Mandate exemptions proliferate to what effect?

The most recent set of rules for the ACA contained two major new individual mandate exemptions:

1) People can claim a hardship exemption if all affordable plans cover abortion
2) People can claim a hardship exemption if their county of residence only has a single insurer.

All of this is only relevant for 2018 as the mandate penalty zeroes out on 1/1/19.

I am not sure how big of a deal this is in terms of how many people would drop coverage this year with these two new exemptions that they otherwise would have kept. The individual mandate penalty this year is 2.5% of income or $695 for a single adult, which ever is greater. There is a current hardship exemption for people who would need to pay more than 8.05% of their income for the least expensive Metal plan.

A good chunk of the country has to earn over $70,000 a year for a single 40 year old to not qualify for that mandate exemption. A family of three has to earn well over 600% Federal Poverty Level in a lot of states to not qualify for a hardship exemption.

Intuitively, a lot of people have an out from the ACA if they wanted to take it already. This makes me lean towards thinking that these two exemptions are primarily messaging rules and not rules with significant pragmatic impact.

Leaving money on the table for coding

Adam Sacarny has a pretty awesome and interesting (to me at least) NBER paper that looks at coding practices in hospitals when there was major money on the line.

Here is the set-up. Before 2008, CMS allowed non-specific Heart Failure codes to be paid at a high level. Almost no hospitals submitted claims with detailed, specific codes as hospitals don’t get paid for more specific documentation. Starting in 2008, the non-specific codes were paid at a low level while specific codes were paid at a high level. There is no change in the underlying characteristics of patients, just a change in the documentation. The total potential revenue swing is large (2%) and fairly straightforward to capture.

These results surprised me a lot.

My prior was that if a claim element drives payment, it will be addressed.

I could expect some learning by doing in the first few months as old work flows are broken and new work flows are implemented. Perhaps new coders were hired? Might some docs be trained to say a bit more in their notes? Maybe there would be a proliferation of billing optimization consultants swarming over hospitals with their Powerpoints with their schemes? I would have thought that by three years, almost all (95%+) of the revenue potential would have been captured.

That was not the case. Hospitals were leaving serious money on the table due to documentation.

His table 6 showed that large hospitals in urban areas with significant teaching responsibilities were most likely to capture more revenue than other hospitals. I am speculating now, but I wonder how much of this type of hidden management/operational expertise drives consolidation as a big, urban, teaching hospital could look at a community hospital and see that there was a major documentation revenue center that could be quickly exploited as a source of “value” (from the Merger and Acquisition viewpoint) while the longer term plans of buying out referral patterns and building a hub and spoke feeder pattern would provide medium term benefits.

I found this paper fascinating as my assumption that if a code gets paid for, it will be coded has been shaken.

The CBO officially sees Silver Loading

The Congressional Budget Office just released their updated budget baseline.  (Yeah, I’ll wait for everyone to stop partying)

Umm, hmmm… the most interesting thing to me is that the CBO is now officially counting Silver Loading of the Cost Sharing Reduction subsidies as part of the budget baseline.


Now that Silver Loading is officially part of the baseline, it can be part of an explicit discussion of policy trade-offs. I don’t expect anything to happen this year for a dozen different reasons but this is a worthwhile thing to note.

Risk adjusting by a controllable variable: bundles and other situations

There is a nifty little paper that looks at length of stay in Skilled Nursing Facilities (SNF) for Medicare beneficiaries**. It finds that BMI is a predictor of length of stay and readmission.


Residents with mild (adjusted relative risk (aRR)=1.16, 95% CI=1.12–1.19), moderate (aRR=1.27, 95% CI=1.20–1.35), and severe (aRR=1.67, 95% CI=1.54–1.82) obesity were more likely to be readmitted within 30 days than those who were not obese. The average difference in LOS between residents without obesity and those with mild obesity was 2.6 days (95% CI=2.2–2.9 days); moderate obesity, 4.2 days (95% CI=3.7–5.1 days); and severe obesity, 7.0 days (95% CI=5.9–8.2 days). Residents with obesity were less likely to be successfully discharged and more likely to become long‐stay nursing home residents.

Obesity was associated with worse outcomes in postacute SNF residents with hip fracture. Efforts to provide targeted care to residents with obesity may be essential to improve outcomes. Obesity may be an overlooked risk adjuster in quality‐of‐care measures and in payment reforms related to PAC for individuals with hip fracture.

I want to think about two thinks about risk adjustment. The first is the need to risk adjust bundles. Secondly, I am a bit leery of risk adjusting on controllable variables as it is theoretically ripe for manipulation.

Read more

Risk adjustment and Silver Loading

Susannah Luthi at Modern Healthcare (subscription only) raises a very good point about Silver Loading; it will make risk adjustment in some states completely unpredictable.

“If told that enrollees are now lower-risk than in the past, which is now the case, plans with a higher number of gold members may get too much compensation under risk adjustment,” Dorn said.

If higher-risk enrollees opt for bronze because they have free or very cheap premiums thanks to the increased subsidies, the carriers insuring those plans could lose money in the risk-adjustment payments. This could impact companies like Security Health Plan, which has seen a shift to bronze plans even though its population, which skews older, does need to use their coverage.

The ACA risk adjustment process is a zero sum game within each state. The formula assigns weights to demographic characteristics and some diagnosis and prescription information. The weights vary by metal bands. Two identical people will produce a different total score if one buys Silver and the other buys Gold.

This is not a problem if the state is a single insurer state.

This is a minor problem if all of the market(s) in the state showed very similar shifts in metal band market share.

This is a minor problem if the changes in metal band composition were a reasonably efficient sorting by health status. If the new Gold buyers who normally would have bought Silver are sicker than average and the new Bronze buyers are healthier, then things wash out.

It is potentially a significant problem in states where there is significant variance in metal share distribution. Pennsylvania and Georgia have a lot of variance happening in their Gold uptake.

In Pennsylvania, every county except Philadelphia and the four county Mainline suburbs had at least 20% of the on-Exchange buyers purchase a Gold plan. There is significant variation by county and insurer. Central Pennsylvania counties in one rating region had over half the buyers go Gold because of pricing advantages. All else being equal, risk adjustment would over score the rest of Pennsylvania and under score Greater Philadelphia. This is counter-balanced to some degree by Greater Philadelphia have a lower Bronze percentage than most of the state. But on first pass, the significant variation and overweighing of Gold plans will pull net risk adjustment out of Greater Philadelphia and Independence Blue Cross and Blue Shield and send that money to everyone else in the state.

Georgia is even more complex of a story. One insurer, Alliant, aggressively Silver gapped and Silver loaded. Most of their service area was heavy on Gold and Platinum plans (Platinum also gets a risk adjustment bump) and light on Silver and Bronze. Other insurers in the state are very heavy on Silver compared to both national averages and Alliant’s share.

The challenge in risk adjustment for the ACA is that the actuaries for a single insurer have to project both their own marketshare and covered population characteristics and all other insurers’ market share and covered population characteristics in the state. Projections of company owned data with a known strategy has inherent variance. Trying to project risk adjustment due to the actions and enrollment of other companies several months before open enrollment starts is a Sisyphian task.

We should expect to see some companies take mid-year charges as variable strategies lead to significant changes in the underlying risk adjustment flows. Real money won’t change hands until the summer of 2019 but capital cushions will expand and contract this summer.

Waiver updates

One of the things that has surprised me is that more states had not submitted 1332 waivers for their ACA markets by the end of February. The Center for Medicare and Medicaid Services (CMS) has, by law, up to 210 days to approve or deny a waiver. CMS intends to sign exchange contracts with insurers at the end of September so late February was the latest that a waiver could be submitted with at least 210 days before the contract signing date. Since I’m so much fun to travel with, I had thought that some states would submit waivers with sufficient time to allow for the entire window to be used.

I was wrong. No state submitted a new waiver application before March 1.

Since then, Wisconsin is preparing a straightforward reinsurance waiver. It has not gone through the entire state based public comment process yet but it is straightforward and should be approvable. CMS has shown that it is willing and able to approve reinsurance waivers fairly quickly once the documents arrive.

Maine is preparing a new invisible high cost risk pool waiver. They will fund the waiver with a $4 per member per month insurance tax for all group and individual market policies sold in Maine. Individuals and their families with 8 chronic conditions will be automatically put into a high cost reinsurance pool so that the selling insurer is never on the hook for more than $50,000 while insurers can cede premiums and risk for other individuals. There are a couple of small technical choices that have me scratching my head but this is a mechanically sound proposal that will lead to lower off-Exchange premiums. It should get approved fairly quickly.

Finally, Ohio submitted their waiver application today. The waiver asks that the individual mandate be waived for all of Ohio. Since the waiver penalty was zero-ed out starting in 2019, this has almost no pragmatic fiscal or coverage impact. It is a messaging waiver.

There is an interesting side note where the Ohio waiver could provide some useful evidence. How much of the efficacy of the waiver is purely derived from a rational cost-benefit analysis and how much of the waiver’s impact is due to the “taste for compliance” where people want to follow the rules? If the waiver is entirely a rational cost-benefit analysis, the Ohio waiver is functionally meaningless. If there is a taste for compliance, wiping a $0 waiver out could lead to lower enrollment. I lean towards thinking that a “taste for compliance” has some effect but this waiver could provide some stronger evidence.

Update 1 Maryland is getting their ducks in a row for a very large reinsurance waiver application.

2018 Open Enrollment report; metal band market share

The Center for Medicare and Medicaid Services (CMS) released the 2018 Open Enrollment report. I am looking at the county level public use files to estimate the market share by county of the different metal bands. My data is here. I made the following assumption that cells which CMS censored due to having less than 10 policies sold have a value of zero. That assumption may skew percentages in a few very small counties but I’m accepting that imprecision. Zero marketshare due to lack of sales or a metal band not being offered in a county is suppressed for each metal band.

The Tableau below is the marketshare by Metal Band by each county. I am starting off on Gold Plans as I was expecting to see significant variation on sales because of Silver Loading and Gold Gap.

And we have massive variation in Gold Plans. Most of Pennsylvania excluding the Philadelphia Main Line had Gold purchase rates that were at between five and ten times the national rate. 58% of all on-Exchange plans in Montour County were Gold plans. This is to be expected. The least expensive Gold plan in most of Pennsylvania was priced underneath the Silver Benchmark plan. From there, Gold purchases were likely a better deal for anyone making over 200% Federal Poverty Level and plausibly a valued trade-off for folks making between 100% and 200% FPL.

Balloon Juice World Headquarters in Brooke County, West Virginia borders Washington County, Pennsylvania. Washington saw 34% of on-Exchange buyers go Gold. Brooke County saw no more than 6.5% of its on Exchange buyers buy a Gold plan (Gold and catastrophic plans were censored as one of those two cells had less than 10 people). Brooke County’s least expensive Gold plan for a 40 year old was $139 more expensive than the benchmark Silver. Washington County’s benchmark Silver was $15 more than the least expensive Gold plan. Pricing matters.

All of Wyoming and New Mexico as well as most of Pennsylvania and Kansas plus chunks of northern Georgia, southern Wisconsin and pieces of Texas and North Dakota had counties where at least 20% of the plans sold were Gold.

The ACA has always been a county by county story. It is even more so this year as aggressive state regulators and the decisions of various insurers could lead to incredible deals being offered on or business as usual.