Open enrollment thoughts

Open enrollment season is coming up really soon. I just got the e-mail from Duke HR that outlined the choices I’ll need to wrestle with. Medicare’s open enrollment starts on October 15th. Covered California opens up their enrollment period on the 15th as well for the individual market. will run their open enrollment starting on November 1st. January 1 is the most common start date for employer sponsored insurance so corporate open enrollments are going into high gear now.

There are a lot of questions that someone going into open enrollment needs to ask as they make their decisions:

  • What are the choices being offered?
  • Am I or my family member(s)  guaranteed to run up a certain level of costs or am I only worried about getting hit by a surprise?
  • How important is our current set of doctors/hospitals?
  • If something bad happens do I have a preference for certain docs and places?
  • How much is the trade-off between monthly premiums and out of pocket costs

Duke has a very small, curated set of choices.  There are four health insurance choices with three networks and fairly low variation in benefit configurations.  There are three dental plans and a vision plan.  This is an easy set of choices.  I am fairly certain that I am going to be perplexed at some point though.

Some counties on the ACA have a single insurer offering sixteen different Silver plans in 2018.  Other counties have over one hundred plans from half a dozen insurers.  Medicare Advantage has dozens of options in some counties from multiple insurers.  Medicare Part D drug plans are frequently numerous.

Choosing insurance is hard.  The best advice that I can give to you is to ask for help and take your time.  Lay out your objectives and constraints.  Lay out what trade-offs you are comfortable making and then knock out the easy failures and then think about the plans that are meeting your minimal standards.  When I buy insurance, I acknowledge to myself that I probably won’t choose an optimal choice but I should be able to choose a satisficing choice.

Picking an insurance plan is tough so accept that fact and don’t be scared of that fact.

Cash for CART

CAR-T therapies are a type of innovative cancer therapy that tweaks a patient’s immune system cells to attack cancer.  The National Cancer Institute explains:

A type of treatment in which a patient’s T cells (a type of immune system cell) are changed in the laboratory so they will attack cancer cells. T cells are taken from a patient’s blood. Then the gene for a special receptor that binds to a certain protein on the patient’s cancer cells is added in the laboratory. The special receptor is called a chimeric antigen receptor (CAR). Large numbers of the CAR T cells are grown in the laboratory and given to the patient by infusion.

This treatment regime has good clinical results and the attack path is expanding to more diseases.  However it is (to use a technical term) wicked expensive.  Novartis charges a list price of $475,000 for their CAR-T regime.  This is one more example of something that we’ve looked at with Hep-C last March.   Hep-C cures generate two true statements with a significant tension.

1) They are really freaking expensive on both a per-patient basis and total spending basis
2) They are really effective and thus high value

JAMA Pediatrics just published a cost-effectiveness study on CAR-T for a particular type of childhood cancer .**  This paper raises the same points as the Hep-C cures:

In this decision-analytic modeling study using deidentified data, cost-effectiveness analysis generated an incremental cost-effectiveness ratio between $37 000 and $78 000 per quality-adjusted life-year gained over a patient lifetime horizon, with more than 40% of those initiating tisagenlecleucel treatment becoming long-term survivors.

Currently, the next best alternative to CAR-T treatment has a survival rate of 5% to 10%.  So there are huge survival gains.  Secondarily, those survival gains are cost effective gains.  Most health policy analysts in the United States assume that a quality adjusted life year price of under $100,00 is a reasonable deal.  Very few analysts will argue that a treatment with an upper bound price per QALY well below $100,000 is unreasonably priced.

A policy problem is that these are huge cash outlays for an insurer that has to assume that a patient won’t be covered by them for the rest of their life.  The insurer will pay for the treatment and won’t get any of the gains of the new, high cash outlay treatment.  If there is perfect risk adjustment with a technological innovation plus-up, this could remove some incentives for insurers to either not cover CAR-T facilities at in-network rates or to try to drown a patient and their family in paperwork and pre-authorization purgatory.

We need to figure out ways to pay for treatments that are both incredibly expensive and incredibly cost-effective within the insurance model that we are committed to.



** Whittington M et al, “Long-term Survival and Value of Chimeric Antigen Receptor T-Cell Therapy for Pediatric Patients With Relapsed or Refractory Leukemia”, JAMA Pediatrics, October 8, 2018


Non-inferiority, creaky knees and cost trade-offs

In 2002, I tore the meniscus in my knee while having a damn good time.

My orthopedist wanted to aggressively diagnosis and treat the injury with an MRI and then surgery but my out of pocket costs for that sequence probably would have been six months of my then limited as a grad student income. That was not going to happen. Instead, he gave me a PT regime that I could do at home and told me that my knee would probably be off for a year but eventually everything would be mostly all right.

And he was right and a recent study in JAMA shows that after a year, PT and surgery have about the same outcomes. Physical Therapy is also a whole lot cheaper than surgery.


