The raw downstream cost claims data fed into Miliman’s “isolation” model were imprinted with Union County’s (likely pro-DPC) model of downstream cost-sharing.


Note. The foregoing post was essentially completed and copied to Milliman in late May or early June 2020. In a footnote to an internet essay at the end of June, two members of the Milliman team presented new material addressing the issues raised below. I will address this new material in a new post.


In its recent assessment of the direct primary care model, Milliman took a two track approach. An employer ROI based approach included comparing claims experience for a group of employees who opted to receive primary care from a DPC clinic versus those using traditional FFS PCPs; in addition, the ROI analysis also assigned costs to employer of the differing cost-sharing structures the employer has required of the two groups. This computation resulted in the conclusion that in the plan under study DPC actually increased employer costs.

In a second approach, however, Milliman looked only at a claim cost comparison between the two groups. According to Milliman, “this cost comparison attempts to isolate the impact of the DPC delivery model on the overall level of demand for health care services“. [Italics in original]. Milliman should be proud of thinking of this approach. It seems likely to work well where both the DPC and FFS cohorts face identical cost-sharing incentives for downstream care. But that was not the case in Union County.

Since Union County offered very different cost sharing landscapes to the two groups, the purported isolation is simply not possible.

For example, all FFS employees (and no DPC employees) had $750 HRAs, which present FFS employees with “use or lose it” choices affecting their decisions on incurring claims. In that situation, FFS employees will be prone to over-consume health services, largely on downstream care, a significant swath of void of cost sharing discipline. By Milliman’s calculation (Fig 12, Line I), the average FFS member does not exceed the HRA-protected level of usage in a normal year.

Accordingly, that everage FFS member never reaches a second feature of the Union County plan, a 20% coinsurance level. On the other hand, all the downstream medical claims of an average DPC members is subject to the discipline of 20% copayment.

Combining the effect of the HRA and the coinsurance, the average DPC group member is likely to under-consume downstream health care services when compared to the average FFS member.

These real decisions by real employees surely affected Union County’s downstream claims experience. A cost sharing plan is not merely a dispensable ornament in an ROI computation. Rather, Union County’s cost-sharing structure presents important “confounders” that Milliman’s isolation model, as originally conceived and executed, simply ignored — with a likely effect of overstating direct primary care’s ability to reduce the downstream care costs of average utilizers.

Because the raw downstream claims cost data fed into the model intended to isolate DPC impact from the vicissitudes of Union County’s cost-sharing scheme had already been imprinted with the vicissitudes of Union County’s cost-sharing scheme to an unknown degree, Milliman’s attempted reinvention of the wheel results in an indeterminately bumpy ride.


To be fully clear, it is certainly possible that, if applied across all members at all levels of downstream care consumption, Union County’s unique and complex cost-sharing landscape for downstream care for non-DPC members, and it’s simpler coinsurance focused scheme for downstream care for DPC members, might combine to have a negligible net effect, or even tilt in favor of DPC. I don’t know. As of the date of its May 2020 report, neither did Milliman.

Until it completes the hard actuarial work needed to quantify the skew of the downstream claims cost data that fueled its isolation model, Milliman should back off its claim to have successfully isolated and determined the health care utilization impact of DPC.


Milliman and Milliman wannabes would be especially well-advised to use caution before attempting to apply the Union County isolation methodology in case where a DPC option is automatically tethered to an HDHP , even though the HDHP may be available on a voluntary basis to a quasi-control group.

Virtually all DPC practices take pains to sign on members with HDHPs, through employer arrangements or otherwise. Once such price-shoppers are signed on, they can be relied on to incur relatively modest downstream costs. The enrolling provider is, of course, happy to claim any reduced downstream costs as attributable to herself and/or to the DPC model. Borrowed valor. Good example here.


For more about Milliman or Union County, use Search.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: