DPC cherry-picking: the defense speaks. Part 2.

Update: In the fall of 2020, KPI Ninja released the first study that relies on it’s new risk information technology. I find it sadly opaque.

Recap of Part 1

The direct primary care community has long tried to support claims that DPC reduces overall health care costs by 20% to 40% with non-risk-adjusted cost-reduction data drawn from employment health plans that allowed employees to elect between DPC and FFS primary care options options. But the first and, so far, only time that independent, neutral, certified professional actuaries looked hard at such a program, careful risk-adjustment showed that the savings claimed were merely an artifact of heavy selection bias. A DPC poster child, the Union County employer program — previously lauded for its claimed 23% cost reduction — was shown by Milliman Actuaries to have had a DPC cohort so young and healthy that it explained away all the observed cost savings.

Any reasonably informed employer or policy maker facing claims about the cost-effectiveness of direct primary care should insist that DPC provider boasts be scrutinized for evidence of selection bias.

In my last post, I noted that the DPC advocacy community has not even bothered to address the simplest of selection bias indicators, the younger average age of DPC cohort members compared to their FFS counterparts. I also noted that Milliman Actuaries have an age/gender risk-adjustment model1 and considered using it for the Union County study.

I further noted that in its Union County study, Milliman relied on a more complex model (“MARA-Rx”) which consider age, gender, and various therapeutic classes of prescription medications used by DPC and FFS group members. Milliman’s Rx risk-adjustment methodology, like its other risk adjustment models, was developed and validated as a health care cost predictor using health care cost data for millions of patients. There is nothing I can see, and nothing in the literature, to suggest that the Rx model has any inherent features that would unfairly penalize a DPC cohort in a studied DPC option employer health plan. As yet, no one in the DPC community has objected to Milliman’s use of the Rx methodology to assess the Union County DPC program, or to its future use in evaluating any other similar program.

There are even more complex and expensive methodologies, like MARA-Cx, that add diagnostic data harvested from payment claims to the factors used in MARA-Rx. In the Part 1 post, I also mentioned an arguably surprising difference in approach to selection among risk-adjustment methodology between Milliman Actuaries, who have no financial interest in the direct primary care movement, and KPI Ninja, a data-analytics group closely connected to the direct primary care movement. Both Milliman and KPI Ninja concurred that risk adjustment methodologies like “Cx” are likely to understate risk score for DPC cohort members because direct primary care physicians do not file claims.

KPI Ninja pointedly laments this “donut hole” in the claims-based data. But there is no anti-DPC donut hole in Rx based risk adjustment methodology.2

Although it possesses a fully-validated claims-based risk adjustment methodology (“MARA-Cx”), Milliman’s common-sense response to the data donut hole problem was to set that Cx methodology aside and determine risk-scores for the Union Count cohorts using only the Rx age/gender/prescription drug methodology. Like Milliman, KPI Ninja has access to risk-adjustment software engines that have equivalents, in a package known as ACG®, to both Rx and Cx. Unlike Milliman, KPI Ninja seemingly rejects Rx methodology and, instead, embraces Cx type methodologies that have the very donut hole KPI Ninja laments.

Why complain about the donut hole in Cx, then reject Rx which has none, and then return to and embrace Cx? Might it be precisely because KPI Ninja knows that Rx based risk adjustment will produce results that are sound, but not happy, for its cherry-picking DPC clients? On the other hand, when convenient, a donut hole can first be performatively disparaged as biased, then filled with custom data products developed by KPI Ninja to tell stories more to DPC’s liking.

Context: DPC docs feel coding makes their patients sick.

Direct primary care practitioners avoid third-party health care payment of claims and the (often digital) paperwork that accompanies it. While the logic of subscription practice renders coding procedures for reimbursement unnecessary and, therefore, a target of scorn, D-PCPs also disdain the type of recording of diagnostic codes that attend claims for third-party reimbursement.

Here’s what Dr Jeff Gold, a co-founder of the Direct Primary Care Alliance, had to say in a post entitled, “ICD-10: It’s Nice Not Knowing You.”

This is nothing more than another layer of bureaucratic red-tape that does nothing to enhance the quality or cost of your care, but rather furthers the disease process. All it does is waste more of your physician’s and office staff’s time – time that should be spent working towards your care. . . .

Luckily for us [Direct Primary Care doctors], we have nothing to do with this nonsense. [Emphasis supplied.]

ICD-10: It’s Nice Not Knowing You

Tell us how you really feel, Dr Gold.

DPC doctors, in other words, not only decline to file claims and code their procedures, they also hold industry standard diagnosis coding in fiery contempt. As a result, the donut hole problem can not be solved by simply collecting DPC patient diagnosis codes from their direct primary care physicians.

Enter KPI Ninja — with balm for DPC’s abstinence

Missing data reduces population risk score. Meaning it will look like you are treating healthier patients in the eyes of those who use these risk scores (employer data vendors, brokers, health plans) … aka they can argue that you are cherry picking.

KPI Ninja Blog Post: Claims vs. EHR data in Direct Primary Care

While the details remain murky, KPI Ninja seemingly plans to meet cherry-picking charges by filling the donut hole with information somehow scoured from such EHR records as direct primary care doctors may have.

But how?

In theory, KPI Ninja could develop a way to reverse engineer a DPC’s EHR entries and other information to generate the diagnostic codes upon which the DPC physician would have arrived had she not thought that participation in diagnostic coding was wasteful. How proceeding “backwards” to arrive at codes would result in any less waste is difficult to image, so that effort strikes me as misguided.