If we are to move towards a system that prioritizes the combination of evidence based care and cost effectiveness, the number of people who get the expensive and non-superior treatment has to fall. In this case, that would mean far fewer meniscus repair surgeries when the injury is straightforward and the patient has no confounding complicating risk factors. And yes, the evidence shows that surgery has a faster return to low pain status than physical therapy but that is a matter of months and the level of pain and dysfunction at a year out are about the same.

From an insurance design perspective, benefits need to evolve. Try to fail which is a common benefit structure in the pharmacy benefit could expand to more physical health interventions where the non-inferior and cheaper alternative is mandated first and then the more expensive alternative is authorized only when the first attempt failed.

An alternative payment mechanism could also be set up where a simple knee meniscus tear triggers an episode based payment that is heavily weighed towards physical therapy. The surgeon would have a strong financial incentive to refer a patient with a simple injury to physical therapy.

Another alternative would be reference pricing where the bundle for meniscus repair is again heavily weighed towards physical therapy but if someone wants immediate knee surgery, they can pay most if not all of the incremental difference in costs.

If we actually want an evidence based medical system where cost is a major decision shaper, then we will need to change incentives to encourage treatment pathways that are less expensive and non-inferior.

Overpriced in 2018, underpriced 2014-2016

The Kaiser Family Foundation has another issue brief on the financial performance of individual market insurers through the end of the first half of 2018. The short story is that insurers are Scrooge McDucking it right now in the individual market as they massively overpriced 2018.

This has been my core analysis since March; insurers got frightened and believed (with little state regulator pushback) that the ACA individual market morbidity would have fallen off the cliff. So far it has gotten slightly worse (4% more inpatient days) but it did not crash through the bottom. So what would “accurate” or “fair” pricing looked like?

I’ve grabbed their Table 3 data for further analysis. This is a quick and dirty analysis with algebra as my only analytical tool. This is limited in that we should expect claims payable in the 1st half of the year to be lower than claims payable in the second half due at least out of pocket exhaustion. But this gets us in the general area.

All I am doing is taking the paid claims and multiplying that by 1.25 to get to a premium level which supports precisely an 80% raw MLR. I then subtract the average premium collected minus the implied premium level. Ideally at the end of the year, the difference should be zero or slightly negative (implying a higher actual MLR than the bare minimum).   As the rest of the year continues to pay out, we will see the implied premiums get larger as claims will increase at the end of the year.

In 2011-2013, the difference at the mid-year is slightly beneath zero.  2014-2016 the premiums were too low while in 2017 (10% to 16% too low).  In 2017 the actual premiums were 3% above the implied premiums.  In 2018, the actual premiums are about 13% above the implied premiums.

These numbers will change in the 3rd and 4th quarter.  But the basic trend is the key insight.  2017 was probably priced close to right where well run insurers could be profitable especially if they could squeeze their admin costs while 2018 has seen significant overpricing of premiums.

Deductibles and distribution

Deductibles are part of the cost sharing structure for plans. A deductible is what a person pays first before they get any additional help from the insurer on their claims. Copays and coinsurance make up the rest of the out of pocket spending portfolio. The design of the cost sharing attributes has significant distributional impacts when we hold everything else equal.

We’re going to look at two plans both with 88% Actuarial Value as calculated by the 2019 CMS ACA AV calculator. We’re going to assume that these two plans have the same network and are priced solely on the actuarial value. That is a simplifying assumption that waives some incentive effects but it is a good first pass.

Plan A: $500 deductible, 20% coinsurance for a maximum out of pocket spending cap of $1,750.
Plan B: $1,200 deductible and no other out of pocket expenses for a maximum out of pocket limit of $1,200.

We’re also going to assume that there are a few classes of people:

  • People who have no cost sharing service utilization in a year (~20%)
  • People who have very light use (for example 1 urgent care visit and 1 generic antibiotic prescription) (30%-40%)
  • People with light to modest contact with the healthcare system ( non-complicated broken ankle treated with a cast and PT, well controlled asthma, well controlled high blood pressure etc) ( 30%)
  • People with heavy contact (knee replacement surgery, Hep-C cures, heart attacks, epilepsy drugs etc)

Folks who have either no cost sharing services in a year or who barely touch the system don’t run up enough charges to exhaust their deductibles under either plan.  Since their premiums are the same, they are indifferent and face no changes in incentives or total personal expenditures.

Now let’s assume that someone has a straightforward broken ankle that runs up $2,500 in charges.  The plan design matters to her:

Plan A: $500 deductible + 20% * $2,000 = $900 out of pocket expenses

Plan B: $1,200 deductible and $1,200 in out of pocket expenses.

A low deductible plan with a comparatively higher out of pocket maximum is in this person’s best interest.

Let’s look at the last case when someone has a $50,000 charge because of a failed assassination attempt by their cat led to significant shoulder reconstruction and a new knee.

