Contributors: Glenna Crooks, PhD and Paul Hambly
To learn more about Glenna and Paul, click here.
Source: Bigstock |
Generative artificial intelligence (GenAI) is poised to be the most transformative technology in medical history, with applications ranging from administrative automation to diagnostic and treatment decision support.
Understanding how physicians perceive GenAI is crucial because integrating it into healthcare represents more than just technological adoption. GenAI will pressure-test fundamental notions of medical professionalism. How physicians view its benefits and risks, regulatory oversight, and accountability frameworks will influence their adoption patterns, the success of health system implementation, and patient acceptance.
Are you a GenAI platform developer who needs your technology to pass muster? Are you a health system administrator who needs implementation plans to succeed? If so, turn to pediatricians for advice. Invite them to become your “partner of choice.”
This recommendation is based on insights gained from a joint self-funded study we conducted, which surveyed 2,739 physicians.
The GenAI Paradox
Our research results summary is not surprising: GenAI offers an unprecedented opportunity, but it also poses unparalleled implementation challenges. The medical community is aware of both.
Physicians are optimistic about GenAI in general, but approach it cautiously. They say it will help them uncover insights they might otherwise miss (87%), make their jobs easier (81%), and more efficient (84%). Most (89%) predict it will be embedded in patient interactions within 3-5 years. Their enthusiasm, however, is tempered by challenges and concerns.
Lack of Clarity About Preferred Applications
The first challenge is this: prioritizing GenAI development projects is difficult because there is little consensus on which platforms physicians prefer. Of the nine benefits we surveyed, only three ranked at the top for more than 10% of physicians: streamlining prior authorization (11%), creating visit and discharge summaries (15%), and improving diagnostic accuracy (21%).
Risks to Patients
The second challenge is a source of concern: the potential risks to patients. Most (75%) agree that GenAI’s warmth, patience, and kindness lead people to trust it more than they should. Regulatory oversight may help prevent patient harm, but 80% lack confidence in regulators to keep up with the rapid pace of development. A majority (65%) worry the technology produces unpredictable results, 80% worry it is a “black box” they cannot see inside, 64% believe bias in training data will exacerbate disparities, 75% predict privacy violations, and 81% are concerned malicious actors will outpace protective security measures.
Risks to Physicians
The third challenge is also a source of concern: the potential risks to physicians. Across every demographic we surveyed, between 73% and 76% of physicians said patients should be informed when physicians use GenAI to assist in diagnosis and treatment recommendations, despite 32% fearing they will lose patients who become aware of it. This convergence did not occur in any other survey question, yielding a striking insight into the complex realities of modern medical economics and physicians’ ethical standards. This commitment to disclosure, despite their fears and associated business risk, demonstrates the profession's moral and ethical foundation.
Addressing risk matters because physicians’ views about accountability for any harm to patients are far from settled.
- 56% would hold GenAI developers and users (e.g., physicians, hospitals, payers) responsible.
- 25% would hold GenAI developers responsible.
- 16% would hold physicians responsible.
The Trust Gap
Unfortunately, GenAI arrives at a time when trust in healthcare institutions seems to be at an all-time low. As one way to demonstrate trustworthiness, we drafted an Oath for GenAI in Healthcare modeled after medical oaths. When physicians were asked if they would take the Oath, 73% said they would. Those who declined said it was redundant, since they’d already sworn to an oath.
What about other healthcare players whose work contributes to patient outcomes? For example, should GenAI developers, payers/MCOs, clinical researchers, and marketing agencies take the Oath? Between half and three-quarters of physicians agreed they should. Would they? In most cases, physicians agreed, probably not. This “trust gap” is the largest with insurance company reviewers, GenAI platform developers, payers/MCOs, and hospital administrators.
Closing the Trust Gap
Closing this trust gap is essential for successful adoption, as the healthcare GenAI market could catalyze a collision between two fundamentally different cultures. Medicine has been intentionally cautious for over 2,500 years. Its conservative approach exists for good reasons. Any "move fast and break things" ethos, which prioritizes rapid iteration, market disruption, and acceptable failure rates, like those found in consumer technology, will be catastrophic in healthcare.
Medical errors affect patient lives, not just user experiences. Healthcare mistakes can be irreversible, unlike software bugs. Physician licenses and patient trust are at stake with each interaction. Physicians will face malpractice liability, and reputation damage can extend to the entire profession. Regulatory responses could be harsh. Failures could damage the healthcare GenAI market, causing physician enthusiasm to evaporate overnight.
Why Sounding the Loudest Alarms Makes Pediatricians a Perfect Partner
Physicians hold themselves to one of the highest standards of any profession. This creates strategic risks for GenAI healthcare ventures, but also an opportunity for companies willing to adopt similar standards voluntarily. Our survey data points to the best partners to help bridge any gaps.
We did not set out to detect protective instincts, but a data pattern suggests we found them among pediatricians, and the pattern is unmistakable. Across every risk category we measured, physicians caring exclusively for children expressed statistically significant concerns greater than those of their colleagues. Pediatricians are more often worried that:
- regulators won’t keep pace with GenAI developments.
- GenAI is a "Black box."
- training dataset biases will worsen health disparities.
- identifiable patient data will "leak" into training datasets.
- malicious actors will disrupt medical systems.
They are also statistically significantly more likely to hold GenAI developers responsible for harms (p < 0.05).
Winning GenAI healthcare companies will position themselves as healthcare-native, with a deep understanding of the ethical, conservative culture of medicine, rather than as technology companies entering a lucrative new market. Their most demanding customers will be pediatricians. Smart companies will engage them early and often.
Contact Glenna (corresponding author) at: [email protected] or 1-610-247-5032.
Contact Paul at: [email protected] or his LinkedIn profile.