For some, artificial intelligence is sharing a cat video to your aunt on Facebook; for others, it’s advanced computation of large datasets to better chart the universe itself. There is a lot of room between cat videos and mapping galaxies. Somewhere in between these opposing AI use cases lies how healthcare providers interact with patients using this technology. Given this vast perception gap, it’s vital that the healthcare industry clearly understands how AI can and is being used, along with ensuring clinicians use this incredibly powerful tool appropriately. Because when it comes to a patient’s healthcare decisions, ultimately, a human bears the responsibility for those decisions, not a computer.
Since AI is such a powerful tool, it needs guardrails. But any regulations applied won’t be as simple as putting on a Band-Aid and forgetting about future problems. To regulate appropriately, we need to learn from the past and act quickly as AI is rapidly incorporated into healthcare.
When regulation works
While it’s easy to say “we don’t want the government making the rules,” there is precedent for effective healthcare regulation. Use electronic prescription regulation as a case study for successful standard implementation. What seemed like a foreign concept in the 1990s, electronically transmitting prescription information, is now commonplace and successful because we have standards. What we’ve seen from e-prescribing should be a lesson on how healthcare should adopt and regulate AI.
- Standardization:
- The Medicare Modernization Act laid the groundwork for providing uniformity across the healthcare industry for the nearly ubiquitous use of electronic prescribing.
- For AI: We will need a similar robust, interoperable standard to ensure clarity across the industry as AI technology quickly becomes adopted.
- The Medicare Modernization Act laid the groundwork for providing uniformity across the healthcare industry for the nearly ubiquitous use of electronic prescribing.
- Safety and quality improvements:
- Ensuring more physicians use electronic prescribing reduces medication errors. A doctor’s handwriting was the butt of jokes because of their prescription pad penmanship. Now, with a few clicks of a mouse, a doctor can eliminate confusion and ensure there aren’t any adverse effects with existing medications.
- For AI: Using AI-powered tools is already improving care accuracy by giving physicians more time and helping them catch potential errors. With standardization comes greater usage and a healthier population.
- Ensuring more physicians use electronic prescribing reduces medication errors. A doctor’s handwriting was the butt of jokes because of their prescription pad penmanship. Now, with a few clicks of a mouse, a doctor can eliminate confusion and ensure there aren’t any adverse effects with existing medications.
- Adaptation:
- E-prescribing regulations are still being updated. Notable changes were made to enable electronic prescribing for controlled substances.
- For AI: As new issues arise, they can be addressed. Any regulation for healthcare AI must be considered an ongoing process.
- E-prescribing regulations are still being updated. Notable changes were made to enable electronic prescribing for controlled substances.
While regulation will be needed, it cannot be so overbearing as to stifle innovation. Misinformation, hallucinations, and biases exist. Some even worry that the potential overuse of AI will erode physicians’ critical thinking. That being said, we cannot let perfection be the enemy of good. AI is imperfect. The potential to help patients by accessing and sorting through the immense volume of patient data while freeing up provider time to research and treat them is too compelling to ignore.
What can the healthcare industry and regulators do now to best replicate the past successes of e-prescribing, all while building trust in AI?
The AI gray area
AI usage in healthcare, especially for physicians dealing with patient information, falls into three buckets:
- Bucket A: The AI output is obviously correct and is easy to use.
- Bucket B: The AI output is obviously incorrect and is easy to ignore.
- Bucket C: The AI output sounds reasonable, but the clinician is unfamiliar with the information. This is the AI gray area.
Bucket C poses the risk of a clinician trusting the AI output without checking references. This is where there is obvious room for improvement, and even regulation.
How does healthcare limit clinician AI interactions from falling into “Bucket C”? The following are strong steps to start.
- Education: Clinicians must participate in continuing medical education (CME) to maintain licensure. Governing bodies should consider requiring AI training as part of the CME process. We can see a similar approach used for controlled substance prescribing education. To use AI as a tool, clinicians must understand how to use that tool.
- Citation: Require that any AI output be cited and that the citation be easily accessible to clinicians. If a physician questions an AI response, the AI must provide a clear path for clinicians to fact-check the output.
When it comes to healthcare regulations, we have the playbook. The only difference is the speed at which the healthcare industry must act. Let’s make sure AI can be as powerful and as trusted a tool as possible. And that means smartly implementing regulations.
Photo: ismagilov, Getty Images
Dr. Michael Blackman is the Chief Medical Officer of Greenway Health, an electronic health records, practice management, and revenue cycle management solution for ambulatory care providers. A primary care physician at heart, Dr. Blackman brings an extensive background in health IT product management along with his knowledge of outpatient and inpatient care. He believes healthcare is a team sport that requires the talents of all contributors working together to succeed.
Prior to joining Greenway Health, he was Medical Director for Population Health at Allscripts. Before that, he served as Chief Medical Officer for McKesson’s Enterprise Information Systems division. He was an early leader in the development of electronic prescribing for controlled substances.
This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.
