This post is part 5 in our series on run charts. This one addresses the run chart rule for outliers. It defines an outlier – subjectively. The video then summarizes this five part series on run charts.
(Duration = 5 min. 0 sec.)
This is part 5 in our series on run charts. In part 1, we learned about run charts in general and why they are so helpful in seeing how process performance changes over time and how to determine if the variation we see in the performance is just common, normal everyday variation or if the variation we see indicates that something different, something special is going on that indicates process performance has fundamentally changed. We learned that run chart rules can be used to help us detect those statistical signals of process change. In part 2, we learned about the run chart rule for detecting trends. In part 3, we learned about detecting a shift and in part 4, we learned about the run chart rule for runs. In this video, we will learn the one remaining run chart rule – for outliers – and then briefly summarize this series on run charts.
The first three run chart rules we looked at were based on the statistical probability that the variation we were assessing was due to common versus special cause variation. This last one for outliers is subjective. It applies to the data points that are so much higher or lower than the others that it’s obvious to most anyone that there is something unusual going on. The data point sticks out like a sore thumb. Every set of data will have a highest and lowest value, but those are not necessarily outliers. Let’s take a look at an example.
We’re looking at some more of the patient contact to 1st 12 lead ECG time data. We see that the centerline is around 14 and the data points are generally falling between around 5 and 25, except for that one point that is much much higher than the others. It’s up around 38. That’s an outlier. Sometimes an outlier might be so far away from the others that you have to question if the data is even valid. With a data error, you might see a value for the patient contact to 1st 12 lead time interval of over 3 days or 3 months or even 3 years.
So, outliers are the extreme high or low values that, again, stick out like a sore thumb. It’s just that simple.
So there you have them – all four run chart rules for detecting special cause variation. Trends, shifts, runs and outliers. Special cause variation is the thing you watch out for. If you are not intentionally changing the process, a statistical signal for special cause variation strongly suggests that something unusual – good or bad – is going on and may warrant investigation. If you are doing a QI project and are changing the process, you look for special cause variation as evidence that your change is actually making a difference in how the process performs.
To summarize, when looking at performance data for a particular process, the run chart is a tool we can use to assess the behavior of that process over a period of time and on an on-going basis. Because processes have variation, the operational question we will have when looking at the data over time is if the variation we see suggests that the behavior of the process is fundamentally changing -something we call special cause variation, or if the variation is just random ‘noise’ that is inherent to the process – something we call common cause variation. The run chart rules can help us tell the difference and thereby give us abetter interpretation of the data, and that leads to better decisions about how to respond. It helps keep from both over-reacting and under-reacting. We want to react correctly and make -appropriate- data-driven decisions.
Notes and Resources
- An excellent summary article on use of run charts and application of run chart rules in healthcare – Perla RJ, Provost LP, Murray SK: The run chart: a simple analytical tool for learning from variation in healthcare processes. British Medical Journal – Quality and Safety. 2011;20:46. e51. doi:10.1136/bmjqs.2009.037895. https://goo.gl/MaEcCL
- Posted October 2018