How The Ottawa Hospital uses AI ambient voice capture to reduce physician burnout by 70%, achieve 97% patient satisfaction

Clinician and affected person satisfaction (or lack thereof) is a big problem in healthcare — sufferers as a result of they’ll battle to get entry to care, clinicians as a result of they merely have an excessive amount of to do. 

The Ottawa Hospital (TOH) got down to deal with this problem with the combination of Microsoft’s DAX Copilot final yr. And it’s already had a huge impact: Early outcomes present seven minutes saved per encounter, a 70% discount in clinician-reported burnout and fatigue, and 93% of sufferers report a greater or equal care expertise.

“Entry to care might be one of many greatest points sufferers face,” Glen Kearns, EVP and CIO of TOH, advised VentureBeat. “If we will enhance the throughput, even a few sufferers per doctor per shift, you multiply that by 10 physicians in a care setting, then multiply that by 365 — that’s not an inconsequential enhance in entry to care.” 

Ambient AI as an energetic assistant

TOH was the primary Canadian hospital to pilot Microsoft’s DAX Copilot, which is immediately built-in with the broadly used digital well being file (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 big says is utilized by greater than 600,000 physicians. 

DAX Copilot, which is out of the field, captures physician-patient conversations through a cellular app and generates draft medical notes in actual time, Kenn Harper, Microsoft’s head of challenge for Dragon, defined to VentureBeat. 

“Clinicians can begin a recording from a cell phone, put their telephone down, look at the affected person, discuss to the affected person, no totally different from what they’d be doing anyway,” he mentioned. 

The system then extracts particulars primarily based on the context of the go to (signs, prognosis, therapy plans, follow-ups), which is straight away accessible within the EHR; all of the clinician has to do is shortly evaluation and finalize.

“As a substitute of getting to creator one thing from scratch and keep in mind all of the minute particulars or kind in entrance of a affected person, that is simply working robotically,” Harper defined, noting that they obtain a “fairly correct” first draft as soon as the go to is over. 

To tune Dragon Copilot, Microsoft makes use of a “large repository of medical knowledge that’s been curated over time,” Harper defined. Engineers constantly fine-tune giant language fashions (LLMs) with that knowledge in order that the system can perceive and reliably summarize medical info. 

To additional enhance accuracy, fashions are optimized by specialty—whether or not that be an emergency doctor, a dermatologist, a heart specialist or every other medical skilled. 

In suggestions loops, groups analyze the primary draft ready by the machine and examine that to what and the way a lot the clinician modified afterwards. 

“It’s ensuring that as a mean over time, as knowledge retains coming in and we study from that knowledge, the modifying goes down,” mentioned Harper. 

How TOH is enhancing visits, decreasing doctor burnout

In response to the Canadian Medical Affiliation, physicians spend round 10 hours per week on administrative duties, corresponding to updating charts after affected person appointments. 

Kearns defined that TOH developed a strong analysis plan for DAX Copilot that features month-to-month updates by way of a Microsoft Energy BI dashboard. This incorporates suggestions from clinicians, affected person surveys and knowledge from Epic. 

“This framework helps us constantly monitor affect and information enhancements,” he mentioned. 

Usually on the finish of the day or shift, physicians should then return and finalize documentation from affected person visits, he defined. However the instrument has diminished after-hours, charting and documentation work for “all classes of physicians.” This not solely saves them time however helps cut back burnout as a result of they’ve much less tedious work to do. 

The instrument has additionally improved clinician cognitive load throughout visits: As a substitute of specializing in inputting affected person particulars and navigating paperwork and kinds, they’re able to “have interaction otherwise and higher,” mentioned Kearns. Moreover, “we noticed an elevated degree of throughput, extra sufferers per shift, per doctor.”

Kearns emphasised that every one sufferers are requested to consent earlier than appointments are recorded and so they have entry to the notes of their MyChart affected person portal. They’re additionally supplied with literature about this system and warranted that their well being data are all the time confidential and safe.

Affected person receptivity has additionally been “very, very optimistic,” with 97% reporting that their expertise with the AI instrument was nearly as good as a typical appointment or higher. “We attribute that to the doctor’s alternative to interact with them otherwise and extra deliberately in the course of the go to, which is commonly compressed by time,” mentioned Kearns. 

Trying forward, he mentioned the instrument might be utilized in eventualities like biomarker detection and social determinants of well being (non-medical points that affect an individual’s well being, corresponding to undernourishment or lack of transportation. Additional, orders, pre-authorization and referral letters might be cumbersome; one aim down the road is to have Dragon Copilot set off post-visit actions. 

“There’s loads coming down the street within the documentation seize house,” mentioned Kearns. 

Digital teammates to beat staffing points

One other space the place TOH is incorporating AI is “digital teammates.” Final summer time, the ability developed use circumstances with Deloitte and launched Sophie, who speaks a number of languages. 

Curiously, she’s in a position to interpret affected person sentiment and behavioral response. 

“I hate to say this, however sufferers deceive clinicians,” Kearns acknowledged. “You already know, ‘What’s your ache scale?’ ‘Oh, I’m doing fantastic, it’s 5 out of 10.’ However then Sophie’s going to have the power to try your face and go, ‘Properly, it doesn’t seem like it’s a 5.’”

She might then go on to ask what the affected person’s definition of 5 is, and might pivot primarily based on goal knowledge. 

TOH will launch one other avatar within the spring to assist sufferers navigate and entry the healthcare system and interact in pre-screening. 

“I don’t suppose anyone’s not conscious of the well being care human useful resource disaster that exists globally,” Kearns famous. “We wish to actually attempt to help and serve sufferers extra robustly than we will immediately.” 

For example, the ability makes follow-up calls to sufferers after sure procedures. Nonetheless, resulting from useful resource limitations, they’ll solely observe up with the highest-risk sufferers. A aim for Kearns is following up with each single affected person, with the avatar offering clarification on any affected person questions and confirming in the event that they understood their discharge orders, had been in a position to get to the pharmacy, or have been following clinician orders. The avatar can escalate to a nurse or the affected person’s medical staff if wanted. 

“One of many issues that healthcare prides itself on is human contact,” mentioned Kearns. “It is a technique to be sure that we’re maximizing and optimizing these human contact assets, but in addition ensuring that sufferers are properly supported by way of their healthcare journey.” 

Nonetheless, he famous that it’s nonetheless early within the recreation. One vital future step is enabling the digital teammate to work together with the Epic info and surroundings. 

“We’ve acquired much more work to do there, we’re nonetheless very centered on the adoption aspect,” mentioned Kearns. “We’re nonetheless a healthcare system that’s very reactive to the well being standing of sufferers. We’d prefer to get to the place the place we’re proactive.”