Strong Improvement Ideas: Eliminate the Denominator

Posted on Feb 20th, 2007 at 9:31am.

Most “Quality Dashboards” contain data on rates of hospital-acquired infections, adverse drug events, falls, and other harm events e.g. “central line infections per 1000 line hours” or “falls per 1000 bed days .” Typically, these rates are shown alongside some sort of benchmark rate for that indicator, usually established by analyzing the rates for comparable hospitals, and then displayed as the 50th, 75th, or 90th percentile. It’s not uncommon for the dashboard to display any rate better than the 50th or 75th percentile as “Green.” Expressing data as rates, with benchmarks, allows the quality staff and executive team to answer a question commonly asked by Boards: “How are we doing compared to other hospitals like ours?” Knowing how you’re doing compared to other hospitals isn’t a bad thing.

But some innovative hospitals have started to ask a different set of questions, and to use a different sort of performance indicator to answer those questions. Instead of asking “How are we doing compared to the competition?” they’re asking “How are we doing compared to the theoretical ideal?” (The theoretical ideal is often either 100% or zero).And to track the answer to that question, they’re eliminating the denominator. (For example, they are simply tracking “total number of central line infections each month” and “total # of falls each month.”)

There are five reasons why eliminating the denominators is a good idea. 1. Neither your basic patient population nor your types of service change that dramatically from month to month, (with some notable exceptions for seasonal conditions such as allergies, and for institutions with large seasonal influxes of “snowbirds.”) So a raw count of the number of people who fall in your hospital, or get infected, or have adverse drug events, is a fairly accurate indicator of the burden of harm over time. 2. Any time we make a measurement more complex (e.g by making it a ratio between two measurements) we add measurement error. How accurately are we measuring things like “ventilator days?” 3. If a measurement is not adding value (many denominators fall into this category) they’re simply adding measurement waste. Somebody has to keep track of “line hours.” Is this value-added activity, or not? 4. In order to get benchmarks, deciles and other indicators of comparative performance, we usually sent off our denominator-based measurements to some national or regional data compiler (e.g. Premier, VHA, State Hospital Association…) so that we can get them to send us back our %tile ranking and position. This inevitably introduces delay. How old are the data you show your Board? Six months? Nine months? This isn’t a timely way to oversee and steer improvement. 5. Finally, and most important, many of these denominator-based measurements lull hospital leaders into complacency, in two ways. First, the ratios make the data fairly abstract e.g. “4.9 infections/1000 line hours.” Compare this to what that abstract really means: “14 people doubled their risk of dying in our care last month, because of a line infection that we gave them.” If we want our Board members to understand our data, and to oversee its improvement with urgency, they need to understand it viscerally. Eliminating the denominators helps. The second way in which denominators cause complacency is when leaders look at their dashboards and say, “Hey, we must be pretty good. All our indicators are Green.” To which I say, “And what, exactly, does it mean to be Green?” Being better than the 50th percentile for hospital-acquired infections, in a health care system where 200,000 people incur serious harm every year from these infections, is not “Green.”

So what do I recommend? Try eliminating the denominator, for many of your performance indicators. Track the number of patients who are harmed, or receive the care they should receive , every month, against the theoretical ideal…either 100% or zero. Your data will be more accurate, more timely, and more viscerally meaningful. And that will give you a jumpstart on improvement.

Note: from time to time, you might still have to answer the question “But how are we doing compared to others?” For this you will need denominators. But if you’ve been working with the theoretical ideal in mind, you just might find something interesting when you check your performance against the competition: you’ve blown right past the benchmark!


The trouble with benchmarks and targets is the tendency for them to be translated into "good enough" Working in a busy UK hospital it is disappointing to hear staff interpretation that the maximum 4hr wait in ED is the length of time it is okay for the patients to be kept there waiting for transfer. Improving the flow of patients is compromised by this type of thinking. Add into that the 98% rule and it is culturally an uphill battle to create the atmosphere of "right patient, right place, right time". Insert instead the promise of 100% patients moved at the point of decision to admit and the issue is clearly defined and challenging.

Posted by elaine hobson (Feb 21st, 2007 10:33am)

looking at absolute instead of,or in addition to, relative numbers is an excellent idea. the most 'viscerally meaningful' truth is that for the unfortunate patient who is the '1/1000' index statistic/victim of a potentially preventable occurrence, his exposure needlessly became 100 percent. a corollary to 'putting a face' on a statistic for board/leadership motivation, would be to present to them a tiny bio/obit with each occurence, not just what they died or suffered from, but who they are/were.

Posted by buol heslin, md (Apr 13th, 2007 7:33am)

Another problem with wasting time with denominators: Even if you have a denominator and determine a rate, you still haven't risk-adjusted, so there is still the tendency to not trust the comparisons to benchmarks. Simplifying data gathering by comparing numerators from one interval to another makes sense in a number of clinical projects. Thanks.

Posted by Chuck Schauberger (Apr 13th, 2007 9:06am)

Please enter the six characters in the image above.