"The Color of Health Care: Diagnosing Bias in Doctors"

Read the full article by Shankar Vedantam, The Washington Post, August 13, 2007.

"A new study by researchers at Massachusetts General Hospital and other institutions affiliated with Harvard University provides empirical evidence for the first time that when it comes to heart disease, bias is the central problem -- bias so deeply internalized that people are sincerely unaware that they hold it....

"It was only when researchers studied physicians' implicit attitudes -- by measuring how quickly they made positive or negative mental associations with blacks and whites -- that they found a mechanism to explain differences in medical judgment.

"'Physicians who had higher biases against blacks were less likely to recommend thrombolysis for blacks,' said Alexander R. Green, the study's chief investigator and a faculty member at the Disparities Solutions Center at Massachusetts General Hospital.

"Thrombolysis is a clot-busting technique given when doctors suspect that a patient is having a heart attack. It is not to be given lightly, which is why a physician's judgment is crucial in telling patients who are merely having aches and pains apart from patients at death's door."

Comment: I think the idea of unconscious bias makes a lot of sense, as does the idea that unconscious bias against minorities results in inferior treatment. Still I am not sure that the association delay method is an adequate measure of bias.

More importantly, the estimate of a posteriori probabilities of disease given the presenting evidence depend on the a priori estimates. So too, the estimation of treatment outcome probabilities should depend on both the a posteriori probabilities of alternative presenting conditions and of the probable impacts of the treatment on the condition.

Race is not a very informative variable for estimating a priori probabilities but it is one. Again, race may not be a very informative variable, but it may carry information on likely impact of treatment alternatives. A physician should use the best available information and if no better information is available and race is specified in the psychologists experimental protocol, then physicians would be likely to use it in diagnosis and prescription, albeit without conscious understanding. So it is possible that the physician behavior was rational, not racist.

But unlikely. Still, the example gave me a chance to blog on a feature of diagnosis that is relevant to this blog. JAD