For all its absurdity, the debate over Obamacare has accomplished something positive: It has educated people that insurance is really about risk pooling — as in you need both healthy and sick people to participate if it’s going to be affordable for the sick.
Some believe that universal government health coverage is the only way to guarantee such risk-sharing. They will be all the more right in the age of big data.
Even if one understands the importance of risk pooling, it’s not easy to transform into policy. As Harvard economist Gregory Mankiw notes, there are a lot of obstacles: Sick people are much more likely to seek insurance, and have better information about their health than insurers do; coverage encourages people to use more health care, and makes containing costs harder.
Yet one of the potentially biggest problems is also among the least recognized: The use of big data to separate the healthy from the sick. Algorithms and artificial intelligence are very good at identifying patterns of behavior that predict certain outcomes. This, in turn, can be used to sort people by the expected cost of their health care.
All kinds of incidentally collected data — shopping history, public records, demographic data — can be repurposed for assessing people’s health. For example, LexisNexis offers a product called a “Socioeconomic Health Score” that seeks to predict costs using such information as education, criminal records and personal finances.
Meanwhile, corporate wellness programs and all kinds of startups — backed by a steady stream of venture-capital money — are working to collect more information on doctor-patient interactions as well as ongoing surveillance of patients themselves. Some estimate that healthcare data is growing by 48 percent a year. The more you have, the better your predictive analytics work.
One promise is that artificial intelligence systems, with access to more data than any one doctor can possibly handle, will be able to offer the most appropriate advice or test to a patient with given symptoms or health status. As a bonus, doctors will be able to see many more patients because their time with each patient will be spent more efficiently.
Just imagine, though, what insurance companies will do with the ability to better predict people’s health care costs. If the law allows, they will increase prices for the riskiest customers to the point where they can’t afford it and drop out, leaving lots of relatively healthy people paying more than they’re expected to cost. This is fine from the perspective of the insurer, but it defeats the risk-pooling purpose of insurance. And in a world of increasingly good predictive tools, it will get progressively worse.
In the context of a free market, the big data healthcare conundrum is this: The better we get at predicting and treating, say, diabetes, the better private insurance companies will get at charging people for their future diabetes care. The asymptotic limit is a subpopulation of poor, sick people who cannot afford insurance and live off Medicaid, and a bunch of healthy people who are well covered but don’t need insurance. As Republicans have learned in recent weeks, high-risk pools are incredibly expensive.
Obamacare took a step toward addressing the conundrum when it forbade insurance companies to charge more for pre-existing conditions. But that’s not enough, because it doesn’t deal with the oncoming tide of “expected health conditions.”
So what to do? One approach would be to handcuff insurance companies in their use of big data technologies, preventing them from using predictive algorithms to assess risks — or even from collecting the data in the first place. If you think that’s utterly inconceivable, I agree.
The other option is universal health insurance, meaning that everyone would be covered by something akin to Medicare. It wouldn’t solve all the problems that Mankiw mentions — particularly the question of which expensive treatments to allow — but it would at least maintain the kind of risk-sharing that insurance was meant to achieve.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Was this article valuable?
Here are more articles you may enjoy.