Some problems have no good solutions as the rest of today’s blog-wall will illustrate. But one of the minor problems in post-racial America (my ass) is how do health insurance policies on the federal Healthcare.Gov exchange get auto-renewed.
Right now the auto-renewal feature is that on December 15th, any one who has a still active 2014 policy and has not selected a 2015 policy will be auto-renewed. The hierarchy of automated choices is the same policy by the same company, then a similar policy by the same company. This is a good 80% solution, but there are plenty of obvious hang-ups. The biggest problem is that the same policy in 2014 will have a dramatically different subsidy most of the time in 2015. Subsidies are calculated by the difference between a family’s expected income as related to the federal poverty line as that produces the expected family contribution, and the price of the second least expensive Silver plan in the market. The gap is the subsidy amount.
The second Silver plan has often changed between 2014 and 2015. The second big problem is that this process assumes no change in income or family size for the 2015 subsidy calculation. That is an unreasonable assumption. Auto-renewed policies will be giving people wildly divergent subsidy amounts. People will be over-paying, and people will be underpaying and the under-payers will get hit with a big tax bill in 2016.
As Adrianna McIntyre fears, this is a significant political problem:
Consumer tendency toward inertia could mean that a lot of subsidized enrollees could see unexpected premium hikes that only materialize when subsidies are reconciled. It’s the reason you’ve (hopefully) read that exchange beneficiaries should shop around during open enrollment…..
Call me a skeptic, but I’m hard-pressed to believe two-thirds of exchange enrollees fully understand the volatile nature of subsidies and want to keep their current plans anyway.
And yes, I worry about political blowback, too. Not only is the underlying problem difficult to explain, it’s also likely to be an unhappy surprise for affected enrollees. For some, that will happen during open enrollment (when there’s still an opportunity to change plans); for others, this issue won’t become clear until taxes are filed. The assumptions implicit in auto-renewal—that an enrollee’s income hasn’t changed, and that she’s entitled to the same level of subsidy
There is no good way of making a universal default decision. The optimal situation is for everyone who has a current policy from Healthcare.gov to go back into the system and make a new choice after they updated their family and income situations. But that will not happen, so then choices have to be made, and negative trade-offs will be incurred. I’ll outline some of the other options and their trade-offs.
Charles Gaba wanted to go to positive action auto-renewal:
— Charles Gaba (@charles_gaba) November 24, 2014
The upside of affirmative renewal is that people are making an active choice to stay on a plan, or to switch. They won’t be too surprised. The downside is that requiring positive action as the only way to re-enroll guarantees significant drops in coverage for January. We see this with 401(K)s where people will engage once with a reasonable but non-optimal decision and forget about it for years. We see this with Medicare open enrollment where people don’t actively renew their plans. We see it with private group insurance open enrollment. Not everyone engages in the process, and the residual is cleaned-up by either not enrolling people or some type of auto-renew schema.
People will forget, especially people who are new to private health insurance. If they only forget to enroll for January, they can still get coverage starting in February or March, but there will be significant gaps. The stories that would have been generated would be that Obamacare was a bait and switch as some mid-30s and very attractive white woman who forgot to auto-renew in November was facing $250,000 in surgery bills that she thought Obamacare would cover in January.
The next option is to develop an optimizing algorithm to auto-renew non-responders. This immediately breaks down into the question of what should we optimize on? Should we auto-renew people into the lowest priced plan at the same metal band? That would save the government subsidy money. That would also require a lot of switching plans as people who were in 2014 2nd Silver are now switching insurance companies, switching networks and switching from PPOs or EPOs to HMOs. If it is a cost optimization problem, the stories will be that Obamacare forced a 63 year old retiree who left the workforce a year or two early to change her doctor who she has seen for 35 years.
If we make the algorithm a bit more complicated than a straight cost minimizer and incorporate claims history into it, this could be an elegant solution. The process could look for the lowest priced same metal band product which has all the providers to which a claim has been paid to in 2014 in network. People will move plans a lot under this schema as it is a double ugly fix. First, the networks that got filed over the summer are slightly different than the networks that are getting sold now. People will be falsely moved as the most recent data conflicts with reality. Mayhew Insurance has added docs and hospitals to some of our networks within weeks of go-live as they saw that there was money to be made and wanted in. There is no national data repository for all networks from all insurers which are updated daily or weekly. There are a few private data repositories and meta directories but they are not all inclusive.
Secondly, and far more importantly, claims data is messy. A claim usually has 180 days from the date of service to be submitted to an insurer. 5% of visits from July have not been billed yet. Insurers then have a couple of months to pay a claim once a clean good to go claim is received.
If we only looked at claims for services performed to April 15, 2014 (6 months run-out plus a month of claims processing time) we’re missing a huge part of the data universe as at least half of the people covered by Healthcare.Gov purchased policies did not start coverage until at least April 1, 2014. Finally, there is no current national claims data base. CMS is building an encounters data system for risk adjustment purposes, but that data lags even longer than claims data. The data simply is not there for this type of scheme.
For either of these optimization systems, I’ve neglected to address the parrallel problem of calculating subsidy level. Do we assume flat subsidy and accept that it will be a good guess but almost always wrong? Or do we data mine IRS records for Social Security withholding records plus attempt to create family units from Social Security data? That is close enough to do-able but it raises significant privacy concerns/paranoid freak-outs for only limited gains in accuracy.
Optimizing on price produces big switches, optimizing on claims data is impossible right now, forcing active selection means half a million or a million accidental drops. There are no good systems of universal, default enrollment decisions that produce no negative consequences. The policy problem is to identify what the actual negative consequences are, and then work to minimize those problems. For the current system, the big issues are people will be assigned to plans that may not be optimal for them (let’s handwave the definition of “optimal” away for the moment) and the subsidies will be wrong in a multitude of quasi-random directions. The policy solution to these problems is to encourage as many people as possible to manually re-enroll early, and then aggressively follow-up with all of the auto-renewers as soon as possible after December 15th to get them to make active choices.
There are no good solutions, only satisficing solutions on mutually conflicting criteria.