By David Shaywitz, MD

The ability to gather, analyze, and distribute information broadly is one of the great strengths of digital health, perhaps the most significant short-term opportunity to positively impact medical practice. Yet, the exact same technology also carries a set of intimately-associated liabilities, dangers we must recognize and respect if we are to do more good than harm.

Consider these three examples:

• Last week, a study from Case Western reported that at least 20% of the information in most physician progress notes was copy-and-pasted from previous notes. As recently discussed at kevinmd.com and elsewhere, this process can adversely affect patient care in a number of ways, and there’s actually an emerging literature devoted to the study of “copy-paste” errors in EMRs. The ease with which information can be transferred can lead to the rapid propagation of erroneous information – a phenomenon we used to call a “chart virus.” In essence, this is simply another example of consecrating information without first appropriately analyzing it (e.g. by asking the patient, when this is possible).
• At a recent health conference, a speaker noted that a key flaw with most electronic medical record (EMR) platforms is that they are “automating broken processes.” Rather than use the arrival of new technology to think carefully, and from the ground up, about the problems that need to be solved, most EMRs simply digitally reify what already exists. Not only does this perpetuate (and usual exacerbate) notoriously byzantine operational practices and leave many users explicitly complaining they are worse off than before, but it also misses the chance to offer conceptually original approaches that profoundly improve workflow and enhance user experience.

• Earlier this week, in a must-read blog post in Health Affairs, Al Lewis (author of Why Nobody Believes the Numbers, my 2012 selection for “digital health book of the year”) and Vik Khanna discussed the methodological limitations of workplace “Get Well Quick” schemes (my related take here), and zeroed in on the exact same fundamental problem: “consecrating standard practices without clear evidence”:

“We agree that workplace wellness is a useful construct, providing morale and productivity benefits. However, where large financial incentives are offered in the hopes that health expenses will decline, measurement of health expense reductions is a critical responsibility that is almost invariably lacking in today’s wellness marketplace. If employers continue to rush to buy workplace wellness programs, they will soon find themselves doing what the health care system itself has done for so long, to its great detriment: consecrating standard practices without clear evidence drawn from sound analytics. This will result in more money spent on services of uncertain value that produce invalid outcomes, and misallocate resources away from more valuable endeavors and discussions.”

Each of these three examples represents a variation of a more general theme: the ability of digital technologies to afford ready access to large amounts of information can be tremendously dangerous, due to the ease with which bad information can be pushed out and widely disseminated. It’s great that a physician in a healthcare system can immediately access the data for a specific patient – but only if that information is actually correct; otherwise, it can actually lead to bad decisions that wouldn’t occur if centralized data weren’t available at all.

The most obvious way to respond to this concern is to focus more effort on getting the process right – making sure we’re consecrating the right things – thinking about how to redesign hospital workflow before we commit to digitizing it, for example, or ensuring we appropriately vet workplace wellness programs before rolling them out. Ditto for clinical decision support tools that encourage “best practices.”

But this response misses a more subtle, and perhaps more important problem: the construction of a system that seems increasingly “fragile,” in Taleb’s sense, with less variation and less redundancy, and more vulnerable to catastrophic failure. A “chart virus” can rapidly endanger a patient in a way far more rapidly than an error in a traditional medical chart, while enforcing strict best-practice guidelines can rapidly endanger entire populations if the guideline turns out to be a mistake (say the prophylactic use of certain antiarrhythmic agents in effort to prevent sudden death in patients who previously suffered a heart attack, to pick but one historical example).

In the case of best practices, as Recon Strategy’s Tory Wolff and I have argued, it may make sense to take a firm line against demonstrably bad behaviors, but permit a greater range of acceptable practices, as well as to ensure guidelines are “living documents” that can be continuously informed and updated by new knowledge and user experience.

More generally, efforts to excessively streamline processes in the name of efficiency must be balanced against some respect for the possibility of unintended consequences, and an appreciation for the robustifying effect of redundancies and inefficiencies (in moderation).

Digital health aspires to elevate human health and the practice of medicine by enabling providers to offer compassionate care with engaged patients, care that is informed by both updated, evidence-driven guidelines and patient data of almost unimaginable richness and depth.

If this vision is to be realized, however, we must also acknowledge the risk these powerful technologies pose, and take care to handle them with the caution they demonstrably require, and with the respect they unquestionably have earned.

David Shaywitz is co-founder of the Center for Assessment Technology and Continuous Health (CATCH) in Boston. He is a strategist at a biopharmaceutical company in South San Francisco. You can follow him at his personal website.