David Anderson

The New England Journal of Medicine has the first trial report out on the effectiveness of remdesivir, an anti-viral medication made by Gilean, for treating hospitalized individuals with COVID-19.

METHODS
We conducted a double-blind, randomized, placebo-controlled trial of intravenous remdesivir in adults hospitalized with Covid-19 with evidence of lower respiratory tract involvement. Patients were randomly assigned to receive either remdesivir (200 mg loading dose on day 1, followed by 100 mg daily for up to 9 additional days) or placebo for up to 10 days. The primary outcome was the time to recovery, defined by either discharge from the hospital or hospitalization for infection-control purposes only.

RESULTS
Preliminary results from the 1059 patients (538 assigned to remdesivir and 521 to placebo) with data available after randomization indicated that those who received remdesivir had a median recovery time of 11 days (95% confidence interval [CI], 9 to 12), as compared with 15 days (95% CI, 13 to 19) in those who received placebo (rate ratio for recovery, 1.32; 95% CI, 1.12 to 1.55; P<0.001). The Kaplan-Meier estimates of mortality by 14 days were 7.1% with remdesivir and 11.9% with placebo (hazard ratio for death, 0.70; 95% CI, 0.47 to 1.04).

This is good evidence that remdesivir is useful as a therapeutic. The study is a gold standard randomized clinical trial without the typical confounding seen in observational studies. It does not have selection effects and the statistics indicate a high confidence in at least recovery effects. Median recovery time is 26% (4/15) faster for people administered the drug than people who received placebo and regular care. Remdesivir does not have proven mortality effects at this time.

So what is its value?

I think remdesivir has two value propositions in the United States. In the current environment where hospitals are currently not overrun, it has real value in reducing hospital stays by 26% and reducing PPE, supportive medication and staff time demands to treat any one patient. This is valuable.

In an environment where a region’s hospitals are being overwhelmed, surging remdesiver into a region has both the baseline value and then an add-on value. A region with an overwhelmed hospital system will have a case fatality rate several times higher than the same region with hospital capacity to spare. The faster recovery time means beds clear faster and the staff that would be committed to helping two patients recover per month would be available to care for almost three patients per month. This would be a fast way to surge effective care capacity into a region that is the equivalent of deploying a hospital ship or an army field hospital with the attached train staff.

In that scenario, remdesivir will likely have indirect mortality effects. Shorter lengths of stay means the effective hospital throughput is higher which means the length of time that a region is in an overload scenario is shorter. Some individuals who would have been admitted to the hospital during a slam period without remdisivir will be admitted to the hospital during normal(ish) operations with far lower mortality risk due to the increased throughput.

Using this logic, remdesivir has both immediate resource replacement value and mortality prevention value in some scenarios. These two value streams can be modeled to be a fairly large number under some modeling assumptions. The value of the mortality prevention by minimizing periods of slam will be a function of modeling both COVID-19’s inherent response to seasonality, our social responses, our public health responses and the length of time to widespread, effective vaccine deployment. The less optimistic an analysis is about any of these parameters, holding all else equal, the more valuable remdesivir will be. But a fairly large number could be justified under standard cost effectiveness models depending on the underlying assumptions about disease spread and regional overload scenarios.

Remdesivir and its valuePost + Comments

49 of 50 states have balanced budget requirements. States are experiencing a massive revenue shortfall due to the COVID pandemic. State budgets have core priorities of education, medical care, criminal justice and civil courts, transportation, infrastructure and recreation on public lands. State budgets also have short run fixed costs of pensions and debt upkeep as well. After those expenses, other add-ons can be included.

One of the add-ons that about a dozen states have chosen to spend money is Section 1332 reinsurance waivers for the ACA individual market. These waivers have a couple of different flavors but fundamentally they are similar. A source of money that is not tied to premiums is injected into the claim paying pool. This external money takes on a share of the claims and allows premiums to drop. The lower premium level triggers lower federal subsidy payments which then gets recycled back into the external source of money. The key here is the external source of money. Some states use general fund revenues, others use a general insurer assessment while others have other dedicated funding streams.

These waivers serve a fairly small number of people in the state. The waivers make insurance cheaper for individuals earning over 400% FPL and slightly more expensive for subsidized individuals who choose a plan that costs less than the benchmark plan. The marginal cost of insuring one extra person is fairly high.

