In the first article in this three-part series I reviewed the findings in CMS’s latest report on one of its “medical home” experiments – the second-year evaluation of the Multi-Payer Advanced Primary Care (MAPCP) Demonstration.We saw that the “patient-centered medical homes” (PCMHs) in that demo have failed to cut costs or improve quality during the first two years of the demo. We also saw that the sloppy definition of “medical home” put the author of the report, RTI International, in a bind: They did not identify a single feature of PCMHs to treat as an independent variable, and were forced to offer an impressionistic, on-the-one-hand-on-the-other-
In the second article of this series I examined the report’s explanation for why the PCMHs have so far been unable to outperform non-PCMHs despite receiving substantial subsidies from CMS and other sources that non-PCMHs don’t get. The report seems to say that insufficient subsidies explains the PCMHs’ failure. I noted, however, that the report contains no evidence indicating how much more money PCMHs might need.
At the end of my second article I characterized the problem presented by the report as a conundrum. On the one hand, PCMH staff and many other observers feel PCMHs are severely underfunded, but on the other hand no one can say by how much or which PCMH services need more money.
So what do we do? Do we just pick a number out of thin air and say that’s how much more money PCMHs need, and pour that money down the PCMH black hole along with the other subsidies PCMHs receive now? That appears to be CMS’s position judging from its endorsement of yet another “medical home” program (CPC+ as an “alternative payment model” in its proposed MACRA rule despite the fact that all three of CMS’s “home” demos are failing.
Salvaging what we can
Throwing more money down the PCMH black hole is not a good idea. I recommend that CMS allow PCMHs to focus, and that CMS do so by radically sharpening and cutting down the definition of “PCMH” so that the concept refers to a uniform set of medical and social services provided to a subset of the chronically ill. 
Once CMS has clearly defined what services it wants “homes” to provide, it can then determine what the extra services cost and make adequate payment for them. It would help if, in addition to paying adequately for the extra services, CMS would let doctors and patients decide when the extra services should be provided rather than stick its nose into the doctor-patient relationship with pay-for-box-clicking schemes. Paying adequately for additional services and eliminating pay-for-clicks schemes would increase the physician “flexibility” that CMS claims it seeks to promote with PCMHs. Eliminating pay-for-clicks schemes would also lower physician overhead and reduce physician burn-out.
Please note the beautiful simplicity of my solution. By treating the problem first as a lack-of-focus problem, we solve the underfunding problem. Moreover, because doctors would now be paid for specific services rather than a fee per attributed patient, we would solve the sloppy attribution problem (see my comment on that problem here . CMS would not have to guess which patients “belong” to which doctors. Doctors would decide which patients they are treating, not CMS.
Compelling evidence CMS’s “homes” aren’t focused
Common sense and a growing body of research supports my insufficient-focus diagnosis and my recommended solution – more focus. The most compelling evidence of insufficient focus by the MAPCP PCMHs is their inability to lower costs or improve quality for what RTI calls “special populations.”
That is the oddest outcome of the MAPCP demo to date. Improving the quality and lowering the cost of care for the chronically ill is the primary goal of the MAPCP demo. As CMS puts it on its MAPCP website , “The care management fee is intended to cover care coordination, improved access, patient education and other services to support chronically ill patients.” (My italics) And yet, according to the second-year evaluation, “MAPCP Demonstration states did not develop unique interventions tailored to special populations….” (p. 2-46) 
It should come as no surprise, therefore, that, with the exception of one state, PCMHs were unable to lower costs or improve quality for those populations compared with non-PCMHs (see Tables 2-12 and 2-14 of the Year Two evaluation). The limited exception was Vermont. In that state PCMHs lowered spending on all five “special population” groups but, oddly, were unable to improve quality for them. In fact, Vermont PCMHs drove up ER visits and hospital admissions for patients with multiple chronic illnesses (see Table 2-15 pp. 2-50/2-51).
On the other hand, a growing body of research supports the conclusion that extra services targeted at the chronically ill improves health and, once in a blue moon, lowers total spending even after taking into account the cost of the extra services. Examples of such research include research on programs that monitor heart failure patients , find houses for the homeless , and deliver primary and specialist services to very sick children
Drastically shrinking the scope of “homes” and defining their services more concretely and uniformly would not only ensure better use of the unknown quantities of money going into “homes,” but would discourage the evidence-free conversation we’ve been having about them and encourage a more evidence-based conversation. Both outcomes would be achievable because analysts could finally publish honest-to-God research – research that examines concrete independent variables rather than warm and fuzzy phrases like “whole-person focus.” With independent variables specified, we could eventually determine why some “homes” fail and some succeed.
Vermont’s success might be an illusion
RTI’s inability to explain why Vermont was the only state in which PCMHs achieved gross savings compared with non-PCMHs illustrates the problem. Why was Vermont the only state in which that happened? Why did every major category of medical spending by Vermont PCMHs drop? Why were Vermont PCMHs unable to improve quality if they were so good at cutting costs? Because of the flabby definition of “PCMH,” no one knows. We can only offer hypotheses. Here are mine.
According to Table 2-9 of the Year Two report, Vermont PCMHs cut every one of the 11 categories of expenditures, with two minor exceptions: “Home health” and “other facilities” (spending in these categories rose a small, non-significant amount). All other categories, including primary care, fell. Something is fishy here. As Al Lewis observed in a comment on THCB about another report which claimed across-the-board savings for Vermont PCMHs, “Every component of cost can’t decline – something has to increase in order to save money elsewhere.”
