EHRs Unreasonable Assumptions Conflict with Reality

For years, we've appreciated that Medicine is “messy” and that the messiness often confounds computing techniques that work in other fields but it was just a feeling. Now the feeling can be formalized and proven, thanks to Pareto’s Principle and Benoit Mandelbrot, noted for developing a "theory of roughness" in nature and the field of fractal geometry.  Mandelbrot studied the noise that affects data communications. Engineers know that some spontaneous noise can never be eliminated. Periods of errorless communication are followed by periods of errors. The more closely the clusters are examined, the more complicated the patterns of errors appear. With the day divided into hours, an hour might pass with no errors at all. Then an hour might contain errors. Then an hour might pass with no errors. Then divide the hour with errors into smaller periods of 20 minutes. Here, too, some periods would be completely clean, while some would contain a burst of errors. In fact, you can never find a time during which errors are scattered continuously. Within any burst of errors, no matter how short, there would always be periods of completely error-free transmission. There is a consistent geometric relationship between the bursts of errors and the spaces of clean transmission. On scales of an hour or a second, the proportion of error-free periods to error-ridden periods remains constant.* This is an example of what is called scale-invariance, a property of fractals. "A fractal is an object that displays self-similarity at various scales. In other words, if we zoom in any portion of a fractal object, we will notice the smaller section is actually a scaled-down version of the big one."]

The Cantor Set is a fractal that demonstrates scale-invariance. It is created by repeatedly deleting the middle thirds of a set of line segments. One starts by deleting the middle third, leaving two line segments. Next, the middle third of each of these remaining segments is deleted, leaving four line segments. This process is continued ad infinitum. The first six steps of this process are illustrated below.

Pareto's Principle - the 80-20 rule - is also a fractal that demonstrates scale-invariance. Pareto's original observation was in connection with population and wealth. He "noticed that 80 percent of Italy's land was owned by 20 percent of the population. He then carried out surveys on a variety of other countries and found to his surprise that a similar distribution applied." Furthermore, the relationship was scale-invariant. It applied to every subset of the income range. "Even [taking] the ten wealthiest individuals in the world… the top three (Carlos Slim Helú, Warren Buffett, and Bill Gates) own as much as the next seven put together."

The typical EHR depends on the assumption that everything is predictable. Being substantially predictable is not the same thing as being completely predictable and this is precisely what the Pareto Principle describes - scale-invariant situations in which things are substantially, but not completely predictable. It is not simply that 80 percent of healthcare encounters or patients are routine and 20 percent are unusual. Rather, 20 percent of any aspect of a patient, their problem or any component of an encounter is unusual - an exception. The details of each exception are almost completely unpredictable. A diagram of this behavior would look different but it would also represent a fractal because of scale-invariance. This is evidence as "hard" as any used to support an evidence-based treatment.

The facility where I see patients is following the vendor's instructions as it embarks on a massive EHR implementation. The first step is to predict every task that anyone might ever need to perform, define what roles will perform the task, etc. Anything that has not been predicted, or has been defined incorrectly, will be disallowed. Will they succeed? The evidence is against them.

This is the evidence:
• The only thing predictable is that in every task, big or little, something will go wrong.

• Those who design "data collection" may choose to ignore the 20 percent but clinicians documenting in the chart cannot. Just the opposite, it is the 20 percent that is most important to capture.

• Eighty percent of a typical EHR is devoted to data collection, 20 percent (or less) to faithfully capturing the details of the encounter that matter most.

• Evidence cited by experts at RAND and elsewhere that today's EHRs are not meeting expectations..

Conclusions:

The typical EHR is least able to help you with those things for which you need the most help. It will interfere with your efforts to capture the most important (unexpected) aspects of an encounter quantitatively and in full detail.

The typical EHR has its priorities backwards. This begins as a conceptual flaw and manifests as a structural flaw. It cannot be fixed with a few tweaks. Find a system that was designed to "thrive and grow when exposed to volatility [and] randomness."

You wouldn't subject a patient to a treatment or procedure that had this kind of evidence against it. Why subject yourself and your patients to a healthcare process controlled by a system with this kind of evidence against it?

There is a solution to this problem and it will be the topic of the upcoming articles.