Walid Gelad and many others have a recent article in the Journal of General Internal Medicine** that looked at the different rate of emergency department physicians prescribing opioids in the Veterans Administration. This is a replication of Barnett’s excellent work from 2017## that examined emergency department prescribing patterns to Medicare beneficiaries who were opioid naive.##
Walid highlights this graph as a key take-away. There is huge variation in physician behavioral patterns when faced with similar patients.
Look how low versus high-intensity ED doctors prescribe opioids to *opioid naive* patients *with the same pain scores*
e.g. among patients with pain score of 1, high-intensity prescribers provided opioids to 11% vs. low-intensity prescribers provide opioids to 2.6%. pic.twitter.com/7XP4fIAx14
— Walid Gellad (@walidgellad) June 13, 2019
I want to go in a slightly different direction than an opioid analysis with this graph. I want to go into network design. The Veterans Administration is a singular entity under, theoretically, singular management control. They have a good to excellent electronic medical record system. The opportunities for integrated care are fairly high. They have a much longer shadow of the future than most health financing entities. There are numerous opportunities present in the VA that most other systems in the United States don’t have. And they still have tremendous variation.
Insurers build narrow networks based on two primary objectives: price per unit control and total cost control. Some insurers will leverage reasonable competitive local clinician and hospital markets to get a dirt cheap rate per unit. The objective is to minimize price per unit without caring too much about how many units are paid for.
The other system is not as sensitive to the price per unit. Instead, the network is designed around variation in provider practice patterns on similar patients. The network if it was optimizing on minimizing opioid exposure in the ER would be built around the docs in the first quartile or at least built to avoid the docs in the most frequently prescribing quartile. This takes advantage of the fact that medicine for common conditions is still a folk art. There can be tremendous variation within the same office much less the same town on following evidence based recommendations. Building networks around providers who practice in a particular way that could conceivably lead to higher quality and lower total costs is a viable strategy.
Communicating the value proposition of a higher per unit cost but lower net cost is a challenge in the ACA domain given the subsidy structures. I think this type of network design choice could work far better in Medicare Advantage and large employer self-insured groups.
Variation in practice is widespread and common and it can be targeted.
** Barnett, M.L., Zhao, X., Fine, M.J. et al. J GEN INTERN MED (2019). https://doi.org/10.1007/s11606-019-05023-5
## N Engl J Med 2017; 376:663-673 DOI: 10.1056/NEJMsa1610524