EHR – Checklists & Bayesian Diagnosis

A very interesting recent article by Shiff and Bates published in NEJM proposes a rather ambitious benefit for EHRs.
The article aptly titled “Can Electronic Clinical Documentation Help Prevent Diagnostic Errors?” and aptly presented as a question rather than a statement, is exploring the possibility that EHRs can reduce diagnosis errors. The authors duly admit that there isn’t much evidence to support such forward thinking for the current crop of EHRs.

“Admittedly, evidence to support the existence of such a benefit is currently lacking, and our hypothesis runs counter to the sentiments and claims of many physicians, who argue that electronic documentation in its current incarnation is time-consuming and can degrade diagnostic thinking — by distracting physicians from the patient, discouraging independent data gathering and assessment, and perpetuating errors.”

However, they are envisioning a new breed of electronic documentation that will be capable of fulfilling such lofty goal. Most of the proposed characteristics of this new generation of software are incremental improvements to collection and presentation of information from various sources. This is in line with the national goals of fostering interoperability and the creation of an Electronic Medical Record for every American by 2014. Computerization is also viewed as a solution for the introduction of mechanical errors due mostly to poor handwriting and poor paper based book keeping, such as the benefit provided by electronic prescribing. All in all, the authors’ suggested “top 10” style list of desired improvements to EHR technologies are very sensible and have the potential to improve usability, collaboration and effectiveness.

An interesting aside is that the authors propose more textual data capture and more reliance on voice recognition as opposed to discrete data capture by “ticking boxes”. I hope the authors understand that dictation and free text do not lend themselves very well to “aggregation, trending, and selective emphasis or display of data so as to facilitate rapid judgments”, which are the cornerstone of the approach suggested in the article.

A more interesting suggestion is the use of Checklists for “Providing prompts”:
“Provide checklists to minimize reliance on memory and directed questioning to aid in diagnostic thoroughness and problem solving”

And to go with the checklists there is “Calculating Bayesian probabilities”:
“Embed calculator into notes to reduce errors and minimize biases in subjective estimation of diagnostic probabilities”.

This, of course is the opinion that decision support at the point of care has to become powerful enough to change the imperfect way medicine is practiced today.

“But we envision a redesigned documentation function that anticipates new approaches to improving diagnosis, not one that relies on the putative “master diagnosticians” of past eras. The diagnostic process must be made reliable, not heroic, and electronic documentation will be key to this effort.”

The question in my mind is at what point does decision support become too powerful and begins crippling both physicians and quality of care? And is there such a point?

For an opinion on this subject, I suggest Dr. Jerome Groopman’s excellent book “How Doctors Think”.