In his latest Direct Primary Care slide-show brag, attributing significant overall medical cost reduction for employees electing DPC over and FFS primary care alternative offered by the same employer, Dr Lee Gross insists that the favorable results do not reflect “cherry picking”.
And yet, Dr Gross fails to compare the health status of the DPC-covered employees and dependents and the actual FFS-covered employees and dependents of the same employer. Indeed, he does not even present the most basic demographic information about relative age of the two cohorts. Keep in mind that in the two most widely-published employer DPC studies the age differences between the cohorts (4.7 and 6.5 years) were large enough to explain a 25% or larger medical cost differential.
Gross’s purported rebuttal of “cherry picking” rests on comparing the frequency of a set of chronic conditions in the studied employer’s specific DPC population to the frequency of the same set of chronic conditions in an unidentified “national benchmark population” — eschewing, for some reason, the chronic conditions profile of the studied employer’s own non-DPC population. What can possibly explain that?
Risk adjustment adequate to the task of comparing population health care cost risks is far more complicated than simply counting the number of chronic conditions. Yet, even if risk adjustment was that simple, a meaningful analysis would require some basis for assuring that Dr Gross’s unaudited methods for identifying the presence of chronic conditions in the patients his practice serves are commensurate with the methods used by impartial analysts to establish the the levels of chronic conditions in the unidentified national benchmark population. There is, after all, a significant history of diagnosis up-coding by self-interest providers.
Whether or not Dr Gross cherry-picked the cherry-picking data (or contaminated it by upcoding), his current self-interested brag is just another in a long line that fails to meet minimal standards of analysis. Yet, again Dr Gross has produced a “case study”, adorned with an at-best naive version of risk adjustment without laying out any methodological detail in an actual masuscript.
Dr Gross’s current presentation notably differs from his 2019 presentation from the same DPC option experiment for a prior year. The 2018 data clearly showed that the employer’s DPC cohort had total medical expenses that were 15% higher than their non-DPC counterparts. Was there a sharp change in his case study methodology?
Also note, Gross’s earlier presentation counted the 2018 DPC cohort as having 35% more chronic conditions than the 21% he currently claims.
Whence the discrepancies between 2018 data and later years?
Outliers? In the current Gross slide show and video, Gross alludes to certain unidentified difficulties in 2021, while showing a slide that notes a $440 PMPM for 2021 which represents a 67% increase in DPC member costs PMPM over 2020. In an asterisked note, Gross links the cost increase to a pair of expensive outlier patients in the DPC cohort. That’s a fair point to make — but only if the analyst is equally sensitive to the possible presence of outliers on either side of the comparison being made; since Gross provides no description of his methodology, we have no idea how the broader DPC/non-DPC comparison might look with an appropriate outlier adjustment.
Outlier years? Note too that, Dr Gross’s spending charts for 2019 show almost identical total spending PMPM for DPC and non DPC patients just over $250, so the
focus of his brag is data from 2020. For that year, more than 60% of the cost difference came from hospital inpatient expenses. That same year, note, the non-DPC group’s hospital inpatient expenses soared to more than 70% above Gross’s selected national benchmark.
Given that 2020 saw a pandemic that, on a world scale, drove hospital admissions and health care needs disproportionately with increasing age, it might be helpful to know the size of the age differential, if any, between the studied DPC and non-DPC populations. More broadly, given that 2019 total costs for the DPC and non-DPC cohorts were substantially identical (and 2018 cost even lower for the non-DPC cohorts), the emergence of roughly 50% savings for the DPC cohort in pandemic times should be examined with through a pandemic-sensitive framing .
In this regard, it should be noted that the employer studied by Gross is a hospital, and all its employees are health care workers, and likely at or near the tip of the pandemic spear. Even seemingly small differences between the DPC and non-DPC cohorts might have been amplified in pandemic conditions. Perhaps hospital employees of different departments have differing inclinations toward DPC, a factor unlikely to matter unless Covid-19 risk expands specifically in those departments.
What we do know is that Dr Gross study was small, less than one-sixth the size of the Milliman DPC employer study. For 2020, based on such numbers as Gross does present, the average number of employees in each cohort was about 150. Dr Gross’s own computations relate savings that varied from 0% to 50% from one year to another; that overall patient costs in a cohort can rise by over two-thirds in a single year; and that in his computations a single pair of patients can account for 25% of an entire cohort’s costs.
As a result of all this, all of Dr Gross’s purported successes can turn on a handful of outlier enrollees and are unsteady year over year. Most particularly, the results during the brag come from two pandemic years featuring mass disruption of world health care patterns; they may mean nothing at all.
Should you work with Dr Gross, be sure to ask him how he knows your first year experience will not better reflect the studied employer’s first year of 15% losses, or the studied employer’s last pre-pandemic year of breaking even, rather than reflect the extraordinary, pandemic years for which he has apparent success. Ask him for demographic data on the studied employer’s DPC and non-DPC populations.
Don’t bet the health of your employees on Dr Gross “case study”.