Plan A: $500 Deductible + 20%*6,250 +0*43,250 = $1,750 out of pocket expenses

Plan B: $1,200 Deductible = $1,200 out of pocket of pocket expenses.

People who have a strong reason to suspect that they will have very high cost years will have a strong preference (holding actuarial value constant of course) towards plans where deductibles are the overwhelming form of cost sharing.

The trend towards higher deductibles in the employer sponsored market  is a distributional and incentive shift rather than a reduction in average actuarial value.  It is shifting more costs due to the benefit design choices onto people who have modest to moderate medical problems than they otherwise would have if the cost sharing design choices had been frozen in amber in 2006.




Example plan designs are here:

I am speculating here, but I have an inkling that this risk shift towards a broader cohort of people is what drives a lot of the frsutration over health insurance over the past ten years as actuarial value has been constant but the combination of growth in medical prices AND more cost sharing in deductible is inflicting more modest to severe pain pokes to more people. This is just an inkling without data to fully support that thought.

Mining risk adjustment for profitability

Earlier this week, I noted that the ACA risk adjustment program is highly likely to overpay for Hep-C anti-viral prescriptions this year and next year due to list price reduction shocks.  The price that insurers pay will be significantly below the prices used to determine risk-adjustment co-efficients.

We had briefly talked about a great NBER paper by Geruso, Layton and Prinz in June 2017 about how risk adjustment and reinsurance mostly worked on the ACA to remove cherry picking incentives.  My thought when I read the paper was that this was a profit-making opportunity for positive risk screening for certain small groups of drugs:

I had my cynical bastard insurance company plumber hat on as I looked at that graph. There are several drugs in that lead to a total cost of treatment between $45,000 and $50,000 but bring in between $55,000 and $65,000 in revenue. They are the massive beneficiaries of risk adjustment.

My big operational question is whether or not insurers were actively designing formularies to attract these patients?

I want to repost a slightly modified version of the graph that I referenced in order to talk a bit more about this idea of attracting profitable risk:

Every circle near the gray line is risk adjustment and reinsurance working close enough to right; total net incremental revenue is close enough to total net incremental costs.  The dots that are noticably above the gray lines are risk adjustment and reinsurance problems where total net incremental revenue is significantly below total net incremental costs.  The dots below the gray line and near the green arrow are profitable drug classes as the net revenue (premiums plus reinsurance plus risk adjustment) is significantly more than the incremental costs of care.

I think that the ACA risk adjustment formula with the currently lagged data is moving Hep-C cures into this group for a couple of years.  Insurers should have strong incentives to make it very easy for anyone that they cover in the ACA individual market to get easy access to the Hep-C cures as it will be very profitable for the insurers to do so.  I think one of the first tests that we should expect to see will be if there is a drop in tiering and pre-authorization requirements for at least one of the drugs in the Hep-C anti-virals therapeautic class.      If that is the case, then we can probably infer that an insurer is more willing to pay-out on Hep-C risk.


Preparing for 2021

New Mexico is preparing for an ambitious future for health policy. Louise Norris notes that New Mexico is looking to move off of and open up their own state based exchange.

In order to reduce user fees, the exchange board considered the issue during a September 2018 board meeting, and voted unanimously to transition to a fully state-run exchange in time for the 2021 plan year.

The exchange will put out a request for proposals in early 2019, as they work to find a vendor to create their state-run enrollment platform. The system will be live by the fall of 2020, in time for the open enrollment period for 2021 coverage (November-December 2020).

There are two good reasons to go down this path. The first is the obvious one: it’s cheaper than using charges 3.5% of premium as an Exchange fee for states that don’t do anything on their own, and a 3.0% of premium Exchange fee for states that manage significant elements of the enrollment process. New Mexico is one of the “partnership” states that uses the front-end but manages a lot of their own back-end. 3.0% of premium is not a good deal. The same fee level would either be used to fund significantly more outreach, advertising and navigators or the same level of outreach and support that provides could be funded at a much lower fee which would slightly reduce baseline premiums.

Secondly, New Mexico is getting ambitious. They are the leading innovators in doing the actual hard work of figuring out how a Medicaid buy-in proposal. This would be effectively a state based public option. This would be a major rejiggering of the New Mexico individual market.

The Center for Medicare and Medicaid Services (CMS) has repeatedly stated under both the Obama and the Trump administrations that they can’t do much with the back-end of to support unusual or aggressive waiver requests. If New Mexico moves towards a Medicaid buy-in model, their open enrollment, subsidy structure and eligibility structures would be unique. The only way that can work within an Exchange framework is if New Mexico can customize the exchange that their citizens and residents see.

So, this is both an effort to reduce premiums through either attracting a healthier risk pool or lower costs of attracting the current risk pool AND a necessary step in building the infrastructure to support a Medicaid buy-in program.