Some areas of AAO.org are temporarily unavailable. We apologize for the inconvenience and are working to restore access.

  • Opinion

    Musings About the Middle: Too Many Regressions?

    Download PDF

    Richard P. Mills, MD, MPH

    By Richard P. Mills, MD, MPH, Chief Medical Editor, EyeNet


    What is it about nature that causes the inexorable tendency of objects to head for the middle? The bed in my college dormitory was a perfect example. Swaybacked to the extreme, the sagging mattress made it impossible to lie anywhere but in the middle. Herds of wildebeests seem consumed with being in the middle of the herd. Crowded, yes, but very safe from lions. In politics, if candidates can get past their party’s primary, they head straight for the middle of the political road. I contend there is something comforting about being in the middle that is a natural attractant.

    For individuals, being in the middle carries an additional, much less appealing connotation: You are average, or close to it. This isn’t bad if you’re a wildebeest or a politician, but if you’re an ophthalmologist, it’s not so great. None of us likes to think of ourselves as average; in fact, like the children of Lake Wobegon, we are all above average—and are always searching for validation of this status. In cognizance of that phenomenon, I recommend that the various doctor-rating schemes that are spreading in the marketplace like crabgrass be on a scale of 50 to 100 (omitting scores 0 to 50). That way, every ophthalmologist could appear to be above average. I know it will never come to pass; some naysayer will point out the fallacy of my 50 to 100 scale, but it’s fun to daydream about.

    Another phenomenon relevant to being average is the statistical fact of regression to the mean. The easiest way to understand this concept is by example. Let’s assume you want to run a clinical study on a glaucoma medication, and you select people whose intraocular pressure is 24 or above. These people, however, have mean IOPs over multiple readings of 22. So you picked them on a day when their IOP was unusually high. Naturally, after treatment is started, on subsequent visits the IOP is likely to be lower (regressed toward the mean), which might lead to the incorrect conclusion that the lower IOP is an effect of the medication.

    Might these musings be relevant to some of the new requirements doctors are facing? Consider, for example, quality reporting. PQRS is now mandated for ophthalmologists to avoid being penalized in future Medicare payment years. Reporting is supposed to occur on measures appropriate to a subspecialty. Presumably, as everyone gets accustomed to the system, the standards for quality will gradually be tightened, requiring a minimum percentage of patients who “passed” each measure. As these processes play out, ophthalmologists will adjust their electronic health records and other office procedures to be sure their performance meets the new minimum standard. The scramble to meet the minimum (taking care not to waste resources by getting too far above the minimum) will create the new phenomenon of “regression to the minimum.” Ophthalmologists will cluster on the passing side of the minimum, which is presumably better than where they had started out. So what will happen when the emphasis shifts, old measures are abandoned, and new ones implemented? Will the gains be sustained, or since the old measures are no longer under scrutiny, will they slide back into the old behaviors? I don’t think we know the answer to that question, and won’t until we’ve implemented quality reporting more fully. Meanwhile, I’m going to call the possibility “Regression to the Way It Was.”