In any event, validation of a reverse engineering model would likely require resources beyond those of KPI Ninja and the DPC advocacy community. It would also likely require the participation of a control group of D-PCPs willing to do extensive coding.

However, for physicians like Dr Gold who have identified ICD-10 coding as “furthering the disease process”, such participation in a coding control group would be an ethical violation; it would “do harm”. Furthermore, if Dr Gold is correct, the members of a patient panel studied in such an experiment would have to give informed consent.

Furthermore, to whatever extent EHR mining for diagnosis codes omitted by a D-PCP produces fully accurate risk codes for members of a DPC cohort, the same mining techniques should be applied to the EHRs of FFS patients to correct any omissions by FFS-PCPs. What’s sauce for a DPC goose is sauce for an FFS gander.

Finally, fair implementation of a model in which diagnosis codes for risk scoring are derived from DPC-EHRs for comparison with diagnosis codes from FFS-claims would require safeguards against the DPCs deliberately larding EHR’s with entries that result in up-coding, just as FFS-claims data is subject to procedures to check up-coding. Again, goose/gander.

Perhaps, KPI Ninja merely has in mind developing a direct method of converting mined EHR data into risk factors that are not directly commensurate with those from diagnosis-based risk models, but that are instead presented to inquiring employers and policy-makers as an alternative model of risk-adjustment.

Precautions would still apply, however. If EHR data on, say, biometrics or family history is brought in demonstrate that the DPC population is less healthy than average, a knowledgeable employer should insist on counterpart data from the EHRs of FFS patients.

A recent addition to KPI Ninja’s website suggests their emphasis may rest on pre-post claims comparisons. It will of course be important to include pre- and post- data on DPC patients and their FFS counterparts. That will be somewhat revealing if the pre-choice claims data for the two populations are similar. But if the results of the Milliman study are representative, that will most likely not be the case.

In the more likely event of a higher level of prior claims in the FFS population, any “difference in difference” analysis of pre-post claims between DPC and FFS populations will still require an attempt to see whether the FFS higher “pre” claims might be accounted for by intractable chronic conditions. Such a finding would impeach any inference that DPC group pre-post gains can simply be projected onto an FFS group. And such a finding seems likely, if the Milliman actuaries were correct to ascribe selection bias to the tendency of the ill to stick with their PCPs; a “sticky” relationship to a PCP seems quite likely to correlate “sticky” chronic health conditions that bind the patient and the PCP.

Further explication from KPI NInja as to how its plans will work was said to be forthcoming. I appreciate the insights they have given in the past, and look forward to learning from them in the future.

There are many risk adjustment software packages available from neutral academics, accountants, and actuaries. They can be expensive; even access to the packages supplied by CMS for use by insurers participating in the Affordable Care Act run a bit over $2 per enrollee per year. Importantly, some of the packages rely on proprietary algorithms, but nonetheless tend to be generally transparent. In most cases, however, these packages come from neutral academic or actuarial sources.

Per se, I fault DPI Ninja neither for its close connection to the DPC industry nor for offering to DPC clinicians the best “risk adjustment” bang that can be generated from the least record keeping buck. To the extent that DPI Ninja delivers DPC data that is generally transparent in its methods and assumptions, that work will speak for itself. Should DPI Ninja lay on data produced by secreted assumptions and methodology, however, it should expect that its business relationship to DPC will affect the credibility of that data.3

Update: In the fall of 2020, KPI Ninja released the first study that relies on it’s new risk information technology. I find it sadly opaque.

For completeness, two parting thoughts.

First, none of foregoing is intended to deny that membership in a direct primary casre practice can significantly reduce usage of emergency departments. Of course it can! The question is what does it cost to avoid such visits. It is very rare for United States Presidents to go to emergency rooms as there’s always a doctor within a few hundred feet. DPC docs are easier to reach at odd hours and are more available for same day visits. They have fewer patients, and that costs a lot more money per patient.

Second, I have discussed in many posts the many sources of selection bias. Use the search item in the menu to find example using words like “selection”, “bias”, or “cherry”. It’s real and the only truly debatable proposition is whether DPC advocates who deny it are unintentionally fooling themselves or deliberately fooling you.

1 A college freshman IT student could probably develop an age/sex risk adjustment engine and apply it age/sex data from paired DPC/FFS cohorts over a weekend. If DPC clinics don’t report age/sex/risk comparisons with their FFS counterparts, it’s because they know full what the results would show.

2 It may even be that Rx metholodogy favors DPCs.

I expect to hear soon that Rx risk adjustment discriminates in favor of “bad”, overprescribing PCPs by making their patients seem sicker than those of “good” doctors who never overprescribe. But this can become an argument that Rx discriminates against DPC if, and only if, it is also assumed that DPC has fewer “bad” overprescribers than does FFS. There is no clear factual basis for assuming, in essence, that DPC doctors are simply better doctors and prescribers than their FFS counterparts.

If anything, DPC self-promotion suggests that Rx data would be skewed, if at all, in the opposite direction. DPC advocates regularly claim that DPC avoids expensive downstream care by better discovering illness early and managing chronic conditions better, by coaching patient compliance and by lowering the cost of medications. Another recurrent theme among DPC advocates is that FFS doctors rely too heavily on specialists; but, if true, FFS cohort patients would be more likely that DPC patients to have over-prescription “wrung out” of their regimes. In these ways, applying a risk adjustment methodology based on prescription medication data, DPC’s own bragging suggests that any donut hole is likely to make the FFS cohort appear healthier.

