Is my quality improvement project actually working?
Paste your numbers in order. This draws the run chart and applies the four rules that distinguish a real change from ordinary variation.
Before and after is not evidence
The most common quality improvement error is comparing a mean before an intervention with a mean after it. Both numbers vary week to week on their own. Comparing two points from a wobbling line tells you almost nothing, and it will show an improvement roughly half the time even when nothing was done.
A run chart fixes this by plotting every measurement in time order against the median of the whole series, then asking whether the pattern is one that random variation could plausibly produce. Four rules do that work. A shift is six or more consecutive points on the same side of the median. A trend is five or more consecutive points all rising or all falling. Too few or too many runs means the line crosses the median less or more often than chance would predict. And an astronomical point is one obviously outside the pattern — a judgement call, which is why it is not automated here.
Points that land exactly on the median are excluded from the counting. They cannot be on either side, and leaving them in inflates the run lengths.
You need enough data. Below ten points the rules have very little power, and with fewer than that you will neither detect a real improvement nor be reassured by a null result. Fifteen to twenty points, collected at whatever interval is natural for your process, is the usual advice.
Frequently asked questions
Is a run chart the same as a control chart?
No. A run chart uses the median and the runs rules and needs no assumption about the distribution. A control chart adds calculated limits based on the variation in the data and detects more kinds of signal, but it needs more points and more care to construct. Start with a run chart.
Should the median be recalculated after the change?
For detecting whether a change happened, use the median of the baseline data and extend it forward. Recalculating across the whole series, as this tool does by default, is the right choice when you are simply looking at a process over time and have not yet intervened at a known point.
How often should I measure?
Often enough to get ten to twenty points in the life of the project, and at an interval that matches how the process actually varies. Weekly is common for ward-level work. Monthly sampling often means the project ends before the chart can say anything.
A rule is broken. Does that prove my intervention worked?
It shows the variation is unlikely to be random. It does not establish that your intervention caused it, since something else may have changed at the same time. That is why the annotation on a QI chart, recording what was done and when, matters as much as the points.
The book behind this tool
The Clinical Audit & Quality Improvement Project Workbook
From a problem you noticed on the ward to a signed-off, portfolio-ready project.