EHRs: Balancing Predictability with Realistic Expectations

Healthcare.gov has flopped. Why? Because, at every step along the way, the unknown, the uncertain, and the uncertain were downplayed. It was sheer hubris to assume that all those details could be predicted AND that the correct logic and linkages could be defined to tie everything together. As you are witnessing, even systems that limit themselves to data collection fail miserably when too many of the design assumptions are based on a predictable universe.

Healthcare is predictably unpredictable. This means that it is impossible to predict what data elements would be needed to describe a patient's condition in detail. In other words, the history and physical can never successfully be reduced to a data-collection problem. Why? Because data collection is deterministic; charting, which involves creating a journal of medical events as they unfold, is not.

What follows may seem highly technical but it's not much more than common sense.

Stochastic means random and/or unpredictable. It's the reason that journal articles present statistics. If every subject responded the same way to the protocol there would be no variation - no need for statistics. The reason for Cochrane meta-analyses is that no investigator every correctly predicts what to control for in their studies. Doctors know that and yet, for some reason, there is no outcry when vendors try to sell us billion-dollar systems based on the flawed assumption of predictability.

The healthcare system is what a stochastic system; patient care and its documentation are stochastic processes. According to Wikipedia, "a stochastic system is one whose state is non-deterministic. The subsequent state of a stochastic system is determined both by the system's predictable actions and by ... random element[s]. A stochastic process is one whose behavior is non-deterministic; it can be thought of as a sequence of random... [events]."

Data collection is deterministic. After all, one must determine which data elements are to be collected and then must design both databases with fields to accommodate the data and a means of collecting each element (such as a data collection form.) Journaling (creating a narrative record that describes what is happening) is stochastic because the precise details of what is happening at any given moment to any patient are far from predictable. The task of the journalist (one who makes entries in a journal) is to describe what life throws at them. Life steadfastly refuses to confine itself to predetermined events that can be completely described with a few (or a few dozen) predetermined elements.

Today's data-collection-oriented EHRs violate several basic principles of database design, namely to eliminate data redundancy (the same piece of data shall not be stored in more than one place, duplicate data not only waste storage spaces but also easily lead to inconsistencies), and to ensure data integrity and accuracy. In this day and age, neither data collection or keeping a faithful, detailed journal is optional. Using traditional methods, each of these becomes a separate task requiring one unit of effort, time, and expense. As a result, many items of information get collected twice (according to different rules) and stored in disparate locations (often with no obvious link.)

The bulk of this double work falls on the practitioner who must not only struggle with a data-collection process that is fraught with a myriad of problems I have described in past articles but who must, in the little time remaining after data entry, create journal entries that may result in sanctions if some "required" information is omitted. Hence the cutting, pasting, use of boiler-plate, the acceptance of templates, and of "all normal" check boxes. It is only by chance that the average EHR note bears more than a tangential relationship to the details and subtleties of each patient's case.

You must be wondering how it is that major EHRs, on which organizations spend hundreds of millions to billions, could be so severely flawed? The explanation is the same one that drives some cancer patients to seek "alternative treatments" - a desperate need to act even if those actions violate every known principle and guideline. The end result is the worst of both worlds.

The seemingly insurmountable obstacle is how to accomplish both, using less than two units of effort? Ideally, a single unit of effort would both document and collect data. This would require that, as care was documented, every element that had a quantitative dimension (that is, it contains elements that could be used as "data") would be captured in such a way that the needs of the deterministic processes could be satisfied by data contained within the medical record while, at the same time, producing a record fit for human consumption.

Until everything about the EHR changes - its orientation, focus, and technology (both hardware and software) - demanding that physicians not only take care of patients and create quality documentation but also act as data-collection clerks will be about as successful as flogging a dead horse. Current EHRs are structurally incapable of being efficient, interoperable, and long-lived - in short - meaningful. Patients will be little better (or perhaps worse off) than if their doctors really talked to them, assiduously maintained detailed paper records, and implemented office procedures to minimize the number of things that fall through cracks. The country would be billions of dollars better off by saving its money until there was something worth spending it on. For those dying to write a check, fund basic research into medical information science.

Of course, my efforts to get this point across may simply be flogging a different dead horse.