I have previously blogged about this topic, here. Now, a more formal version has been published, first as an online version, here. Due to copyright rules, I cannot post this version (for a year). It is a cleaned up and somewhat expanded version of the blog version.
The point of this paper is not so much to criticize Six Sigma, but more on how it is used (or misused). That is, certain proponents of six sigma use it as a sole measure of quality – which in clinical laboratory terms, means a sole measure of potential patient harm. One is reminded of the manufacturer customer dilemma. Thus, a manufacturer, must either accept or reject product, often based on a single limit whereby accepted product is equated with full value and rejected product is scrap. But to a customer, product which is just meeting or just not meeting specification is similar in value – hence the manufacturer’s model, while necessary, doesn’t transfer well to the customer. The manufacturer has a dichotomous model and the customer has a quadratic model.
Now manufacturers can’t use a quadratic worth model to make decisions – it’s impractical, but they can do better than a single acceptance limit. Hence, having two sets of acceptance limits, one to contain a high proportion (such as 95%) of results and the other to contain few or no results – helps to have a more customer oriented specification and correct specifications are an ideal starting point for most projects.
I had proposed using a process capability metric (Cpm) to evaluate laboratory assays (1) but here, I had also proposed that in addition to calculating Cpm, one must estimate the proportion of values that fall outside of medically acceptable limits.
1. Assay Development and Evaluation: A Manufacturer’s Perspective. Jan S. Krouwer, AACC Press, Washington DC, 2002, pp 96-101.