Tuesday, January 19, 2010
Practice variation: Real data
That there is some variation in practice patterns among physicians, even for comparable patient populations, is inevitable. That its range is so wide is not, as often noted by Brent James. To the extent variation is not based on scientific evidence, it presents an impediment to process improvement that could reduce overuse and underuse in the delivery of medical care, or the amount of harm caused to patients. Why? Without some standardization, it is impossible to have a baseline against which to collect evidence as to the effect of proposed process improvement measures.
With help from friends at Blue Cross Blue Shield of MA, I offer an example. The issue here is the percentage of times that physicians choose to endoscopically examine and conduct a biopsy on patients with GERD, gastroesophageal reflux disease, which often presents as heartburn.
The top chart shows how the average cost per episode varies among the four quartiles of all cases. Note a variation of almost 100% in costs between the bottom and top quartiles. As noted by BCBS, the procedure cost is the single most important source of variation.
The second chart shows the variation, doctor by doctor, for use of endoscopies with biopsies. The charts shows that 74 of the 331 gastroenterologists have a significantly higher than average use of this procedure.
The question that follows is whether this degree of variation is accounted for by the variation within the patient population. That is, if one were applying standards of evidence-based medicine, would the distribution look like this? Or, is the distribution skewed by habit and predisposition of doctors? Is it influenced by a fee-for-service payment regime that encourages more procedures than are necessary? Are some doctors more fearful of malpractice suits and engaging in defensive medicine?
I often hear doctors say, when they are presented with these kinds of data, that "my patients are different," and that the data don't prove anything. But that assertion usually has no quantitative support.
BCBS is providing a valuable service in sharing these data with the hospitals in Massachusetts. The BIDMC data indicate that our doctors, like all others, vary within and across practice groups in management of the conditions at hand. We are finding this to be a useful tool in evaluating our practice patterns, both within our own practices and in comparison to others. In the face of these kinds of numbers, it is important to ask the questions.
Posted by Paul Levy at 1/19/2010 04:38:00 AM