Health Insurance and the Net

Individuals are risk averse. We buy insurance to mitigate our personal risks. One of the most important expenses in an individual's budget, whether it be out of pocket or as an employment benefit, is health care insurance. With health care costs rising, estimates place the the total amount spent on health care in the United States to reach 20% of GDP by 2017 (See NCHC Report), the importance of health insurance is rapidly increasing.

The value of available health insurance policies may be diminishing because of activity on the net. Activities, such as net browsing, filling out health surveys, and use of internet medical advise sites, are invaluable resources for insurance companies when designing policies. If insurance companies begin to data mine the net activity of an individual before offering an insurance policy, there will be a significant effect on the type and price of policies available to individuals. Two factors play into this change. First, the insurance company is making less of a "bet" on each individual as they have far more information on each person applying for health care insurance. Second, the use of data mining could expand the insurance companies' ability to exclude specific coverage due to pre-existing conditions. Neither of these issues are simply solved, but the importance of health care makes this an issue worth investigating.

More information = more accurate actuarial assessment

Although there are many problems with the ways insurance companies operate today, they are theoretically a system to improve social welfare. Insurance companies are useful because they pool the risks of individuals so those individuals are not responsible for large immediate costs. Most people end up paying more in premiums than the insurance company pays out in claims, but some pay less. Even if insurance companies paid for all legitimate claims, they would still be designed to make a profit off the difference between premiums paid in and claims paid out. This is a socially desirable result for a risk averse populace because they pay risk premiums so as to not be subjugated to the risk of a catastrophic charge.

With the increase in information an insurance company has regarding an individual, there will be a decrease in risk pooling. Through data mining, an insurance company can learn invaluable information about its potential subscribers. It can learn how well an individual takes care of himself, how often he feels sick, any significant indicators for future health risks, and close to anything else it would possibly want to know. What this leads to is a more accurate actuarial assessment of how much the insurance company will have to pay out for each individual. While this helps low risk individuals (insurance companies will be able to offer lower rates to low risk individuals), it will increase the price of insurance for high risk individuals, likely making that insurance to expensive to be purchased. Thus, through information gathering, insurance companies will account for risks through more accurate risk assessments rather than risk pooling.

This decrease in risk pooling will drastically reduce the social welfare of the country. High risk individuals will not be able to access many necessary health care services until much too late. Emergency departments, which are places of last resort, will become primary care providers for more individuals. As this occurs, not only has the health of those individuals become significantly worse, but the rest of the country has to pay for this health care through taxes. Instead of having risk pooling where low risk individuals pay slightly more for their health insurance than it will cost in the aggregate and allowing high risk individuals to access care when they need it, low risk individuals will end up paying far greater sums to support the government's provision of medical care and high risk individuals will be in much worse health.

The expansion of pre-existing conditions

One way insurance companies limit the coverage they provide is through pre-existing condition exclusions. Insurance companies will not cover health conditions that an individual had prior to enrolling with the health insurance company. This is limited by federal and state law to be a condition for which an individual has received medical advice or treatment in the past 6-12 months, and for which the individual did not have previous insurance coverage during that period.

With an increasing amount of available medical advice in the net, this protection against pre-existing condition exclusions disappears. Often the first action a person will take when feeling a certain way will be an internet search. Does the availability of medical advice in the net make this search the "receiving of medical advice/treatment"? It is unclear at this point, but if it is, then we may see a large reduction in coverage because many of our ails will be considered pre-existing conditions.

The problems with an expansion of pre-existing condition exclusions are the same as the problems mentioned above. Those individuals who need coverage the most will either not be able to afford the policies they need, or will not be able purchase those policies due to lack of availability. Once again, the burden will be upon the state, and thus tax payers, to pay for the health care of these individuals, and these costs will be greater because the care will be applied retroactively, which is a larger sum than if preemptive care were administered.

Solutions?

We run into a problem trying to prevent these negative effects by not allowing insurance companies to perform this data mining. We encourage the freedom to learn, and a prevention of data mining directly opposes this ideal. One way to limit the consequences would be separating identifying information from these net activities. This will make the actuarial assessments for insurance companies come from pool of risk, rather than an individual's risk, and will make it so information solicited in the net cannot be used to define a pre-existing condition.

This decoupling is a difficult task. Anonymity is not easy to come by in today's net, and this is especially true when the vast majority of individuals using the net are not interested or willing to put in the required work to be anonymous. If the protection is not on the user side, it will have to come from regulation. We could require all data acquired by insurance companies to be aggregated and disconnected from identifying information before being mined. This seems like a possibility as it allows insurance companies to still take the information and make their policies more accurately account for the risks of the population, while disallowing specific information on individuals applying for insurance.

Health insurance is an important enough element of our social welfare that we must account for these considerations. Regardless of what a perfect health insurance scheme may look like, we have to examine the current situation and at least guarantee that the insurance available not deteriorate.

-- MattDavisRatner - 12 May 2009

I'm not entirely convinced by the argument for pre-existing conditions being fulfilled by internet searches. While I cannot say I have verified with all medical information sites, a couple of the larger ones I visited (WebMD? , for instance) clearly state they are NOT providing medical advice, likely to avoid liability.

Beyond that issue, though, I think you make a convincing point that using information to decrease risk pooling is ultimately harmful, even to low-risk individuals. For the proposed solution: where would the regulation come from (Congress or agency)?

-- JonathanBonilla - 23 May 2009

This is a minor point, but insurance companies aren't just betting that insurance premiums will be higher than payouts. They are also taking advantage of the ability to invest the premiums at high rates in large amounts.

-- KateVershov - 30 May 2009

• I think you need to reconsider a premise here. Most people still have employer-provided health insurance, and they buy insurance from insurers who know nothing whatever about them as individuals. Group health insurance rates for large employers (and also for very small employers, as I know being a very small employer) are related only to the numbers of employees, and are not the result of experience-rating. So the insurer can't make ratings use of any information it can buy about my employees, or IBM's employees for that matter. And for medium-sized businesses, whose plans are experience-rated, it's the experience, that makes the rate, not a prediction.

• This is related to Kate's point, which is right in asking us to look at insurers as they are, and unfortunately not quite right in thinking of health insurers as economically closely related to life and casualty insurers, who collect premiums inorder to reserve against low-probability risks with high payouts, and invest their premiums in a diversified, lower-risk portfolio. Health insurers are essentially profit-making health care managers, running an unbelievably poorly coordinated system in which they and other parties in a badly-organized market purchase too many services from providers who are not exposed to actual market incentives, and sell the too many services to a poorly-defined collection of public and private buyers who are not allowed to make decisions based on actual real-time information about costs and benefits. Everyplace this horrendous system contains inefficiencies, the insurers try to be one of the consortium of thieves that goes rent-seeking. Understanding their economics is difficult because they aren't transparent, and because the complexity of our system inhibits incremental policy design. You're on the right track to consider how changes in the cost and distribution of information affect their business, but I think you might have more success considering how their vulnerabilities might be affected, instead of their powers.