States will have to make very hard choices in the next several weeks as updated revenue and expenditure projections come in. 1332 waivers can be either reduced or eliminated to free up the funding stream that went to the waiver. The freed up funds could be redirected to other, higher priority programs.

Below is the language from the Colorado 1332 waiver approval letter that outlines early termination procedures. The key is a notification by July 1st.

If a waiver was to be terminated, we should hear rumbles in the next week or so and then comment periods in two or three weeks.

This is worth keeping an eye on.

1332 waivers and state budget crunchesPost + Comments (9)

Once I feed the cats and finish posting this post, I am going to donate blood. I am O+ without CMV which evidently makes the hematologists very happy to schedule me on a regular basis. One part of the e-mail campaign for this donation was a promise that my blood would be tested for COVID-19 antibodies and I would be notified of the results in a week.

I can understand why the blood bank wants to screen for COVID as a means of insuring a safe blood supply. However we can’t think that the numbers will tell us much about a region’s COVID exposure. This is attractive but almost useless for systemic surveillance.

Blood donors are a self-selected population. Before donating blood, a donor has to have answered over forty questions of a basic current health and long term medical history. One of the knock-out questions is if you have had the cold or flu recently. Screening blood for COVID antibodies might allow for a plausible lower boundary of regional infection history. This population is likely to notable underestimate disease prevalence.

The other challenge that we need to be aware of if any blood donor infection and recovered figures are released is the problem of false positives. If we assume that the blood donor population that is screened is likely to have a lower than population average incidence rate, than we also have to assume that a higher proportion of the reported positivity rate is false positives. If we assume that a regional cumulative COVID prevalence rate is 10% and the test returns 1% false positives and no false negatives, than the false positives will be about 8% of the total reported positives. If we assume prevalence is 5% then the false positives are 16% of the reported positives. If we assume prevalence of 2.5%, 29% of the total reported positives are false positives. The Positive Predictive Value of the tests that have Emergency Use Authorizations (EUA) from the FDA are frequently below 90% if we assume a 5% population prevalence.

Since we are assuming that blood donors are less likely than the general population to be infected, the value of the testing result being reported back to the individual donor is going to be low.

And this is problematic. If someone receives notification that they have COVID-19 antibodies, they are likely to act on that information and engage in behaviors that are more likely to increase spread risk. If they are truly immune, this is not a big deal. But if they are not immune, they are more likely to be an unsuspecting spreader during a non-symptomatic phase of a future infection.

Communicating these types of results for a low incidence but high impact disease is tough as the information is likely to change behaviors.

Blood donors and biased samplesPost + Comments (15)

Families USA has preliminary analysis of new Medicaid enrollment due to COVID-19:

As of May 15, 2020, at least 28 states publicly reported Medicaid enrollment data through March 2020 or later months. The analysis compares states’ March, April, and May enrollment levels against their January enrollment levels. Findings show enrollment increases beginning in March or April and then rapidly accelerating. Most states saw modest or no enrollment growth from January to March (0% to 3%), significant increases from January to April (0% to 6%), and massive increases from January to May for those states that have reported (3% to 10%). As a point of comparison, Families USA also analyzed all states based on the Centers for Medicare & Medicaid Services (CMS) monthly enrollment data between January 2019 and May 2019. During that time, a calmer economic period, only two states exceeded 2% growth (Idaho at 2.39% and Vermont at 2.34%), and nearly every other state remained relatively static.

This enrollment increase is happening at a time that state budgets are collapsing.
CBPP reports massive declines in state tax revenue:

CBPP estimates that state budget shortfalls will ultimately reach about 10 percent in the current fiscal year (which ends on June 30 in most states) and more than 25 percent in fiscal year 2021 based on recent economic projections.

Medicaid is a joint federal-state financed program. The federal government does not have a balanced budget constraint and can cheaply access extremely deep capital markets. States can’t do run large operating deficits. They are pro-cyclical accelerators while the federal government can, if our political leadership so chooses, be a counter-cyclical balancing force.

We don’t know how expensive the new enrollees will be. We know that people who, in normal non-pandemic times, experience an insurance transition have a short term utilization spike. We can’t assume that will hold in COVID-19 as both the medical system effectively closed for six weeks and is still down 30%. We can also suspect that the characteristics of new enrollees in the COVID environment is systemically different than new enrollees in 2019. This is an aim of a project I’m drafting at the moment.

