The Myth of the Mean                                         Joseph C. Mello                                     joe.mello@davita.com                                    Chief Operating Officer                                             DaVita Inc.

Over the past number of months I have been in a quandary about something that I would love your opinions about. That is, “the myth of the mean”. I’ve been calling this out more and more lately with my direct reports. In essence: I believe that the mean/median/mode has only minimal power in true evidence based management. Increasingly the use of “average” data never gets to the essence of what is happening with any discrete population that drives performance. Let me give you but one simple example.

In measuring clinical performance for DaVita we pay attention to lots of indicators. One of them is “average hematocrit level”. We’ve been patting ourselves on the back for moving this indicator higher over the past 3 years. I am certain that is a good thing. However, in any given month when that number goes up, I have no window on whether we really improved the aggregate population, we got the patients that were already high higher, or we got the low patients up --which of course would be the best thing.

I could provide you with many other examples that are relevant to managing a portfolio of anything: reducing turnover is a good thing, but how many of our centers have zero turnover? How much of the reduced “average” turnover comes from the worst performing centers going from 110% to 80%? We’ve reduced our nursing vacancies from 548 to 392 in the past 12 months. If you’re at a center that has 2 nurse openings you don’t really care about that statistic. I’m not saying that averages don’t matter. I just don’t think they are as useful for managers to make decisions/take actions as other indicators. As you know, we at DaVita are relentless at only measuring stuff that we believe will require a management action at the other end.

I thought at first that “the myth of the mean” idea was only relevant for looking at the big picture. Stuff at my level of the organization. So I decided to test it a bit. I visited a few of the better Facility Administrators (our dialysis center managers) and asked them how they drove improvements in clinical indicators. They, of course, said that they focused on the bottom performing patients and worked on improving them first. So, what value can I add as a senior management schmuck by looking at the mean? Not much. I can say “improve those averages”. Or… I can say “look at these 22 centers that are the outliers”.

So…while we have intuitively always done a bunch of managing at the tails of the distribution I am focusing much more on outliers. Systematically measuring the “less than” and “greater than” populations. Looking at data on the best performers and the worst performers. Gaining insights from the best and sending SWAT teams to the worst. This seems to be working and I can almost generalize it to any population. About four years ago, we coined a moniker here that we refer to as “B-52’s”. At the time we had about 500 dialysis centers so really we were looking at the bottom 10% performers. Now that we are 1,300 centers, we still look at the bottom 10% (and we still like the B-52 label). Driving performance in this population drives aggregate performance up and if all else stays constant, the mean will improve as well. I no longer trust the rest of the population, so we now use averages or means very infrequently in our metrics, particularly our in process metrics.

Food for thought…

 

Posted on February 19, 2007

 

Comments on this column?  You are welcome to create a post on the EBM Blog.