Thoughts on the Robert E. Lee Statue Removal from Duke Chapel

I became chair of the Academic Council at Duke on July 1, 2017, and was chatting with the Provost a few weeks ago and we agreed there “weren’t any hot button issues” on tap for the Fall semester. That, of course, is no longer true.

I support President Price’s decision to remove the statue of Robert E. Lee from the Duke Chapel Entrance, as well as with his commitment to preserve the statue in a way that will allows students, faculty and all of us to learn from it. The creation of a commission of faculty, students, staff, trustees and members of the broader Durham community to help guide the next steps shows that this will be an ongoing process. The removal of the statue is best understood as the beginning of a new chapter, not an ending.

Duke is a relatively young University, and I think this new chapter may provide a book end of what I often think of as the beginning of modern Duke–the Bassett Affair of 1903 (the University was then called Trinity College). Professor Bassett wrote that Booker T. Washington and Robert E. Lee were the two greatest Southerners of the previous 100 years:

….he inserted a sentence praising the life of Booker T. Washington and ranking him second in comparison to Robert E. Lee of Southerners born in a hundred years. Such a sentiment invited controversy at a time when race baiting was commonplace due to the revival of bitter partisan politics with the division of the Democratic Party, the rise of the Populist third party, and the revival of the Republican Party. State Democratic leaders in nearby Raleigh who were also represented on the Trinity College Board of Trustees demanded that Professor Bassett be fired. When parents were urged to withdraw their children from the college and churchmen were encouraged not to recommend the college to prospective students, Bassett offered his resignation.

The Trinity College Board of Trustees did not accept Bassett’s resignation, a founding chapter in Duke’s history that made clear the critical principle of freedom of speech for the University. Of course, it was non-controversial to so-praise General Lee in 1903, while today his likeness is the source of controversy.

The common denominator then and now is the life of the mind that should be the heart of the University, demonstrated first and foremost by the faculty and students engaging in scholarship, teaching and learning. This is a big moment for Duke and our country, and I believe that the University has special process and educational responsibilities not only to the members of our community, but to society at large. We have a chance to model that difficult issues can be navigated truthfully, respectfully and openly, and if done well, we can help to make our world a better place.

May we be up to the challenge.

N.C. Medicaid Managed Care Proposal

The N.C. Medicaid Managed Care proposal was put out today by the Cooper Administration NCMedicaidManagedCare. Don’t look now, but North Carolina looks to be engaging in bipartisan health care reform. You may recall that the Republican-controlled Legislature dictated that the Republican Governor develop and submit a 1115 Medicaid waiver by June 1, 2016, which he did. I submitted a comment on their plan with Aaron McKethan–some good ideas, some not so good. However, it was mostly an aspirational document and was not really detailed enough to warrant a CMS response.

Governor Roy Cooper’s Secretary of HHS Mandy Cohen (and former COO of CMS under President Obama) was confirmed last Spring and immediately set out to lead a process that is putting the meat on the bones of the Republican Medicaid reform aspirations. She fully committed to moving it ahead, demonstrating that both political parties are searching for common ground on Medicare reform.

It will take both sides to do something this big.

Counterproductive cost sharing

The theory behind cost sharing is that it should deter ineffective care. In reality, it is a blunt and crude instrument that places barriers in front of effective care as well as ineffective care. I want to highlight two papers that look at the misaligned incentives of cost sharing in a pragmatic context.

The first examines IUD utilization. It found that the pre-ACA up-front cost-sharing deterred utilization:

Results: Overall, 5.5% of women initiated an IUD in 2011. After adjustment, IUD initiation was less likely among women with higher versus lower co-pays (adjusted risk ratio = 0.65; 95% CI, 0.64–0.67). Women who saw an obstetrician/gynecologist during 2011 were more likely to initiate an IUD (adjusted risk ratio = 2.49; 95% CI, 2.45–2.53).

IUD’s are fix it and forget it contraception. Once an IUD is inserted correctly, it is very difficult for it to fail. It gives significant female autonomy and over the long run of its effective use span, an IUD is less expensive and more reliable than oral hormonal contraception. The big problem pre-ACA with insurers discouraging IUD utilization is that there is a high initial up-front cost which means that a significant portion of the women would experience significant benefits from the IUD including the net cost savings compared to oral contraception after they had stopped paying premiums to the insurer that paid for the IUD. It is a churn problem.

