No measure found - zone,

Using data for improvement

Improvements happen over time. Determining if an improvement has occurred, and if it is lasting, requires observing patterns in data over time (Source: IHI).

Plotting measurements over time turns out, in my view, to be one of the most powerful things we have for systemic learning.
– Dr. Donald Berwick, 2004, Escape Fire: Designs for the future of healthcare

The real world challenge is that everything varies to some extent. So, we need a tool to help determine when results are varying naturally and when meaningful changes have occurred. Statistical Process Control (SPC) is that tool. SPC charts (i.e., control charts) can help decision-makers understand variation and make informed decisions to take the most appropriate action.

Understanding variation

Variation can be separated into two types: (1) expected or ‘common cause’ (e.g., day-to-day discrepancies in patient volumes, patient acuity, etc.), and (2) unusual or ‘special cause’ (e.g., under-staffing, an influx of many patients due to a nearby emergency department temporarily closing, a new improvement initiative, etc.).

Expected variation is natural to the system and outside of your control. It is normal, so investigating and taking action on expected variation is not helpful and can be wasteful. On the other hand, instances of unusual variation, which result from things outside the system that can be influenced – are worth investigating further and potentially taking action on to deliver or sustain improvements.

You may have heard that SPC is complicated – the practical application of it is not. For most healthcare decision-makers, understanding how to react to data is the most important thing.

Questions to reflect on

If the chart shows ONLY expected variation:

  • Is the centre reference line (the mean) result acceptable? If not, you  could consider implementing changes to move the system toward a more desirable mean result.
  • Is the range of expected variation (between upper and lower limits) acceptable? If not, you could focus on understanding why this is and then look to minimize the variation and make your system more predictable.

If the chart shows ANY unusual variation:

  • What may have caused the results for this period of time to be different than the results at other times?
  • Is there an action that could be taken to resolve or eliminate a cause of deteriorating results?
  • Is there an action that could be taken to learn from, sustain, or spread an improvement?

Rules for detecting unusual variation

Unusual variation is found by applying five rules based on probability.

1 – A sudden change has occurred: 1 point outside the control limits
2 – A change has occurred: 2 out of 3 points are near the control limits (in the outer one-third of the chart)
4 – A shift has occurred: 8 or more points in a row above (or below) the centerline
5 – A trend has occurred: 6 or more points in a row increasing or decreasing
6 – Reduced variation has occurred: 15 or more points in a row close to the centerline (in the inner one-third of the chart)

View our FAQ page in the Resources section below for more details about these rules and control limits.


One of the biggest challenges in improving quality of healthcare is reducing variation.

Uncontrolled variation is the enemy of quality.
– W. Edwards Deming

To reduce unusual variation, you need to identify the causes of these changes and then take action to address them specifically. To reduce expected variation, you need to fundamentally change the process that is producing the expected variation you’re seeing. In general, unusual variation should be addressed before applying an improvement strategy to impact expected variation.