Texas Ebola Case Highlights Larger EHR Failings

October 27, 2014

As displayed in Dallas, you can't expect EHRs to interoperate when they can't even intra-operate in their current structure.

In medicine, nothing is predictable. Every unpredictable event reveals the conceptual flaw in those EHRs that require the customer to predict in advance what they expect to encounter. It creates a rigidity that virtually guarantees that the system will fail when confronted by unusual situations.

The Dallas hospital that initially sent home Ebola patient Eric Duncan experienced such a failure. The hospital initially claimed that "a lack of interoperability" between the nurse and physician work flows in their Epic EHR system was the reason. Clearly they expected that all recorded information would be transparently available to all but the system apparently did not perform as expected.

When a system is designed, as many are, with modules that lack access to a single, consistent view of patient information, it is not really an interoperability issue. The poor design stems from treating the EHR as a gaggle of isolated data collection systems instead of a single documentation system. Data starts out being application-specific. It is not automatically shared with other applications unless the need was predicted. Clearly, the definition of interoperability is vague and misunderstood.

Nevertheless "lack of interoperability" remains a lightning rod. Don Fluckinger of SearchHealthIT cites a New York Times article critical of the lack of interoperability of Epic and other EHRs.

The Times said, of one physician's expectations: "His system, made by one company, cannot share patient records with the local medical center..." "Records" suggests a document-centric view of interoperability. A data-centric view, involving the exchange of discrete items of data, is the norm however. The Affordable Care Act mentions information sharing between competing systems while meaningful use requires demonstrating the ability to "share some patient data." The certification criteria are similarly imprecise.

People play fast and loose with the terms "information" and "data" not realizing that:

• Data is not informative without context. Today's EHRs typically concentrate on defining and collecting discrete data elements, devoid of what I call contextual metadata. Should data need to be transmitted, there is no context to send along. As a result, the data are subject to misinterpretation, or are simply meaningless. The context necessary to understand the data, which may be obvious to the sender, is missing.

• Data cannot typically be collected unless a decision has been made to do so. This requires anticipating (predicting) what will occur and deeming it sufficiently likely to warrant the expense.

• Physician's notes, on the other hand, can be open-ended. They present no barrier to capturing what happened, whether it was expected or was a complete surprise. Good notes include contextual material, making them potentially informative to others.

• Clearly, if a system attempts to store the record of events as data, some relevant material can't or won't be captured. If the data can't be collected, it can't be sent to another system. No matter how many data elements are defined, there will never be enough, ever.

If interoperability was treated as a record- (document-) sharing problem rather than a data-sharing problem, interoperability could have been easily achieved at any time during the past 20 years. That's how long the eXtensible Markup Language (for structuring documents) has been an international standard.

The real failing of the Dallas hospital was, most probably, to shirk their management responsibilities which are:

• understand their goals and expectations;

• define a broad strategy for managing information and define what parts of that strategy are to be provided by vendors; and

• insist that any system implement their policies and procedures rather than allowing the system to impose vendor-defined policies and procedures

Managers typically buy into the sales hype that these steps are unnecessary because the EHR will "handle all that stuff." Managers rarely comprehend what "all that stuff" is. Vendors rarely document the decisions they have made nor do they transparently disclose them to the customer. Management remains happily in the dark until forced to confront a major policy failure or procedural breakdown.

To paraphrase David Newman's comments about the "number needed to treat": People tend to think that if it's a computer system, especially a medical computer system like an EHR, there's science behind it. Unfortunately, that's rarely the case. It is a lie to tell people to adopt and rely on an EHR without telling them: You should know that we're not good at anticipating the unpredictable nor have we  done much research on EHR, but the evidence that we do have is that it produces very little benefit while introducing a number of identifiable risks including the tendency to cost much more than anticipated. Used intelligently, you may find a way to benefit from an EHR. To leave the important decisions to a vendor is to gamble with your institution's future.