There is another set of problems that come from inefficient cost sharing designs.  It is the non-adherance of effective and high value treatment problem  From one of the same authors, an examination of medication adherence for a long term cancer (Chronic Myeloid Leukemia (CML)) treatment regime that has impressive survival effects.

Over the study period, the mean copayment required for a 30-day supply of imatinib was $108 (SD, $301). Mean copayments varied widely, with patients in the lowest 25th percentile paying $17 and those in the upper 75th percentile paying $53. Importantly, the median copayment required across all years was approximately $30 per fill, but copayments ranged from $0 to $4,792. Over time, monthly copayments increased from an average of $55 in 2002 to $145 in 2010….
Among new users of imatinib, approximately 10% of patients with relatively lower copayment requirements and 17% of patients with higher copayments discontinued therapy during the first 180 days following treatment initiation (Table 2). There was a 70% increase in the risk of discontinuing TKIs among patients with higher copayment requirements (upper 75th percentile; aRR, 1.70; 95% CI, 1.30 to 2.22). Similarly, approximately 21% of patients with lower copayments were nonadherent to their TKI therapy ( 80% of days with drug available) versus 30% of patients with higher copayments. Patients with higher copayments were 42% more likely to be nonadherent to their TKI therapy (aRR, 1.42; 95% CI, 1.19 to 1.69)…..

Okay, there is a lot going on in that results section.  The important take-away is that cancer patients with higher co-payments are far more likely to be non-adherent.  This is financial toxicity, a concept advanced by one of my Duke Margolis colleagues, Dr. Yousuf Zafar.  The cost of treatment forces people away from getting the care that they need.

People don’t elect to have cancer.  It is not an episode that is immediately deferrable.  Once you’re told that you have cancer, treatment usually needs to get started quickly.  And the effective drugs are expensive.  At this point the idea of cost sharing as a means of shaping choices and avoiding preventable conditions goes out the window.  It is merely a burden sharing between the insurer and the individual.  It is just a financial shock with no positive incentive effect when an individual is hit with a dozen other life altering shocks at the same time.

If we are trying to design a system that minimizes non-adherence for financial reasons, we want to change the benefit design.   A theoretically attractive solution would be to offer no cost sharing on effective but expensive courses of treatment for diseases that are not immediately preventable. This would be a value based insurance design approach.  VBID encourages people to get effective care by making it comparatively less expensive than ineffective care or care for preventable episodes.

There is a problem with a VBID approach.  It is an ugly selection problem.  Let’s imagine that there are two types of plans offered.  The first is a standard plan design with high cost sharing for imatinib (Gleevac).  The other is a low or no cost-sharing design for imatinib.  Individuals who are healthy at the start of open enrollment period will choose the standard plan.  Individuals who have CML will go to the low cost sharing pool.  An individual who was healthy at open enrollment and then is diagnosed with CML during the first policy year will switch over.  This incentive structure makes the VBID pool extremely sick.

Sick pools are not show stoppers.  There are tools that can be used to compensate insurers with sick pools.  The most common is risk adjustment where money from either other insurers with healthy pools or the general government funds are sent to the insurers with disproportionate number of CML patients to cover the high and recurring costs.  And given that the VBID design is meant to encourage more people to take more of a very expensive drug, the costs will be higher than the baseline average CML patient cost.  Reinsurance tied to specific diagnosis could work as well.

The other option is to mandate benefit designs that are based on VBID principles for high cost diseases. States could do this on their individual and fully insured group markets but they can’t touch the ERISA regulated employer sponsored self-insured segment.  We would still have a serious churn problem as people with expensive chronic conditions would still be moving between ESI and Exchanges and given the differences in after treatment incomes, these people will be far stickier to their preferred option than healthy people.

Benefit design right now is complex, confusing and ineffective at promoting health in the CML case.  Any changes will also be complex and confusing but hopefully more effective at enabling better health.



** Pace, L. E., Dusetzina, S. B., Fendrick, A. M., Keating, N. L., & Dalton, V. K. (2013). The Impact of Out-of-Pocket Costs on the Use of Intrauterine Contraception Among Women With Employer-sponsored Insurance. Medical Care, 51(11), 959-963. doi:10.1097/mlr.0b013e3182a97b5d

*** Dusetzina, S. B., Winn, A. N., Abel, G. A., Huskamp, H. A., & Keaton, N. L. (2013). Cost Sharing and Adherence to Tyrosine Kinase Inhibitors for Patients With Chronic Myeloid Leukemia. Journal of Clinical Oncology. Retrieved August 4, 2017.

