We’re Still Waiting For Searchable EHRs
I’m feeling reeaalllly lucky.
All Things Considered on NPR had a lovely snippet yesterday entitled Big Data Not A Cure-All In Medicine. Admittedly, I rolled my eyes at one of the opening lines:
“…Those [Electronic Medical Records*] are available to professionals and importantly, they’re searchable.”
The fact that non-healthcare people, journalists included, think software technology works like every other piece of software in the world would be funny if I didn’t so badly wish that it were true…but I digress.
The story profiled a use-case at Stanford Hospital where they leveraged EHR (They use Epic) data to determine medication dosing for a particular patient based on existing patient outcomes in their population. The individual case was a success. However, the hospital isn’t confident that the results could ever be replicated for any other situation so searching their existing EHR data is just something that they won’t do again anytime soon. But why?
The tech world has been moving forward operating under the assumption that as we approach Big Data, we will merely have to search the massive data sets to find all of our answers. Wired even went as far in 2008 to say that the scientific process, the asking of questions and testing of hypothesis, would soon come to an end because we would have all of the information so experimentation wouldn’t be necessary. Yet, herethat is a seemingly ideal use-case that is shunning the big data process. The problem is that in order for Big Data to work, the data must be accurate and complete. These are two attributes that are absolutely never used to describe data in an EHR. Even if the data is accurate and complete, we also have to be looking at the correct metrics.
Current Big Data algorithms can do a pretty good job at identifying the most impactful metrics (again, assuming data is accurate and complete), but how many hospitals are using Big Data algorithms on their data? It’s not because they haven’t thought of it, it’s just not practical. Of the research hospitals that I have heard or seen conduct studies utilizing their EHR data, they generally extract a subset of data from patient records first before performing an analysis. Public health institutions do the same thing in the form of registries. By definition, the data is no longer complete and that’s assuming it was complete in the first place.
Additionaly, Medical Knowledge is still not complete no matter what Dr. Oz might try to tell you. We don’t fully understand what causes every disease and we are only just beginning to collect the data of our genetic code and microbiome. Therefore, until we become “All-Knowing”, the scientific process will have to be the tool of choice.
*NPR still hasn’t gotten on the EHR bandwagon yet.
Aaron Berdofe, MHI - Health Informatics 