3 The following fairly new statement on KPI NInja’s website does not bode well, suggesting both secrecy and a predetermination of an answer favorable to DPC.

“We analyze historical claims data from self-insured employers by utilizing our proprietary algorithms to identify cost savings and quality improvement opportunities. Get in touch to learn more about how we can help show your value to employers!”

DPC cherry-picking: the defense speaks. Part 1.

Jump to Part 2.

Within days of the Milliman report warning of the “imperative to control for patient selection in DPC studies [lest] differences in cost due to underlying patient differences [] be erroneously assigned as differences caused by DPC”, the first rumbling of resistance from the DPC advocacy community emerged. This was a suggestion, addressed to one of its authors, that the Milliman study may have treated the direct primary care employer unfairly.

KPI Ninja also reached out to me. After some initial misunderstanding on my part, and subsequent examination of KPI Ninja’s online published material on this subject, I reconstruct my understanding of KPI Ninja’s argument in the immediately upcoming section. For reasons discussed in the next following section, I have concluded that KPI Ninja’s argument, although not without insight, simply does not apply to the risk-adjustment methodology used in the Milliman study itself. Thereafter in this post, I begin to respond more broadly to the KIP Ninja critique and to the not-yet fully visible “remedy” it is apparently developing. I will conclude in a subsequent post.

KPI Ninja’s Donut Hole Concept

Some data used in some risk-adjustment methodologies are diagnostic data harvested from claims, including both primary care and downstream care claims. The risk that manifests itself in downstream claims properly counts in evaluating a patient’s risk, whether the patient is a DPC or an FFS patient. But a problem can arise (a”donut hole” in KPI Ninja’s lexicon), in situations where some PCPs have success at averting downstream claims. If the patient is in an FFS practice, the practice is scored as bearing any risk factor appearing in the primary care claims submitted by the FFS PCP. FFS practices, in essence, get fair credit for their good work in avoiding downstream claims. On the other hand, D- PCPs do not file primary care claims; as a result DPC patients risk factors go unrecorded when not reflected in downstream claims; DPC docs therefore do not get proper credit for their good works.

In the upshot, DPC patients look relatively less risky than they really are. On this analysis, risk-adjustment by claims dependent methods of two equally risky panels of DPC and FFS patients will inevitably disfavor DPC.1

The Milliman study of DPC used risk-adjusment that had no donut hole.

There are ways of evaluating population risk that do not depend on harvesting diagnostic data from physician claims, methods for which there is no claims associated donut hole. Milliman itself, based on experience with millions of people, developed and validated at least two methodologies that do not require claims data: one (age/sex) determines risk factors based only on the age and gender characteristics of the populations being compared; a second (Rx) adds, to age and gender, additional patient information about the usage of different therapeutic classes of prescription drugs. Under the Rx methodology, claims are not looked at; DPC and FFS docs get the same “null credit” for every prescription drug avoided through their primary care work.

Of course, Milliman also has a claims based risk-adjustment methodology (Cx). For its study of Union County’s direct primary care option, Milliman carefully considered using the Cx methodology and, precisely because they identified the donut hole, expressly rejected it. Milliman also considered using its age/gender risk adjustment methodology , but decided to use its more granular RX methodology.2

Simple, lost cost risk-adjustments are affordable for modestly sized employers

The beauty of an age/gender risk-adjustment is that it is straightforward. Nor is it doomed to failure by its simplicity. Over three-fourth of persons under the age of 65 lack any diagnosis consider significant for risk adjustment carried out under the Affordable Care Act. Nor is it likely to be expensive; it is definitely not rocket science. A college freshman just learning how to use a spreadsheet could fire it off in an hour, if pointed to a readily available table, and a data set with each employee’s age and gender.3

In its own published “research”, one of the larger DPC groups (Nextera) used raw total claims from an immediate pre-DPC period as a basic risk adjustment methodology for employer DPC and non-DPC groups claims data for the period following the creation of the DPC option. Those choosing Nextera had vastly lower prior claims.4

Dan Malin, a TPA exclusively serving small employers, addressed a 2019 DPC Summit forum on employer DPC with about 100 attendees present. He claimed that his firm could calculate the medical consumption of that group for a coming year to within five or six percent by having each fill out an ordinary health insurance application.

Has an employer-option DPC cohort ever been OLDER than its FFS counterpart?
Or ever entered a DPC option with a HIGHER prior claims experience?

But DPC advocates could go some of the way long way toward reassuring ordinary employers by demonstrating measurable similarities between cohort members being studied. Say, for example, that a DPC group had virtually the same claims as an FFS group in a claims year prior to creation of a DPC option. Unfortunately, for some reason, whenever these type of comparison have been made, the DPC cohort has turned out to to have had a history of smaller pre-DPC claims. This might be because, as Milliman warned, direct primary care is a cherry-picking machine.

As one DPC thought leader has pointed out, “larger direct primary care companies like Qliance, Paladina, and Nextera have repeatedly reported 20% plus savings for employers using the DPC model. Many smaller employers have found similar savings [].” If those selecting DPC were equally likely to be older or younger than their FFS counterparts, the chance that even as few as ten5 such studies would each have had younger DPC populations would be less than 0.1%.

Alas, for some reason, there are no known reports of a DPC cohort being older than its FFS counterpart. Also, perhaps for some related reason, it often appears that employer DPC cohorts are younger than their FFS counterparts.