But with all of that said, Medicaid is going to be more expensive for states at precisely the time that states have no uncommitted free cash flow.

The Federal government is helping to some degree. The 2nd COVID relief bill increased the federal share of legacy Medicaid spending by 6.2% points. However state budgets will still get squished and that means even as need increases, services are likely to be cut.

Medicaid enrollment expands as state budgets contractPost + Comments (3)

Kent asked a great question yesterday about the mechanics and money flows of large group health insurance:

Say you are a big Portland-area employer with thousands of employees and your employee-provided health plan is through Kaiser.

If you lay off 2,000 employees during the pandemic and those 2,000 employees drop off of the Kaiser health plan because they are unemployed, isn’t that a direct hit to Kaiser’s bottom line?  And doesn’t COBRA prevent that from happening

 

There are two basic flavors of insurance products a large group employer can buyer. The choice of which insurance flavor changes the answer to Kent’s musings.

Fully Insured (FI) means that the insurer takes on all risks of medical claims in exchange for collecting a fixed premium from the employer. The insurer makes money when claims and administrative expenses are less than the premium. The insurer will lose money if there is a cluster of unanticipated high cost claims that exceed reinsurance. In this case, 2,000 employees getting laid off and losing their insurance will be a significant revenue hit for the insurer.

Fully Insured is how the Medicare Advantage, and the ACA small group and individual markets work. Fully Insured is unusual to see in large groups (100+ covered lives).

Self-Insured (SI) is the other flavor of large group employer insurance. Here, the employer pays the claims as they are incurred. The insurer merely performs Administrative Services Only (ASO) functions such as network construction, benefit design, and claims processing in the stead of the employer. The employer pays the insurer a per member per month fee to use these services. The insurer is fundamentally indifferent if there is a cluster of multi-million dollar claims or if there are not claims. The employer takes on all of the risk.

Most large employer group insurance plans are self-insured. Self-insured plans are regulated under ERISA which keeps state regulators far removed from the administration of these plans while state regulators are actively involved in fully insured plans. Self-insured plans are also systemically cheaper than fully insured as there is risk sorting and selection.

So when an employer that is self-insured lays off 2,000 people, the ASO insurer loses a small per member per month fee otherwise the insurer is almost indifferent. The change in claims being paid will be immediately interalized by the firm, not the insurer.

Two flavors of large group insurancePost + Comments (5)

The House passed the HEROES Act, a $3 trillion dollar relief bill. One of the components is a nine month program to fully subsidize COBRA, as VOX explains:

COBRA subsidies: The bill funds approximately nine months of full premium subsidies for the existing health insurance program COBRA, which allows laid-off or furloughed employees to stay on their health insurance plans. COBRA is typically prohibitively expensive, but this bill would make it more affordable for millions of workers losing their health insurance along with lost jobs.

I think this is effectively a cash flow injection for businesses that have laid off a lot of folks but still remain a going concern.

To get there, I think we need to talk about mechanics first. I will primarily be thinking about group insurance through an Administrative Services Only/Third Party Administrator arrangement where the insurer merely does the back-end administration and the employer takes on the full cost of paying actual incurred claims.

COBRA allows for people who lose qualified employer sponsored coverage from entities that are still a going concern to buy back into the group coverage for 102% of premium or premium equivalent. Group insurance provided by an employer has pure community rating so a 21 year old with no medical history pays the same premium or premium equivalent as a 64 year old whose medical chart makes War and Piece seem like a light read. COBRA also allows for carry-over of out of pocket spending so if you have paid $1,000 to your $1,500 deductible while on the employer sponsored plan, your COBRA plan variant still only has $500 left to spend on your deductible. COBRA is also guaranteed issue.

These characteristics of COBRA make COBRA very attractive to older, sicker and more expensive employees. People sign up for COBRA when the price of the coverage is less than the expected benefit. 102% of community rate premium is what an employer collects, but the expected claims outflow can be significantly above the individual premium. A COBRA beneficiary is likely to cost the employer more money than incoming premium.

