
DrAfter123/DigitalVision Vectors
PCPs seek additional training to fully trust healthcare AI
New survey data show that while using healthcare AI is common in primary care, clinicians desire additional training to fully trust AI tools.
While half of primary care providers (PCPs) use AI tools in their workflows, many feel excluded from AI purchasing decisions and desire additional training, according to a new report from The American Academy of Family Physicians (AAFP) and Rock Health Advisory.
Researchers collected survey responses from over 1,200 family physicians and other PCPs between September and November 2024.
AI use cases
Half of PCPs reported having used healthcare AI tools in their workflows for at least one use case. Clerical support was the most popular application, with 32% percent of physicians using AI for tasks like message drafting and clinical documentation.
About one in four respondents reported the use of AI-powered information management (24%) and AI-enabled clinical decision support (23%) tools.
Notably, over 90% of respondents have tried or are open to trying each one of the AI tool categories referenced in the survey:
- Clinician clerical support
- Information management
- Clinical decision support
- Patient support solutions
- Administrative functions
- Population health management
- Behavioral health support for patients
Little involvement in AI purchasing decisions
AI's long-term success in primary care will depend on physician buy-in. However, most PCPs (65%) reported having limited or no influence over the AI technology available in their practices.
"This gap isn't entirely surprising considering how the practice landscape has shifted from physician-owned and -led practices toward more of an employment model," the report authors wrote. "Given the high costs, inherent risks and desire for consistency across organizations, top-down decision making is the norm."
However, the report also noted that when physicians don't feel like their perspectives are valued, the effects can be long-lasting.
For example, many providers have historically felt left out of EHR implementation and optimization, and the survey found that 55% of PCPs indicated EHR charting and documentation was their top operational challenge at work.
Clinicians’ current lack of involvement in AI purchasing decisions underscores the need for organizations to foster clinician trust in AI tools. However, the survey found that 81% of respondents desire more training to fully trust AI solutions at work.
Other key findings include:
- Almost 70% of PCPs want medico-legal protections before they trust AI tools.
- 64% want education on legal, liability and malpractice risks.
- 68% are looking for ethical guidelines on how to best use AI in their practice.
Mixed outlook on AI's impact
Many respondents reported optimism that AI will drive clinical efficiencies by helping them improve time to diagnosis (73%), diagnostic accuracy (66%) and treatment plan appropriateness (66%). Overall, 70% of PCPs expect AI to improve their wellbeing, and 66% believe it will improve their workload.
However, despite anticipated benefits, respondents were doubtful that patients would experience the same improvements. About half of respondents said they don't expect gains from AI to impact patient satisfaction (47%), care personalization (52%) or patient-provider relationships (55%) over the next three years.
PCPs also noted concerns about AI in the workplace. For instance, 18% of clinicians fear that AI will have negative or very negative impacts on their job security, while 52% are still unsure.
"Family physicians believe AI could reduce administrative burdens, allowing more focus on patient care," Steven E. Waldren, MD, chief medical informatics officer at AAFP, said in a press release. "However, concern remains about AI's ability to replace human judgment, with many clinicians stressing that AI should serve as a support tool, not a substitute for a well-trained physician."
Hannah Nelson has been covering news related to health information technology and health data interoperability since 2020.