At my annual physical exam last week, my primary care doctor employed a widely used web-based calculator to plug in cholesterol levels and other risk factors to estimate my likelihood of having a heart attack during the next ten years. I thought this was a neat idea until it produced an answer of 8%.
Wait, you mean I have a one in twelve risk of a heart attack over the next decade? That sounded really high. She calmly and thoughtfully explained that the main value of the algorithm was to help make a judgment about prescribing statins or other interventions that could lower risk. She also noted that anything under 10% at my age was a very good number.
So, I was going to write this post to tell this story and to make the point that these kinds of estimates can be shocking for the uninformed unless we have a context within which to interpret them.
I was also going to assert that the estimates give an impression of precision that may not be valid. What is the standard deviation around the estimate? How often is the actual estimate found to be true?
And, then, like a deus ex machina, the New York Times published this story about the very heart risk calculator that we had been using. The pertinent excerpt:
A new study finds that a widely used version of the ubiquitous heart attack risk calculator is flawed, misclassifying 15 percent of patients who would use it — almost six million Americans, of whom almost four million are inappropriately shifted into higher-risk groups that are more likely to be treated with medication.
Wow. So I revert to my doctor's excellent advice about diet, exercise, and other life style factors as the main things on which to focus over the next ten years.