Six Sigma and the Clinical Laboratory, Part 2

Welcome to Part II of our discussion of six-sigma and the laboratory. (Remember that sigma refers to the SD of the population (SD, as we use it, usually refers to a portion of the population – a sample.) We can think of six-sigma as a concept, an idea, a goal. “We’d like 30% of our methods to be ‘six-sigma methods.’ We can use sigma as we use SDs – to measure the distance our mean is from the TEa (or whatever we use for QC limits).




from mean
























Setting 6-sigmaas a goal gives the QC plan. It may or may not be achievable. The table below shows the increase in faulty parts as we move from a system with 6–sigma up to an undesirable method with only a 1-sigma method. The table shows defects per million opportunities, and since few of us turnout a million hemoglobins a year, we have included defects per 1000. These numbers refer in QC to the number of runs exceeding the SD from the group mean. We have illustrated how to adapted any of your methods to this approach in the examples below.

Keep in mind that the defects are not patients but runs. In other words, the number of patients affected if a control exceeds the TEa maybe be 0 or many. There is no easy way to quantify that number. If at all.

The Figure below may help you understand how reducing the size of your measured SD will increase the sigma label of a method (i.e. 3-sigma (left) vs. 6-sigma (right). Again, the smaller the method SD (sigma) the more error budget and the smaller the chance of releasing ‘bad’ data.

Figure 1

How could this concept be used in the clinical laboratory? As we have intimated in the laboratory during the analytical phase of providing the physicians with patient data is one and we will explore that now. There at least two areas where the concept is worth thinking about. Here are three examples that illustrate how you can use Excel to determine what your sigma is for any method. In these examples we have used the lab mean rather than the group mean in the math since you need to measure your distance from TEa (which will be your goal). That is to have a 6-sigma method the numbers of your SD from your mean to the TEa must be 6 SD. As you see in Example 1, the laboratory’s current SD is less than the SD for ‘6-sigma.’ It has already achieved the goal – given the bias and SD as they are now. They may change with a new lot of reagent or instrument maintenance or wear and tear. This is true for the protime also in Example 2. For Example 3 the method is a bit short. The laboratory can decide to attempt to reduce the SD (or bias or both) or accept a very good ‘sigma’ at this point.

Figure 2

As we have said, six-sigma stature is a goal. It may be possible. It may not. Let’s continue the discussion of reducing the size of the SD to achieve six-sigma. Starting with the variables that contribute to the imprecision of the method (the SD) we could list the thermistor that regulates the temperature of the instrument (it does fluctuate somewhat), the pipets for reagent(s) and sample, the tubing that moves the reagent(s) and sample, and the lamp that measures the reaction. (What can you add to this list?). Suppose we begin our search for a six-sigma method by changing the tubing. Doing so leads to shutting down the instrument for some time. Changing the tubing is not without cost. Before we venture further, we submit that these two negative aspects be considered with the possible good aspects of achieving a six-sigma method. (If the Enzyme method at this point is a five-sigma method, does the gain offset the cost?). Let’s imagine that changing the tubing reduces the SD and we have a six-sigma method. Obviously the tubing will age and we will need to change it again once the SD has increased to the point that we no longer have a six-sigma method. Or imagine that changing the tubing has no effect on the SD. What shall we try next? And so on. Again we must consider the pros and cons.

In a sense the work to a 6-sigma method is not over when the math is done. It may just beginning.

In the next (and last) of this series, we will explore another, perhaps more rewarding application of 6-sigma. We will share some books that may be of interest should you want to explore 6-sigma and its cousin LEAN.

About The Author