This has been true since COBRA started. It is especially true now as the ACA gives low income and healthy workers who just lost coverage access to insurance that is extremely likely to be far less costly to the worker than an unsubsidized 102% of premium even if the worker does not qualify for a subsidy. Under current configuration, COBRA is very attractive as an option to an 63 year old individual with an ongoing cancer treatment who makes well over 400% Federal Poverty Level(FPL). COBRA is a cruel and expensive joke to 35 year old who earned 250% FPL.

So with this set of mechanics, COBRA right now is a loss center for employers. More cash goes out the door than comes in as the COBRA selecting population is significantly adversely selected.

However, this dynamic changes if there is 100% federally funded COBRA. All of a sudden, COBRA looks extremely attractive to young, healthy and likely to be low cost individuals compared to the alternatives of being unemployed, ACA coverage where deductibles reset and learning costs would be incurred on navigating a new network and potentially a new employer, or paying premiums for short term underwritten plans. COBRA also beats other alternatives in that it is simpler with fewer administrative frictions. There are fewer chances for paperwork to create cracks to fall through compared to any other alternative.

100% COBRA subsidies moves a lot of people from the Exchanges, Medicaid or uninsured back to the employer sponsored group books. These people are likely to have below community rated average costs so the net difference between COBRA spending and COBRA revenue will shrink as the expensive people who would always have been on COBRA are balanced out by a healthier and cheaper risk pool. Total COBRA claims will increase far slower than COBRA revenue. For employers that offer self-insured health benefits that are COBRA eligible, most of the subsidy incidence falls on their bottom line as a significant net cost reducer and thus improves short term cash flow.

COBRA subsidies in HEROES as a compensation subsidy to employers?Post + Comments (7)

Seanly asked a good set of questions in comments yesterday:

1. If we’re mostly staying at home, how is the virus still spreading so much? Is it that even with restrictions there are still some people spreading it? Seems the odds of that happening would decrease? Is it possible there are some people who are asymptomatic carriers (like for weeks)?
2. at some point we will have to open things up a bit more (not now, not so fast). I’m not sure what that will look like. We’re going to still need some food production, still need to get teeth cleaned, and may need pants. How & when can we do this?
3. Weren’t there some experts who thought the way to proceed with effective testing, tracing, etc is that we’d have the country mostly open (with masks & social distancing) and then if a hotspot bloomed only that area would need the stricter lockdown?

#1 Yesterday, Scott Gottlieb in Congressional testimony estimated the (R)eproductive number is currently about 1.1.  1.1 R means that for every 100 people who are currently infectious with COVID, they will infect, on average about 110 new people.    Other sources have the national R at or a smidge under 1. All of these estimates have large confidence intervals.

Infection is not occurring randomly. Instead we are seeing clustering of new infections and clustering of non-infections. New infections are happening disproportionally in places where isolation and distancing do not readily or easily occur. This means nursing homes, assisted living centers, slaughterhouses, prisons, jails, detention centers and aircraft carriers are seeing virgin field, rapid spread outbreaks. Indoor spaces with prolonged proximity and lots of aerosol exchanges are prime locations for hot spots.

#2 Building on #1, a recognition that indoor spaces with high density of interactions is a very bad thing allows us to shape the environment and our monitoring to lower risk. North Carolina is slowly re-opening in phases with the intention of keeping density of interactions and personal proximity low even as stores that have been closed for 7-8 weeks restore some degree of service. Florists can probably have a decent chance of safely re-opening while sit-down dining at Red Lobster is far more questionable in terms of risk management.

Duke University has been slowly planning to bring a small cadre of workers back on campus for a few weeks. I fully expect that I won’t be allowed on campus until July. I expect that I will be encouraged to work at home for the rest of the year as a means of minimizing on-campus interaction density. No one needs a life saving blog post or a critical revise and resubmit.

#3 Yeah, that is my boss’s plan. The hope is that the baseline rate of spread is low enough that a hot spot can be quickly identified and then smothered. South Korea is doing that now. They had a micro-hotspot in Seoul’s bars. They tested 2,000 20,000 or more people and isolate plenty of folks so that the hot spot could not light the rest of the city on fire again. That is the suppression model that has a chance in hell of working until there is a vaccine while keeping the economy mostly open.

Current COVID statusPost + Comments (57)

COVID is a generator of new unknowns.

