Control Charts

If we don't monitor our processes, the variation mentioned earlier will creep in beyond that which we would normally expect and throw these processes out of control. When a process is not well controlled, making improvements becomes complex. In this case, there are just too many different ways of accomplishing a goal, and trying to implement one method of doing so will be very difficult. For change to be successful, it is necessary to ensure that the process being changed is being done in the same manner by everyone involved. It is only then that the outcome will be predictable and have the effect that is desired.

Control charting is an excellent method of identifying variation within the process. These charts use mathematical calculations to identify whether or not the variation is normal or abnormal. When this variation is identified and addressed, the process is ready for change. If day and night shifts are practicing differently, what process should be changed? If, on one occasion, the ECG was misplaced causing a longer than normal door to ECG time, should the entire process be changed? It would be better to remove that one data point and recalculate the average. Only if you repeatedly identified a lost ECG machine would you show a process in need of change.

The following illustration is an example of a control chart. Note that the control chart displays the average, 16.5 minutes, and an upper and lower control limit. Any data points that fall around the average in a "non-random" manner, or that fall outside of the upper and lower control limits, could indicate a process that is not well controlled. For instance, it would not be normal for one data point to be below the line and the next above the line, repeating in this fashion for a number of times. If these non-random data points represented different shifts, it could mean that day shifts and night shifts had two different ways of doing something. Data points that fall in a "random" manner are considered normal variation, or that amount of variation that will normally occur if a process is in control. Note in the example that there is one data point outside of the control limits. As previously stated, in addition to looking at non-random patterns, there is a need to look at those data points that fall outside the upper and lower control limit. Statistically speaking, one would expect almost all data points to occur within plus or minus three standard deviations from the average (the upper and lower control limits, or UCL/LCL). This data point deserves a second look as it is outside of those parameters indicating that some variation crept in and caused this out of control behavior.

One important point to remember regarding the selection of a control chart to identify variation is that processes change over time. In order to see what a certain process is doing presently, you need to get as close to the process as possible and have at least 15 data points before constructing the chart. For instance, you would not choose 15 ECG times done over the past six months. In the course of six months, the process of obtaining an ECG may have changed many times. Staffing differences, age of the equipment, etc., all could produce variations. It would be better to select the last 15 ECGs that were done. This sample is closer to the process you need to understand, the one that must be addressed. Control charts that span long periods of time can certainly show improvement in a process but do little to identify variation that is currently occurring. If you are doing 36 primary PCIs for STEMI each year, that would amount to only three per month. Control charting over the year will display trends up or down with a recalculation of the average but may not be as effective in showing current process variation given the long span of time in which the data were collected.