Statistical Process Control and Run Charts

Statistical process control (SPC) and run charts are analytic tools for understanding how a process behaves over time. They plot data in time order and apply simple rules to distinguish ordinary, expected variation (common cause) from variation that signals a real change (special cause). Because improvement is about changing processes, these time-series methods are central to judging whether a change actually made a difference.

Definition

Statistical process control is a set of time-series methods — including run charts and Shewhart control charts — that plot a process measure over time and use defined rules to distinguish ordinary common-cause variation from special-cause variation indicating a genuine change.

Scope

This topic covers run charts and Shewhart control charts, the distinction between common-cause and special-cause variation, the probability-based rules used to detect signals, and the evidence on their use in health care. It is a methodological reference and does not interpret data for any specific process.

Core questions

• What is the difference between common-cause and special-cause variation?
• How does a run chart differ from a control chart?
• What rules signal that a process has genuinely changed?
• How are these tools used to judge whether an improvement worked?

Key concepts

• Common-cause variation
• Special-cause variation
• Run chart and the median
• Shewhart (control) chart and control limits
• Shifts, trends, and runs
• Centre line and sigma limits
• Time-ordered (longitudinal) data display

Key theories

Common-cause versus special-cause variation

Shewhart's distinction holds that every process shows inherent common-cause variation, while special-cause variation signals something genuinely different; acting on common cause as if it were special (or vice versa) leads to mistaken conclusions.

Run-chart rules for detecting non-random signals

A run chart applies probability-based rules — such as shifts, trends, and runs about the median — to detect non-random patterns over time without requiring control limits, offering a simple first analytic tool for improvement.

Mechanisms

Data are plotted in time order. A run chart uses the median as a reference line and applies rules — a shift (a run of consecutive points on one side of the median), a trend (consecutive increasing or decreasing points), too few or too many runs, and an astronomical point — to flag patterns unlikely to arise by chance. A control chart adds a centre line and control limits (commonly three sigma) calculated from the data, with additional rules such as a single point beyond the limits; points inside the limits with no pattern indicate a stable process subject only to common-cause variation. Distinguishing the two kinds of variation tells a team whether an observed change is signal or noise, and therefore whether an improvement intervention had a real effect.

Clinical relevance

SPC and run charts let teams track measures such as infection rates, waiting times, or readmission rates over time and judge whether a change produced a real improvement rather than ordinary fluctuation. This entry describes the methods and the evidence about them; it is a reference, not guidance for interpreting any specific clinical dataset or for patient care.

Evidence & guidelines

Methodological accounts establish SPC's logic and chart selection for health care (Benneyan 2003) and explain run-chart rules as a simple first tool (Perla 2011), with practical guidance compiled in The Health Care Data Guide (Provost & Murray 2011). A systematic review found SPC is widely and feasibly applied in health-care improvement, though the rigor of application varies (Thor 2007).

History

Walter Shewhart introduced the control chart and the common-cause/special-cause distinction in manufacturing in 1931 (Shewhart 1931), and the ideas spread through Deming's work on quality. Health care adopted SPC from the 1990s onward, with methodological papers adapting chart selection and rules to clinical data (Benneyan 2003) and run charts promoted as an accessible entry point (Perla 2011).

Debates

Run charts versus control charts as the starting tool

Some argue the simpler run chart, requiring only a median and a few rules, is the right first analytic tool for most improvement teams, reserving control charts for when control limits add value; the choice depends on data type, volume, and the team's analytic capacity.

Key figures

• Walter Shewhart
• James Benneyan
• Robert Lloyd
• Lloyd Provost
• Rocco Perla

Seminal works

• shewhart-1931
• benneyan-2003
• perla-2011

Frequently asked questions

What is the difference between a run chart and a control chart?

A run chart plots data over time against the median and uses simple probability rules to spot non-random patterns; a control chart adds a calculated centre line and control limits, allowing additional rules such as flagging points beyond the limits.

Why does the common-cause versus special-cause distinction matter?

It tells a team whether an observed change is ordinary process noise (common cause) or a genuine signal (special cause); confusing the two leads to reacting to noise or missing real change, both of which undermine improvement decisions.