The New York Times reported on the experience of some Italian citizens who have been infected by COVID and are no longer infected. However these folks are not seeing their health restored to their pre-COVID states:

Studies also point to kidney, heart, liver and neurological damage, often from secondary infections, and no one knows what the long-term prospects are for those patients.

But even some of the infected who have avoided pneumonia describe a maddeningly persistent and unpredictable illness, with unexpected symptoms. Bones feel broken. The senses dull. Stomachs are constantly upset…..

On the weekend of March 7, Mr. Cirielli ran a fever and suffered cold-like symptoms, and became convinced he had picked up the virus by touching an infected surface in the Parliament chamber. He tested positive that week….After 40 days of feeling lousy, he tested negative for the virus, but his eyes still burned and bouts of diarrhea continued, he said….

“It never finishes,” said Martina Sorlini, a 29-year-old math and physics high school teacher who has been running a low-grade fever since the beginning of March. She said that the cough and sore throat eventually went away, and that after three weeks she got her sense of taste and smell back, and even found enough energy to run and tend vegetables in her garden.

Then came stomach aches, fatigue and the return of fever. And it hasn’t left, making it extremely tiring to teach her online high school classes….

We don’t know what “typical” post-COVID looks like. We just don’t have the data as the first people who we know were infected with COVID and survived now have four months worth of follow-up.
We know enough to know that there are likely long term secondary effects from a COVID infection even after the virus has been rendered ineffective by the patient’s immune response. But we can not readily project in a year how many additional dialysis sessions are needed per 10,000 COVID infections stratified by age and controlled for pre-COVID observables. We can’t project how many physical therapy/rehab sessions will be needed.

One of my curiousities this summer will be tracking how insurers are trying to price COVID into their 2021 rates. Right now we know so little that almost any assumption is facially plausible. I expect that there will be huge variance among insurers in May and June and a decrease in variance by the time rates are finalized near Labor Day as the few extra months of knowledge generation will help us figure out what COVID actually does to people over the long term.

Right now, we just don’t know, but the early knowledge is that we can expect a lot of weird things to happen.

You’ve recovered, now what?Post + Comments (61)

Vermont’s insurers filed their initial rates on Monday, May 11. Blue Cross and Blue Shield is asking for a 6.3% average increase.  MVP Health wants an average of a 7.3% increase.  Those two are fairly tightly clustered in the initial asks.  I was more curious about how they dealt with COVID.

MVP makes a lot of assumptions that are facially plausible:

Line 17- Adjustment for COVID-19 Immunization Cost
MVP is assuming that a vaccine to prevent the novel coronavirus (COVID-19) will be tested and widely available in 2021. To account for the costs an immunization would add to claim cost, MVP is assuming that an immunization would be covered in full at the cost of $75 per dose. MVP is also assuming that 80% of the population would obtain the vaccine (based on an analysis published by Wakely Consulting), which corresponds to a PMPM claim cost of $5.00 PMPM ($75 per dose times 80% utilization PMPY divided by 12 months). This adjustment is reflected on line 17 of Exhibit 3.

Line 18- Adjustment for COVID-19 Pent-up Demand
As a result of the COVID-19 pandemic, elective surgeries and associated services have been postponed for effectively all of MVP’s service area. As of the time of the filing, both Vermont and New Hampshire have announced that they are allowing the ramp up of elective procedures again. Therefore, MVP is assuming that 2 months of elective surgeries have been canceled in 2020.

MVP analyzed its entire commercial population for 2019 and found that the claim cost related to elective services was $45.09 PMPM. To value what the delay of these elective services will be worth in 2021, MVP made the following assumptions:

• – Resumption of a normal level of elective services will begin in mid-May 2020
• – Twenty percent (20%) of elective services deferred during the 2 months will be eliminated
• – In order to recoup lost revenue, beginning in August 2020 providers will perform 110% of their prior elective service volume until deferred services were fully performed.

These assumptions combine to add $4.51 PMPM in claim cost for the time period of January to April 2021 (at which point all deferred services will be made up and providers will return to normal utilization levels). MVP is reflecting $1.50 PMPM ($4.51 PMPM for 4 months, converted to an annual PMPM) in this filing to account for the increased utilization.

To translate this from actuary speak, MVP thinks that COVID is going to cost them about $10 per member per month (PMPM) in 2021. That is facially not bizarre!

