Marshall on Dershowitz; or is he talking about me on DPC?

To put it baldly, if it’s a topic and area of study you know nothing about and after a few weeks of cramming you decide that basically everyone who’s studied the question is wrong, there’s a very small chance you’ve rapidly come upon a great insight and a very great likelihood you’re an ignorant and self-regarding asshole.

11% claims reduction, with no adjustment for selection bias, is pretty tame.

Paladina Health maintains a news and information page on its website. As of the start of 2020, Paladina’s most recent entry of favorable cost reduction results is entitled “Paladina Health gives Akron schools a cost-saving model” and links to this Crain’s business report of an 11% reduction in claims. There was no adjustment for selection bias.

This is less than half the more-heavily promoted 22%-26% savings claimed for Paladina’s better performing clinics.

Amazingly, a years-old interim report of 38% claims cost reductions from a Paladina clinic is still being presented in a pitch to employers by one of Paladina’s multi-state competitors.

A single-post critique of AEG/WP's recommendation on direct primary care.

In “Healthcare Innovations in Georgia:Two Recommendations”, the report prepared by the Anderson Economic Group and Wilson Partners (AEG/WP) for the Georgia Public Policy Foundation, the authors clearly explained their computations and made clear the assumptions underlying their report. The report’s authors put a great deal or energy into demonstrating that billion dollar savings could be derived from direct primary care under certain assumptions. After what I believe was careful examination, I concluded that those assumptions were unsupportable.

Here, I summarize my opinions, linking to about twenty individual posts. The posts themselves contain numerous supporting citations and data, as well as access to spreadsheets that can be used as templates for the reader’s own calculations.

AEG/WP made two questionable assumptions about direct primary care fees. One assumption was that appropriate direct primary care would have a fixed monthly fee of $70. My analysis shows that $70 lowballs the fee considerably. A second assumption was that monthly direct primary care fee would remain flat for a decade.; I noted that these fees were likely to track medical cost inflation. I recomputed the possible savings based on using a more accurate monthy fee and the same medical cost inflation number AEG/WP used. And I left in place AEG/WP’s assumption, discussed below, that direct primary cuts 15% off downstream care costs. Correcting only AEG/WPs two assumptions about $70 fees caused the billion dollar purported savings to fall by 85%.


The most central assumption in the AEG/WP analysis is that direct primary care reduces the cost of downstream health insurance by 15%. Direct primary care needs to show significant reduction in downstream care costs to justify the fact that even $70 direct primary care monthly fees would exceed expected fee-for-service primary care payments — by about $350 per year in the individual market. While the AEG/WP’s 15% assumption corresponds to a downstream care cost savings in the vicinity of $660 per year, there is no clear evidence to show that direct primary care can even cover its own $350 annual upcharge,

I noted that AEG/WP supported its 15% assumption only by referring to undisclosed research internal to its own team. I noted that the marketplace had already demonstrated skepticism about similar claims. I noted that a DPC practice founded by one of the authors of the AEG/WP reports authors had made similar claims, without producing supporting data. I further noted that selection bias had infected the best documented argument that direct primary care reduced downstream costs.

I contacted AEG/WP and learned the 15% assumption was based on three reports, available on the internet, about different DPC clinics. I was able, therefore, to carefully examine the information available to AEG/WP. In a single post, I addressed the experience of two clinics, which together were both the two largest and the two most current examples used by AEG/WP; I concluded that these both examples failed to address selection bias adequately.

The third example, the Nextera clinic, deserved its own posts. Their report noted obvious selection bias, while revealing modest evidence that Nextera cuts cost for some of its patients. But the data set was skimpy and contaminated by results for fee for service patients. The patient data that showed downstream cost reductions for patients served by Nextera included both significant numbers who paid only Nextera’s fixed monthly fee and significant numbers who paid Nextera only on a per visit basis. This may be an adequate method for measuring the positive value of Nextera. It is hardly sufficient as a yardstick for the positive downstream value of fixed-fee direct primary care. In a separate post, I noted that Nextera’s experience showed only a $72 PMPM overall claims cost reduction, an amount that would barely exceed the $70 monthly fee.

I pointed out that even if the foregoing criticisms of the source data on which the AEG/WP relied were in error, their further assertion that the 15% assumption is a conservative one is incorrect.

I also pointed out that AEG/WP’s source material for the 15% assumption consisted only of marketing information, and I suggested that a few brags from a few DPC companies is not a sound basis for public policy decisions.


