 |
| Click image to enlarge. Source above |
Today’s NY Times has an article focused on high risk patients – who in the under-65 population represent about 1% of patients, and 25% of all medical costs – over $100,000 per year. Full Report Here (Registration Required)
The concentration of the cost of care is an important observation – the bottom 50% of Americans represent about 3% of total medical costs – so interventions to lower the cost of care for these least expensive health plan members are bound to be unsuccessful. This is also why programs to convince Americans to go see the doctor raise costs. (Good Op-Ed on this topic from Dartmouth University today, too).
The concentration of the cost of care is an important observation – the bottom 50% of Americans represent about 3% of total medical costs – so interventions to lower the cost of care for these least expensive health plan members are bound to be unsuccessful. This is also why programs to convince Americans to go see the doctor raise costs. (Good Op-Ed on this topic from Dartmouth University today, too).
The robust and publicly-available Medicare claims database has been studied extensively – but many of the conclusions from the over 65 population are not applicable to the younger cohort insured through employer-based plans
- Medicare beneficiaries are highly likely to be readmitted to the hospital (almost 20% in 30 days). Readmission rates are much lower in the non-Medicare population (5-8%) , and many of these readmissions are planned, like followup inpatient chemotherapy.
- Medicare spends a quarter of its dollars on patients in the last 6 months of life. Commercial plans actually spend a very small portion of their claims on end of life care, since death is much less common in this population
- Medicare beneficiaries stay with that insurance plan for the remainder of their lives, while commercial health plans “churn” membership at a rate of 15% or more a year
Care management interventions should not assume that the commercial population is akin to the Medicare populations.
Health plans have hired platoons of nurses to do outbound calls to high-risk beneficiaries– and they charge employers large fees to perform this work. However, most of the evidence of efficacy of this intervention is highly anecdotal. From Reed Abelson’s article:
When Wendy Meath, a 59-year-old with diabetes, was hospitalized a year ago, she was identified by HealthPartners as someone who needed help to control her disease. She had been admitted for kidney stones, one of many possible complications of diabetes. Although she had insurance through her husband, she was unemployed.
Since leaving the hospital, where she was admitted for 12 days for a series of complications from the surgery to remove the stones, Ms. Meath has been in regular contact with one of HealthPartners’ nurses, who serves as a case manager. The nurse calls at least once a month and checks in after she goes to the doctor for any developments. The health plan also assigned a social worker to help her with the cost of medications and other obstacles that were preventing her from taking better care of herself. “It makes me feel like I’m not alone,” Ms. Meath said.
“They’re trying to prevent the big things from happening, which is great,” she said.
But the iron-clad evidence of the effectiveness of this intervention is still lacking.
Health plans tend to cite compelling anecdotes – but here’s what we should look at to assess efficacy of these programs
Structure
a. What intervention is in place? Are the elements evidence-based? Is there a measurement plan?
Process
a. How many of the targeted high risk members actually participate? Many health plan programs have half or more targeted members refuse to participate. That sharply limits potential effectiveness
Outcome
a. Are there changes in clinical course due to the intervention?
b. Are there beneficial financial outcomes as a result?
There are two major problems with the outcome evaluations I’ve seen – most of which have focused on only program participants.
The first is regression to the mean. The likelihood that Wendy Meath will be hospitalized again over the next year is low. This doesn’t mean that the intervention has been successful – it’s what you’d expect in the year following such an admission.
The second problem is selection bias. The half of patients who eagerly participate in the intervention are fundamentally different than those who refuse. Even efforts to do “propensity matching,” to try to adjust for known differences between participants and nonparticipants, aren’t effective for adjusting for these differences.
We should only attribute cost savings to these programs if there are cost savings over the entire population. Further, we should only find claims of success believable if enough members were engaged to credibly lead to the claimed cost savings.
I believe that high risk programs performed in the provider realm are likely to get more patient engagement – and could lead to more success in preventing bad outcomes and lowering costs. Health plan leverage in the future might have more to do with payment reform than with hiring legions of remote nurses.
Of course, we will have to measure that too!
