RCTs have long been considered the gold standard in medical research, especially when it comes to determining if pharmaceuticals and medical devices work better than placebos. In an RCT, neither the physicians nor the patients know which trial arm they are in. The results are considered definitive, at least until the next RCT comes along.
Is such a thing even possible for testing delivery system changes? Imagine trying to test whether the concept of a medical home works.
The experiment's designers would have to randomize both physicians and patients. Most doctors either want to participate in a medical home project or they don't. They would bring their biases with them to the trial. That would undermine randomization since it is impossible to blind them to which arm of the trial they are in.
The patients, meanwhile, would have to agree to stick with only their own doctors for the duration of the trial. The outcome would be compromised if patients switched between medical home and non-medical home physicians.
Beyond those technical difficulties, the idea that the CMS should focus its innovation money on RCTs runs counter to some of the more promising research approaches now being used in healthcare. Rather than engage in lengthy and expensive prospective RCTs to conduct comparative effectiveness research, these approaches mine the mountains of data that are being amassed in healthcare claims and patient records warehouses.
Last week, Modern Healthcare reported on the expansion of one such project called SharedClarity, which is a joint venture between UnitedHealthcare, the nation's largest insurance company, and a handful of large healthcare systems. They are evaluating the relative performance of about 30 cardiovascular devices. Initial results are expected in a few weeks—a little more than a year after the project was first announced.
SharedClarity's goal is to compare the effectiveness of different devices. This information isn't just valuable to patients and their physicians. It's also a boon to cost control because it allows purchasing officials to substitute cheaper alternatives when they can show their outcomes are as good as pricier items on the list of evaluated technologies.
Drug research is also starting to move away from RCTs in certain circumstances. It has long been known that some drugs work better in some people because of their genetic makeup or the genetic makeup of their disease. The positive responders may not be identified until part way through the trial. By adjusting who gets recruited—it's called adaptive clinical trial design—an experimental drug headed for failure can be turned into a useful therapeutic, albeit for a targeted population.
Adaptive approaches are precisely what is needed as the Innovation Center continues to test delivery system reforms. If some of the Medicare Pioneer ACOs made money by reducing hospital readmissions while others failed, then program designers need to figure out what worked for the former and transfer that learning to the latter.
The same is true for quality measures in Medicare's value-based purchasing program. Medical best practices are always subject to change based on new clinical evidence, whether it comes from RCTs, claims databases or registries. Quality measures need to be adjusted to reflect that knowledge.
Delivery systems looking to achieve better outcomes at lower cost need to constantly incorporate the latest evidence into their processes and procedures—to become what the Institute of Medicine a few years ago called rapid learning systems. Insisting on lengthy and expensive RCTs for testing delivery system reforms would slow the process of change, and make rapid learning impossible.
Follow Merrill Goozner on Twitter: @MHgoozner