In some cases, DPC advocates make an effort to show that a presumably large percentage of a DPC cohort have multiple chronic conditions, but reliably matched data from the corresponding FFS counterpart cohort has not, to our knowledge, been reported. Indeed, a DPC advocate promoting the very poster-child clinic studied by Milliman once sarcastically dismissed the idea of cherry-picking by pointing to the chronic conditions of the DPC’s patients — while noting that “control group data is not available”. Really! Since Milliman presented that very control group data in May, we’ve heard no comment by that author.

Does any of these methods have any significant built in distortion against direct primary care? Does direct primary care make younger? Did receiving direct primary care in one year lower claims costs for the preceding year?

Link to Part 2.

Footnotes

1 A masterful extention of that argument came in a tweet from one DPC thought leader, who was asked about the lack of risk-adjustment in a just-published savings claim for his own DPC practice. He invited the inquirer to consult DPI Ninja “or spend $100,000 to prove them right or wrong.” Is this how he plans to answer employers who’ve read Milliman’s advice for employers?

2 Interestingly, Milliman recorded that adjustment using Rx in this particular case was more favorable for DPC than using age-sex adjustment.

3 Virtually every direct primary care clinic in the country has chosen an even simpler path when using age only to establish a variable, usage-anticipation- component for its monthly fee DPC schedule.

4 The employees who chose Nextera’s direct care had, on average, historical prior claims that were 30% less than those who declined direct primary care. Interestingly, Nextera sought to rely on a “difference in difference” analysis, but intertemporal claim comparisons of that sort fail if, for example, the declining groups higher “pre” claims can be accounted for by intractable chronic conditions, something that seems entirely likely.

5 At the moment of commencement of DPC Summit 2020, a least ten such studies of employer option DPC have been publicized. At least one additional study seems likely to appear on July 18, 2020. I wonder if the study results reported will include a comparison of the average age of the two cohorts.

Link to Part 2.

Milliman: A $60 PMPM DPC fee buys an employer a zero ROI.

An actuarial study brings employer direct primary care to a turning point.

Milliman’s actuaries insisted that DPC cost reduction data without risk adjustment is essentially worthless. A second prong of Milliman’s analysis suggested that the direct primary care model is associated with a 12.6% over-all reduction in health services utilization*. Then, working from that number, Milliman went on toward a third suggestion: that an employer who buys into DPC at an average price of $60 PMPM would likely have an ROI of zero.

That is not a very good deal for either an employer or a DPC practice.

For an employer, entering into a break even deal would mean foregoing other investment opportunities, while incurring inconvenience of change and probably a loss of good will from a number of employees who might be, or feel, forced or financially pressured into a narrow primary care network. Why bother?

For a DPC provider, $60 PMPM works out to $432,000 per year for the patient panel of 600 members that DPCs like to brag about. But an average PCP compensation package in, say, Anderson, South Carolina comes to about $276,000. That leaves about 36% of revenue for all overhead. That’s not much.

The American Academy of Family Physicians tells us that overhead for family physician practices runs around 60% of revenue. The most recent peer reviewed study of the subject (2018) indicated billing and insurance (BIL) costs for primary care came out 14.5% of revenue. A somewhat earlier peer reviewed study came near this, finding that physician practices had BIL costs of about 13% of revenue. Even if, say, 15% of revenue can be eliminated in direct pay, D-PCPs could still expect overhead to be 45% of revenue.

In the direct primary care employer clinic, moreover, billing costs do not fall to zero; patient rosters need to be kept up to date, matched to employer records, and processed. In addition, most direct primary care physicians also provide a significant measure of separately paid goods and services for which an employee, employer, employer’s insurer, or a TPA may require documentation and billing. Moreover, much of the data attributed to billing and insurance costs in an FFS setting has a counterpart in direct primary care collection of metrics needed to demonstrate value to an employer.

Accordingly, even for a non-insurance direct primary care clinic, overhead of 36% of revenue is scanty.

At the same time, DPC practitioners have — and regularly express — a very high opinion of themselves and the care they give. As group, they seem distinctly unlikely to settle for merely average levels of compensation.

The most profitable path forward for direct primary care is to persuade employers that paying PMPMs over $60 will do more than break even. Historical brags, based on data that was not adjusted for risk, claimed employer savings from 20% to 40%. A DPC thought leader recently published a book on employer DPC that collected references to such seven studies, including one that claimed to have reduced employer costs by 68%.

Now that real actuaries have weighed in, those days are numbered.

So now they just fake it.

* I believe that Milliman’s 12.6% figure vastly overestimates the reduction in health services utilization association with direct primary care. As explained here, DPC may even result in an increase in health services utilization. Is it really plausible that taking a scarce resource — the time of PCPs — and spreading it thickly over tiny patient panels would NOT result in net economic loss?

Ultimate DPC goal: capitation without accountability (rewrite)

Allowing an HSA holder to use pre-tax dollars to buy subscriptions only gets DPC operators so far. The HSA holders would still notice that their paid subscription fees will not actually make a dent in meeting their insurance deductible. DPC advocates will then reprise their perennial theme song, “Insurer’s conditions on payment interfere with the doctor-patient relationship.”

It is interesting to note how deeply this idea is invested with emotions: specifically respect for physicians, and disdain for insurers. But why, after all, should any physician have anything to say about an individual’s contractual relationship with an insurer?

This will play out in a multi-pronged DPC effort to interfere with the insurer-insured relationship. The effort will be tagged “free market reform”.

It will, of course, be demanded that insurers credit capitated subscriber payments on the same basis that they credit fee for service payments. Insurers might or might not accept that principle. DPCs will, if necessary, seek legislation to interfere with the insurer-insured relationship by requiring insurers to accept capitation, citing interference with the doctor-patient relationship and “free markets”.

