Editor's note: SearchHealthIT received this feedback in response to the article "Health data interoperability in sight … via Epic?" The letter has been edited for style but not for content.
I appreciated your story about where we are with interoperability. Your story is on target. The current status represents a regulatory and technical failure. Interoperability is important, really important. It tends to be a tech focus for things we need to achieve on behalf of our patients, but it is not more important than other major limitations of EMRs, including the vendor product (Epic) which you mentioned. Here are other EMR hurdles in the way of improving patient care.
1. Customizing order sets to patient or place is hard within one package: To change the order set to improve patient care within one organizational setting, an EMR is likely to provide all-or-nothing choices. Programming situational choices is remarkably difficult. So, if we are content that the patient in the ICU should have the same CPOE [computerized physician order entry] options as the patient in the general ward or the patient in the psychiatric hospital, then the current EMR is likely fine. But when there are important, patient safety reasons to feature different order sets for the clinicians, the EMR seems to create real limitations and requires major efforts to do what should be simple. An example is the daily orders dilemma, a featured topic in the national Choosing Wisely campaign. The functional periodicity should be different in the units and on the wards, but that simple task requires major efforts.
2. Creating data marts is way too hard: The creation of data marts should flow naturally and easily from any modern database, which ought to be fully relational at the outset. But, this is a huge task for any EMR owner. Healthcare organizations need to improve quality assurance in multiple domains, not just the handful designated as the federal flavor of the month. But this work is a costly and time-consuming enterprise. Why is that? Will our community hospitals be able to routinely do good analyses in the near future? I am skeptical that is on the horizon for any but the biggest and best integrated systems who have laboriously built their parallel data warehouses and useful data marts for years. Yet the patients need us to be able to do that anywhere care is provided.
3. Even simple text searches can be really hard: ICD-9 or later codes will still fail to uncover all the things we need to know about our patients. So, why are text searches now harder to do than when all we were doing was dropping [information] in word docs? Why does the clinician need to be an expert in architecture before the medical record information is useful to others? For this problem, I am less sure of my ground than for the others, as maybe my expectations have simply increased without commensurate ability to get the job done.
Alan Ducatman, M.S., M.D.
Professor of Public Health
Professor of Medicine
West Virginia University