Variation in treatment of patients has attracted a lot of research interest. Some of the variation is geographic but much is between physicians practicing in the same market. Without knowing exact patient characteristics, it is often hard to even assess whether there is true variation in treatment of patients with supposedly the same disease or condition. Several reasons have been given for variation in practice patterns, including training differences, local peer pressure and fear of malpractice. A new study in one of the Journal of the American Medical Association publications examines the possibility that physician preferences for what kind of patients they treat may be at least a partial explanation for observed variation. (JAMA Article) What is really being proposed under this theory is that the patients being compared for treatment patterns aren’t identical, that they vary in some way that is relevant to the specific doctor who ends up treating them. So what appears to be practice variation is actually just a difference in patients that justifies different treatment approaches. The authors used EHR data from one large academic medical center to exam patient selection effects in the emergency room context. Adult patient visits to the ER between January 1, 2010, and May 31, 2015, were studied. Substantial information was gathered about patient characteristics and prior medical treatment. About 62 physicians and 294,000 ED visits were included in the analysis. An obvious limitation is the use of one facility with a small number of doctors.
On the assumption that there was some freedom among ER doctors to select which patients they saw, the analysis examined whether certain doctors saw more patients with certain characteristics. The researchers also performed a cost analysis to see if the characteristics of the patients selected affected the physician’s relative apparent intensity of practice and spending. Certain physicians tended to select older patients on average. Some selected patients with higher or lower comorbidity or health status scores. Some selected more patients with certain conditions, such as car accidents, bleeding or chest pain. There appeared to be no selection effect on sex of patient. There appeared to be an association among selection for age, comorbidities and urgency of need. These variations in selection did not appear on the overnight shifts, when typically only one doctor was on call and selection opportunity was limited. Looking at spending, if you ranked doctors without regard to the selection effects, and then did so accounting for selection, the rankings changed substantially. Over half of the doctors moved up or down at least 25 percentiles, with very signficant shuffling among quintiles. Interestingly, patient preference by doctors also appeared to change over the course of a shift, with doctors selecting less complicated cases near the end of a shift. The explanation for this is obvious, the physicians don’t want to get hung up with a patient that extends their shift time. As to the other selection effects, it may be due to doctor interest, but also likely has a financial explanation, with some physicians looking for patients on whom they might be able to bill more. Since most variation research already tends to adjust for patient health status, I am not sure how this new research relates to that existing body of studies. It may be that using EHR data allows for greater discernment of differences in patients. But I don’t think it would surprise anyone to see that certain physicians may prefer to see more or less complex patients.