Insurers that decide, or are forced, to accept capitation will naturally want to negotiate price and quality measures to protect themselves from the inherent incentive of providers receiving capitated payments to do too little and to push costs downstream. A satisfied insurer will declare some compliant DPC providers to be “in network”.

Among the non-compliant, a familiar cry will arise: “Insurer conditions and limits on payment interfere with the doctor-patient relationship.”

“Freedom” and “free markets” will be invoked: “only doctors and patients are needed to determine primary care value and pricing”. The freedom of an insurer and an insured to agree to the shape of an insurance policy will not play a part. In the name of “health care freedom”, DPC advocates will insist that insurance companies give full credit toward deductibles of every subscription dollar paid to any provider.

The foregoing thought exercise also gets to a question raised by certain DPC thought leaders’ statements that DPC clinic are naturally disinclined to employer purchases of DPC subscriptions for their employees, and would much prefer that employers simply offer their employees HDHP/HSAs in a post-PCEA environment.

But in such an environment, neither an HDHP carrier nor a self-insuring employer, would be obligated to credit the DPC subscription fee toward the deductible. Some HDHP/HSA holders might still resist DPC unless the DPC subscription fee counts toward their deductible. On the other hand, a direct DPC contract with the employer removes that barrier to DPC enrollment. Whatever they may say, DPC contracts with employers profit DPC substantially.

It has been evident for many months that a great many DPC providers are pushing hard for employer contracts despite a decade of complaint that third party payment is the source of huge evil. It is noteworthy that the tip of the spear is pointed at landing employers that are not likely to have truly savvy benefit managers — the type who might actually demand real accountablity through properly risk-adjusted performance data.

Downstream consequences when employers fall for non-risk-adjusted data brags.

Do you remember when Union County’s three year DPC commitment for 2016-2018 was claimed to be saving Union County $1.25 Million per year? So why did Union County’s health benefits expenditure rise twice as fast as can be explained by the combined effect of medical price inflation* and workforce growth?

Union County Budget and Bureau of Labor Statistics

For the first year or two, a clinic owner in an employer DPC option may get away with presenting the employer with data brags that package selection bias artifact as DPC cost-effectiveness. The wiser employers will figure selection bias out before making the mistake of tossing all their employees into direct primary care based on non-risk-adjusted data.

For those employers who do not figure this out, it will be interesting to watch the DPC clinics adapt to the influx from the sicker, older population. Better, it will be outright fun to hear DPC clinic owners explain why having more employees in their clinic led to an increase in PMPMs for DPC patients. Since it will be hard for them to suddenly come to Jesus on the cherry-picking issue, where will they turn? Probably to blaming a combination of Covid-19, insurance companies, and Obamacare.

*For medical cost inflation I used the Bureau of Labor Standards medical goods and services marketplace statistics. I thank economics student D.S. for pointing out that in the employer insurance sub-market the rate of increase was higher. Kaiser Family Foundation’s annual Employer Heath Benefits Survey data indicate that for the period 2015-2018, costs increased 11%. Substituting 11% from KFF for the 8.5% from BLS, I compute that the 47% increase in Union County was not 200% of that expected, it was a mere 175% of that expected.

Of course, I might have attempted a narrower focus. I fully realize that the average increase might have been different for North Carolina employers, or for county employers, or for county employers in North Carolina. Then, too, there counties named Union in a total of 17 states. I confess that I deliberately refused to determine the employer health coverage cost inflation rate for the other 16 Union Counties. Life is short.

It might also be the case that bringing DPC to Union County created such great opportunities for affordable health care that DPC employees married at a higher rate, and decided to have more children. And, of course, the presence of DPC might have helped Union County recruit a large tranche of new employees with enormous families.

Medicare, dual coverage, and opt-out. The cherry on top of the cherry-picking machine for employer-based direct primary care.

In 2016, the share of people between 65 and 74 who were still working was over 25%. Any of them working at employers with more than twenty employees covered by group health plans are required by law to be included in the employer’s plan. They may also enroll in Medicare Part B. Some employer plans even require their Medicare-eligible employees to enroll in Part B. When optional for the employee, the choice to add Medicare Part B and have dual coverage is typically made by relatively heavy utilizers looking to meet cost-sharing burdens by having Medicare as a secondary payer.

In any event, elder employees with dual coverage will have low, or even zero, out-of-pocket expenses, whether for primary care visit fees; for the types of services, like basic labs, often included in an employer DPC package; or for downstream care. These elders have relatively little incentive to join DPCs, especially in cases where the employee pays more for a DPC option than for a non-DPC option (such as Strada Healthcare’s plan for Burton Plumbing).

Many with dual coverage will have even more incentive to avoid DPC. A large majority of DPCs, including DirectAccessMD, Strada Healthcare, and many Nextera-branded clincs, have opted out of Medicare. Medicare-covered employees who receive ancillary services that the DPC performs for separate charge will be expected to see that the DPC gets paid, but will receive no Medicare payment for those services. A Medicare-covered employee in Nextera’s St Vrain Valley School District plan, for example, would be denied the ability to have Medicare pick up his cost-share for Nextera’s in-office labs and immunizations, Nextera’s on-site pharmacy, or Nextera’s on-site cabinet of durable medical equipment. Were a dual covered employee to choose the Nextera clinic, she would have to make a point of declining to have Nextera draw her blood work or put her in a walking boot.

