A physician I know in Merced County, California, has been practicing medicine for the greater part of three decades, and he was recently obliged by his medical group to transition from paper to electronic medical records. The medical-records software he uses mandates that he indicate on a pull-down menu the ethnicity of each of his patients. He views his patients’ ethnicity as non-essential to his practice, and so – pressed for time like most health-care providers - he clicks on “Albanian” for all of his patients, not because they are Albanian, but because “Albanian” is the very first item on the pull-down menu and therefore the fastest to access. The result is that over 90 percent of his patients appear to be Albanian, when in fact they represent the usual constituents of Merced County, with few if any Albanians among them.
It's a funny anecdote that also has some important implications for health policy. If researchers rely on data like this, it's not going to provide an accurate representation of which patients doctors are actually treating.