A life-changing project

Rachael Pells profiles Medica Reporting’s teleradiology AI project (named 2021 Overall Project of the Year by APM), which is making a vital difference to stroke victims.

When it comes to strokes, there’s a well‑known saying among medical professionals: time is brain. The quicker a stroke is recognised and the quicker the patient receives emergency care, the better their chances of recovery. Just like a serious traffic accident or a blazing fire, speeding up response times by just a few minutes can be the difference between slow recovery and good recovery, and even life and death.

It’s no wonder, then, that when presented with a new tool that could help to reduce that waiting time for patients – and ultimately save lives – teleradiology provider Medica Reporting took a risk and seized the opportunity. Last year, that risk paid off: Medica became the first company in the UK to use artificial intelligence (AI) to help identify patients suffering from an intracranial haemorrhage, a life‑threatening form of stroke causing bleeding within the skull.

Teleradiology is the sharing of patient images such as X‑rays, CT scans and MRI reports from one location to another. It allows medical specialists to provide expert advice in the form of a report, even in cases where a patient is thousands of miles away. Previously, Medica’s overnight emergency support service, NightHawk, was able to get radiology reports back to trauma wards in just 23 minutes on average. Since successfully introducing the algorithm qER into its existing workflow, that average has been reduced by one minute – which really matters when it comes to brain bleeds – and in some cases, reporting time has been brought down to around 17 minutes or less.

It’s a huge achievement, and one that was recognised at the 2021 APM Project Management Awards, where the Medica team won both Technology Project of the Year and Overall Project of the Year. The judges praised the team for their outstanding execution of the project, but also their passion and supportive team ethos in achieving their goal.

For chief information officer Marc O’Brien, whose team led the project, rapid reporting times have always been a key priority where intracranial haemorrhages are concerned, and so adopting machine learning to help facilitate that felt like a natural step. “We are always looking at ways of improving our service, and AI is something that’s been talked about more and more in the sector,” he explains. “We decided to examine the market and see what AI could do for us – if there was something that could make the process faster and the doctors’ lives easier. All of those things make a big difference to patient care.”

For a haemorrhage patient to receive emergency treatment, first a detailed scan of the affected area of the body must take place using an MRI or CT scanner. These images are sent to radiologists, the experts who can diagnose and determine the severity of the problem, which will in turn guide doctors on which emergency procedure or treatment to undertake. For many NHS hospitals, where resources may be stretched and radiologists aren’t always on hand, it’s the hours between 5pm and 8am that can be especially challenging for tackling emergency trauma cases.

This is where Medica plays a role: its NightHawk service connects NHS trusts to out‑of‑hours teams of radiologists, who operate remotely from around the world to support emergency care staff by identifying hard‑to‑detect haemorrhages and providing a report for the doctors on the ground. Part of that service includes a vast digital platform for the rapid sharing of patient scans: these are data‑rich and can contain hundreds, if not thousands, of individual images of each scan, rendering email attachments or other web transfers wildly insufficient.

The service has been improved even more thanks to the new AI program, qER, which uses a carefully tuned algorithm to recognise when a patient scan shows signs of head trauma, such as intracranial haemorrhage, and these time‑sensitive cases are pushed to the front of the reporting queue. The effect is that highly skilled NightHawk teams don’t waste time determining priority, and the highest‑risk cases are dealt with first.

A cutting‑edge tool to support clinicians

The word ‘groundbreaking’ is used so often when describing new technology that it borders on cliché, but in the case of qER, it seems especially fitting. Debate around the potential value of AI and automation in healthcare has heated up in recent years. For many, the creeping inclusion of AI within medicine poses a threat – the idea that machines could replace any aspect of human expertise.

This is why O’Brien and his colleagues make a clear distinction: the tool is designed to be something that can undeniably support experts, but never replace them. “We only ever investigated AI from a patient prioritisation perspective, never from a diagnostic perspective – and that decision was influenced by the conversations we’ve had with doctors over the years,” he says. How does he respond to those still sceptical of AI in healthcare? “If, god forbid, anyone in my family were to ever have an intracranial haemorrhage, I’d be really glad that AI was there to push them to the top of the prioritisation list.”

The qER program was designed by independent technology developer Qure.ai, which specialises in AI solutions for healthcare. O’Brien and colleagues first met the Qure.ai team at a major conference hosted by the Radiology Society of North America in Chicago in 2019. The algorithm had already been created and peer‑reviewed in The Lancet medical journal, which helped assure Medica of its quality. After a series of positive meetings between the two organisations, Medica set about testing the product before confirming the partnership.

The necessity of a clear business case

“One of the first things we needed to do was a proof of concept,” explains David Evans, functional architect for Medica. “This was to make sure that the tool would fit easily into existing workflows and facilitate what our teams do, that it wouldn’t slow them down. This is especially important when trying to report CT head scans with potential intracranial haemorrhages, where every minute counts.”