Most employer workforces have a relatively small percentage of employees over age 64. But provider health coverage for these elders is apt to be very costly. The employees likely to be most costly are the very ones that will find Medicare Part B’s annual premium of less $1750 a good bet for avoiding cost-sharing burdens like those in Nextera’s SVVSD plan – a $2000 deductible and a $4000 mOOP.

Accordingly, those with dual coverage are likely to be high utilizers of services with nothing to gain from DPC. Or, worse: some will pay more in employee contributions; some will have added costs and/or inconvenience owing to Medicare opt-out by the DPC provider.

These high-cost, dually-covered employees will disproportionately end up in the non-DPC cohort under most employer DPC option plans. And every one of them will skew non-risk-adjusted claims data, contributing to a selection bias artifact masquerading as DPC savings.

Much the same reasoning will apply to other employees who have a secondary coverage, such as being a covered spouse. Dual coverage usually comes at a price, such as a premium add-on for spousal coverage. But the price will often be worth it for high utilizers whose primary coverage has high cost sharing burdens that can be brought to negligible levels. For these high utilizers, the incentives to select a DPC option are minimal, even negative if the DPC option comes with a larger employee contribution.

Finally, whatever the source of secondary coverage, the heavy utilizers for whom it is particularly desirable are also the very people most likely to cling to particular PCPs who have served them well in the past, rather than sign on with a DPC clinic offering a narrow choice.

Two new DPC brags failed to show bona fide risk-adjusted savings; together, they make clear that DPC brags rely on cherry-picking.

Two recent DPC brags fit together in a telling way.

Nextera Healthcare reported non-risk-adjusted claims data indicating that employees of a Colorado school district who selected Nextera’s DPC option had total costs that were 30% lower than those who selected a traditional insurance option. But that employer’s benefit package confers huge cash advantages (up to $1787) on risky, older patients if they choose the non-DPC option.

Comes now DirectAccessMD with a report of non-risk-adjusted claims data that employees of a South Carolina county who select a DPC option have percentage cost “savings” that are less than half of those shown for Nextera, a mere 14% lower than those who selected a traditional insurance option. But that employer’s benefit package design has an opposite effect to Nextera’s, conferring a significant cash advantages on risky, older patients if they choose DPC option. (And, good on the County and the DPC for it!)

Nextera’s program works to push risky, older patients away from DPC; DirectAccessMD’s program welcomes them. Pushing the sick and old away helps boost Nextera’s corporate brag to more than double that of DirectAccessMD who, proudly, invite the same people in — even if it makes them appear less successful.

Even without a fancy actuarial analysis, or even a basic one based on patient demographics, it’s apparent that most of Nextera’s brag is merely selection bias artifact masquerading as Nextera DPC cost-effectiveness.

Is DirectAccessMD’s clinic free of selection bias? Not at all. Selection bias can also arise when older and/or riskier patients make enrollment decisions based on differences in access to physicians of their choice. Older, riskier patients tend to cling to a long-standing PCP rather than select from the relative few offered by DirectAccessMD. Think of this as a primary care form of cherry-picking by narrow-networking .

We have learned that Anderson County’s DPC patients are 2.7 years younger than their non-DPC counterparts. Applying CMS risk adjustment coefficients for age and sex reduces DirectAccessMD’s “savings” from 14% to 2% .

If you would like to bet that the age difference between the two Nextera school groups discussed above is less than 2.7 years, please use the contact form. I’ll be right there.

Nothing huge, but a possible small win for DirectAccessMD cost reduction claims.

The DirectAccessMD clinic that serves the employees of Anderson County, SC, is run by a tireless advocate for, and deep believer in DPC, Dr J Shane Purcell. Here the employer, with Dr Purcell’s apparent support, has taken steps that seems to have somewhat mitigated the selection bias that is baked into most other direct primary care option arrangements. Specifically, the competing benefit plans for county employees have both a lower deductible ($250) and a lower co-insurance maximum ($1250) for DPC patients than for non-DPC patients ($500, $2500). Where other benefit plans structures, like the Nextera SVVSD plan reported here, push higher risk patients away, the Anderson County plan is more welcoming to those patients. I applaud the County and Dr. Purcell.

In fact, a high risk Anderson employee can see more than $1500 per year in added costs if she declines DirectAccessMD. A patient expecting the average utilization seen for the FFS cohort (~$4750) in Anderson County would likely incur about $375 in added costs by declining DPC, where a similar patient in Nextera’s SVVSD plan would have saved $925 for doing the same. Again, this important difference is a feather in Dr Purcell’s cap.

UPDATE: 11/28/2020

In November 2020, we applied CMS’s actuarial value calculator to compare the County’s plans. The traditional plan had an actuarial value of 82%, the DPC plan of 87%. More of this update is set out below.

Yet, as the recent Milliman study suggests, high risk patients may be reluctant to disrupt standing relationships with their PCPs, and may choose to resist other incentives if it means having to select a new PCP from a small panel at a given DPC clinic. Consider also that older employees, even those not at high risk, are more likely than younger employees both to have deeper attachments to their long-standing PCP and to have more disposable income to spend on keeping that relationship going. On average, therefore, we would expect employees who eschewed the direct primary care package to be an older and/or riskier group. Let’s go to the tape.

Not surprisingly, raw data — without any risk adjustment — from the employer indicates a noticeably smaller percentage of purported savings, not adjusted for risk, than has been bragged about by other DPCs in the past. Anderson County’s net cost for DPC members came in at 9% less than for non-DPC members, but the employees in DPC had “OOP” only about half of what their non-DPC counterparts did. Combining both employer and employee costs, the average total spend for Anderson County DPC patients came to about 14% less than for non-DPC patients, purported savings of $56 a month.

