Physicians’ Variation in Care: The Practical Balance of Warranted Versus Unwarranted Variation
Warranted versus unwarranted care variation is a hotly debated topic (4). In our work, many physicians and advance practice professionals suggest that care variation is appropriate for accommodating local practice patterns, accounting for institutional capabilities and adapting evidence to individual patients. While not without merit in some situations, this argument is often a defensive posture, by us the providers, that avoids transparency and shortcomings in our own practices. The results by Peltan et al (3) are trenchant and give us an opportunity to elevate the debate on types of variation toward an alignment around practical performance improvement.
Care variation was first raised into our national conscious in 1973 when Drs. Wennberg and Gittelsohn (5) published regional variations in care, which they attributed to a variety of causes including, “factors intrinsic to the operation of the healthcare system.” Four decades later, we still defend our variations in care arguing delivery systems, and poor data are alternatively the source of the variation and/or that our variation is driven by the patient, social, fiscal factors, and other variables out of our control. In essence, we say since the data are not capturing all variables and care is complex, “my variation” is warranted.
Our research and practice with health systems show that this unfruitful debate can be circumvented. We use a balanced measurement method and customized feedback tool that removes patient variability and focuses on our own clinical decision-making. Fully controlling for the countless patient variations in real-world practice is impossible, but simulations, which have revolutionized training and variation reduction in other fields like aviation, hold the same promise for medicine. Simulations, administered to groups of providers, eliminate case mix differences found in real patients and make it possible to address unwarranted physician variation and drive learning and standardization.
When simulation measurement is done serially, learning occurs and group consensus emerges. An insight taken from our experience doing this is that most variation is unwarranted; this is revealed through peer-to-peer benchmarking. This is a powerful tool for behavioral change and practice improvement. Measurement efforts must be serial and regular, done in a collaborative space, without blame, and include transparent and detailed performance evaluations that carry evidence-based and readily translatable actions for actual practice. Where Peltan et al (3) explored the narrow (important) issue of unwarranted variation in antibiotics administration, simulations have the flexibility to explore a broad range of areas including history taking, diagnostic evaluation, and therapeutic interventions. With high sepsis mortality and the enormous cost consequences, new approaches uncovering, isolating, and addressing unwarranted variation are key to advancing our care quality and cost efficiency.