Al Franken Explains Selection Effects

Al Franken is quickly becoming my hero. Below is a clip (h/t Effect Measure) where he not only makes a very prescient case for health care reform and the impact on "medical bankruptcies," but also shows that he is unwilling to be bullied, knows what selection effects are, and is able to explain them so that they can be understood in such a way that he uses them to make the case for health care reform.

I can admit that I was a little bit weary of Al Franken running for the Senate. I used to listen to his radio show on Air America and thought that he was a little arrogant and showed a lack of nuance in some of the opinions that he expressed on that show. I was also worried that he didn't have the discipline to stay on message and not make needlessly disparaging comments that would come back and haunt him later. But, he handled himself with incredible grace during the recount and subsequent court battles in Minnesota and has shown an incredible acuity for both politics and policies in the Senate.

Medical bankruptcies are bankruptcies that are related to medical or health-related issues. A 2005 study showed that medical bankruptcies affect between 1.9–2.2 million Americans in 2001 alone. As if filing for bankruptcy wasn't bad enough, the authors found that many of those who filed for bankruptcy had problems finding a future apartment, job, or cars because of their credit unworthiness. Sen. Franken points out that there are zero of these bankruptcies in France, Germany, and Switzerland.

Then, when the right-wing shill asks Sen. Franken if he knows the international disparities in cancer outcomes, he already has an answer that every intro stats class should watch for an explanation of selection effects. If, because we only treat people socioeconomically advantaged enough to not have serious complications for cancer, of course we would expect that our cancer outcomes would be much better than those who treat difficult cases as well. In other words, the design of the study selected on the dependent variable — likelihood to survive cancer treatment — to receive treatment; it is therefore, unsurprising when cancer patients here survive more.

As is becoming frequent of late around here, I congratulate Senator Franken for his outstanding representation of this country.

The Unconscionable Math of Insurance Fraud by Insurance Companies

E. and I were driving home from a family function last week and caught the end of the This American Life episode on "fine print." Among the stories was a heart-wrenching tale about the problem of rescission, which is basically where insurance companies take premiums for years and, right when you need an expensive procedure, drop your insurance because you "lied" on your application. This really affects people who get their insurance on the private market. During the episode (described in more detail by James Kwak at Baseline Scenario), the health insurance companies being grilled during congressional testimony argued that rescission affects about 0.5% of their clients.

But, in a truly amazing post, Taunter explains how conditional probability can explain why this 0.5% can sound so small but amount to an unconscionable fraud on the part of insurance agencies. Using the Monty Hall Problem (and, implicitly, Bayesian reasoning), he explains why the insurance executives essentially bank on rescinding one in two claimants for those in need of service! He points out that you have a three times better chance to survive a game of Russian Roulette. It is an amazing lesson in conditional probability and Bayes Theorem that could be an incredible teaching moment.

Until the patients needs an expensive procedure, patients are paying their premiums into the coffers of the insurance company and the pockets of their executives. The expensive procedures, of course, are why people need insurance. The insurance companies, though, get to keep all those years of premiums for all of those years even though they don't follow through on their commitment to pay out, getting rich off those years and years of premiums. The procedure goes something like this: take money from a group of people and promise them a return in the future, use the money to pay off claims of others (minus a heavy for the service), and when it comes time to pay out, renege. Taunter points out that this seems vaguely Madoffian.

The justification of this procedure is that, without it, insurance prices would rise for everyone in the system and that people should be punished for fraud. Of course, many of these people did not commit any sort of fraud and instead got confused by the leagalese and convoluted medical language which were intentionally designed to get people to slip up. It is probably true that insurance premiums would probably go up, but it doesn't really matter if people aren't getting a service that they pay for (in other contexts, that's called stealing). But, it is also an argument for why the "free market" does not always know best and why health care reform is so important. I hope that in the coming days and weeks, the Obama administration and Congress get their footing and actually start to stand up for reform.

Jamming about Stats

I have a hard time imagining anything that could be better for the Pragmatic Idealist household. E.'s favorite band jamming about statistics! No, I'm not kidding. Really, I'm not. Check it out (h/t Mark Blumenthal):