Mon, 06/18/2018

There is little doubt that one of the hottest trends in healthcare information management is predictive analytics. Some of the most common research topics are predicting the likelihood of sepsis in a hospital patient, understanding who is most likely to readmit and getting insight into co-morbidities and their likely progression for certain chronic conditions. These advanced informational algorithms get our attention because they can have such a direct impact on patient wellness and our ability to deliver cost effective quality care. But if we believe that (and I am sure of us most do), why aren’t we doing it already. 

In order to be able to effectively predict the future we need to have a firm understanding of the present and the past. The nature of healthcare data and its tendency to be voluminous while often living in segregated siloed systems often doesn’t allow us to have a firm understanding on what is actually going on. Sure, I can tell you everything we did while the patient was in the hospital, every test we ordered, the results and other observations, what we instructed the patient when they left and what meds we prescribed. But I don’t necessarily know if they filled their prescription, whether they reliably took their medicine or followed any of the instructions. We do outreach but I am not exactly sure what we learned from it. And predictive models can be heavily influenced by the lack of a complete picture of what happened, when and under what conditions. 

But that simple example represents a real opportunity for technology to help understand what happened. Every process I describe above has supporting systems that contain important information that we need to understand the present and the past. By acquiring and harmonizing the data, we can break down the barriers that fragmented systems create. We can then develop a solid understanding of the patient, their interactions and experience with us and get a complete understanding of all aspects of the patient encounter(s) represented in the data. This “new” data can now form a reliable basis from which to predict the likelihood of future events.