I spoke with the actuary who was on the AEG/WP team; he made clear that his role did not include validating the 15% assumption.


I noted that the AEG/WP sourced low monthly fees to a set of direct primary care providers who had sharply lower fees than the providers to whom AEG/WP sourced its claim of downstream cost reduction. I suggested that an analyst seeking to establish cost-effectiveness would be well-advised to draw both cost data and effectiveness data from the same sources.


Not a penny of the savings in the AEG/WP report can be achieved unless direct primary care will significantly reduce downstream care costs. There is no sound evidence in the sources on which the AEG/WP authors relied that direct primary care can even manage to cover its own added cost, even if direct primary care even if were priced at $70 and would stay at the level for a decade.


1/13/2020 Update. See this post for some cost-adjusted data that suggests that direct primary care has net positive effects.


Here’s a chronological list of posts relating to AEG/WP’s “Healthcare Innovations”.

A few brags from a few DPC companies is not a sound basis for public policy decisions.

Leave aside the specific critiques of the last twenty or so posts. The support for direct primary care in the report Healthcare Innovations in Georgia: Two Recommendations ultimately turns on the source material from which the report authors drew the key assumption that direct primary care reduces downstream care cost by 15%. That material comprises three self-reports of claimed successes in medical cost control by three direct primary care operators seeking to extend their reach by open marketing efforts.

There is, of course, nothing wrong with self-service so long as it is done in good faith. We may assume, without deciding, that Nextera’s relationship with Digital Globe, Paladina’s with the city of Arvada, and CHI ‘s individual care clinic shows some real cost-reduction success, and still ask, “Who reports the failures?

There are over 1200 DPC practices. Surely there are a great many marketing directors that would be happy to brag about their DPC firm’s demonstrable success in reducing downstream care costs. Just as surely, there are few who would make an effort to publicize a failure to reduce downstream costs.

On that score, note that Paladina could report cost control results for many more of its clinics than it does; it relies on two. Nextera, now with over 50 physicians, a presence in seven states, and patients in the thousands, still relies on the 200 patient DigitalGlobe study from 2015 to prove its cost-effectiveness.

The reality is that marketing involves reporting stuff that makes the marketed firm look good — and burying anything else. Scouring marketing materials and collecting press releases does not get representative results.

Any serious investigation of the effect of direct primary care on downstream care claims costs requires recognizing that collecting isolated, self-serving reports of good results tell us next to nothing. What is needed is comprehensive, objective research that detects broadly representative results of whatever stripe they may bear.

When was the last time you went to a financial services company’s website and read that the firm’s mutual funds have among the lowest returns in the industry? Or seen a city put up a sign that the local high school football team went 1-9?

Total claims cost caution: when DPC is implemented primary care claims vanish. AEG/WP's 15% estimate is not conservative in the least.

When the direct primary advocates toss out figures about overall claims cost reductions, it’s important to carefully separate overall cost, downstream care claims costs, and overall claims costs.

For example, the authors of the AEG/WP pitch for DPC in Georgia, have assumed a 15% reduction in downstream care costs and claimed that it “represents the low end of possible savings.” At the same time, they have relied on sources like the Arvada County clinic and the Nextera clinic that have show overall claims cost reductions of 22% and 20%, respectively. Properly understood, the reports from those two clinics suggest that 15% is anything but a conservative estimate of downstream care costs. Here’s why.

In analyzing potential Georgia savings, the AEG/WP authors correctly recognized that direct primary care enrollees would cease to incur claims costs for primary care. To amend their calculations, they relied on widely available medical claims cost data that indicated 7.0% of all medical claims were claims for primary care. Accordingly, they reduced overall claims cost by 7% to determine the base amount of downstream claims.

The same common sense needs to be applied to the Arvada and Nextera overall claims cost data. Since direct primary care members pay a fixed fee instead of making claims for for primary care, enrolling in direct primary directly reduces overall claims cost by 7%. So, for example, when the brochure for the Arvada clinic notes 22% lower claims costs for DPC-members vs FFS member, 7% is due to elimination of primary care claims, leaving only 15% that can be fairly attributed to downstream claims cost reductions. For Nextera, which sees a 20% lower claims cost for Nextera members, an even smaller downstream care cost reduction is demonstrated,

Even if the Arvada and Nextera claims data were entirely free of selection bias or any other errors, they come nearer to refuting than to supporting AEG/WP’s assertion that a 15% reduction in downstream care costs “represents the low end of possible savings”.