But note these warning from the Milliman study: “We urge readers to use caution when reviewing analyses of DPC outcomes that do not explicitly account for differences in population demographics and health status and do not make use of appropriate methodologies.” Or this more recent one: “It is imperative to control for patient selection in DPC studies; otherwise, differences in cost due to underlying patient differences may be erroneously assigned as differences caused by DPC.”

A risk analysis of the health status of all the county’s patients, fully detailed as to all chronic coniditions, may not have been financially feasible for a modest operation like Dr Purcell’s. But a sensible population demographic methodology is at hand: comparing the ages of the two populations and using that as a predictor of utilization. This is certainly a “rough approximation”. But, not only is a rough risk adjustment likely to be better than no risk adjustment at all, the reasonableness of using age as a proxy for predicted utilization is affirmed by the fact that nearly all DPC practices use age-cost bands, and no other risk-based factor, in setting their subscription rates. Basic demographics are at the core of risk adjustments used by CMS for the ACA; over 75% of ACA enrollees in insurance plans under 65 have no adjustment-worthy chronic conditions; they are risk-adjusted on demographics alone.

The coefficients for age/sex risk adjustment used by CMS for ACA plans in 2020 can be seen here. Dr. Purcell’s slide pointing out the DPC cohort of Anderson County employees was 2.7 years younger than the traditional cohort is here. Going back to the tape, I estimated the risk adjusted overall medical costs for the DPC membership to be about 7.1 % lower than for traditional primary care group.

UPDATE: 11/28/2020

A second adjustment points the other way. All other things being equal, richer plans are known to produce “induced utilization”. CMS’s risk transfer machinery applies an induced utilization factor to adjust for benefit richness. As shown on the calculation sheet also linked above, this adjustment would increase employer costs by 3.8%.

This brings a tentative measure of savings, pending more definitive risk adjustment, to about 11 % overall, about $40 PMPM.

Maybe it is not what what Dr. Purcell hoped, but his results are more promising than most others.

The great caveat, of course, is that proper risk adjustment could turn this estimate up or down. In the Milliman study, the difference between actuarial value of the benefit packages of the DPC and FFS programs was modest, yet the Milliman team still found massive adverse selection into the FFS. Milliman accounted for this as the result of sicker people clinging to the trusted PCPs who had served them in the past. I think of that as adverse selection by narrow primary care panel. Whatever the explanation, Milliman found that the selection bias required an overall upward adjustment of 8.3 % of DPC costs; and they predicted that most employer DPC option clinics would see similar. On the other hand, a fairly large difference in benefit packages favoring DPC members, as in Anderson County, is something the Milliman team appears not to have contemplated, and it must surely drive some of the risky into the DPC pool.

I am still not betting on DPC saving big money. But, if you call me with your proposed wager, I’ve shortened the odds.

Do bears sh. . .ake cherries out of trees? Selection pressure is built into DPC choices for any population with a normal deductible.

At last, it dawns on me. Selection bias is baked into virtually every DPC cake.*

Direct primary care usually comes with a significant price and a package of financial incentives revolving around primary care (and, sometimes, around some downstream care). For some, the game may be worth the candle. The incentives, typically the absence of primary care visit cost-sharing and free basic labs and generic drugs, have their best value for those who expect total claims to fall near but still short of their deductibles. These people are relatively low risk.

For those expecting to have total claims that will exceed their deductibles even if they receive the incentives, the dollar value of those incentives is sharply reduced — usually to a coinsurance percentage of the claims value of the incentive. These people have risks levels that range run from a bit below average risk to well above average risk.

The least healthy people have the highest claims. At a next level, and all the way up to the stratosphere, are insured patients expecting to hit their mOOP in an upcoming year. For a typical employer contract, however, these people are not necessarily extreme; for an employee with a $2000 deductible and a $4000 mOOP, this represents a $12,000 claims year. That’s not even a single knee replacement at the lowest cash price surgery center. For these, DPC’s financial incentives have essentially zero financial value.

Higher risk patients have significantly less incentives to elect direct primary care. DPC patient panels are enriched for low risk patients while higher risk patients tend to go elsewhere.

Financial considerations apart, the higher risk patients are also likely to be the ones least interested in replacing established relationship with particular PCPs with primary care from a narrow panel DPC practice. A second reason why DPC patient panels are enriched for low risk patients while higher risk patients tend to go elsewhere.

The upshot: Virtually any employer-option DPC clinic can trot out unadjusted claims data that shows employers having lower PMPMs for DPC patients than for FFS patients. After risk adjustment, however, not so much.

*I recently came across an employer health benefit system that included both a DPC option and cost-sharing features that apparently mitigated selection bias somewhat. But note, in that program, employees who chose to retain relationships with PCPs not affiliated with the DPC clinic paid up to $1250 per year for that privilege. Layouts of that order seem likely to correlate with either profound health impairments or advanced age. I have learned that the non-DPC population at that employer is, on average, two years older than the DPC population. On a standard age-cost curve of ~4.6 to 1, every penny of the difference between the groups can be full accounted for.

Epiphany. Dr Gross’s risk adjustment metholodogy for direct primary care stands contrary to contemporary understandings of how to assess the relative expected costs of differing populations.

Dr. Lee Gross’s Epiphany Healthcare provides DPC services for some of the employees and some members of of some of their families at a hospital in Florida. Some hospital employees decline Epiphany; they and some members of their families receive instead traditional insurance based primary care. Unusually for such arrangements, a recent assessment of the dollar value of the claims experience for the DPC cohort exceeded that of the FFS component by about 15%.

