AI-driven online patient triage introduced at Priory Medical Group in York proved to be a ‘game changer’ – patients were able to get urgent care much faster and it led to £300k savings. Beth Gault finds out more
In 2019, Priory Medical Group in York reached a critical point when it came to its routine health care. The waiting time for appointments was four weeks plus, which had led patients to ‘game’ the system, according to managing partner Martin Eades.
‘They would ring for a routine appointment, have to wait for 20 minutes to get through, and then be told there was nothing available before four or five weeks,’ says Martin.
‘So, the patient would put the phone down, ring again, and then ask for an urgent appointment,’ he adds. ‘The knock-on effect was that urgent sessions were sometimes 50% full with routine care concerns.’
This had an impact on everyone. Patients, who Martin believes did feel bad about playing the system, were frustrated and desperate because they felt there was no alternative if they wanted to be seen more quickly. Clinicians were also dissatisfied because they were dealing with lots of routine cases during urgent care sessions. And finally, the practice’s call handlers and receptionists were stressed because it fell to them to deal with a ‘significant number’ of complaints about the whole process.
It was a turning point for the 58,000 patient single-practice PCN. ‘We decided we had to reinvent patient access to our clinical services,’ says Martin.
The first step was a review of all the practice’s clinical pathways and where patients were directed, which led to the discovery that many queries and concerns just ‘bounced around’ the system.
‘A request would come in and go to a receptionist, who would then set up a GP appointment only for it to be bounced back to reception and then sent to a musculoskeletal practitioner, for example,’ says Martin. ‘The patient would see the right person eventually, but the process was very inefficient.’
The practice had already hired a number of roles under the additional roles reimbursement scheme (ARRS), including advanced nurse practitioners, allied health professionals, paramedics, and musculoskeletal practitioners, due to pressures on GP recruitment and retiring doctors. However, the team was not yet making best use of these newer roles, Martin explains.
‘It was affecting morale internally. GPs were overloaded while there were other clinical roles that were underutilised, particularly some of the nursing and HCA roles.’
The practice turned to the power of tech to find a digital solution to streamline pathways, boost efficiency and make better use of staff. In the end, Priory opted to introduce AI-driven online triage software called Klinik.
The tool, widely used in Finland, aims to reduce the burden on practice staff and ensure patients get to the right clinician. It works by directing all patients requesting an appointment – whether they call in, walk-in or come via the practice website – to an online system that asks a series of health questions about symptoms. The algorithm then uses this and other indicators such as a person’s age and gender to compare with its database of 1,000 conditions to determine urgency and avenue of treatment.
Reducing the strain on GPs
‘The system should be able to make an assessment like a doctor, take a full history and direct a patient to the right person.’ says Dr Rony Lindell, GP and director of medical development at Klinik.
‘Based on our case studies, around 40% of patients should be seen by a GP,’ explains Rony from Klinik. ‘And around 60% shouldn’t. A lot are directed to admin staff, but a big proportion go to physiotherapists or mental health workers.’
Without the triage in place anything between 80 to 100% of patient enquiries go to a GP, he adds, so the difference is dramatic and frees up a lot of time for clinicians.
The practice worked with Klinik to map out the services and staff that were available, to ensure patients were directed to the relevant healthcare professional. It then went live with the new tool during the pandemic (November 2020) when there were fewer face-to-face appointments. Within two months, 70% of enquiries went directly via the digital tool, says Martin. Previously, 99% of contacts came through via the telephone and 1% through the front desk.
Martin says he was ‘amazed’ with how quickly patients switched to using the tool. He believes its success was partly due to the timely response patients received. For routine cases, they would be sent a text message or email to let them know when a clinician would be in touch but for urgent cases, they would receive a phone call asking them to come in that afternoon if needed.
‘The patients got the help they needed in a timeframe that was appropriate to their issue, which meant they were happy to keep using the system,’ says Martin.
Now, more than two years on, the practice has 86% of enquiries coming directly through the digital tool, bypassing the telephone systems or the front desk.
