Thinking about linear assumptions in healthcare, thanks to Jordan Ellenberg's blogpost. Theoretically, we should have dodged that bullet of unthinking linearity. We know that blood pressure too-low is as bad as too-high (or even worse). The extremes of blood sugar are also bad. This should be easy stuff: non-linearity should be in our bones, since we (health care providers and patients both) encounter it often enough.
Why, then, when it comes to quality metrics do we seem to think that more is better? We need to get EVERYBODY MAMMOGRAMED, EVERYBODY SCREENED, or the world will end, amen.
There's got to be a sweet spot. It's gotta be non-linear. If we don't screen ANYBODY, that's a problem. People get cancer and die - we should stop that. On the other hand, if we screen EVERYBODY, that's not good either. People get false positives, get biopsies, feel terrible, and don't live any longer. We should avoid that.
The graph should look lumpy. For certain groups of people, their screening rates should be as high as possible - for others, low; and, for still others, it depends.
Now we have to operationalize that curve, get it into the hands of doctors and patients, make people understand it, and help our computers remind us to do the right things at the right time. Without skiing headlong down a linear graph and ending up in a confused heap at the bottom.
Why, then, when it comes to quality metrics do we seem to think that more is better? We need to get EVERYBODY MAMMOGRAMED, EVERYBODY SCREENED, or the world will end, amen.
There's got to be a sweet spot. It's gotta be non-linear. If we don't screen ANYBODY, that's a problem. People get cancer and die - we should stop that. On the other hand, if we screen EVERYBODY, that's not good either. People get false positives, get biopsies, feel terrible, and don't live any longer. We should avoid that.
The graph should look lumpy. For certain groups of people, their screening rates should be as high as possible - for others, low; and, for still others, it depends.
Now we have to operationalize that curve, get it into the hands of doctors and patients, make people understand it, and help our computers remind us to do the right things at the right time. Without skiing headlong down a linear graph and ending up in a confused heap at the bottom.
RSS Feed
