You're reaching for your keys in the parking lot after a doctor's visit when it hits you, "What did my physician say I wasn't supposed to take with this new medication?"
Visits can feel like a slalom for patients and providers. Patients can struggle to take in what feels like a mountain of medical information. Health care providers face their own challenges, from patient loads to paperwork.
UC San Francisco Clinical Associate Professor of Medicine Sara Murray , MD, MAS, studies how technology and data shape health care. She is also the chief artificial intelligence (AI) officer at UCSF Health. For more than a year, she has been working with UCSF Faculty Practice Organization President Inga Lennes, MD, MPH, MBA and team as well as other UCSF Health information technology leaders and teams to launch a new AI tool called AI scribes. She tells us how the technology currently sweeping the nation is reducing the mental load for patients and providers alike.
Major health care providers are increasingly adopting AI scribes, what are they?
AI scribes are specially trained tools that leverage artificial intelligence to record and transcribe clinical encounters with UCSF patients and then draft notes and patient instructions for doctors as part of the electronic health record.
With these tools, a draft is ready in minutes. But they still need to be reviewed and edited by physicians ensure everything's accurate. After that, these become a part of the patient's medical record and are accessible to patients MyChart app.
Why are so many providers starting to use them?
For every visit a physician has with a patient, the doctor has to write clinical notes. These notes are often long and quite detailed, describing everything you as well as your physician have discussed and your doctor's assessment and plan for every part of your care.
It's why many patients likely remember visits where their doctor was furiously typing on the computer as the talked. Doctors also often have to spend time on the nights and weekends finishing up their notes.
So my doctor can focus on speaking with me now?
Yes. Our clinicians have become very good at multitasking, so they're often able to type furiously while conversing but that's a lot of cognitive burden, or what you might call mental load. What clinicians have told us is that with AI scribes, they feel relieved of that cognitive burden because they can focus more deeply on talking to and thinking about their patients.
What else are clinicians saying about AI scribes at UCSF?
Nationally, we know that many health care workers are overworked and struggle with burnout. In survey data, physicians report they are more likely to complete their notes on the same day and feel like their clinic workload was manageable after they adopted AI scribes.
What are patients saying?
One of the most beautiful quotes we received from a patient was, "My doctor was testing this new AI program which allowed him to speak directly to me: No typing, just eye-to-eye [contact] - simply spectacular."
Why do you love AI scribes for patients?
I'm an internal medicine specialist who takes care of patients in the hospital. I tell my patients that it's always good to have someone with them for important conversations who can take notes and advocate for them. While I still think this is important, I like that AI scribes can help draft detailed instructions for patients to take with them. We have also heard from patients that they feel more connected to their doctor when they are about to speak with them without having to simultaneously type.
But are AI scribes always listening?
No. A clinician must first get verbal consent from a patient to turn on the AI scribe, which records audio using a secure tool. Even though we use a phone to do the recording, it isn't stored on the doctor's phone. These tools comply with state and federal laws to protect patients' health information, including the Health Insurance Portability and Accountability Act or HIPAA. Patients can also ask for the scribe to be paused at any time during the visit.
How does UCSF ensure that patients' data remains protected?
UCSF has extensive and rigorous information technology security processes for all our clinical tools, including AI scribes. As part of safeguards, we require AI scribes recordings to eventually be destroyed after use to ensure patient privacy.
How many UCSF Health doctors are using AI scribes right now?
We've made the signup process and training available to about 1,700 eligible physicians and about 575 have completed training. We will be making this technology available to other care team members as the technology permits.
Is the AI UCSF uses trustworthy and fair? How does it do that?
Yes. Years ago, we developed a robust AI oversight process at UCSF as we started to see AI tools come onto the commercial market that we weren't sure would provide fair value to all of our patients. We wanted to ensure that the tools we developed were trustworthy, meaning they were safe, ethical and secure, for instance. As part of these processes, our AI governance committee brings together experts, including health system leaders, data scientists, researchers, ethicists and privacy specialists, to evaluate and vet all AI tools before they touch patients - whether the AI comes from vendors or from our own researchers.
In collaboration with the Division of Clinical Informatics and Digital Transformation (DoC-IT) we are building and refining our AI monitoring infrastructure. Once a solution is deployed, like AI scribes, we continue to monitor it and give feedback to - in this case - the vendors to ensure the tool and the implementation adhere to our values. In this way, we're shaping the tools.
What do you think the future holds for AI in the clinic?
As the technology evolves, AI scribes will become AI assistants, doing more and more to help clinicians with tasks that are needed to deliver safe and effective care to patients. For example, they may help by drafting orders for tests or to better summarize and incorporate past visit information into new documentation. Importantly, these tools are assistants and not replacements for your doctor. They are meant to make it easier for them to deliver high-quality care while always reviewing AI outputs. We call this "human-in-the-loop" and it is a critical part of ensuring these tools are trustworthy.