Three explanations come to mind:
(1) The Medicare beneficiaries attributed to Vermont PCMHs were healthier than the beneficiaries assigned to non-PCMHs;
(2) “leakage” was greater among Vermont non-PCMHs than PCMHs, which artificially raised non-PCMH costs; and
(3) PCMH patients received services that were paid for “off the books,” and those services either substituted directly for medical services or improved the health of PCMH patients enough that they needed fewer medical services.
I’ll explore each hypothesis briefly.
Healthier patients. RTI states that patients assigned to Vermont “homes” were “three percent less costly” than the “average Medicare FFS beneficiary.” (p. 5-7) This may be a consequence of the convoluted plurality-of-claims method RTI used to “attribute” patients to PCMH and non-PCMH clinics.  That method assigns more “loyal” patients to PCMHs, and loyalty may be correlated with health. RTI also informs us that “Medicare expenditures are not risk adjusted” but instead are subjected to “entropy-balanced weighting.” (p. 1-19) The combination of healthier patients and failure to adjust accurately for health differences might explain the apparent ability of Vermont PCMHs to lower CMS’s gross (claims only) costs.
More “leakage.” RTI’s plurality-of-claims method might have resulted in more “leakage” for non-PCMHs.  By giving PCMH’s Vermont’s least transient and most loyal patients and, conversely, handing the non-PCMHs a pool of less loyal and more transient patients, RTI might have guaranteed that the non-PCMH attributees (yikes, what a word!) would seek more care outside the clinic to which they were attributed. For example, if non-PCMHs were assigned more snowbirds who sought treatment in the Sunbelt, non-PCMHs would have been dinged for those treatment costs, but there would have been nothing the non-PCMHs could have done to control those costs.
Enron book-keeping. Finally, as I have noted already, RTI has not reported CMS’s net savings in Vermont – savings as measured by claims minus the “care coordination fees” CMS paid to PCMHs. Those savings might well be negative. But CMS’s payments are not the only subsidies that must be subtracted from the savings Vermont PCMHs allegedly achieved. PCMHs in Vermont, as in the other states, received numerous other payments and in-kind contributions which RTI did not measure. Non-PCMHs received none of those subsidies, measured or unmeasured. It seems likely that some of the subsidies PCMHs received funded services which either took the place of or reduced the need for medical services, thereby allowing expenditures on traditional medical services to drop. It is possible more money was spent on Vermont PCMH patients than non-PCMH patients when all subsidies are included.
The most important fact about Vermont’s apparent success is that neither RTI nor anyone else can explain it because the variables are so numerous and so poorly defined. RTI did not even try. I understand.
More focus, less mumbo jumbo
In this three-part series, I have discussed three defects in RTI’s report on Year Two of the MAPCP demo:
(1) RTI’s inability to say anything concrete about what “homes” are doing that run-of-the-mill clinics don’t do and, conversely, RTI’s heavy reliance on impressionistic commentary;
(2) RTI’s inability to describe the independent variables they plan to investigate, and, most importantly,
(3) RTI’s inability to explain the failure of MAPCP homes to work as advertised and, in Vermont’s case, to explain why PCMHs in that state appear to be successful in lowering gross costs but unsuccessful in improving quality.
All these problems can be traced back to the condescending mumbo jumbo that PCMH advocates use to describe PCMHs – “patient-centered,” “whole-person focus,” “meaningful use,” “coordination across the medical neighborhood,” etc. An entity defined by such happy talk cannot possibly focus, and researchers who investigate such an entity can’t possibly define concrete, measurable factors that might explain the entity’s successes and failures.
“Medical home” proponents claim they are motivated by a desire to pour more resources into primary care and expand the primary care work force. The “home” fad is doing the reverse. It is sucking resources out of the primary care sector and contributing to burn-out among primary care doctors, which is already at epidemic levels. It should be clear now even to CMS, the American Academy of Family Physicians, and others who have exhibited such blind faith in “homes” that improving primary care requires pouring more resources into primary care, not “homes.”
Ideally, CMS and the insurance industry will do that by raising reimbursement rates for primary care physicians. That seems unlikely to happen in the near future. The next best solution would be to redefine the “medical home” so that it adds to rather than depletes resources available to primary care doctors. The way to do that is to define the “medical home” as a clinic that provides specific, uniformly defined services for a subset of the sick, and to pay that clinic adequately for those services.
 An example of such a service set would be all those offered by the Diabetes Prevention Program http://care.diabetesjournals.
 RTI reports two “exceptions” to their statement that PCMH programs did not focus on “special populations.” New York focused on rural residents by default (the PCMHs were all in rural areas of the state); Vermont’s Support and Services at Home program, which began in 2011, served 3,000 elderly people in subsidized housing by the end of 2013.
 The plurality-of-claims method RTI chose required RTI to examine CMS claims filed over a “look back period” (one to two years, depending on the state) and then assign patients to PCMHs and control clinics for the “performance” period based on which clinic filed the most claims for that patient in the look back period. For a discussion of RTI’s attribution method for each state, see Appendix B https://downloads.cms.gov/
 Attributing patients was apparently a messy process in all eight states. According to RTI, “All eight states reported challenges with health information technology … and the quality and timeliness of data, which was associated with challenges with patient attribution.” (p. 2-7)