Doctors at Stanford Medicine have introduced a new artificial intelligence tool to help send messages to patients about their test results. This technology drafts lab tests and interpretations of test results, explaining them in plain language messages for the doctor to review and approve.
This technology is increasingly offering AI-based tools that allow doctors to spend less time on administrative tasks and more time on more meaningful work, such as interacting with patients.
“Artificial intelligence has tremendous potential to improve the experience for both patients and clinicians in healthcare settings,” said David Entwistle, president and CEO of Stanford Healthcare. is one of the many ways we can unlock that potential.” “At Stanford Medicine, we are proud to be at the forefront of responsiblely implementing AI into clinical care, with a focus on improving and improving the health of all people.”
The tool was built in-house and leverages Anthropic’s Claude Sonnet LLM hosted on Amazon Web Services’ Bedrock generative AI platform to leverage AI to create responses to patient inbox messages for healthcare providers. It follows a similar model to the draft message generator.
“It’s important to continue to give back to the patient-provider relationship,” said Michael Pfeffer, M.D., chief information officer for Stanford Health Care and Stanford School of Medicine. “We hope this will reduce administrative tasks, facilitate better and faster communication, and free up time for that.”
Testing the tool
When doctors order tests such as blood tests, X-rays, and biopsies on patients, the results are displayed as medical data without being interpreted. By law, these results must be shared with patients as soon as they are ready, but the explanations are often highly technical and can be difficult to understand without a medical degree. This means that the physician is tasked with interpreting the results for the patient.
“Physicians order millions of lab tests for their patients every year,” said Aditya Bhasin, vice president of software design and development at Stanford Healthcare. “By drafting responses for physicians, we not only support their workload, but also provide timely and comprehensive comments that help patients understand specific outcomes.”
In a pilot test of the technology, 10 primary care physicians used the tool for one month, and after incorporating their feedback, the team ran a second pilot with a cohort of 24 physicians for an additional two months. I did. After the tool creates a draft, it is not sent directly to the patient. Each draft is reviewed and edited as necessary before the physician sends the note.
“As a clinician, I love that I don’t have to start with a blank page and that the draft is written in language that patients can understand,” said Christopher Sharp, M.D., chief medical information officer at Stanford Medicine. Ta. “I’ve had patients say to me, ‘Doc, Mr. Sharp, you always write comments about my results, which makes me feel so much better.'” Create these notes in a clear and empathetic way. Although it takes effort and time to do so, I think this tool will make it easier and more efficient to provide the interpretations that are so important to patients. ”
Pfeffer said the goal of the tool is not to replace messaging interactions between doctors and patients, but rather to provide doctors with drafts ready or ready to send. Of course, doctors are not required to use it.
The principles of the RAISE Health Initiative focus on the development and deployment of Stanford’s AI technology and FURM (Fair, Useful, and Reliable Model) evaluations, with accountability and safety at the forefront, and Bhasin and his team evaluated multiple models. To find the best fit for Stanford Health Care and to ensure that the tools do not introduce unintended biases or kinks into clinical workflows.
The team received positive anecdotal feedback during the initial implementation of the software.
“So far, we’ve had a great response from doctors,” Bhasin said. “Our physicians say they appreciate the concise and accurate nature of the draft, that it is customized to the patient, and that the message is reassuring when communicating normal results.”
Supporting primary care and beyond
So far, primary care physicians across Stanford Healthcare are benefiting from this tool. The plan is to roll it out to professionals later this year. The team also plans to measure the tool’s success by analyzing data showing how much time is saved and how often AI-generated drafts are used, among other metrics.
“We are excited to have the opportunity to bring these technologies to physicians, study them, and learn from them,” Pfeffer said. “This is still fairly new, but as we develop it with feedback from physicians, it will become even better and more valuable.”
