Statistical Process Control (SPC)

Statistical process control (SPC) is the application of various statistical tools and techniques to data acquired from a process, with the intention of process improvement. The term is related to statistical quality control (SQC), an older term that includes techniques such as acceptance sampling, which are no longer in widespread use. At the heart of statistical process control is the control chart, and its use for process understanding and improvement. All processes manifest variation in characteristics (measurements) of its output, and understanding and reducing this variation results in process improvement. Variation that is observed in a process can either be random variation inherent to the process (called common cause variation) or sporadic variation that is not part of the normal process (called special cause variation). Control charts provide the ability to distinguish between these two types of variation. More generally, control charts can be used to identify and characterize different sources of variation that act on a process. This comes about by the application of rational sampling and rational subgrouping. For example, control charts can be used to separate variation arising from the measurement process itself from inherent process variation.

Various kinds of control charts have been developed to handle different types of data and different methods of data collection. Charts designed to display data involving counts of attributes include p charts, np charts, c charts, and u charts. A p chart could be used to chart the percentage of medication orders that had one or more errors. A c chart could be used to chart the number of errors per 100 medication orders. A u chart would be used to chart the number of errors per week (where each week would have a potential different total number of orders). Charts designed to display data involving measurements of process variables include X-bar and R charts, X-bar and s charts, [Page 521]and individuals' charts. Various financial ratios like net operating margin would be an example of a measure from a financial process. Laboratory turn-around time would be an example of a measure from a medical production process. Statistical process control also includes the assessment of the ability of a process to meet requirements, often expressed as the ratio Cp or Cpk. Cp is the ratio of the allowable range of values to the range of values that the process actually yields (determined statistically). Cpk is a modified version of Cp that also measures how well the process is on target.

Another chart used for statistical process control is the cusum chart, which is used for maintaining the status quo, and is designed to detect small shifts of known magnitude. Also, the exponentially weighted moving average chart serves a similar purpose. However, these charts are more complex and harder to use, and usually fare not much better than the classic control charts when they are used with additional tests for runs.