Sunday, August 8, 2010

Behavioral Biases in Medical Testing

As evidence continues to mount that established models of rational decision-making are dangerously out of date, behavioral science has embraced human irrationality in all of its deceptively predictable forms. At the forefront of the field is Duke University professor Dan Ariely, whose simple experiments into human bias have shed light on everything from the fallacy of supply and demand to the problem of procrastination.

In a recent interview with NPR, he turned his gaze toward the growing debate about rampant health care costs and their potentially behavioral origins. To contextualize the issue, he describes an experiment in which different groups are asked to order pizzas. One group is presented a menu where the default is an all-dressed pizza and toppings have to be taken off if they aren't wanted. Another group is presented with a menu featuring a cheese-only pizza where any preferred toppings have to be added. It turns out that those presented with the all-dressed pizza menu were more inclined to order more toppings, and those with the cheese-only menu were more likely to order fewer toppings.

If this outcome is truly representative of a natural human bias, the implications for ordering medical tests is painfully obvious and potentially quite costly. As Ariely confirms:

If you go to the hospital these days, or to visit your physician, you will see that they have these electronic order forms. And they basically use those to order tests for you. And sometimes these order forms are empty, nothing is selected for them. The default is nothing, and they have to pick what they want to order. And sometimes some tests are preselected for them.

So we created scenarios in which we described to physicians some patients who arrived at the ER, and we asked them to decide what test to give them. And to half of the physicians we gave the fully-loaded options, like the pizza. And for the other half we gave them one that were empty, and they had to check which one they wanted to do.

The basic result was that in the empty set, physicians chose an average of five tests. And in the full set, they chose an average of 13 tests....the difference was about $1,300 per patient. So now if you think about it, these information systems are going to roll out into hospitals in all kinds of ways and I think they have tremendous influence on what the physicians will decide.

The experiment didn't use actual patients but the intuition should be fairly obvious: given a natural human bias to anchor to a default selection when presented with a menu of options, more attention should be paid to what that "default" set of tests will be for a particular set of symptoms, and how that decision will ultimately impact both the quality and the costs of care.

1 comment: