Administrative costs

I just saw two interesting (to me at least) things about back end administrative costs for health insurance.  The first is from Dan Gorenstein at Marketplace.  He writes about a ride-along for a major Medicare/Medicaid fraud bust.

The federal government estimates it loses 12 percent of Medicare spending and about 10 percent of Medicaid to fraud and other improper payments.

Last year that totalled more than $70 billion dollars….

I believe fraud does continue to run rampant and remains largely unchecked,” said Harvard Kennedy School Professor Malcolm Sparrow.

Sparrow doubts much will change until controls are put in place that make it harder to exploit the current automated billing system which processes claims, in most cases, by computer with no human involvement.

“These are beautiful targets from a criminal’s perspective,” he said.

“You can test the system and then you can just submit tens of thousands of claims, absolutely sure because it’s automated that they are all going to be paid.”

 Medicare has very low administrative costs.  One measure has their administrative costs under 2%, another has a total administrative spend of roughly 6%.  Part of the way that those costs are kept as low as possible is that Medicare claims payment is extremely automated.  Most claims are never systemically reviewed until long after the check has been cashed.  Automated claims payment is extremely cost efficient.  A claim that is never touched by humans will cost CMS under fifty cents to process.  As soon as a claim is seen by an individual with the simplest verification, the claim costs $5.00 to process.  Complex adjustments and systemic investigation of oddities dramatically increase processing costs.

Since the overwhelming majority of CMS claims are paid without being touched, gaming the system by submitting a set of mildly fraudulent claims to see what set of codes get rejected and what set get paid is fairly simple.  Once a solid claim template is found, it is not hard to run thousands of bogus or at least extremely aggressively coded claims through in a year before anyone may even take a look at oddities.  The system is extremely efficient at paying claims, but not so good at paying only valid claims.  To get better at not paying odd claims without justification means spending more money and hiring a lot more people.  It is fundamentally the same problem the IRS has in that hiring more agents to crack down on fraud and aggressive claims will bring in massive amounts of new revenue at fairly low cost.  But that would mean upping the administrative ratio and hiring more bureaucrats.

Adam Cancryn at SNL Financial has an interesting tidbit on the co-ops after they finished whining about getting taken to the cleaners with risk adjustment.  Some of the co-ops are forming a combined back-end entity:

the co-ops have doubled down on alternative efforts to stabilize their balance sheets. The nonprofits are negotiating the formation of a management services organization that would merge much of their back-office operations, two sources said, cutting down on the individual co-ops’ administrative costs.

Adam and I went back and forth on Twitter to figure out what the cost savings could be.  I think we are both in the ballpark of a couple dollars per member per month at most.  It is not nothing (remember, I’ve been on projects to reduce postage costs by five cents per member per month and the project lead got promoted when we saw six cent PMPM gains) but it is not a huge amount of money.

The co-ops are in a tough position in that they operate in a dozen states and they don’t have scale like an Aetna or a UHC.  Each state has their own regulatory environment and their own network requirements.  Typically insurance companies’ back-end can be split between customer service, actuary, network/provider, product development,  pharmacy, regulatory compliance, data geekery, sales, medical management and claims.  I would think that some of the regulatory compliance and data geekery that supports compliance could be a pooled resource (national accreditation, some state regulatory burdens etc), customer service might have a shared call center in a low cost state, while medical management could share a call center but that is not that effective.  Most of the knowledge needed to run an insurance company is state specific so building a shared back-end sounds great but it is realistically limited.

 

UPDATE 1: Anon in Comments has a very good point on Medicare fraud:

Medicare does have a very robust fraud prevention unit, they’ve always been able to get money for that. My Big Government Agency that does something with social services supports them with data exchange that’s complex, and always being sliced in different segments for more data mining. We and they do run analyses for different kinds of fraud, every day. But once we find something, we need to prepare a case and follow due process, and for that you often need cooperation from the states and localities where the fraud occurred.

Data mining and looking for the obvious and not so obvious cases can be down fairly cheaply and fairly well.  The issue is the next step.  Once an oddity is detected, human intervention is required.  And that takes a while.  Sometimes the oddity is legit, sometimes it is borderline but not fraudulent, sometimes it is fraud.  Making that determination takes expensive expertise.






