found that machine learning algorithms produced fewer decision-making diagnostic errors than trained psychologists suggesting that clinicians could benefit from using such tools in the practice of their profession. In the U.S. where 290,000 to 330,00 people die each year because of medication and diagnostic error combined, the utilization of such novel technology is bound to become more widespread if we want to reduce the death toll associated with these causes. ChatGPT which was developed by OpenAI and recently released to the public appears to present promising applications to the practice of medicine and pharmacy. The platform works like a giant encyclopedia capable of generating full answers to any questions regardless of the field of study, in a matter of seconds. ChatGPT can be used as clinical decision support to practicians by providing real-time evidence-based treatment recommendations; as medication management coach for patients by helping them manage their medications through reminders, dosage instructions, drug interactions and contraindications alerts; as patient triage support by determining patients’ urgency and condition severity. Even though concerned parties have raised red flags as to the potential dangers of relying on a still developing technology to make clinical decisions that could affect patients’ lives, the number of healthcare professionals who have chosen to embrace this new technology is steadily increasing. The decision to utilize AI and its numerous applications such as natural language processing, speech recognition, and machine learning in patient care is obviously up to the practitioner. However, AI should not be seen as a threat to the practice of medicine but instead as a useful tool to improve patient care. The bottom line is that the era of AI in patient care has arrived and is here to stay whether we decide to fight it or embrace it.     

Serge Afeli, PhD, MSHA

Dr. Afeli is the director of innovation and entrepreneurship and tenured associate professor of pharmacology at Presbyterian School of Pharmacy.

Source: “Machine learning to analyze single-case graphs: a comparison to visual inspection,” by Marc Lanovaz and Kieva Hranchuk, was published July 15, 2021 in the Journal of Applied Behavior Analysis.