Primary care providers embrace AI-generated imaging results explanations
Primary care providers are sharing positive feedback following a pilot program testing the utility of artificial intelligence-generated imaging results explanations.
Stanford Health Care has developed an electronic health record-integrated tool for helping draft comments on certain test results. The California institution recently surveyed PCPs to gauge whether the outputs are meeting expectations, sharing their findings in JAMA Network Open.
Nearly 85% of responding clinicians found the tool straightforward and user-friendly, labeling it as particularly beneficial for both lab (72%) and imaging results (63%).
“This study demonstrated the utility of a generative AI tool for drafting test result explanations, highlighting ease of use, improved efficiency, and higher-quality explanations,” Shreya J. Shah, MD, an internal medicine specialist with Stanford, and colleagues wrote Aug. 22, adding: “Barriers to adoption included content accuracy and completeness.”
Stanford selected Claude 3.5 Sonnet (Anthropic) as the large language model for the pilot, because of its “superior response time, fidelity to prompt instructions, and ability to generate outputs resembling result comments from clinicians.” The university created the novel program to draft result comments across lab work, imaging and pathology. Its AI software operates similarly to other Stanford tech, helping physicians automatically generate responses to patient portal messages. With this new iteration, when a physician orders an X-ray or other test, AI assists with distilling the findings into layman’s terms, to be relayed to patients.
Primary care clinicians were invited to participate in the pilot, with surveys administered after 4 and 8 weeks of using the AI tool. Out of 244 clinicians who used the program at least once, about 38% (or 93) completed surveys. Respondents also said AI appeared to improve efficiency (71%) and led to higher quality explanations (72%). More than half reported using it frequently (58%) and that the tool was ready for broad implementation (54%). Most anticipated using the program long-term (83%), while a smaller portion felt motivated to send test results to patients more frequently (42%). Average perceived time savings was about 1.1 minutes, with a range from 5 minutes saved to 3.3 additional minutes spent.
In free-text responses, PCPs’ positive sentiments centered around the AI program’s utility and opportunity for patient engagement, while negativity stemmed from content accuracy and completeness. Respondents suggested possible modifications such as including more patient context from visit notes and improved workflow integration, the authors reported.
“While these early results suggest that AI-generated draft result comments could help reduce clinician burden and enhance patient experience, further optimization grounded in clinician feedback is needed to improve accuracy and completeness of draft explanations,” the authors concluded. “Additional improvements should focus on optimizing prompts, updating the LLM, incorporating patient-specific context, and streamlining workflow integration. Future evaluations should quantify impacts on clinician inbox burden (time spent, message volume) and consider patient perspectives.”
Stanford further explained the pilot in a news item published in January. It noted that the AI program helps physicians release results more quickly to conform with the 21st Century Cures Act.
Marty Stempniak has covered healthcare since 2012, with his byline appearing in the American Hospital Association's member magazine, Modern Healthcare and McKnight's. Prior to that, he wrote about village government and local business for his hometown newspaper in Oak Park, Illinois. He won a Peter Lisagor and Gold EXCEL awards in 2017 for his coverage of the opioid epidemic.