Martin stresses that all forms of access (phone and walk-in) are still available to patients in order to be digitally inclusive. But whatever the method they use to make an initial enquiry, patients all go through the same AI tool. So, if they attend in-person they are asked to fill in the online form themselves with support from a receptionist or if they call in, the receptionist will input information into the tool on the patients’ behalf.
Improvement in patient satisfaction
While AI has made a difference to the day-to-day running of the practice, there have been important added benefits – it has contributed to increased morale among staff and soaring patient satisfaction.
‘When we were still using the old model, patient satisfaction was around 27%, which is pretty awful,’ Martin says. ‘But now it’s over 83%.
‘One of the biggest improvements has been that queries are no longer bouncing around the system, and we’ve reduced the number of tasks that land on receptionists’ desks by 50%. In a practice our size, that could equate to around 20,000 tasks a month.’
This freeing up of time for clinicians and non-clinicians has meant the practice can be more proactive in the care of their patients, such as ringing up about QOF issues, chasing patients to attend flu clinics, and helping people to access Covid vaccinations.
It’s hard to measure whether overhauling the patient access system can be directly linked to retention of staff, but Martin says staff turnover has been ‘incredibly low’ for the last two or three years.
‘The model of care has absolutely made a difference,’ he adds.
Though the results have been striking, they haven’t come without a cost to the practice.
There was a ‘huge investment’ in the whole process – £70,000 in total – a sum that included the AI tool itself, the cost of the practice creating its own training materials for staff, as well as for project management expertise and business analysis.
It’s not just about paying for the software, the full outlay goes beyond that, Martin explains. ‘This wasn’t just a digital tool that we bolted on. It was absolutely ingrained in what we were doing.
‘If the system hadn’t been set up properly, or patients if had an inconsistent experience, we knew they wouldn’t use it again. It had to work first time, so there was a huge investment in training and education of the team,’ adds Martin.
That extra work and cost paid off. Analysis by York Health Economics Consortium has suggested that the move has resulted in around £300k in annual capacity releasing savings for the practice. This figure was based on comparing the 12-month period before Klinik was introduced (December 2019 to November 2020) with the 12-month period after the system was put in place (December 2020 to November 2021).
Using data to reshape working practices
With tech systems come availability of data. As the AI software is used more and more, it gathers a growing amount of information for the practice.
‘It can be seen a as a bit of a risk because we are now really reliant on that data for analysis’, says Martin. ‘It dictates how we organise our workforce, our working hours, and our premises.’
For example, as the tool is a digital one, patients can use it to put through their requests 24/7 – unlike traditional telephone queries to a practice. After implanting the AI software, Priory identified that a lot of prescription requests were coming through during the weekend. Based on that new insight, the practice felt it had to adapt, recruiting more prescribing team members and extending their working days to seven days a week.
‘They work a modest number of hours on a Saturday and Sunday, but it means that by Monday morning, the team hasn’t got a huge spike of prescribing work to get through, which previously made Mondays awful,’ says Martin.
What also became much more visible was the level of ‘true patient demand’ the practice faces. It was an ‘uncomfortable’ truth, says Martin.
‘Once your phone lines are busy, that’s it. The unmet demand sitting behind it is hidden,’ says Martin. ‘That’s not the case with online you can see a truer level of demand. It might not be comfortable viewing for practices but it helps shape your services around meeting the needs of your population.’
Martin isn’t concerned about the tool potentially opening the ‘floodgates’ to patients getting in touch with the practice.
‘If you don’t know genuine demand, how can you shape your services around your patient’s needs?’
However, currently, the practice is trialling switching the AI system off at the weekend to see what the impact is – if the demand goes elsewhere or waits until the practice reopens on Monday morning. This is still being assessed, Martin says.
Despite the costs and the risks involved, the practice says it would not go back to the old model – not just because patient satisfaction is higher but because the insight it provides is invaluable.
‘The data we are getting at the back end of the system goes beyond anything we’ve had before from our medical record system. The level of understanding of what presents when and how is absolutely vital from a managerial decision-making process,’ says Martin.