Insurance Irony

There's an old joke about banks that says that in order to get a loan you first have to prove that you don't need it. Health insurance is similar in many ways, and likely to get moreso.

At my physical exam last week, my doctor expressed his opinion that over the next decade or two we will probably find ourselves doing more genetic testing to predict predispositions for cancer and other diseases with strong hereditary links. He further opined that this naturally raises a host of issues with insurance companies and privacy, since people who have marker genes for some nasty illness will either be charged outrageous rates for insurance or else will be denied insurance altogether. Or, an alternative consequence would be that people would avoid having these tests done in order to avoid these outcomes - which seems to me to be an even worse outcome, frankly.

I think that either outcome would be a shame, of course, but I think it is also unnecessary. In fact, I think there is possibly an opportunity for a whole new business model to grow.

Let's step back for a moment and understand why people buy insurance and why insurance companies would charge higher rates (or refuse to offer coverage) to people with known, quantifiable risks. At the simplest, insurance companies are betting that you'll stay healthy while policyholders are betting that they will get sick. The word "bet" here is not accidental - it's entirely a statistical process. The insurance company makes money by charging rates that cover the expected rate of claims, plus a small margin for profit. In exchange, the value for the patient is that the financial downside potential of getting sick is capped.

This, of course, only works because of three reasons:

1. The risk for a given person gets spread among many people, including healthy people
2. Information about who will get sick (and how sick they will get) is imperfect
3. Insurance companies have good oddsmakers figuring out the right premiums to charge each policyholder based on known risks (e.g., smoker, overweight, etc.)
   

The level of predictive quality around those known risks is still pretty low, which is why it must remain a statistically-driven process. However, with genetic testing for serious hereditary diseases, of course, the insurance companies can make far better predictions about expected rates of illness.

Let's take this to its logical extreme to see why my doctor's fears are not unreasonable: suppose that the insurance companies could predict with 100% certainty the future healthcare needs of a given individual. What premium would they charge? Precisely the cost of that healthcare plus some margin (hopefully reasonable!) for profit, of course. On the other end, if a person knew that they would remain perfectly healthy until they die by getting hit by a bus, they would not bother buying any health insurance. In this world of perfect knowledge, therefore, there really is no need for an insurance product at all, since the only people who need insurance would end up having to pay more (by the profit margin) than the health care would cost out-of-pocket. Obviously, that's not good for the insurance companies, nor, frankly, is it good for customers because they have not been able to cap their downside expenses.

In other words, in a world of perfect information, insurance collapses because conditions 1 and 2 above fail. Of course, perfect information is not now nor will it ever be possible, but we will continually get closer.

So how do you prevent these fears from becoming reality as we get closer to perfect information? I think one possible answer (and the potential business opportunity to which I alluded) is by providing an anonymizing buffer to the insurance companies.

Here's how it might work. A buffer company collects customers in groups of, say, 1,000. It collects all of the information it can about the customers so that risk assessment can be accomplished as accurately and completely as possible. It then contracts with an insurance company to insure the whole 1,000 person lot.

If there are, say, 200 smokers in that group of 1,000 and 150 people who are likely to get Alzheimers, the insurance company is told these facts. However, the insurer is not told which members of the group of 1,000 are the smokers or the Alzheimers candidates, so it cannot cherrypick the healthy patients - it must cover the whole group of 1,000 or not. The premium it charges, of course, must reflect the fact that (in this example) 15% will likely get Alzheimers and 20% are strong candidates for lung cancer. However, because the group is insured as a block, this model forces the risk to be shared/spread among healthy and unhealthy, and the insurance company is forced to cover all using statistical models.

Of course, premiums should not be the same for everyone. People with discretionary risk factors such as smoking should pay higher premiums, young healthy people typically should pay lower premiums than older people who are almost certain to make greater use of health care services. The buffer company could adjust the premiums as needed to make sure that, in the example above, the smokers pay more than the non-smokers. But by anonymizing/grouping patients, it becomes possible to make sure that everyone can get insurance, to make sure that people with non-discretionary health issues (i.e., inherited) can have reasonably priced coverage, and to make sure that there is no dis-incentive to collecting the most accurate and most complete information possible.

There are a thousand business model details that would need to be worked out to make this new intermediary business work, including the obvious one of "how would it make money?" I suppose that if I were interested in acting on this, I'd be diving in to these details. But I'm not; I leave execution of this idea as an exercise for the reader. I merely offer it as a possible way to solve the "problem" of constantly improving medical information.