Now let’s take a look at what Blue Cross and Blue Shield are thinking:

The various impacts of the COVID-19 pandemic will impact 2021 costs, and they will do so in both an upward and a downward direction. Based on information currently available, it is difficult to so much as predict an overall directional impact, particularly because future government actions will drive the timing and magnitude of many of the above considerations.
We believe that it is entirely reasonable to forecast that the ongoing pandemic will increase 2021 claims costs beyond the levels projected within this filing. However, because of the margin
guidance provided in Attachment C (as discussed in section 3.8.7.2), the addition of a COVID related factor of any magnitude would be offset by a reduction in CTR of equal and opposite
magnitude, resulting in no change to the premiums presented herein. Furthermore, based on information known at the time of filing, we see no reason to believe that the best estimate of
the pandemic impact is a decrease in 2021 claims costs. We therefore include a COVID-19 impact of zero within the 2021 premium rates.

BCBS-VT is attributing no part of their rate increase to COVID. They expect COVID to have no material impact (up or down) on premiums for 2021. And it is not a crazy assumption.

What we are seeing is two sets of actuaries looking at an incredible amount of known unknowns and probably several unknown unknowns and making good faith guesses (they call them projections as that justifies their hourly rate). BCBS-VT is systemically making a slightly more optimistic projection than MVP. BCBS could be right, they could be wrong. Given how ACA subsidies are structured, BCBS-VT optimism is giving them a slight pricing advantage and thus enrollment gain compared to a universe where there is a mandated set of uniform assumptions or a universe where the BCBS actuaries were as pessmistic as the MVP actuaries.

$10 Per Member Per Month is real money if BCBS is wrong and MVP is right. But it is not lose the company wrong.

I expect that more information will emerge over the summer and one way or another, the COVID rate setting assumption will converge to a closer level of optimism or pessimism.

ACA pricing during COVID uncertaintyPost + Comments (4)

In late March, I put out a post expressing concern about the shape of the claims curves for states with 1332 attachment point reinsurance waivers.

These are not normal times. States with attachment point reinsurance programs could be facing an unusual and unprojected claims curve…
States built their program budgets on the assumption of a normal experience year with lots of low and medium dollar claims and few big claims. Insurers priced their premiums on similar assumptions regarding reinsurance payments. A bad COVID19 month could put a lot of people into the ICU with high cost claims that don’t displace too many other expected high cost claims as those claims are likely for cancer, or other must treat conditions…

So far, I have been wrong. (I am glad to be wrong).

I’ve talked to some nerds in the know and they are not seeing a ton of ACA insured individuals landing in the ICU. At least they are not seeing more ACA insured individuals landing in the ICU that has not been compensated for by the incredible reduction in almost all other utilization, elective and in normal times not elective. The cost curves have moved down and to the left.

There was one other factor that I had not considered in March that is worth considering. Special Enrollment Periods (SEPs) occur when there is a significant life changing event. They allow people to buy insurance outside of the normal open enrollment period. The most common SEP is loss of insurance through job loss. There has been incredible amounts of job loss. Several state based marketplaces have also opened up a SEP for anyone who fears COVID-19. The Urban Institute estimates millions of more Americans will be insured through the ACA Exchanges this year and next year. Some of those individuals are enrolling through the SEP system.

SEP enrollees have traditionally been more expensive than full year enrollees but once you pull out mid-year newborns (those adorable expensive balls of poo and giggles), SEP enrollees have historically tended to be well risk-adjusted. There is a decent reason to expect in 2020 that SEP enrollees who start coverage on May 1 or later are going to be very low cost enrollees as far more of the SEP enrollees just had coverage through work. And the small group ACA market has lower average risk scores than the individual ACA market segment on a state by state basis.

So there is likely to be a big infusion of relatively healthy folks into the ACA individual market. These folks will be paying premiums either in full or with APTC assistance. The 1332 waivers are structured so that the federal government passes through to the state programs the incremental extra amount that would have been spent without the waiver. There are a lot more people getting a little bit less subsidy which means the federal government is generating more pass-through savings to send back to the states.

My March worry neglected the increased pass-through revenue when I was initially worried about the shape of the claims curve.

1332 and SEP enrollment?Post + Comments (1)