BPM+ Health Virtual Coffee sessions
How CQL fits into BPM+ Health

Presented by

Denis Gagne, CEO & CTO at Trisotech

Good day everybody. Welcome to this virtual coffee. Today our topic is how does CQL fits into BPM+ Health. Fundamentally, we want to address OMG’s BPM+ namely BPMN (the business process model and notation), DMN (the decision model and notation) and CMMN (the case management model and notation), and how it relates to HL7 CQL (which is the Clinical Quality Language). So, how does CQL actually fit into all this mix? I’m going to try to address this from different perspectives to generate some conversations after.

So, looking at it from a language perspective, BPMN, DMN and CMMN are what I would call visual low code languages, where CQL is a literal low code language. They are all low code in the sense that they are amicable to Subject Matter Experts. The difference is that with CQL you have to actually type the expression where in BPMN, CMMN and DMN you are more drawing pictures.

From an industry perspective, BPMN, CMMN and DMN are industry agnostic, where CQL is healthcare specific. Now, there’s a lot of value for CQL to be healthcare specific because it means that it brings together some already recognized notions. For example, CQL knows of the notion of a population versus a single patient, and it knows things that are closer to this industry. Where BPMN, CMMN and DMN don’t have at their core these notions. You have to bring them in somehow with your models.

From a main purpose kind of perspective — and that’s where I’m going to ask for some leeway — BPMN is mostly for orchestrating activities, DMN is more or mostly about deciding based on some inputs, CMMN is more about reacting within a certain context, reacting to events, and CQL, I would argue that it is more about computing healthcare measures and ratios or creating some healthcare queries, and more specifically about FHIR. But CQL is broader than just FHIR.

Bringing all this together: within the BPM+ collection of languages there is already a lot of overlaps. There are things that can be done either in BPMN or DMN or either in CMMN and BPMN. And there is a lot of overlap. But by now, we know that within the BPM+ Health community, that our sweet spot is really where we can efficiently jointly use BPMN, CMMN and DMN together. And that is where we get our best value. When we bring CQL into this mix, again we have a lot of overlaps that exist and that are possible. Here I would argue that the sweet spot for healthcare is really in that center piece where you make the best use of BPMN, CMMN, DMN and CQL all together into a combined environment.

But then the question is: Okay, but what about all those overlaps? What do we do there? Within the BPM+ community, or the model driven architecture community, we are big on the notion of separation of concerns. Which is if you have things that could be modeled in DMN or in BPMN or in CMMN and BPMN, then if it is about orchestration of activities, you should really move these into a BPMN pure sense. And when you have things that are more about reacting to events within a context, then these would be better addressed within the pure CMMN area of this diagram. And the same goes with deciding based on inputs. You could write decision in BPMN or in CMMN but it’s really not efficient, you are better to do that within the pure environment. And finally, I would argue that the same goes with CQL. Bringing CQL into the mix, your best bet is to, whenever you are computing healthcare measures or dealing with any kind of queries of healthcare, you are better to use CQL to express these. So, I would then argue globally that the sweet spot for healthcare is by combining both the separation of concern and that sweet spot where we have the best of all worlds all together. In a more practical way, we have the orchestration activities which is done in BPMN type diagrams. We have the computing of healthcare measures and ratios and other related healthcare activities that we can do in CQL. And then we have decision that we can do in DMN and we have reacting to particular events within a context that we can do in CMMN. This way, we have a single visual knowledge artifact for both clinicians and the automation. So, I would argue, that the best of all worlds is obtained by combining all these together. And, oh! Did I forget to mention about FHIR? Naturally with BPM+ combined with CQL, the underlying data information access would be all relying on FHIR, wherever you are into these processes.

View the slides