Clinician and patient satisfaction (or lack thereof) is a significant issue in healthcare — patients because they can struggle to get access to care, clinicians because they simply have too much to do.
The Ottawa Hospital (TOH) set out to tackle this challenge with the integration of Microsoft’s DAX Copilot last year. And it’s already had a big impact: Early results show seven minutes saved per encounter, a 70% reduction in clinician-reported burnout and fatigue, and 93% of patients report a better or equivalent care experience.
“Access to care is probably one of the biggest issues patients face,” Glen Kearns, EVP and CIO of TOH, told VentureBeat. “If we can improve the throughput, even a couple of patients per physician per shift, you multiply that by 10 physicians in a care setting, then multiply that by 365 — that’s not an inconsequential increase in access to care.”
Ambient AI as an active assistant
TOH was the first Canadian hospital to pilot Microsoft’s DAX Copilot, which is directly integrated with the widely used electronic health record (EHR) platform Epic. In March, Microsoft bundled DAX Copilot with Dragon Medical One (DMO) into the embedded AI assistant Microsoft Dragon Copilot, which the tech giant says is used by more than 600,000 physicians.
DAX Copilot, which is out of the box, captures physician-patient conversations via a mobile app and generates draft clinical notes in real time, Kenn Harper, Microsoft’s head of project for Dragon, explained to VentureBeat.
“Clinicians can start a recording from a mobile phone, put their phone down, examine the patient, talk to the patient, no different from what they’d be doing anyway,” he said.
The system then extracts details based on the context of the visit (symptoms, diagnosis, treatment plans, follow-ups), which is immediately available in the EHR; all the clinician has to do is quickly review and finalize.
“Instead of having to author something from scratch and remember all the minute details or type in front of a patient, this is just running automatically,” Harper explained, noting that they receive a “pretty accurate” first draft once the visit is over.
To tune Dragon Copilot, Microsoft uses a “massive repository of clinical data that’s been curated over the years,” Harper explained. Engineers continuously fine-tune large language models (LLMs) with that data so that the system can understand and reliably summarize medical information.
To further improve accuracy, models are optimized by specialty—whether that be an emergency physician, a dermatologist, a cardiologist or any other medical professional.
In feedback loops, teams analyze the first draft prepared by the machine and compare that to what and how much the clinician changed afterwards.
“It’s making sure that as an average over time, as data keeps coming in and we learn from that data, the editing goes down,” said Harper.
How TOH is improving visits, reducing physician burnout
According to the Canadian Medical Association, physicians spend around 10 hours per week on administrative tasks, such as updating charts after patient appointments.
Kearns explained that TOH developed a robust evaluation plan for DAX Copilot that includes monthly updates through a Microsoft Power BI dashboard. This incorporates feedback from clinicians, patient surveys and data from Epic.
“This framework helps us continuously monitor impact and guide improvements,” he said.
Typically at the end of the day or shift, physicians have to then go back and finalize documentation from patient visits, he explained. But the tool has reduced after-hours, charting and documentation work for “all categories of physicians.” This not only saves them time but helps reduce burnout because they have less tedious work to do.
The tool has also improved clinician cognitive load during visits: Instead of focusing on inputting patient details and navigating documents and forms, they are able to “engage differently and better,” said Kearns. Additionally, “we saw an increased level of throughput, more patients per shift, per physician.”
Kearns emphasized that all patients are asked to consent before appointments are recorded and they have access to the notes in their MyChart patient portal. They are also provided with literature about the program and assured that their health records are always confidential and secure.
Patient receptivity has also been “very, very positive,” with 97% reporting that their experience with the AI tool was as good as a typical appointment or better. “We attribute that to the physician’s opportunity to engage with them differently and more intentionally during the visit, which is often compressed by time,” said Kearns.
Looking ahead, he said the tool could be used in scenarios like biomarker detection and social determinants of health (non-medical issues that impact a person’s health, such as undernourishment or lack of transportation. Further, orders, pre-authorization and referral letters can be cumbersome; one goal down the line is to have Dragon Copilot trigger post-visit actions.
“There’s a lot coming down the road in the documentation capture space,” said Kearns.
Digital teammates to overcome staffing issues
Another area where TOH is incorporating AI is “digital teammates.” Last summer, the facility developed use cases with Deloitte and launched Sophie, who speaks several languages.
Interestingly, she’s able to interpret patient sentiment and behavioral response.
“I hate to say this, but patients lie to clinicians,” Kearns acknowledged. “You know, ‘What’s your pain scale?’ ‘Oh, I’m doing fine, it’s five out of 10.’ But then Sophie’s going to have the ability to take a look at your face and go, ‘Well, it doesn’t look like it’s a five.’”
She may then go on to ask what the patient’s definition of five is, and can pivot based on objective data.
TOH will launch another avatar in the spring to help patients navigate and access the healthcare system and engage in pre-screening.
“I don’t think anybody’s not aware of the health care human resource crisis that exists globally,” Kearns noted. “We want to really try to support and serve patients more robustly than we can today.”
For instance, the facility makes follow-up calls to patients after certain procedures. However, due to resource limitations, they can only follow up with the highest-risk patients. A goal for Kearns is following up with every single patient, with the avatar providing clarification on any patient questions and confirming if they understood their discharge orders, were able to get to the pharmacy, or have been following clinician orders. The avatar can escalate to a nurse or the patient’s clinical team if needed.
“One of the things that healthcare prides itself on is human touch,” said Kearns. “This is a way to make sure that we’re maximizing and optimizing those human touch resources, but also making sure that patients are well supported through their healthcare journey.”
However, he noted that it’s still early in the game. One important future step is enabling the digital teammate to interact with the Epic information and environment.
“We’ve got a lot more work to do there, we’re still very focused on the adoption side,” said Kearns. “We’re still a healthcare system that’s very reactive to the health status of patients. We’d like to get to the place where we’re proactive.”
The post How The Ottawa Hospital uses AI ambient voice capture to reduce physician burnout by 70%, achieve 97% patient satisfaction appeared first on Venture Beat.
