Did you think that Process Mining was only for businesses? Not at all! Process Mining also plays a major role in the optimization of processes in the Healthcare sector from waiting times in the emergency room to resource planning in surgery to hospital bed allocations. In fact, Process Mining in Healthcare has become one of the most actively researched fields within the area over the last few years and acts as a very good example of how Process Mining can be used to do good and improve crucial patient care and allocation of critical resources when you most need it. Owen Johnson is a Senior Fellow at the University of Leeds and has focused his research activities on the application of Process Mining technology to healthcare using electronic health records from the UK NHS. Current projects include process mining cancer care, musculoskeletal pathways, the detection and management of complex and rare diseases, the progression of frailty in relation to polypharmacy and falls and evaluations of accident and emergency, triage, pediatric care and an AI-based hospital command center at Bradford. Owen is an active partner of the Celonis Academic Alliance and one of the key drivers for Process Mining research within the UK. He will explain how Process Mining can create important value for society and improve major health systems and managerial procedures in place.
Everyone appreciates that data mining can have huge business value. Most businesses are concerned with how they can improve their business processes and process mining is the data mining approach that directly addresses that urgent need.
I actually discovered process mining by accident. I was working with my local hospital extracting data that would build a visual simulation of chemotherapy care and we hit on replaying all the actual events that each patient experienced. It was only after we had finished (with a really great model) that it occurred to me to see who else had done this and I discovered a huge European and global community of people doing the same sort of thing. I had found Process Mining.
Healthcare is like any other business in that it has business processes for its core operations like procurement, expenses and audit. But, of course, healthcare is mostly about caring for people so the most important processes are about delivering this care - doing it right, doing it quickly and of course keeping cost under control. In healthcare we call these “care pathways”, these are the standard pathways of care that are embodied in clinical guidelines, clinical practice and evidenced-based healthcare. Tailoring these pathways to best meet the complex clinical and personal needs of each individual is what makes a great healthcare system.
In my research we see care pathways as just another type of business process. That means we can leverage all the same process improvement approaches and technologies that businesses use. Sadly, many healthcare organizations just don’t have the tools, experience or patience to take a helicopter view of their care pathways. And in reality, things like messy data further complicate the process.
Healthcare data is a mess, that’s the biggest pain point. But there’s more to that story. The reason their data is a mess is that healthcare has been slow to adopt enterprise scale electronic health record systems and the reason for that is that healthcare is inherently more complex in multiple dimensions than typical businesses. In business, issues can be reduced to money, in health the issues are literally about life and death. The patient is both a co-producer and a “unit of production” in their own care, medicine and health are complex and fast evolving fields and care is increasingly delivered by multi-disciplinary teams working across multiple settings and organizations. Untangling the data is as much about untangling these complex relationships as it is about the data itself.
Well that’s hard. The databases behind most Electronic Health Records (EHR) systems will have 30 to 200 relational tables and anything coded with a case ID and timestamp could potentially be seen as an activity. Even limited selection for event logs will produce spaghetti models. I have two top tips that I hope might be helpful. Firstly work with your clinical reference group to produce a normative model of what they think the pathway should look like and then agree which “markers” in the database will be used to mark that that activity took place. Secondly, and this may sound simple but it will really help, apply the 80:20 rule –80% of cases will follow one of a limited number of variants (these are standard patient pathways) but the last 20% of cases may have hundreds, even thousands, of variants (these are complex patients). You can ignore this last 20% of cases – the care for complex patients with highly complex personal needs is something best left to medical experts. What we can do is make lives easier by improving the effectiveness of care for the 80% who have standard, more predictable needs.
My advice is to start simple. Introducing a process view is already novel. With some justification, healthcare can see itself as too complex for process oriented data science approaches but, collectively across the world, we are building up an impressive array of case studies and published literature that demonstrates that process mining of care pathways can be done, and can deliver innovation and fresh insights. With colleagues from other universities across the globe we set up Process Oriented Data Science for Healthcare (PODS4H – www.pods4h.com ) and will be running our third international workshop in October. With the COVID-19 crisis that will be a virtual conference so hopefully open to many more people interested in seeing how Process Mining is driving innovation in healthcare.
We have recently been working on using Process Mining to model the progression of diseases over time. That’s quite a radical re-interpretation of “a process” but the power of Process Mining is that there is a large community of interest and powerful software tools so huge scope for research and innovation. I see Healthcare as the new frontier for Process Mining. We are tackling more complex challenges than many other application fields so forcing more innovation, new methods and pushing the boundary of what Process Mining can achieve.
I am always interested in new case studies. We are launching an international comparison of emergency room/ accident and emergency pathways so interested to hear of others working in that field. A lot of recent work is around temporal changes in pathways from internal changes like software upgrades or external events. The biggest external event in healthcare, and everywhere, is the COVID-19 pandemic. Process mining of how healthcare has reacted to and adapted to the huge stress and complex demands of the pandemic will keep us busy for some years I suspect.
The COVID-19 pandemic puts healthcare center stage globally. Humanity needs better, smarter, more intelligent healthcare that embraces modern business management methods and Process Mining is one key technology that will help. It’s never been more urgent.
If you want to learn more about Process Mining in healthcare check out our blog post Can Process Mining be applied to Unconventional Processes?