RCT of Palliative Care in Heart Failure

A big team from Duke lead by Joseph Rogers has a new Paper (PAL HF) reporting the results of a RCT of palliative care in late stage Congestive Heart Failure, published today in the Journal of the American College of Cardiology. Patients who got palliative care had better quality of life, higher function and reduced anxiety and depression (with no survival difference, as hypothesized). Costs were not a primary endpoint, but will be analyzed in later work.

CHF_7-11-2017 8-56-05 AM.

After the publication of a RCT of early palliative care in stage 4 lung cancer in the NEJM in August 2010 showed improved quality of life, lower costs and longer survival, there was a move to see if such findings could be replicated in other diseases. This NHLBI funded study is part of that effort.

Real world evidence is provided imperfectly, especially with respect to policy. One example from this paper. This studies’ exclusion criteria ruled out patients with anticipated heart transplant or LVAD implantation within 6 months. Arguably, these patients aren’t ready for palliative care because they are seeking such aggressive treatments, but from a policy perspective, it is arguable that they definitely need the goals of care type of discussions that are a key part of palliative care. The successful conclusion of a trial of CHF without patients on the way to heart transplant or LVAD implantation leaves an evidence hole, even as we add to the evidence base.

Should patients thinking about an LVAD or a heart transplant get palliative care? I think the answer is yes, but this trial provides no direct evidence to inform this question. So, there remains an evidence gap (as always–good studies suggest more questions than they answer). However, the health care system often doesn’t and can’t wait for such answers, and in the case of palliative care there has been a tremendous increase in its provision at Duke, and across the nation since this study was began. The process and timing of research is slow, and the answers are often muddled, while the pace of change in addressing the needs of seriously ill patients is much more rapid. When one’s interest is policy applications, even an RCT–the gold standard of research–cannot answer all of the most pressing questions.

The Senate bill confuses already confused consumers

The Senate’s Better Care Reconciliation Act (BCRA)[1] is a significant modification to the current Patient Protection and Affordable Care Act (PPACA)[2] exchange structure in a variety of ways.  One major change is the designation of the benchmark plan which determines the level of subsidy that the federal government provides to individual buyers on a health insurance marketplace.  There are two elements of note.  The first is that the benchmark plan is now a plan with a calculated actuarial value of 58%.  This is a significant change from the current benchmark plan with a calculated actuarial value target of 70%.  More importantly for this post is the benchmark plan in the BCRA is the median qualified plan.

Research has shown that consumers can be overwhelmed by too much choice[3].  Dominated choices can and often will be selected.[4]  Buyers have frequently confused the value proposition of Gold and Bronze plans based on prioritizing either more out of pocket maximum protections or lower monthly premiums. [5] Buying and using insurance is a complex task with significant uncertainty and cognitive demands.  The BCRA subsidy attachment system creates incentives for insurers to further increase complexity.

Within PPACA the incentive for insurers to clone plans with minimal meaningful differences between them only applies to the insurer which controls the least expensive Silver plan in a county.  This single, low cost, insurer faces a decision as to whether or not to design a second plan with the same actuarial value in a slightly different cost sharing structure in order to guarantee that this single carrier controls both the least expensive and the benchmark Silver point.  All other insurers plan offering decisions are made independent of subsidy attachment point manipulation purposes.

The BCRA creates a broader and more complex strategic design problem for carriers.  Every entry in the benchmark category influences the benchmark price.  Once an insurer has built a network and performed the basic actuarial calculations, modifying a basic plan design by altering co-insurance slightly or decreasing deductibles while increasing co-pays to achieve a constant actuarial value is a fairly low cost action.  A high cost carrier can introduce an isomorphic plan design to increase the benchmark and asymmetrically decrease the relative post subsidy price of its preferred offerings.  Low cost carriers have an incentive to offer one more plan to lower the benchmark and make its offerings more attractive compared to higher cost peers.

Counties with multiple insurers will face an unstable equilibrium.  Every insurer will have an incentive to add a marginal, incremental plan to be offered on the BCRA exchanges in order to move the subsidy attachment point closer to their preferred position.  Once this arms race begins, consumers will be asked to differentiate miniscule differences between dozens of plans offered by two or more insurers.  Common heuristics such as evaluating a plan first on the inclusion of a specific doctor or hospital in network and then examining a subset of plans based on maximally acceptable premium with the final decision step based on minimizing deductible will be corrupted.  Plans can be offered with low deductibles but higher maximum out of pocket exposure with a cost structure that significantly advantages or disadvantages certain types of consumers.  Insurance buyers will face decision fatigue and be overwhelmed with almost meaningless choice.

