Run chart
A run chart plots a measure in time order against a fixed median to look for patterns suggesting a process change.
Fit
When to use it, and when not to
Use it when
- You collect the same process measure at regular times while improving care.
- You want an early view of whether a change persists over time.
Not when
- You need a causal effect estimate. Use an interrupted time series with a credible comparison where possible. Read: Segmented regression for interrupted time series
- You need limits based on a probability model for the measure. Consider a control chart. Read: Control charts for process monitoring
- You have only a before and an after value. Collect more time points rather than calling them a run chart. Read: Uncontrolled before-and-after studies
Before you start
What you need
- The measure and time intervalKeep the numerator, denominator and collection process stable; retain the actual time order.
- A defined baseline periodChoose the baseline before judging change. State how its median is calculated.
- One published set of signal rulesState how ties and missing time points are treated. Different rule sets have different false-alarm rates.
- Each point measures the same process in the same way.How to check: Annotate changed definitions, patient mix and collection gaps.
- Serial dependence is considered when interpreting patterns.How to check: Seasonality and autocorrelation can create runs without a new intervention effect.
- Missing time points have a stated explanation.How to check: Keep gaps visible. If high-workload periods are missing more often, examine how those gaps could alter the pattern.
Plain language
How it works
Plot each value against time and draw the baseline median. Apply a prespecified rule set to runs or crossings of that line. For example, the Perla rules identify a shift from six consecutive points on one side, ignoring points on the median. Name the rule set; thresholds are not universal.
Annotate interventions and changes in data collection. An apparent shift can reflect seasonality, changed patients or missing periods. Perla and colleagues describe practical run-chart use; Anhøj and Olesen evaluate rules through simulation. Their results caution against treating every rising or falling sequence as a useful signal.
Worked example
A worked example
- Setting
- A clinic’s invented baseline weekly waits are 8, 9, 10, 11 and 12 minutes.
- Question
- Do later waits stay on one side of the baseline median?
- Approach
- Fix the baseline median at 10 minutes. Later weekly waits are 8, 7, 9, 8, 7 and 9 minutes.
- Result
- All 6 later points are under 10 minutes. Their median is 8 minutes, a decrease of 2 minutes from the baseline median.
- What it means
- Under a prespecified six-point shift rule, this is a signal to investigate. The very short baseline is illustrative; it does not establish that the clinic change caused the improvement.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
A run-chart signal suggests a change in the process, not its cause. Keep the baseline median fixed while assessing a change; repeated recentering can hide a shift.
- Moving the median after every favorable point.Instead: Keep it fixed for the assessment period. Rebaseline only after a documented sustained process change.
- Trying many rules until one signals.Instead: Choose one published rule set first and state its handling of ties.
- Calling the intervention effective because waits shifted.Instead: Investigate other changes, case mix and seasonality; a chart alone cannot identify cause.
Inside Mallard
How Mallard uses this
Mallard includes this article in Learn search and the Atlas. The plan mapper does not link to this article yet.
analysis.run_chart, the id a plan cue records.Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- The run chart: a simple analytical tool for learning from variation in healthcare processes (opens in a new tab) · BMJ Qual Saf · doi 10.1136/bmjqs.2009.037895
- Run charts revisited: a simulation study of run chart rules for detection of non-random variation in health care processes (opens in a new tab) · PLoS One · doi 10.1371/journal.pone.0113825