Recent guidance from the Institute of Medicine talks about the need for 90 percent or more of point-of-care clinical decisions to be evidence-based. The problem is that we don’t have evidence for many of the decisions clinicians makeand in situations where the evidence does exist it is not used consistently.  

Fundamentally, there is a huge gap in availability of evidence and the ability to use it across the board, all the time for all of our patients.

While the clinical community works hard to generate evidence-based guidelines where it does not exist and update where it does, we have to look at other options that leverage the big data, analytics and super computing technologies and capabilities that are becoming more pervasive.

One possible solution I am very excited about is the concept of “practice-based evidence.” Instead of spending all our energies to understand pathology and physiology, we need to leverage the information that is collected via transactional systems (read: EHRs). These systems daily are collecting a gold mine of information on millions of patients. Add to this the variability of practice that exists and you now have a pseudo clinical trial going on across the nation.

If we start data mining this rich source of information and start looking at problems where there is no available answer or evidence, we can group actions for a particular clinical question and analyze the outcomes to help guide our decision making. At the last AMDIS Physician-Computer Connection Symposium, a presentation from Lucile Packard Children’s Hospital showed how they took this approach in trying to decide between administering anticoagulation to a young adolescent with a new diagnosis of lupus. Experience within the system showed that giving the anticoagulation led to a favorable outcome and was based on best available “practice-based evidence.” Imagine the possibilities once we combine datasets across institutions--the subject of just one recent Patient-Centered Outcomes Research Institute grant.

I’m big on making clinical decision support actionable. We need technology to help get over nonspecific, inert information that is available in clinical decision reference tools being touted as decision support. Complex event processing (CEP) engines is one example of an exciting new capability currently under implementation at the University of Chicago. We will leverage the technology to take best practices based on the evidence published or practice-based evidence from clinical databases and deliver the information to the clinician at the point of care or, as I like to say, at the point of decision making within the continuum of care. CEP engines “listen in” on data and events generated from various systems as they pass through the enterprise service bus and combine these pieces of data and events looking for complicated patterns and actionable sequences. The engine then has the ability to take certain pre-defined actions once a particular event(s) and data combination takes place in real time.

More to come on this exciting new venture and how we all work together to generate value out of our health IT investments and leverage lessons learned from other industries.