Caveats Regarding the Use of Control Charts

Abstract

Control charts are essential tools in the statistical process control armamentarium. Although designed for ease of use in highly structured production settings, they are in fact quite complex in nature; their proper use depends on numerous conditions that may not be clear to newer users. This article briefly contrasts the use of control charts in production and healthcare settings and highlights common sources of problems among novice users. These include misunderstandings of the type of data being charted, nonindependence of samples, the influence of natural cycles, ignoring known special influences, overlooking clear signals that a process is out of control, and calculating control limits before adequate data are available.

Les méthodes de contrôle statistique du processus permettent d’analyser et d’améliorer les soins anesthésiques

“…Statistical process control methods, and 'control charts' can be used to accomplish this. [6][7][8][9][32][33][34][35] If a new data point on the chart has a higher value than the previous one, but both points are within the 'control limits', this reflects natural variation within a stable process. Were this 'increase' to be acted upon as if the process had fundamentally changed, the analysis and action taken could be wrong (as the process probably is unchanged).…”

Bugs Are Not Funny Syndrome

“…2 I found it instructive that the problem rates in the data he used to illustrate the p chart (his Figure 3) and u chart (see top of right column on page 208) conform with the rates (actually rules of thumb) suggested in my Table l. 4 Further, as the denominator for data suitable for a « chart increases, the c chart can be shown empirically to give a good approximation to the result one would get from the u chart. Not that Dr. Benneyan encourages the use of approximate methods; he most certainly does not, even while acknowledging the similarity of results that may be found when using them.…”

“…2,3 Had it appeared before I submitted the final draft of my article, I would have included it among the sources I recommended for those interested in learning more about these important and deceptively complex tools. 4 …”

Catheter‐related bloodstream infections in hematology

“…[112][113][114][115] One way to classify these is according to the nature and expected probability distribution of the measurement data. Determining which chart is applicable in particular situations can be dependent simply on how data are collected.…”