A possible 11% reduction in overall care cost, adjusted for risk, is suggested by Union County's 2018 report.

Here’s some data that shows plausible overall cost reduction from direct primary care even after adjusting selection bias. It comes from the Paladina-operated clinic in Union County, North Carolina, the principal subject of two prior posts.

The county employees choose either a high-deductible HSA under which primary care is received on a fee for service basis or a plan under which they receive primary care at a direct primary care clinic. About 1000 persons were enrolled in each group. The county reported that for the plan year that ended in 2018, overall costs to the employer for the DPC group were 26% lower than those in the FFS group.

The county also reported relative medical risk scores for the the two groups, and these were distinctly different: 1.11 for the FFS group and 0.92 for the DPC group. Based on these risk factors, the DPC group could be expected to have total medical costs that were 17% lower than the FFS group.

The expected costs for the DPC based on these risk factors would 83% of those for the FFS group. Instead, the costs were 74% of the FFS group. When the difference of 9% is divided by the 83% base figure, it suggests that the beneficial effect of direct primary care was just under 11%. That’s well short of the 26% that appears in a first glance at the County’s report; reasonable risk adjustment makes a big difference.

Seemingly, however, direct primary care can make a significant difference as well.


Bear in mind that, however valid the 11% improvement may be, it will not necessarily hold up when extended from the the relatively healthy (“0.92”) group to a significantly less healthy (“1.11”) group. See this prior post.

To learn how much direct primary care can do, try it first in the ACA-compliant, full-benefit individual market.

If Georgia must mandate the availability of direct primary care, here’s how.

For some future open enrollment period, the individual market will offer paired plans that differ only by how primary care is paid for and how it is received. Bigco, for example, offers Bigco Silver FFS and Bigco Silver Direct ; MajorCo probably offers MajorCo Silver FFS and MajorCo Silver Direct. There will be similar paired offerings from Aetna, Cigna, and LargeCo. Each company has contracted with primary care practices to provide fee for service, primary care services at in-network rates; each company has also contracted with one or more direct primary care practices to provide in-network primary care services at an agreed monthly fixed fee.

The scope of direct primary care services to be covered by the fixed fee is defined in a list. The sole compensation to direct primary care practices by the insurer is the monthly fixed fee. In addition, the direct primary care practices may, if they wish, perform unlisted services at the insurer’s in-network rates.

BigCo, MajorCo, and the other large insurers will have sat down with competing direct primary care providers and will have been presented with data purportedly showing potential downstream care reduction performance . The competing insurers will each have a team of actuaries and econometricians make an independent evaluation of these performance claims; the insurers will then determine which direct primary care providers, if any, are likely to deliver cost-effective results if contracted to the insurers network.

Some insurers may accurately estimate the performance of direct primary care and produce appropriately priced DPC policies for the market, relative to FFS policies. Some insurers may make inaccurate estimates of DPC performance, resulting in relative underpricing or overpricing of DPC policies. Any underpriced policies will spell losses for the insurer on a per policy basis; moreover, these loses will be magnified because the underpriced policies will attract a enhanced disproportion of policy sales.

While subsequent enrollment cycles approach, however, insurers’ econometric/acutarial teams will adjust prices based on accumualting performance data. The payments insurers are willing to make willing to make and the prices DPC providers are able to demand will move as needed toward pricing that reflects whatever actual downstream cost reduction performance direct primary care has delivered.

The resulting premium differential, if any, between DPC and FFS coverage will come to reflect strongly, if not entirely, the real difference, if there is any, between DPC and FFS in cost-effectiveness. However that turns out, there would likely remain some willing to pay a premium to remain in particular plans to retain the FFS or DPC primary care physicians they prefer.


At least in the individual and small group markets, moreover, selection bias would not play a role determining how an ACA marketplace comparison of DPC and FFS delivery would turn out. The Affordable Care Act includes a risk adjustment process under which plans with riskier populations receive funds from plans with less risky populations. The very intent of risk adjustments was to get providers to compete by innovating in healthcare systems instead of competing by risk selection. Accordingly, a mandate that Georgia insurers make direct primary care available for a serious market test might be relatively safe in the risk-adjusted market.

