When it comes to clinical quality measures, “exclusions” are a surprisingly tricky concept. The basic idea is simple - some patients should be excluded from a quality measure. Think adult weight screening measure and pregnant patients or the breast cancer screening measure and patients’ who have had bilateral mastectomies. But as with everything quality measure related, when we dig into the details of how this gets implemented, things quickly get complicated.
Take a look at the breast cancer screening eCQM definition:
These specifications can look a bit daunting, but this one essentially boils down to finding all appropriately aged women who had a qualifying encounter and determining if they had a mammogram. And, of course, don’t look for mammograms for women who’ve had mastectomies. Technically, when we calculate this measure, we’re coming up with three separate numbers - the denominator population (DP), the exclusion population (EP) and the numerator population (NP). For the sake of a simple example, let’s pretend we got the following:
- Denominator Population - 110 women of appropriate age and with an encounter
- Numerator Population - 75 women who had a mammogram
- Exclusion Population - 10 women who had a mastectomy
The quality measure performance rate here is 75%, calculated by doing 75/(110-10), so the equation is NP/(DP-EP). Now let me ask you this, what’s the “denominator” here? Is it 110 or 100? This is where we need to distinguish between the mathematical denominator and the Denominator Population. This distinction is something I’ve always found frustrating, and I think the eCQM’s would be better off renaming what they call the “denominator” in their specifications because the mathematical denominator is a different number. In Azara DRVS we always display the mathematical denominator, so in our example above you would see 75, 100 and 10 in the numerator, denominator and exclusions columns in our scorecards.
One thing that’s often overlooked is that a patient cannot be in both the Exclusion Population and the Numerator Population, the math just wouldn’t work out, and you could end up with a negative compliance rate. So, if a patient qualifies for both the numerator and exclusion criteria, we need to pick one. According to the eCQM logic guidance, the exclusion criteria always wins, so if a patient had a mammogram a year ago, but ended up having a bilateral mastectomy a few months ago, they would be in the Exclusion Population, not the Numerator Population.
This is where “exceptions” come in to play. Some measures define an exception population which functions just like the Exclusion Population, except priority goes to the numerator. A good example of this is the flu shot measure which defines an exception population for patients who have refused their flu shot. If they refused a flu shot at one visit, but later received their flu shot at a different visit, they would still be included in the numerator. The idea here is that patients who refuse a flu shot are “freebies” who cannot hurt your performance rate but can help it if they also happen to be compliant.
This distinction between exclusions and exceptions is clearly very subtle, and only applicable to the CMS eCQM’s; the difference doesn’t exist in most other measure specifications like HEDIS, UDS and more. In Azara DRVS, we are always trying to simplify things so that you get straight to the numbers you care about without getting caught up in these nuances, so we present both the Exclusion and exception Population counts rolled up into a single column called “exclusions”.
Personally, I find some of these nuances of the CMS eCQM’s well-intentioned, but ultimately too complicated. I think the measures would be much easier for users to understand if they were specified regarding “numerator” and “denominator' because a performance rate is simply a numerator divided by a denominator. While you do gain some extra information explicitly reporting an Exclusion and exception population (as well as the “initial patient population” which wasn’t even covered in this post), I think the extra reporting detail comes at a heavy cost of increased confusion and, ultimately, lower confidence in the quality measurement. At Azara, we’re constantly trying to unravel these technical complexities while presenting a simple and intuitive interface, and exclusions are just one example of that battle.
I welcome your thoughts…