For the first time, aggregate patient data from an electronic health record (EHR) has been used to help make a real-time diagnosis and treatment decision in the context of a rare pediatric condition; and in an article published online on Nov. 2 in the New England Journal of Medicine, three physicians from Lucile Packard Children’s Hospital at Stanford University, Palo Alto, Calif., describe how the process worked, and what the implications of that development are for patient care more broadly. (The article will appear in the Nov. 10 print issue of NEJM.)
Jennifer Frankovich, M.D., Christopher A. Longhurst, M.D., and Scott M. Sutherland, M.D., authored the article, titled “Evidence-Based Medicine in the EMR Era.” The authors, a pediatric rheumatologist, the hospital’s CMIO, and a pediatric nephrologist, write in the article, there are frequently situations in pediatric care in which there is not enough—or even any—evidence in the medical literature that can guide unusual patient cases. “We recently found ourselves in such a situation as we admitted to our service a 13-year-old girl with systemic lupus erythematosus (SLE),” the authors report. “Our patient’s presentation was complicated by nephritic-range proteinuria, antiophospholipid antibodies, and pancreatitis. Although anticoagulation is not standard practice for children with SLE even when they’re critically ill, these additional factors put our patient at potential risk for thrombosis, and we considered anticoagulation. However, we were unable to find studies pertaining to anticoagulation in our patient’s situation and there were therefore reluctant to pursue that course, given the risk of bleeding.”
Instead, leveraging a platform called the Stanford Translational Research Integrated Database Environment (STRIDE), which acquires and stores all patient data contained in the EMR at LPCH and provides immediate text-searching capability, the article’s authors were able to perform an automated cohort review and come to a diagnostically valuable conclusion about the patient in question.
Dr. Longhurst, LPCH’s CMIO, spoke with HCI Editor-in-Chief Mark Hagland on the eve of the publication of the NEJM article and shared his perspectives on the implications for automation-assisted patient care from this development. Below are excerpts from that interview.
This is was a fascinating case, and what appears to be a perfect example of the gains in patient care quality that can be made by fully leveraging electronic health records [EHRs].
What this is really about, from my perspective is that lots of people talk about secondary use of the EHR for clinical research. And a lot of talk is taking place around the use of business intelligence and analytics. This is the first instance in the literature in which aggregated data culled from the EMR has been used for real-time clinical decision-making. That’s the key concept. What I would call it is the first sign of a shift from evidence-based practice to practice-based evidence.
Christopher A. Longhurst, M.D.
The patient case that you and your colleagues described in your article involved just too rare a situation for any clinical trials-based literature to turn to, correct?
Yes, that's right. And another way to think about it is that we’ve made this remarkable progress in curing childhood cancer. In the 1970s, Hodgkin’s lymphoma had a mortality rate of 90 percent; now it’s down to 10 percent. And the reason that happened was that every kid with lymphoma was involved in a trial, and we learned from every patient’s case. With adult cancer, fewer than 10 percent of patients are on trials. But what we’ve learned is that what is key is not only collecting information on every patient, but also using the data from the warehouse. And we have collected data using the Cerner and Epic EMRs.
There are a lot of clinicians who still tend to think of the EHR as a one-way repository, right?
Yes. And the concept of the EMR/EHR as simply a digital repository is already outdated. What’s more, everyone who’s seen by a doctor, their data should be available for clinical decision-making.
Some physicians still resist losing the free-text narrative. So there is this tension between allowing physicians to produce unlimited free text in their documentation, and moving too fully toward drop-down menus, to the point where the “patient story” is lost. What are your thoughts on that issue, in this context? Searchability was clearly an element in this mix.
Yes. We could identify all the kids who had lupus through discrete data. But what we couldn’t do was, we didn’t have the data discretely documented around clots. So we did a search across charts for a variety of terms associated with clotting. It was a partially automated search, searching for terms. So I agree with you, that is an issue. And in fact, there have been some really good publications out of Vanderbilt around what they call the tension between free-text and discrete documentation. And most people are coming down on the side of, give people free-text when possible, use discrete data when necessary, and then use tools like natural-language processing to fill in the gaps.
When clinicians and clinical informaticists read this article, will it provide a kind of “light-bulb” moment for them?
Well, I hope so. And the quicker we get to this, the more lives will be saved. We need to think about providing tools within these systems to allow for using local data to help improve decision-making. It’s the holy grail—the ability to facilitate having the average doctor use this combination of tools to make good decisions on a real-time basis.
What would your advice be to CIOs and CMIOs, based on this?
I would tell them that, to whatever extent is feasible, they should decentralize and federate their analytics tools.
Because that will help spark a certain kind of creativity or ingenuity, right?
Yes, because the more you keep these tools restricted to a centralized group of experts, the less innovation you’re going to get.