Since Georgia is not a command economy, however, it seems wise that any mandated test be of limited scale, say something limited to an urban county, a rural county and one in between. A limited test may also be wise given that there are relatively few direct primary care providers in Georgia. Upscaling for a full statewide test could stretch direct primary care resources too thin. Relatedly, some attention might be needed to assure that neither negotiating insurers nor physician practices willing to provide direct primary care are forced by the mandate’s design to make bargains that would prevent the real value of direct primary care from being correctly determined and reflected in the market.


For the large group market, the ACA does not have a built in risk adjustment procedure. Without that backstop, it would make sense to allow direct primary care prove its cost-effectiveness in an individual market experiment, prior to mandating the availability of direct primary care in the large employer market.

On the other had, allowing the large group market to test direct primary care would be dandy.


The best reason to limit and, even, to eschew entirely any mandate is that insurers can figure all of this out on their own. Some large insurers seem to be moving on their own toward incorporating something like direct primary care. See, e.g., this piece which ties together direct primary care, Aetna’s merger with the drug delivery giant CVS, and CVS’s experiment in offering somewhat DPC-like “Health Hubs”. For another example, large insurers are working alongside both and established DPC providers on coordinated arrangements like these involving Boeing, Iora Primary Care, Aetna and Cigna.

At same time, the features or results of direct primary care that are touted as being most effective for reducing downstream costs are already known to insurance companies through past and/or present experience: a set number of primary care visits included without any charge; primary care visits for a small copay; “welcome” physicals; wellness programs; fees-for-service that pay primary care physicians more money for longer visit durations; and, of course, primary capitation itself, including that with rates tied to results.

Rather than a new mandate, why shouldn’t the health insurance marketplace work out the role of direct primary care on its own?

Three bad ways to bet the health of Georgia citizens on direct primary care.

Every published claim that direct primary care makes a significant dent in necessary health care spending is dubious at best. See, for example, here, here, here, here, here, here, here, here, here, here and here. When the data from the Union County clinic — a Georgia Public Policy Foundation favorite — is age-adjusted, it indicates that the direct primary care option is at least as likely to have increased that county’s overall costs as to have reduced them. There is no proven performance that would justify the State of Georgia mandating that insurers offer direct primary care.

The major attraction of direct primary care to some advocates surely is its potential for limiting the public subsidies health care, whether for Georgia’s indigents or Georgians of modest means. Even if direct primary care could be shown to have some ability to lower overall healthcare spending, that ability needs to measured against a realistic expectation of what health care costs.

The recent report on healthcare innovation prepared on behalf of Georgia PPF by AEG/Wilson Partners was based on annual claim cost to insurers for the year 2020 of just under $5000 per person per annum for the relatively inexpensive large employer insurance market. The figure did not include a penny for insurance company administration and profit. It was also based on policies with an actuarial value of under 80%. That puts the annual average cost of healthcare goods and services well over $6000.

The annual cost per person of Medicaid expansion has been with sight of that number. For 2018, the 94% federal share and the 6% federal share worked out to a little over $5200 a year. That’s also almost exactly what Union County paid for each person covered by its direct primary care plan in 2016; but those county beneficiaries paid on average an additional $379 out of their own pockets to bring the average 2016 cost for Union County insureds in its much-praised DPC clinic over $5500. ACA Medicaid expansion does not look too bad in that light.


Bad idea #1. Direct primary care as an alternative to ACA Medicaid expansion.

A 2017 GPPF proposal for Medicaid is built around a direct primary care model claimed to be so cost-effective that the healthcare of indigent Georgia citizens can be safely be budgeted at $2500 per person per annum; and the same GPPF proposal would, concurrently, relieve Georgia public hospitals of all obligations to the indigent. That’s well under half the amount that the GPPF’s vaunted Union County spends on its direct primary care members. It’s magical thinking, or cynical.


Bad idea #2. Massive cuts to public employee benefits based on migration to direct primary care.

The aforementioned GPPF-sponsored healthcare innovation report, without a word of explanation, identifies the State Health Benefits Program as an “obvious candidate to pioneer” direct primary care. How long before GPPF proposes huge budget cuts for the SHBP built around wholesale migration to direct primary care? Hopefully, not until direct primary care has proven its worth in a fair, significant marketplace test.


Bad idea #3. A §1332 waiver implementation, or any other scheme, that encourages the seemingly healthy to enroll at direct primary care clinics without maintaining an insurance plan that offers comprehensive, catastrophic coverage.

Here’s what the people running direct primary care have to say about maintaining wrap-around health insurance.