There are two possible solutions.  The first is for the Department of Health and Human Services to update stringent meaningful difference regulation.  Current regulation[6] allows carriers to offer multiple plans built on the same network ID and same plan type (HMO, POS, EPO, PPO) with fairly minor differences in cost sharing.  Stronger meaningful difference regulation would restrict carriers to offer only a single cost sharing design per network ID and plan type dyad.  The second is for the Senate to modify the benchmark.  Every participating carrier would place their lowest cost plan in the benchmark set and the median plan from the benchmark set would be the benchmark plan for the county.

These design modifications lower consumer confusion and will minimize strategic manipulations of the subsidy formula which will lead to more effective and efficient markets.

Read more of this post

Does Health Insurance Improve Health?

An old question has gotten some new evidence. Does health insurance coverage improve health? This is a simple (and important) question that is complicated to answer definitively due to methodological reasons, but the evidence base has grown by two important papers in peer review journals in the past few weeks:

  • Ben Sommers has a recently accepted paper in the American Journal of Health Economics that is available online as an eprint. He finds that Medicaid expansions in the early 2000s in 3 states saved lives, and he documents the costs of doing so. The point of the paper is better methods of determining if Medicaid expansions did in fact save lives and looking at the cost of coverage. His top level findings:
    • 1 life saved for every 239-316 persons newly covered by Medicaid (the range represents the uncertainty in the estimate–could be as high as 316 lives saved, or as low as 239, for each newly covered person).
    • The cost per life saved was $327,000-$867,000
    • He discusses the other things that could be done with this amount of money and notes these amounts are less than often paid per life saved for many interventions
  • Ben Sommers, Atul Gawande and Kate Baicker have a paper in the most recent NEJM that looks more comprehensively (not only mortality as an outcome, and at different types of insurance, public and private) and reviews the recent peer reviewed work on the broad topic of does health insurance improve health. This is a nuanced paper that focuses on the current health reform discussions that our nation is having and goes directly to how important losses in recent coverage expansions could be to public health. On the most basic question at hand they conclude

    There remain many unanswered questions
    about U.S. health insurance policy, including
    how to best structure coverage to maximize
    health and value and how much public spending
    we want to devote to subsidizing coverage for
    people who cannot afford it. But whether enrollees
    benefit from that coverage is not one of the
    unanswered questions. Insurance coverage increases
    access to care and improves a wide range
    of health outcomes. Arguing that health insurance
    coverage doesn’t improve health is simply
    inconsistent with the evidence.

The evidence base changes weekly. These two important papers need to make their way into the ongoing policy debates.

A cheat sheet to read the Senate version of AHCA

The Senate healthcare and tax cut bill is expected to drop soon. Here is a cheat sheet on how to read it.

1) Reconciliation places severe constraints on the bill

a) The Parliamentarian is most likely going to be stripping out significant, non-germane to the budget, items that were in the House bill.
b) $1 billion in savings must come from each of two committees (HELP and Finance).
c) Anything the Senate passes must meet or beat the $119 billion in budget window deficit reduction that the House AHCA was scored at.

2) Three major pots of money

a) Tax cuts
b) Individual market changes
c) Medicaid cuts to pay for tax cuts

3) Follow the money
Any extra dollar used to pay for a slower Medicaid termination has to come from either Medicaid on the back-end, fewer tax cuts or lower individual market changes. Anything used to up subsidies on the individual market has to come from itself, faster/steeper Medicaid cuts or fewer tax cuts. Anything that ups the tax cuts must come from the individual market or Medicaid…etc.

4) Index rates matter
Slower terminations but lower index rates on per capita caps is a budget gimmick. It gives a little bit of money in the 10 year budget window but leads to massive cuts in the out years against the current counterfactual.

5) Market design and incentives matter

a) Look at where the work disincentives apply

a1) Medicaid expansion where the FMAP disappears once a person churns out once
a2) Medicaid expansion to individual market transition without CSR as people move from high AV low premium insurance to low AV high premium insurance if they earn a dollar too much
a3) 350% FPL instead of 400% FPL

b) How does the individual market function without a mandate and without the patient and state stability funds?