Which is Worse? Dumping the Sick? Or Denying Them Care?
A New York Times editorial complains that insurance companies cancelled 20,000 insurance policies over a five-year period – usually because of withholding information about pre-existing conditions. But at least patients got the care.
The Times' solution: Prevent insurers from dropping anybody for any reason, but leave them with even stronger incentives to underprovide care.
Ever since Peter Passel left the New York Times, the paper has displayed stunning ignorance of the science of economics.
The ability to cancel a policy for a material omission keeps costs down and is a sensible way to run medical underwriting. Rather than pull the records and order a medical exam of every single applicant in the individual market, insurers rely on statements by an applicant and a centralized data bank of information on past claims.
If someone has routine claims for some period, say two years, the policy then becomes guaranteed issue if premiums are paid and no fraud is committed. The assumption then is that initial representations were accurate.
If someone gets a policy and immediately has a big claim, he gets investigated more thoroughly. Did he see a physician and get a diagnosis that he didn’t report? Did he withhold information? If so, he has no right to expect other people to help pay for his medical expenses and the policy is cancelled.
An estimated 18 million people were covered by direct purchase health insurance in 2007, much of which is medically underwritten. In this context, the cancellation of 20,000 policies over a 5 year period is equivalent to an rate of something like 217 defects per million covered people.
For comparision, a 2006 GAO study of improper SCHIP enrollment estimated that its enrollment procedures had a 1 percent error rate, or 10,000 defects per million covered people. Other states report higher rates.
were these people denied care or did the insurnace company refuse to pay for it? There is a significant difference.
These people went to a physician, got care, and had their insurer billed. The insurer looked at the bill and said it wouldn’t pay. Furthermore, it moved to terminate the policy on the grounds that the insured misrepresented his condition at the time he bought his policy.
…But good luck finding other coverage. It may be that the patient chose lower-cost underwritten coverage in the belief that it was more secure than employer-provided group coverage. But now group coverage (or COBRA or HIPAA coverage) is no longer an option. Whoops.
I occasionally see the figure that “only” 0.5 percent of policies are rescinded. But if, say, the rescissions were concentrated in the five percent of policies with the largest claims, then people with large claims actually have a 10 percent probability of rescission. It’s impossible to know how much of a problem this really is without more transparency.
When you apply for insurance, you make your medical records available to the insurance company. If underwriting were as accurate and cost-effective as it’s often portrayed, then it should be possible to complete the process when the policy is sold or soon after. The fact that post-claim underwriting even exists is evidence of the limitations of risk-rating.
I’m not privy to the actual procedures used in underwriting, but judging from the questions asked on insurance applications, it seems to make use of statistical correlations that would be considered junk science if used elsewhere. The fact that insurance premiums jump in 25-percent increments in response to rating changes should be evidence that underwriting is an extremely blunt instrument. I’m not saying that underwriting doesn’t have its uses, but it seems to be over-applied in our system.
As for community rating as an incentive to over-provide to the healthy and under-provide to the sick, how does that compare to pure risk-rating? A healthy individual may think “My insurance is really cheap, so I might as well upgrade to the low-deductible plan with all the extras,” while a high-cost individual may look for bare-bones coverage because that’s all he can afford. How do you find the net difference between the underwritten and non-underwritten scenarios?
I think the 20,000 figure was only for the three insurers who testified before Congress. My 0.5 percent figure from the same testimony (see third paragraph from bottom): http://trailblazersblog.dallasnews.com/archives/2009/06/health-insurance-companies-tak.html
I think most insurance companies now require pre-approval for expensive procedures.
My mistake, thanks for the correction. It does look like the 20,000 figure comes from the three insurers who testified. Assume a higher recession rate of .005. This suggests that the private sector still has a lower rate of error than the .01 reported for SCHIP.
Most private insurers use experience rating to estimate claims. Yes that’s correlation, in that they figure that people with XYZ condition are likely to generate ABC in claims next year, but I’m not sure why you say that correlation is junk science. Have you got a better way to do it?
The point is that these articles are pot-boilers that allege unscrupulous conduct by insurance companies without providing much information about whether unscrupulous conduct has or has not occurred. We don’t know if these recissions are warranted or not. We do know that controlling fraud is a big problem.
Is it not possible that the companies are using a low cost method of combating fraud and that appeals processes that exist already provide sufficient protection against unjustified cancellations?
There is academic work showing that community rating does more harm than good because it raises rates for the vast majority of people who are standard risks. Other work shows that underwriting as it exists typically combines risk groups so that people in high risk groups pay less than the expected value of their claims.
While underwriting may be a blunt instrument, it is a lot less blunt than the government run price controls that masquerade as community rating.
I don’t say that correlations are junk science, just that using them in another context might be. Insurance companies are just doing what they have to do to survive in a competitive market. But from an insurer’s prospective, underwriting doesn’t have to be especially accurate or treat customers fairly, it only has to be as good as the the competitors’ methods.
No, I don’t have a better way to establish risk; that’s why I’m skeptical of the process. As I said earlier, I don’t deny that underwriting is useful, but given its limitations there should also be some way to limit the stakes involved.
I agree that we don’t know how often rescission is warranted. I don’t think the insurers are eager to share that info. But I did notice that the insurance execs in the Congressional hearing refused to limit rescission to cases of intentional fraud. It seems that post-claim underwriting is being used to do more than combat fraud. At best they’re using it as a way to cut corners in new-issue underwriting.