There appears to have never been a case in which a DPC provider applied risk-based adjustment to validate raw claims cost differences that seemed to reflect well on DPC. When, on the other hand, raw data went the opposite way for him, Dr. Gross decided the time had come to became a data analytics pioneer and develop a risk adjustment model for direct primary care.

Conveniently, widely-used models developed by professional actuaries in and for CMS (in making patient population risk adjustments for use in Medicare Advantage plans and under the Affordable Care Act) and Dr Gross’s share a common feature – each adds up, in certain circumstances, the total number of a patient’s chronic conditions to develop an adjustment factor. This facilitates a comparison.

To assess the precise role of the common feature of the two models, respectively, I will summarize the more widely used CMS model and then explain some key differences between it and the Gross model.

In CMS’s method, depending on the case, up to four different components come into play. [FYI, the lettering system is mine and is intended to simplify the explanation.]

A. There is, of course, a significant demographics component scored to reflect the historical utlilization by persons, specified by ages, gender, and other factors. Basic demographics are at the core of risk adjustments used by CMS for the ACA; over 75% of ACA individual enrollees under 65 have no adjustment-worthy chronic conditions and are risk-adjusted on demographics alone.

B. There is a specific conditions component in which a patient is given a score for each condition she has off a list of over eighty broad health conditions, each of which has a specific score reflecting that condition’s historical correlation with the use of medical services – so much for asthma, so much for porphyria, CHF, etc.

C. There is also an “interactions” component in which the patients can receive additional scores for as many as she may have off a list for certain specific combinations of conditions from “B” above that are historically correlated with enhanced need for services when the those health conditions are combined.

D. a recently added number of chronic conditions component reflects considerations generally similar to those in Dr Gross’s index

(D)(1) there is a specific list of about two dozen chronic conditions of the eighty condition from “B” above

(D)(2) if a patient has FOUR or more of the conditions in the list in (D)(1), then patient is given an additional score that reflects how having that many of those conditions has historically been correlated with an enhanced need for services.

Component (D) scoring is non-linear: the adjustment for five conditions reflects historical correlation between costs and having five conditions; scoring is computed separately for four, five, six, etc. Eight conditions are not simply given twice the score given for four conditions.

Now let’s look under the hood of the Gross model to see how these factors play out in his risk adjustment model for direct primary care.

Gross’s model removes the demographic component.


Under Gross’s model the number of different specific conditions being scored is reduced from a hierarchy of over eighty condition categories down to a single data point: a “chronic condition”.

.
Gross’s model removes the interactions component.


Gross’s model counts all chronic conditions equally and assumes a linear relationship between health costs and the raw number of chronic conditions.

Gross’s model starts the count of multiple chronic conditions needed to trigger the multiple chronic conditions factor at one (1), while the CMS model believes that number of conditions requiring a complexity factor should start at four (4).

A theoretical advantage of Gross’s method is ease of application.

On the other side, Gross’s is untethered to any historical utilization data whatsoever. Within the one component where there is any measure of similarity between the Gross model and the CMS model, Gross’s approach expressly contradicts CMS’s actuaries’ explicit determinations that a linear correlation did not accord with historical reality and that the count should begin at four (4). (CMS actuaries did agree that there was an additive effect for each additional condition, but unlike Gross they denied that the additive effect was necessarily linear.)

On every other component Gross’s model is in even sharper contrast to the views of the professionals. Notwithstanding Dr Gross’s omission of demographics information, demographics strongly correlate with costs. As noted, over 75% of patients under 65 have their risk adjustments based only on demographics as they have no chronic conditions among the many dozens that the professionals felt needed to be includes. Notwithstanding Dr Gross’s model in which all chronic conditions count the same, long experience shows that the costs of different chronic conditions vary profoundly from one condition to another. Notwithstanding Dr Gross’s model, multiple conditions do interact with each other.

Gross’s risk adjustment model for direct primary care looks like something reverse engineered to explain adverse data.

As with all other aspects of his otherwise relentless promotion of the results at Desoto Memorial Hospital, Dr Gross has expressly declined to make public any details of his methods of gathering and processing data. Even if his risk methodology was not at war with those of his analytical betters, important questions would remain.

For example, what criteria did he use to identify and count chronic conditions in the DPC cohort (his own patients) and those in the non-DPC patient cohort (who get primary care elsewhere). Especially for gathering chronic condition data on those who are not his own patients, what if anything did he do to validate data accuracy and consistency?

For these reasons, I report only this: (See update below.)

The raw data shows that non-Epiphany patients at Epiphany’s largest employer clinic have lower claims costs.

Dr Gross has produced a slide show that includes, as far as I can tell, all the public information about the Epiphany experience at that employer.

The slide show utilizes the Gross risk adjustment methodology for direct primary care discussed above.

On Slide 22, Dr Gross made an error of double adjustment. Here is what appears.

Chronic Conditions846 per 1,000623 per 1,000
Relative Chronic Conditions1.350.74
https://dpcconference.com/wp-content/uploads/2020/01/DPC2019-Pres-Gross-sm.pdf

There is at least one other, similar error.

UPDATE: 12/4/2020

After more careful analysis, and having learned even more about risk adjustment, I am sure that Dr Gross’s approach is complete bullshit. And yet it still may be a bit more justifiable than the bullshit of the KPI Ninja/Nextera report; Gross, at least, is not falsely claiming the imprimatur of a prestigious academic research team.