From SALTA, a direct primary care firm founded by one of the principal authors of the GPPF direct primary care proposals, says: “We recommend you continue to carry private health insurance to cover the additional medical services not covered in the program”

From CHI Health Clinics, one principal sources relied in the GPPF direct primary care proposal: “Do I still need health insurance? Yes, by all means keep your insurance to cover specialty care, hospitalizations, high-cost imaging, medications and true emergencies.”

From Nextera, another main source relied on in the GPPF DPC proposal:”Patients are generally best served by combining Nextera Healthcare membership with a high-deductible insurance policy to cover specialist physician and hospital services.”

From Evans Direct Primary Care (Evans, GA): “You need healthcare insurance for hospitalizations, ER and specialist visits, X-rays, Labs, Meds, etc. . . .”

From Southern Care Direct (Calhoun, Georgia): “[O]ur affordable fees make our high level of service ideal for individuals with low cost, high-deductible insurance plans (with or without accompanying health savings accounts). . . .”

From Direct Primary Care of Marietta (Georgia): “All patients should maintain current health care insurance coverage or purchase a high deductible health care plan to cover emergent, urgent, hospital, specialist, radiology, laboratory, pathology, home health, Physical or Occupational Therapy or any other medical service not provided by [us].”

Let’s follow what direct primary clinics actually preach about wrap-around coverage, rather than combine claims of direct primary care efficiency with a §1332 waiver to support bad health insurance choices.


It is also likely that diversion of significant numbers of the relatively healthy into direct primary care plans with no other insurance would create a very large adverse selection problem. That in turn would tend to convert ACA-compliant, full benefit plans in the individual market into high-risk pools with premiums unaffordable to the unsubsidized middle class.


Instead of these three bad ideas, let’s look at this somewhat better idea.

AEG/WP's chosen actuary did not validate the assumption that direct primary care reduces downstream care costs.

AEG/WP report declares that “[Nyhart, an independent] actuary determined that “(1) the modeling assumptions are reasonable for this type of analysis and (2) the illustrative projections and savings are reasonable outcomes based on the modeling assumptions and data inputs selected.” This statement sounds like powerful support for report’s key assumption that direct primary care brings huge reductions in downstream care. It seems intended to convey the idea that the actuary from Nyhart engaged in an independent, probing analysis of past claims data to make an actuarial sound prediction of future claims. But that’s not what Nyhart was asked to do.

As the actuary also noted:

Nyhart relied on AEG and WP for the accuracy of the data provided to us and accepted it without audit. To the extent that the data provided is not accurate, the projections provided in this report would need to be modified to reflect revised information.

<SNIP>

Claims Reduction Due to Direct Primary Care Model: A 15% reduction in total claims (medical and prescription drug) was assumed. The factor is based on research and case studies prepared by Wilson Partners and represents the low end of possible savings.

Taken together these passages suggested that the actuary, Randy Gomez of the Nyhart firm, had concluded no more than that the 15% was “on the low end” of the specific material presented to him by Wilson Partners. Mr Gomez has confirmed that understanding to me in an extended telephone conversation.

We have previously examined the source data used by Wilson Partners: a 2018 20% claim of possible involving CHI; a 2016 claim of possible savings for 22% for Paladina/Arvada; and 21 % possible savings for Nextera in 2015. Mr Gomez could appropriately sign off on Wilson Partners suggestion that 15% represented a safe low end figure for possible savings after assuming 20%, 22% and 21% were accurate possible savings data points. Gomez intended no more than that.

Gomez, in other words, had accepted but not validated the cost reduction claims presented to him by Wilson Partners. Specifically, Gomez was not asked to undertake, and did not undertake, an actuarial examination of medical risk data or historical claims data from the clinics identified to him. He was not hired to, and did not, in any way address selection bias or any similar factor. When asked about the clinical and risk data supporting the cost reduction assumptions, Gomez referred me back to Wilson Partners.


Reach out to Mr Gomez at randy.gomez@nyhart.com or (317) 845-3595. He was interested in the subject, candid, and a pleasure to talk to.


But what if AEG/WP had found an actuary who claimed to have actually validated the 15% downstream cost reduction claim for direct primary care? In that case, the appropriate response would have been, “Show me.”

The reality is that neither actuaries nor Wilson Partners nor anyone else has as yet demonstrated that direct primary care lowers downstream claim costs after adjustment for selection or other biases — by any amount.