This week, a RAND research team published a report in the Journal of the American Medical Association (JAMA) on a three-year evaluation of Patient-Centered Medical Homes in Pennsylvania, funded by the Commonwealth Fund and Aetna. Since, like almost everyone else, we believe in the potential benefits of enhanced primary care, this is an important paper.
However, its implications are very complicated. I believe the results confirm concerns I had from the beginning that this kind of project wouldn’t work. Why? Because it wasn’t focused enough on the complexly ill and it didn’t incorporate enough special expertise in their care.
The evaluation was led by Mark Friedberg, MD, and used a pre-post, matched practice/patient design where 32 practices worked to become medical homes. The outcomes of the practices and the patients were tracked over three years. A non-random comparison group was created by looking at other practices and patients followed over the same time period to serve as a contrast group not implementing medical home elements.
One piece of good news is that the intervention “pilot” practices were fairly successful in putting in place the “structural” elements of the NCQA (National Committee for Quality Assurance—one of our new grantees) framework for Patient Centered Medical Homes (PCMH). By the end of the three years, 13 achieved level 1 status (lowest level), three achieved level 2 status, and 16 (50 percent) achieved level 3 status (most comprehensive change).
This transformation is not to be undervalued, as many have described PCMH status as too ambitious for most practices. All of these practices achieved their changes with only $20,000 in upfront capital payments and bonus payments tied to stage of change that averaged a modest $32,000 per provider per year.
This means, for example, that practices set up computerized registries of patients and their health data that would allow them to search for and identify patients in the practice with a chronic disease, such as diabetes, who either didn’t have up-to-date testing or whose tests were out of target range (from around 30 percent at baseline to around 85 percent of practices by the end of year three).
Other good structural changes that we think can help people were also put in place. Practices increased their referral of people to community resources, such as for physical activity or diet change, from around 55 percent of practices to 77 percent. They also increased regularly assessing patients’ self-management needs around chronic illness, from around 11 percent to 57 percent.
Unfortunately, because only intervention practices were asked about how they changed their capabilities, we don’t know if the comparison practices might also have made some of these changes. I think it is also unfortunate that these changes were only measured by asking practice leaders rather than by asking patients or observing practices directly. In my experience, there is a lot of room for divergent views on these kinds of issues.
And even worse, the study found little or nothing in terms of quality improvement in processes of care (e.g., rates of Hemoglobin A1c testing among diabetics or screening rates for colorectal cancer) and no change in utilization of services like emergency room visits or hospitalizations, despite a lot of digging and clever analyses.
So what should we conclude? Melinda Abrams, senior program officer at Commonwealth, counsels patience and writes “the study’s findings underscore what we already knew about this team-based model of primary care: we need to continue to improve how care is delivered, how providers are paid, and how the model is implemented in different settings.”
To my mind, I think there are two possible interpretations of the results that merit some real discussion: First, it may be that structural change for enhanced primary care is necessary, but not sufficient, to achieve different outcomes—the content of care within the structure matters. And second, not withstanding my recent argument against too much faith in targeting, the population one is focused upon matters a great deal as to the kind of results one is likely to get—you can’t get blood from a stone and you can’t get cost reduction where costs/utilization are already low (Melinda highlights this point as well).
So first, we do know that interventions such as IMPACT, our depression treatment in primary care model, and Care Management +, our IT enabled care management model, both WORK. IMPACT and CM+ use much of the same kinds of PCMH structural capacities, such as registries, but also carefully specifiy what one does with those capacities (e.g., how you support patients to adhere to their medications). And they produce improved quality of care, patient function, and lower costs.
We also know that comprehensive programs such as GRACE Team Care or the Mass-General Care Management Program demonstration model also produce better outcomes and lower costs when they incorporate both these kinds of structural capacities and specific content protocols and geriatrics expertise.
Which leads to my second interpretation: It really matters what the patient population needs and how much opportunity there is to help them improve their health. Patients who are too sick and patients who are too well make it hard to show an impact of even the best-designed care. While the paper makes it hard to tell exactly, I don’t see any effort to really focus attention on the “just right” population.
The patients participating are described as either in commercial insurance plans or in Medicaid plans and their median age is 44 (with the 75th percentile at about 57 years old), so I don’t know if there were any Medicare beneficiaries—the age group with the highest levels of chronic illness–in this demonstration at all. Their total costs of care were about $4,800 per year, less than half of the annual cost for Medicare beneficiaries. Fewer than 2 percent of the patients had two or more hospitalizations and only about 12 percent of the population had two or more emergency department visits in a year. This is not a “frequent flying” population of high utilizers such as targeted by CM+ or GRACE.
So I think there are two arguments worth considering in future primary care redesign efforts. First, the content of care matters. Just having a registry doesn’t mean that patients who are found to be having trouble taking care of themselves get an evidence-based care management practice such as motivational interviewing or the Stanford Chronic Disease Self-Management Program.
There is a lot of missing expertise that needs to be incorporated in enhanced practice that does not come from structure alone.
Second, practices that don’t focus their new efforts where they can do some real good are going to spend a great deal of effort on patients who never had a reducible risk of running into trouble. And they probably will never be able to put enough well-planned systematic effort into patients with the potential to benefit.
When we have results that take content of care and patient population into consideration, we will have a far clearer picture of the true benefits of Patient-Centered Medical Homes.