Unclogging the information arteries of the NHS

How two Trusts are taking a data-driven approach to optimise processes

If you live in the UK, the following will be all too familiar. Your phone pings and you see an SMS confirming the date of your NHS GP appointment. Then a letter arrives saying it has been postponed. A lengthy wait on hold ensues as you try to talk to someone to find out which is correct – the text or the letter.

"The NHS can be frustrating to the point of insanity," said Jake Koris, an orthopaedic surgeon and National Clinical Fellow working in Oxfordshire. "My hairdresser probably has better systems in place for making sure appointments happen when they should."

To be fair, a hairdresser has a much simpler operational model than the NHS, one of the world's largest and most complex organisations. This complexity is at the root of many of the issues that cause frustration to citizens and health workers alike.

In complex systems, a change in once place can affect other areas in ways that are very hard to predict - the butterfly effect in action.

Optimising processes in such systems, so that messages arrive in the right order, medicines are delivered on time and operations do not regularly overrun because of supply chain or scheduling problems is difficult. But it's not impossible, said Koris, a practitioner with a long-time interest in continuous improvement.

Koris has worked with National Medical Director Steve Powis, as well as the Getting It Right First Time NHS programme, and is a strong believer in the power of data. "We do evidence-based medicine in terms of how we treat people with individual medicines," he said. "How we move patients through our system, how we organise our system and how we improve our system should also be evidence-based."

Patients flow through the NHS, from referral to discharge. But these flows, or pathways as they are known, cross multiple functional and organisational boundaries, which is where things can start to go wrong. Decisions made by clinical experts in one area don't always take account of what's happening in other areas. No-one is seeing the big picture. Which is what led Koris to look at process mining.

"I realised that, actually, you can start to put numbers on it."

With those numbers, you can ultimately create graphical, interactive dashboards, which allow you to see—at a glance and across boundaries—where the bottlenecks lie.

"The whole pathway can only operate as quickly as its slowest step. So being able to quantify that and say ‘look here's where everyone's getting stuck' means that we can then introduce a change to unclog the system."

As an example, process mining could reveal that operations are often overrunning because of the time taken to sterilise the instruments: "Maybe we need to invest in a second sterilisation machine, or maybe it's something else," Koris said. It provides a starting point for further interventions.

"We've been through an era of putting money behind things when there's little real evidence that it's going to be beneficial. We don't have that privilege any more".

What is process mining?

Process mining is a field of data science that involves trawling event log data to analyse business processes, looking for patterns and hidden connections. The log data can come from enterprise software systems such as ERP, CRM or BI, or in the case of the NHS, pseudonymised data from the electronic patient record (EPR).

Major enterprise software vendors such as IBM, Oracle and SAP have data mining solutions in their portfolio. There are also specialist vendors, including the German firm Celonis, which was the provider adopted by Koris and his colleagues.

Robotic process automation (RPA) vendors, such as UIPath, also offer process mining functionality. Indeed, process mining can be considered a front end to RPA. The underlying causes of bottlenecks uncovered by process mining can be eased using RPA to de-complexify problem areas.

"We've got the insight that here's where the problem is," Koris explained "We can automate the process after the fact to improve it. And then you could essentially re-audit [the automated process] using process mining again."

University Hospitals Coventry and Warwickshire: an NHS centre of excellence

Oxfordshire wasn't the first region to adopt Celonis. That was up the M40 at the University Hospitals Coventry and Warwickshire (UHCW) NHS Trust, where director of performance and informatics Dan Hayes and his team introduced it, in part to tackle the Trust's historically high waiting times. At the time, the Trust's post-Covid waiting lists had soared from 35,000 in 2019 to 72,000 people in 2022.

"Obviously, we couldn't carry on as we were," said Hayes.

UHCW adopted Celonis through its existing management partner, IBM, in order to gain some insight into the throughput problems. As a "data mature Trust," Hayes said UHCW had the right mix of expertise to make it work, and importantly strong backing from the top.

"Our chief executive and chief medical officer are very driven by the technical side of things," he said. "Having that support from the get-go was very important."

Equally essential, the Trust has several data scientists and analysts who are adept at putting the results into context and presenting them to various stakeholders. UHCW is now an NHS process mining centre of excellence.

The first intervention was a deceptively simple one. The SMS reminder schedule was adjusted from four days prior to an outpatient appointment with a one-day-before follow-up to 14 days with another message four days before the appointment, giving patients more time to prepare. At a stroke, this reduced non-attendance rates from 10% to 4.4%. This change may have been simple, but it wasn't obvious. "Without process mining we'd never have seen it," said Hayes.

The additional capacity released by this simple change saw waiting lists start to fall, week on week, eventually reducing by 5,000 patients. "That was a massive step in the right direction," Hayes remarked.

Process mining unveiled several other surprising statistics, too. While most patients follow a standard outpatient pathway (referral, diagnosis, intervention, prescription, release), a huge number do not.

"We found we had 130,000 different variations of how patients come into the hospitals," said Hayes. "That was a bit of an eye-opener.

"To be able to improve our patient experience and our performance and reduce our waiting lists and provide better care we needed to reduce that variation."

In another revelation, 14,000 patients were found to be on multiple pathways.

"There might be a patient who's waiting for a cardiology appointment, waiting for a dermatology appointment and waiting for a diagnostic test," Hayes explained. "What invariably happens is they're all scheduled by separate teams, which means that those appointments could be on three different days.

"Celonis has helped us pull those patients out," he added. "Our next step of the journey is to align them all together, so that they can come in for their dermatology, cardiology and have their diagnostic tests on the same day."

Process mining has also made it possible to schedule families to attend appointments together, avoiding people having to take time off work unnecessarily. And operating theatre operations are starting to be rescheduled using object-centric process mining, a modelling technique that can map many-to-many variables, to minimise overruns.

Next up will be an analysis of the prostate cancer patent pathway, followed by urgent and emergency care, from which department patients follow a huge number of pathways to different wards and for different treatments. "We think that the data that we could pull from that would be really intuitive and a good use of process mining," Hayes said.

It's not just about technology

The NHS has suffered more than its fair share of expensive, top-down IT failures, not least NPfIT and care.data. Because of this, new proposals can be met with suspicion and resistance, said Koris. "There are always 12 reasons what we can't possibly do that particular thing in this particular hospital."

As an evangelist for data-based decision-making, Koris has learned that the bottom-up, show-and-tell approach is far more effective than trying to impose change. Rather than charts and mindmaps, he gets one of the engineers to present a live demo to the finance people or managers.

"I've changed the way that I approached those conversations, and actually, the results from Coventry now speak for themselves. When you're talking to the correct person, you just see their eyes light up as they realise the potential.

"Immediately they see here's how patients are moving through the system. And then you zoom out and show them the layers upon layers, and then you zoom back in. People suddenly understand: it's not guesswork, it's a real pathway, it's what's actually happening in their hospital. And then you can start talking about what you need to change."

We asked Celonis about the value and duration of the process mining contract but the company had not responded prior to publication. UHCW said it did not know the value of the contract as it is funded by NHS England.