HTN was joined by a panel of experts from across the health sector for a focused webinar on the use of ambient scribe technology in NHS trusts. Panellists included Lauren Riddle, transformation programme manager at Hampshire and Isle of Wight Healthcare (HIoW); Ynez Symonds, CNIO at HIoW; Dom Pimenta, co-founder and CEO at Tortus AI; and Stuart Kyle, consultant rheumatologist and clinical lead for outpatient transformation at Royal Devon University Hospital.
Our panel discussed the practicalities and considerations for ambient scribe implementations, from operating procedures and policies, integration and functionality, through to best practices around patient-practitioner interactions.
Each of our panellists offered a brief introduction and overview of progress to date, with Ynez noting how one of HIoW’s main ambitions is to support its clinical staff with a reduction in admin burden and time spent on clinical notes outside of their working day. “Lauren and I kicked things off with an AI workshop in January, to understand from a trust perspective what our priorities were,” she told us. “Rather than going straight to market, we spent a lot of time working with our services to understand what we were trying to fix for them.”
The trust has now been piloting ambient voice technology (AVT) for just over six months, Ynez continued, “and from an organisational point of view, I don’t think we’ve struggled with clinician input – all of our clinicians have been really keen”. The pilot has gone out to seven divisions and 26 different profession types and has received some excellent feedback, “but one of the things I’m personally most proud of is the feedback we’ve had from patients on their experience and what it feels like for them to go to a clinical consultation with a clinician using AVT”.
Lauren followed on from Ynez with some of her own experiences and insights from HIoW’s journey to date. “We were already looking at reducing wait times across our services,” she said, “so it wasn’t like we’d seen something shiny and wanted to make it fit; it was very much that there was a need and we wanted to try and solve it.” After being “blown away” reading about a pilot of Tortus AI with several trusts at the end of last year, the decision was made to look into the technology, she went on. “We started small, with 30 people or so, in children’s, and it grew very rapidly from there. We were a victim of our own success, if you like, because the engagement was so good we could have rolled out to 16,000 people there and then.”
Royal Devon became early adopters after seeing the potential of AVT and deciding to pilot several different products in several different areas, Stuart noted. “We set out our prerequisites and went through a formal procurement with G-Cloud,” he explained. “We procured Tortus, and the business case was predicated on productivity based on some of the stats coming out of the US and from Tortus, looking at the template of the clinics where we thought there would be that productivity gain, with five specialities, including cardiology in our initial implementation phase.” So far, the trust has had over 500 consultations, with plans to roll the technology out across all outpatient specialities by the end of the financial year.
Dom talked about his 16 years of working in the NHS as an internal medicine physician and cardiologist and making the move to found Tortus three years ago to tackle the “tremendous amount of work” that the digitisation of medicine was adding for clinicians. “It’s been about how we reduce the friction between the computer and the human clinician,” he shared. “That’s what we’ve been doing with AVT – we were the first to launch that in the UK about two years ago, the first to become a medical device, and now the first to deploy at scale in several hospital sites.”
Trials of Tortus’s solution to date have yielded a 20 – 25 percent reduction in terms of time taken to document and complete a patient consultation, according to Dom. “But imagine what we could do if we got that down to 50 percent and suddenly you’re doubling capacity and time with patients. I think that’s why we’re seeing such a pull around AVT, because for the first time in history, here’s a technology that clinicians actually want, that the system actually gets benefits from – the two things don’t often align.”
Implementation, challenges, workflow
Rolling out with multiple suppliers for HIoW’s pilot was “really, really difficult”, Lauren reflected. “It was the most complex pilot I’ve ever been involved in, but I wouldn’t change that – it means we have some very rich data, a strong understanding of what our clinicians need, and an understanding of the different features and functionality that can be provided.” Now that other trusts are moving ahead with their own pilots, Lauren’s team also has a lot of learning to share, she said. “I echo what Dom was saying about the time saving, but we’re actually seeing more like 40 to 50 percent time saved in our sample group, sometimes even higher.”
HIoW is now at the stage where it needs to proceed with the writing of its business case and consider what it might look like to scale across the organisation, Ynez shared. One of the major challenges is in developing a model suitable for the trust’s diverse range of services, she went on, “and whilst it’s great we’ve got 40 – 50 percent productivity gains in some areas, in others it’s quite minimal, so we need to plateau that out and understand what it looks like across the entire organisation”.
There’s a balance to be found, according to Ynez, between increasing productivity and looking after staff as a result of the technology provided to them to utilise as part of their skills. “One of the things we’ve found is that it’s really highlighted the discrepancy in digital literacy across our organisation, and it’s important to look at how we train our current and future workforce to support those who are fluent in using technology and those who are not. Lauren has been a huge asset to this, and I can’t thank her enough because our pilot has gone really well.”
Looking back on Tortus’s pilots across the country in a variety of different care settings, Dom referred to making “a bunch of mistakes” in the early days, including stretching too thin and failing to support the workflow. What every successful implementation has had in common, he offered, was that it had a team behind it, tending to have a clinical champion and an operator to help drive it through the organisation. “The best implementation we’ve seen is St George’s A&E, where 40 clinicians all trained together, worked together, and built the template together. I was on the WhatsApp group during their pilot, and they were doing their own tech support, and that’s very powerful because suddenly everyone’s in it together.”
