Data Science in action

Processmining a gamechanger in Healthcare

Ten years ago I made an interesting discovery. I worked as an advisory physician at a large health insurance company in The Netherlands. Health insurers agree with providers on waiting times and access times for interventions and surgery. These intervals should not be too long, after all. Theory and practice are not always tracked. That was clear from the calls for help to the company. Usually it was then resolved on an individual level. But situation at the macro level remained unsatisfactory. I found out that the databases that were provided by hospitals as part of the declaration procedure, were a source of more information. Unfortunately, these were anything but real-time data.

The data were far too late which had to do with the flaws of the (DBC) declaration system. Anyway, at the first outpatient contact at the hospital a DBC is opened on a certain date (timestamp). The (planned) surgery followed later and which was also linked to a certain date. The period in between could be described as a waiting time. Of course this may be biased because the patient would first like to go on vacation before surgery. That means there is always a certain bandwidth, but the differences between hospitals are obvious. To my disappointment it remained at a fun exercise. The time was apparently not ripe for an active feedback from insurer to the provider on these data.Nowadays indicators are more important than ever before. 

Big Data

After my involuntary departure from the healthcare insurer company my interest remains strong. Meanwhile, the term big data has become commonplace. I made contact with the research group (Eindhoven) that deals with mining processes and develops expert software. The Netherlands plays an important role with Professor Wil van der Aalst (TUe) in processmining. He has numerous publications to his name including a manual about Process Mining (Publisher Springer). Wil van der Aalst was also the driving force behind a recent Open Online Course (MOOC) on the topic. For me it turned out a bridge too far, but the PetriNet I will never forget.

Drilling down into the data


I would like to draw your attention to an (English) handy booklet from the series Springer Briefs entitled Process Mining in Healthcare. The book of about 100 pages, is written by three authors (Mans, Vanwersch and van der Aalst). It has six chapters. I make a jump to the second half of the book, which explains the importance of the quality of the data entered (garbage in, garbage out). An enumeration makes that clear. It explains what information is relevant. On a timeline the arrival of the patient, the consultation by the doctor, the end of the consultation, departure of the patient are all important issues. In order to bring theory and practice together, an exercise was done by the authors using large databases. Maastricht (MUMC) yielded files over the period 2008-2013 with 1206 eye surgeries (cataract) and 296 bowel surgeries (carcinoma).

The AMC produced a file over the 2010-2013 period with 232 surgeries (cataract). This produces a number of decision trees which make it very clear once again how complex healthcare can be. The same can be seen in something relatively simple like a medication order in a hospital (intramural). The number of variables is huge as you can read in tables supporting the text of the book!


In an epilogue the authors pointed out how important this approach can be. They point to the behavioral care processes. They see an important role for this process mining technique to detect bottlenecks, inefficiencies, compliance issues, variations, performance and ultimately result in better healthcare processes.

Mijn gekozen waardering € -