So, does the medical home "save money?" A recent publication in Population Health Management about Community Care of North Carolina (CCNC) says "yes." That's important, because CCNC program has had more than its fair share of controversy.
You can read more about CCNC here. According to the Commonwealth Fund, Raleigh pays CCNC's 14 non-profit regional networks $3 per member per month (PMPM) for medical home services for over a million Medicaid and CHIP beneficiaries. In exchange, the 1300 clinics provide preventive care services, 24 hour coverage services, coordinating access to specialty services, care management and quality improvement. To do all that, CCNC uses a "medical home model" with "specialized chronic care programs" staffed by teams of docs, pharmacists and care managers.
The quasi-experimental evaluation published in PHM used "hierarchical modeling" to evaluate the impact of CCNC on two different samples of non-elderly (ages 0 to 64) disabled Medicaid patients who had no other insurance:
Model 1: compared the medical home patients' claims expense within and outside the enrollment periods "after controlling for other covariate values"
Model 2 created matched cohorts of enrolled and non-enrolled patients to compare pre-post differences in insurance claims expense. Matching was based on pre-enrollment pharmacy use, race, age, enrollment duration, clinical risk and behavioral health burdens. For every enrolled patient, ten non-enrolled comparison patients were selected.
The study period was January 1 2007 through Sept. 30, 2011. Any single months of disenrollment were "filled in" if there was enrollment 2 months per and 2 months post.
Model 1: This used insurance claims data for over 169,000 patients with an average age of 35 years. 52% were male with a 24% rate of mental illness and an 8% rate of chemical dependency. Compared to the time of not being enrolled in a program, claims expense was statistically significantly $190 per member per month (PMPM) cheaper in the first year; that declined to $64 PMPM cheaper in the last ear of study. Persons with a higher burden of illness had even greater savings.
Model 2: This studied claims from approximately 102,000 enrolled patients with pretty much the same baseline characteristics in Model 1. Savings achieved statistical significance in the 3rd, 4th and 5th years of study: $81, $73 and $121 PMPM, respectively.
The DMCB's take:
While it can get lost in the sublime minutiae of hierarchical modeling, the DMCB finds the methodology and the numbers to be credible. It has used the same Model 2 style of matching in its own studies. Since a pristinely conducted prospective randomized control clinical study is functionally impossible in a state-wide Medicaid program, quasi-experimental study designs like this are a good window into figuring out what happened.
And what happened is that they saved a lot of money. Assuming CCNC was paid $3 PMPM or $36 million per year for a about a million beneficiaries, avoided claims expense appeared to be well north of that.
While CCNC has a lot of moving pieces, the DMCB believes the key success factor was based on identifying the most vulnerable patients and then using nurses to intervene on the them.
The average caseload per nurse ranges from 150 to 200 patients. As the Commonwealth Fund summary describes.....
"Case managers... work with primary care providers (“medical homes”) to identify patients who will benefit most from targeted care management interventions, such as patients making repeated ER visits; patients diagnosed with asthma, diabetes, or heart failure; and patients who have two or more chronic conditions (including mental health conditions) with high service use or activity limitations indicating complex care needs. Care managers identify high-risk patients through the CMIS and from case-identification lists provided by the CCNC central office, notifications of admissions provided by hospitals, and physician referrals."
CCNC is to be congratulated for moving from opaque actuarial studies to the harsh glare of peer-reviewed publications. While some critics may pounce on some of the weaknesses inherent in any retrospective analysis of subpopulations, the observations from two "Model 1 and Model 2" vantage points are sufficiently positive to believe that North Carolina's taxpayers got their money's worth.
The DMCB would point out two caveats:
The disabled Medicaid beneficiary population is a notoriously high utilization group that is a classic example of the return on investment from "low hanging fruit." A little coordination goes a long way in a population with a baseline of high utilization. The same approach may not work in other populations with different patterns of claims expense.
Unfortunately, this gives us little insight about the potential impact of a similar medical home model in commercially insured populations or among Medicare beneficiaries. That's doubly true for fee-for-service beneficiaries who are outside of any managed care networks.