Underestimating workflow changes and how they derive productivity was another mistake Dom highlighted. “For example, in outpatients we saw a big time saving, but you don’t really get any productivity from that from a system perspective unless you plan for it and bring more patients in. In A&E, however, demand is infinite, and being able to cope with that, reduce stress on the shift, and reduce waiting times moves the needle without the need to think about it.” Future challenges when looking at scaling include hallucinations, emissions, and error rates, Dom shared, “because a one percent error rate across 100 consultations per day is manageable, but that error rate for 40,000 consultations becomes 400 errors per day going into your clinical systems”. Tortus is currently working to address this with enterprise-grade safety and security.
Royal Devon is the first trust to run a fully integrated AVT through its EPR, Stuart said. “We launch from within Epic, and then everything is pulled back into the EPR note. That’s great in principle, but getting the product exactly how you want it back into the EPR is quite tricky.” One issue has been that formatting can disappear, leaving everything looking “a bit piecemeal”, he elaborated.
Stuart agreed with Dom’s point about needing a champion to drive engagement, noting how losing that individual to sickness for the cardiology implementation meant it hadn’t been pushed as much as it should have been, and how gynaecology has gone “much smoother” with a champion at the helm. “Another key relationship is with your supplier because if they’re not responsive, the technology is moving so quickly that you won’t be able to keep up with it – it’s key to have a good structure in place, linking in with the providers you’re using.”
“We’re seeing exponential advances in accuracy, speed, cost, latency, and capability,” Dom agreed. “And having your own clinical AI team is the feeling you should be getting when working with your provider of choice. Once you’ve got there, people actually love it. They lock in and do the same practice day in, day out, 30 – 40 times per day. So it’s well worth the front-end energy of making that work.”
Lauren picked up on Dom and Stuart’s points around workflow issues, recommending having change management or transformation support in place working with specialities and services. “That’s something we didn’t particularly do, but we absolutely will do going forward,” she noted. “As a result of that, we did learn a lot. In a lot of our specialities, for example, they literally lifted out dictation and swapped it for AVT, which meant more work for business support colleagues as the generated summaries were longer than before. We also saw the same issue as Stuart around formatting, and it took us a long time to learn there’s a little button that allows you to copy with formatting, so it was showing as a big chunk of text.”
When implementing, understanding the clinical workflow is key, Stuart told us, not only because it helps with the clinical safety case and hazards workshop when talking about potential impacts and mitigations. Having the trust’s clinical safety officer, as well as Tortus’s clinical safety officer, engaged from the beginning to walk through that process made it much easier, he said. “You can also use it as a real opportunity to properly delve into the clinical risks around what you’re doing.”
Risks observed at HIoW have included inaccuracies and omissions, as Dom mentioned, Laura highlighted. “The mitigation for that is having the clinician check every summary. I always say that it’s the first draft – you’re not going to have the perfect summary generated first-time to go straight into your EPR.” Omissions tended to be down to the template being used, which again comes back to understanding the workflow, she added. “The template and an understanding of the workflow are the two key things. One of the things we saw was that although clinicians were asking about consent, that was not being accurately recorded in the AI-generated summary. When looking back on that, we realised it was because the template didn’t know to look for it, so we made a very small tweak.”
Key takeaways and the road ahead
A key takeaway for Dom would be not trying to do too much or go too far, too fast. “When you’re trying to do ten times more, you’re going to struggle if things start to go wrong, and the scale of the errors is going to be a much bigger problem. It also creates a lot of downstream tasks, so ironically, slow and steady is the best way of deploying. That’s why we’re called Tortus, because we believe in building value slowly, evidencing it, and then deploying it.”
“We’ve had some lovely patient feedback, including one patient telling us that for the first time they felt their clinician was really engaging with them,” Ynez shared. “But there are still patients who have said they don’t understand what AVT is, so we’ve talked about how we can ensure we’re sharing literature beforehand with their appointment information, to let them know they will be asked about their clinician using AVT.” The team has also worked closely with those from community engagement and PALS to help drive this forward. “One of the things we’ve tried to encourage is clinicians going through the consultation documentation with patients at the end of their appointment, so patients can see what’s in there. That also means you can produce referrals at the end of the appointment, rather than having to come back.”
For the next 12 to 18 months at HIoW, the focus will be on productivity gains and how to scale those across the organisation safely, Ynez noted. The challenge for procurement is that the solutions on the market today aren’t going to be the same in 12 to 18 months, she added, “and we don’t want to be stuck in that cycle of committing ourselves to a three-year contract only to find there are better solutions out there”. Improving digital literacy across the organisation and digital champions are also on the agenda. “Good for me looks like a healthy workforce with high levels of retention, with IT and digital not cited on exit interviews as part of the reason people are leaving,” she continued, “along with a reduction in some of our waiting lists.”
For Royal Devon, the roadmap for the next 12 months and beyond includes outpatients and ED by Christmas in some workflows, according to Stuart. “We’ll probably adopt the workflows that are most adaptable to what we’ve already got – probably triage in ED, but perhaps not more complicated trauma majors, where people tend to move around a lot more – that will come later. We’re keen to get to the point where we have integration of orders, and we think that’s where the biggest productivity for an EPR system will be, because data shows we spend most of our time in consultation putting all the orders in our EPR systems.”
We’d like to thank our panellists for taking the time to join us to share their insights on this topic.