The team also wanted to make sure that they were clinically happy with the results they were seeing through the tool. “Because we’re a clinical company, we usually take a sceptical approach and like to prove the value and accuracy of tools ourselves,” O’Brien says. For this, Medica analysts took hundreds of studies and ran them through the algorithm. “We also ran them past some of our clinical reporters and compared the results,” says O’Brien. “Even when we were very satisfied, we took a historical look back over around 500 cases. We were very happy with what the algorithm did and that it matched with what the humans told us, which gave us the confidence to pilot it.”

From here, a dedicated, cross‑discipline project team was put together, including a clinical director, IT directors and radiology reporters, led by project manager Samantha Davey. Reflecting on that time, Davey cites “having a clear and informed business case” as one of the key factors that drove her team to be successful. “We knew exactly what our success criteria were from the outset and that helped to inform our pilot,” she adds. “There was no point in having something that was 100 per cent perfect but took 25 minutes [to get reports back].”

Strict criteria and benefits realisation strategy

The tool also needed to fit within the NightHawk workflow – which was already producing a rapid turnaround for acute cases – but there was no value in implementing something fast that produced errors. Similarly, Davey adds, “there was no point having something that demanded a level of quality in terms of the images we received from clients that was at a level above what they could actually produce, because that would not be workable. This was the range of criteria which we identified early on and also linked to our benefits realisation strategy.”

In early 2020, Davey and Evans put a timetable in place with a strict deadline of 7 December that year (“avoiding Christmas annual leave challenges and other pressures on the system”) and remained determined to hit that deadline even when the Covid‑19 pandemic hit. “Working backwards, we could be clear where our milestones were, where there was a little bit of flexibility and where there was absolutely no flexibility at all,” says Davey.

She also deployed task‑and‑finish groups for specific technical and qualitative outcomes – this was useful, for example, when the algorithm was ready to be tested but not ready for use in the real working environment. “This involved releasing the algorithm into the live environment, but we didn’t want anyone to use it, we simply wanted to be able to look at it and to comment on it,” Davey explains. “So a small group was tasked with considering how we were going to do this, deciding at which point in the calendar we should do it, who we were going to engage with it and who was then going to communicate this information out there.”

Keeping the team motivated

Even for a company with roots in remote technology, Covid threw up challenges. “We went from office‑based to fully remote in three days,” says O’Brien. “All our staff were carrying screens and machines down to their cars. It was a huge operation, but we got it up and running very quickly, which was a huge achievement for the team.”

Meanwhile, Medica’s ambitious growth strategy meant a number of other projects were taking place at the same time, including a major overhaul of the technology platform used by Medica for its medical image sharing. “One of the real challenges was the constant prioritisation of resources – determining which aspect we needed to be allocating our human resources to at what point, and ensuring that those people didn’t get exhausted or become completely confused because of differing priorities. It was also important to keep the enthusiasm and excitement about the project going.”

Motivating a team through a long‑term intensive project can be difficult at the best of times, but Covid‑19 lockdowns exacerbated the challenge during a time when employees were stuck at home and life became repetitive. The solution to this, according to Davey, was “clear communication” between colleagues from start to finish: “Not waiting for scheduled meetings to check in to see how something was going, to make sure everything was on track, making sure that people were acknowledged. We spent quite a bit of time just updating people so that they didn’t start to feel disenfranchised.”

Delivering something with real purpose

Ultimately, what also helped to pull team members together was the shared end goal: that once the algorithm was successfully implemented into the system, Medica would be helping its clients to save lives. “Everyone involved knew that they were impacting on and influencing the delivery of something important,” says Evans. “There was energy and excitement for this, which extended from the company executives downwards.”

Speaking to O’Brien, it’s abundantly clear that nurturing a supportive and friendly working atmosphere is central to Medica – and he believes it’s also a key factor in the company’s success. This is especially important in the medical healthcare sector, he says: “What we do is stressful. We are moving at a hundred miles an hour and growing at around 20 per cent per annum, which sounds great, but it does create issues, because you’re always slightly short‑staffed.

“We needed ambition and drive, but I also wanted people to be kind to each other,” he concludes. “Getting on and working well with your colleagues is really important and motivating in itself, and I feel proud to have that ethos in place at Medica.”

Lessons learned

One of the biggest learning curves was the realisation that the technology itself would not pose the biggest challenge, but rather how and where to fit it into the existing service in a way that felt seamless, Medica’s functional architect David Evans reflects. “From a technical and design perspective, deploying AI is not all about the algorithm. Understanding your workflow and how the technology is going to fit into your workflow is just as important as the actual algorithm itself.”

From a managerial perspective, “understanding what you’re hoping to achieve from the project” should be foremost in a project manager’s mind, says Medica project manager Samantha Davey. “Ask: what are your stakeholders’ expectations? What are the benefits that you’re hoping to deliver and how are you going to measure those?”

When putting a team together, think personality as well as skills: “You’ve got to have people on board with the right skills. These people should be identified up front and you need to have them in the room from the start. But something else I try to build in my team is having staff who are personable,” says chief information officer Marc O’Brien.