AI should target administrative burdens first, physicians say

 Artificial intelligence may have the most immediate applications in alleviating administrative burdens, and should be applied more broadly to reducing paperwork load such as prior authorizations, clinical documentation, and patient communications, according to a new survey released by the American Medical Association (AMA). 

More than half (56%) of physicians participating in the poll, which was conducted during the summer of 2023, indicated that AI can best help with administrative burdens by automating cumbersome tasks.  

While respondents also expressed keen interest in using AI for diagnostics, they tended to rank administrative use cases as more relevant to their immediate needs and plan to implement a variety of workflow improvement tools first over the next five years. 

An even split between excitement and concern

Just over 1000 physicians responded to the survey, and were fairly evenly divided into self-described “early adopters” and “tech averse” individuals. While two-thirds agreed that there is at least some degree of advantage to using AI – called “augmented intelligence” by the AMA – they weren’t all agreed on their attitudes toward the technology. 

Thirty percent stated they were more excited than concerned, while 29% were more concerned than excited. The remaining 41% expressed equal concern and excitement over the new technology. 

A focus on administrative use cases to improve workflow 

Despite the general ambivalence, physicians are interested to see AI being applied to high-value use cases, mostly in the administrative arena. When asked about their enthusiasm levels for different uses of AI, 54% ranked “documentation of billing codes, medical charts, or visit notes” as their top pick, with automation of pre-authorizations coming close behind at 48%.  

In addition, approximately three-quarters of respondents said that these two applications would be highly or somewhat relevant to their practice, with specialists and primary care providers both agreeing on the need for AI to assist with day-to-day workflow tasks. 

Respondents also showed enthusiasm about using AI for the creation of discharge instructions, care plans, and/or progress notes (43%), generation of chart summaries (42%), summaries of research and standards of care (40%), and translation services (39%). 

Interestingly, less than 40% prioritized purely clinical tasks, including assistance with diagnoses and prediction of health risks, despite being most familiar with these technologies.  This may indicate a divide between the hype around AI from developers and what providers are actually requesting to improve their daily practice. 

Few respondents are currently using clinically focused AI tools, with about 10% stating that they are actively engaging with predictive analytics and assistive diagnosis tools. In contrast, 14% are using AI to create discharge instructions, care plans, and/or progress notes, while a similar number are leveraging AI for documentation of codes, charts, or visit notes.  

What do physicians need to accelerate AI adoption?

To move the needle on AI adoption, physicians are seeking assurances that automated tools will work as expected – and that they won’t be held liable for mistakes.  

Data privacy was top of mind for 87% of respondents when asked about the most important attributes of AI tools, but “I am not held liable for errors of AI models” and “AI is covered by my standard malpractice insurance” garnered the same number of responses. 

Easy integration with existing systems and practice workflows were also highly desirable traits, as were the proven safety, validity, and effectiveness of AI algorithms. 

Respondents also showed a desire to be somewhat involved in the process of deciding what AI technology to bring into their practice, with about half wishing to be consulted and 36% saying they’d like to be in charge of the project. To help with decision-making, 35% of physicians said they would like to see clinical evidence of a tool’s effectiveness, including peer-reviewed research, evidence of physician involvement in development, and the results of US-based clinical trials.  

With this data at hand, and a relatively active role in the adoption process, physicians may be more likely to embrace AI tools in the administrative space and elsewhere. This buy-in will be crucial for the success of AI models in the real-world care setting, so executive leaders may wish to concentrate on including physician staff in upcoming discussions about AI and encouraging clinical champions to get involved in selection, implementation, and education for their peers. 

Jennifer Bresnick is a journalist and freelance content creator with a decade of experience in the health IT industry.  Her work has focused on leveraging innovative technology tools to create value, improve health equity, and achieve the promises of the learning health system.  She can be reached at jennifer@inklesscreative.com.