12 replies
  1. 1
    Exurban Mom says:

    Two thoughts: if the thing is so automated, the government ought to be able to run algorithms that reveal tendencies towards fraud, which could be investigated by humans. Is this not the case?

    And secondly: I think I probably know the answer to this, but the new Paul Ryan, aka Zombie-eyed granny starver’s health care plan: how bad/off the mark is it? I haven’t had time to look for analysis and I trust your judgment more than any of the journalists who try to cover health care…if you’ve seen an analysis you think is good, please share.

  2. 2

    @Exurban Mom: Re: Ryan — it is worse then the typical Republican plan — if you are healthy and wealthy, it is good for you; if you are healthy, wealthy and male, it is great. Everyone else it is a POS. The biggest difference this year is the proposal to raise the Medicare age to 67 which costs the country (but not the Federal government) more money for worse results.

  3. 3
    Face says:

    @Exurban Mom: This was my thought — why would a computer program blindly allow a hundred claims a day from the same provider without any kind of “hold-stop” function kicking in? Would seem easy to program; any number of claims above “x”, with “x” being a closely guarded secret and perhaps diff for diff states (ahem, FL) triggers a human to investigate.

  4. 4
    Anon says:

    I used to work in the same building with people who contracted for Medicare’s computer systems.

    The problem is, as with many government systems, Medicare’s billing and fraud prevention systems run on some pretty old computer programming infrastructure. There is a fairly massive technical debt left over from years of underinvestment during Republican administrations.

    Medicare does have a very robust fraud prevention unit, they’ve always been able to get money for that. My Big Government Agency that does something with social services supports them with data exchange that’s complex, and always being sliced in different segments for more data mining. We and they do run analyses for different kinds of fraud, every day. But once we find something, we need to prepare a case and follow due process, and for that you often need cooperation from the states and localities where the fraud occurred.

    It’s an arms race between us and fraudsters, however, and the fraudsters outnumber us and often have more resources.

    Hatch Act disclaimer, I am not writing this as a representative of the government this is personal opinion on a nonpolitical post.

  5. 5
    Richard Mayhew says:

    @Face: that is expensive but eminently practical

  6. 6
    Redshift says:

    @Face: And even without the cost of investigating fraudulent claims, it seems like the technique of submitting rapid claims to find out which ones are accepted could be slowed down enough to make it ineffective. I’m thinking something similar to locking logins for a time after to many failed password attempts.

  7. 7
    dr. bloor says:

    The fraud rate would drop to near-zero if Brooklyn and Queens ever decided to secede from the Union. I did some record reviews of worker’s comp and motor vehicle accident claims on the side a number of years ago, and the same one-stop-shopping strip-mall clinics showed up over and over again. I don’t know how they managed to keep the con going.

  8. 8
    jake the antisoshul soshulist says:

    Any thought on Kentucky Governor Bevin’s Medicaid exemption. Seems to me that he is screwing people while pretending not to. In other words typical Republican governance.

  9. 9

    @jake the antisoshul soshulist:
    Thoughts on Kentucky waiver are here:
    https://www.balloon-juice.com/2016/06/23/ky-and-a-dry-screw/

    TLDR: It is a ratfucking without lube

    @dr. bloor: Also exclude everything east of I-95 south of Boynton Beach in Florida

  10. 10

    @Redshift: Rapid in computer terms is different than rapid in claims terms — by this I mean sending a batch of 8-10 claims per day for a week through the system and see what gets denied, held and paid. That is not a suspicious volume for most docs. Each claim would be for nearly identical services with one or two modifiers, procedure or diagnosis codes tweaked.

  11. 11
    amygdala says:

    Is fraud a significant problem in the health care systems of other advanced economies? I’ve always wondered about that.

  12. 12
    qwerty42 says:

    @Anon: … that does something with social services supports them with data exchange that’s complex, and always being sliced in different segments for more data mining. …
    SAS Data Miner? When I went to the SAS conferences, there were a lot of feds there, but (other than, say CDC or BLS) not doing a lot of presentations (that I saw). Been a while though.

Comments are closed.