Do Not Underestimate The Power Of…The Control Phase

By:  DM Kashmer, MD MBA MBB (@DavidKashmer)



Keep The Gains You Made!


Have you made any gains with your quality improvement project?  If you made meaningful improvements, it is every bit as important to make sure those improvements are sustained. This is the last step in our quality improvement project: the Control Phase.



Do Not Underestimate The Power Of…The Control Phase


The importance of the control phase is difficult to overstate. It is one thing to work through all the steps of the process and to make meaningful, measurable improvements. It is entirely another thing to be able to sustain those improvements as you continue to move ahead. Make sure that routine maintenance is performed so that you can sustain improvements and have something on which you can build the rest of your program or other processes.


The control phase has several important toll gates as pictured here. One of the important steps of the control phase is to transition the project to the project owners. That means that after extended quality team has gathered around the table and measured data from the trenches, it is important to make it very clear in the control phase regarding who the process owners are. The process owners will be responsible on an ongoing basis for maintenance and reporting if there becomes an issue with the end points of the project. If, on repeated measurement at different intervals, there is a new issue with the process the process owners send out a signal that there is a problem and the end points need to be revisited.



What Tool Should I Use For My Data?


From Villanova University's Lean Six Sigma Black Belt Course, 2010
Figure 1:  How To Choose A Control Chart Type.  (From Villanova University’s Lean Six Sigma Black Belt Course, 2010)



Just as important is the question of what tool to use for the control phase. There are multiple tools available for the control phase and here we will focus on one important tool that be easily utilized in healthcare. That is the ImR chart, where ImR means “individuals moving range” chart. The ImR chart is one of the multiple types of control charts that can be performed with your data. Please review Figure 1 regarding how to choose a control chart type for your data. In healthcare processes where individuals come through the system one at a time, the ImR chart works best. This chart demonstrates where individuals fall along a continuum over time. We will discuss some important highlights of the ImR chart now.


CAUTION:  There are some things the diagram above does NOT tell you.  For example, did you know that control charts generally require normal data? In other words, if your data for a certain endpoint are non-normal, you cannot apply the straightforward, continuous data driven control charts that the diagram lists. If your continuous data are non-normal, that clean Figure 1 just doesn’t apply.  For more information regarding how to determine whether your data are normally distributed look here.


Ok, now you have established whether your data are normally distributed.   If they are non-normal you will need to create a control chart based on non-normal data. We will delve into how to do this in another blog entry. For now, let’s just highlight how ImR and other control chart types are based on normal data.



Why, Here’s An ImR Chart Now!

Beneath please see a typical ImR chart (Figure 2). Notice that if you compressed all the data points to one end or the other of the control chart and remove the factor of time you would create a normal data plot. So, now to some definitions to let you know more about ImR charts and how to know whether your process is still in control or has gone out of control.


Figure 2:  An ImR Chart Example.  Time To OR (in hours)



Before we move on, look at the top plot in Figure 2.  The green bar represents the mean/median/mode of these data.  The red bars represent the 3 standard deviation mark beyond the mean.  Any data beyond three standard deviations is considered to be MORE than the routine amount of variation expected in this system…and that makes sense.  After all, the probability of a data point being more than 3 standard deviations beyond or beneath the mean in a normal data set is less than about 1%.  So, data points outside the red bars make us ask:  “What happened?” Setting the 3 standard deviation mark as what we use to identify outliers balances the risk of type 1 vs. type 2 error here.  Note that identifying cases to review this way doesn’t mean anything was done wrong in those cases by any particular doctor or healthcare provider.  We just have to ask how did the system produce such an outlier?  For more info on type 1 vs. type 2 error, look here.


Now pretend that there was some requirement (perhaps made by your state) that trauma patients needed to be brought to the OR within 2 hours of arrival for severe abdominal injuries (if they need to go at all).  What would you think about that top date in Figure 2?  Maybe draw another line at 2 hours to represent that barrier.  What do you notice?  Well, PLENTY of points on the graph show patients as being brought to the OR outside of the 2 hour mark.  Interesting!  Those points at 2.5 hours, 2.7 hours, are completely “in control” from what we told you earlier YET THEY ARE COMPLETELY UNACCEPTABLE.  Interesting, right?  This tells you that your system is performing at its routine levels, and that routine level of performance is NO GOOD.


This is why we should NOT apply control charts until the end of a quality project:  the control chart can tell us when the system is performing routinely yet lull us to sleep.  It can tell us the system is performing routinely…yet that routine may be NO GOOD!


More On How To Use The Tool


Control charts can be further evaluated with what are called the Westinghouse rules. These have been modified over time but are still a good basic starting point. Here those Westinghouse control chart rules:


1. The most recent point plots outside one of the 3-sigma control limits.

2. Two of the three most recent points plot outside and on the same side as one of the 2-sigma control limits. 

3. Four of the five most recent points plot outside and on the same side as one of the 1-sigma control limits.

4. Eight out of the last eight points plot on the same side of the center line, or target value.

5. Six points in a row increasing or decreasing.

6. Fifteen points in a row within one sigma.

7. Fourteen points in a row alternating direction.

8. Eight points in a row outside one sigma.


However, let me tell you that not everyone exactly agrees with the Westinghouse Rules, and there are different rule sets out there.  That one is pretty standard, however.  One thing we all agree on:  a point outside of the 3 sigma (3 standard deviation red bars) control limits is just not good.  Look at those cases.


Now let’s discuss the other part of the ImR chart, the range chart. The range is a measure of variance between data points. In other words this shows us how wide the swings are in our data. It is possible to have a range which is unacceptably high and which demonstrates that there is a significant amount of variance in the system that was unexpected. This can happen even when the data point itself is in control. If we see an unusual amount of variance between data points the question becomes “Why such a wide swing and why such an unusually wide swing for our data set at our institution?”





At the end of the day the ImR chart is a very useful tool in healthcare for the control phase of the project. Remember, just as important as the particular tool is the fact that we engage in a control phase. The control phase helps us receive feedback from the system when something has gone wrong, something needs maintenance, and the “weeds need trimming”.


There are many options for the particular tools you can use during the control phase for your project. Some are listed above. As usual the question of whether your data are normally distributed is central to being able to apply a control chart in a straightforward manner. Not only are there many types of control charts, but there are also some that do not require this focus on normal data and we will discuss these at another time. Best of luck with your quality improvement project.  Please leave any thoughts or comments beneath!