Some of S2N’s recent work has gotten me thinking about variation, sometimes extreme, in the practice of seemingly routine medical care. Take the interventional treatment of peripheral arterial disease, for example. We recently spoke to physicians about their approaches to stenting; some stent about 10% of lesions, some stent everything, and about 1 in 10 had practices that were really out of the norm, such as using TPA or inordinate numbers of covered stents. Peripheral interventions being somewhat new on the scene, one could expect, and accept, a certain amount of variation. Not long after, though, we were speaking to neurointensivists about the management of neurogenic fever in the ICU and came across wildly different beliefs about the efficacy of acetaminophen (a.k.a. Tylenol), definitely not a new therapy, in reducing fevers (somewhere between 0% and 100% apparently). What are we to make of this vast variation in medicine?  What would Darwin say?

Natural selection, the greatest algorithm of all time, addresses variation by filtering variants based on strength vs. weakness; the strong variants survive while the weak die off.  The application of selection methods to medical practice has had some successes over the years (e.g. the sterile field in surgery, lobotomies not a thing), and there is a substantial industry around scientific study and evidenced-based medicine.  Despite all this science, however, medicine remains more of an art. Shifting the balance toward replication of optimal care and reduction in counterproductive variability will require streamlining of both the selection and dissemination of best practices (unless we want to off the weak practitioners, which no one here is recommending). 

Selection: Simply put, medicine needs a good way to pick winning practices.  Maybe you are thinking, silly, we can do that now with randomized controlled trials (RCTs), the pinnacle of rigor in determining the superiority of approach A vs. approach B. Even with medical devices, we’ve figured out a way to conduct sham-controlled studies using fake surgical sounds and video reels. A key limitation of the RCT, however, is the huge investment in time and money required for just one such selection process; data quality is prioritized over cost-efficiency and speed. Perhaps medicine could learn something from the tech industry, where web-based companies are constantly running real time experiments to evaluate page performance and optimize particular metrics in the form of A/B testing. In comparing medical practices that are thought to be similarly safe against near-term endpoints such as length of stay, patient tolerability, or acute complications, this type of rapid, scaled testing might be feasible.  In any event, there needs to be some innovation in picking winners in medicine outside the traditional RCT paradigm if we are going to keep pace with patients’ needs and really impact the variation problem.

Dissemination: According to Darwin’s theory, it is crucial not only that traits best suited for their environment confer survival, but also that they be heritable - reproduced through the passing down from generation to generation. As Atul Gawande explains, the problem in medicine is “...good ideas still take an appallingly long time to trickle down.” “Scaling good ideas has been one of our deepest problems in medicine,” laments Gawande. Currently, good ideas in medicine are circulated primarily through a complex web of hundreds of peer-reviewed journals and scientific conferences.  The timeline from completion of science to publication or presentation is a year at lightening speed, and it can take 10 or 15 years for a truly good idea to be incorporated into standard of care and benefit the majority of patients. While I don’t think dissemination in medicine will ever be as fast as pushing new code out to an autonomous surgical robot (the FDA being one major hurdle to this vision), health IT solutions leveraging the ever increasing capture of real-world healthcare data could greatly facilitate the identification and dissemination of incremental advances in care.

There is no doubt that variation will be an enduring, and essential, aspect of medical practice. While our current system may be burdened with too much needless variation that detracts from quality, some level of experimentation is necessary to prevent stagnation at a “local optimum” (like reaching the top of a hill when there is a mountain beside it). Along with natural selection, variation is what drives evolution. We will always need fresh, new experimental ideas / techniques / care pathways to run against current best practices and challenge ourselves to find even better solutions. Given human nature, and physician nature in particular, I am confident there will always be a wealth of new ideas. We just need a better way to know if they are good ones, and if so how to put them to work asap.

S2N Analyst Rebecca Noble contributed to this blog.  Thanks, Rebecca!