Video Transcript: Static vs. Dynamic Data Bob Lloyd, PhD, Executive Director Performance Improvement, Institute for Healthcare Improvement Now, a key aspect of your improvement efforts is being able to take data and analyze it and be able to understand the variation that lives in your data. But you really have two choices. Many people have been acquainted with static displays of data. What we want to talk about is the difference between static and dynamic. If you’ve had any course in basic statistics, you know that there are measures of central tendency: the mean, the median, and the mode. And there are measures of dispersion, characterized by the min, the max, the range, which is the difference between the min and the max, the absolute difference, and finally the standard deviation. The thing is that these are all statistics, trying to capture some static aspect of a distribution of data. The problem is that often times, summary statistics that are static in nature, don’t give you the appropriate picture of the variation that lives in your data. For example, let’s say that you had two time periods. Time one, and time two, and you had an average of something – a time one and time two, it doesn’t really matter at this point. What do you conclude if I showed you in a meeting these two dots? You’d conclude that basically time one is the same as time two. But then if I plotted data over time and showed you this is year one, and this is year two, and I plotted the twelve months of each year, and you saw that year one actually looked like this, and year two actually looked like this, you’d get a very different picture. Obviously, movement downward, movement upward, yet the means are the same. This happens every time you have static displays of data – you have to ask yourself, does the statistic, mean, median, mode, standard deviation, really reflect the variation that lives in the data? Here’s another example: let’s say that you have an average mean of something, let’s say that its seventy-six. How could you get that average of seventy-six? Well, you could have data that looked like this, you could have data that looked like this, you could have data that looked like this and you’d get an average of seventy-six. Like this, or data that just randomly array themselves. Which one do you want? If you were looking at patient satisfaction you might say, “Oh, I like this one, it’s going up.” If you were looking at reducing infections, you’d say, “Oh, give me this one.” If you were trying to show that your intervention worked and you were here, and then an intervention was here, you might say, “Oh I like this one because we had infections up here, we made our change, and then they went here.” But the point is that they all produce exactly the same average. I have data sets that demonstrate that all four of these different arrays, or five depending on how many you draw, the max, the min, the standard deviation, could all be the same. And yet they display very, very different patterns of data. Let me show you one final example, because if you’ve been in health care for any length of time you’ve probably seen this happen. “Thanks for coming to the quality meeting today, I have great news. Here is time one bar graph, and time two bar graph. Average of seventy minutes wait time, average of thirty-five minutes wait time, and here ladies and gentleman is where we made our intervention. Look at the big difference between time one and time two.” Now the question becomes, how did the three units that were part of this initiative fair from time one to time two? And you say “Oh, well I can show you”, here is unit one, here is unit two, here is unit three. Now, each one has time one measure, time two. Here’s where the intervention occurred, between time one and time two. Here’s the data for unit one – it was here and then it was here so here’s the average of seventy, average of thirty-five. Unit two’s data looked like this – average of seventy, average of thirty-five. And finally, unit three, its data looked like this – it went like this but now it has gone like this. Average of seventy, here average of thirty-five. Only one of these units actually can be characterized by these two little bar graphs in relationship to the intervention here. Which one is it? Unit one, two, or three? I often get people saying unit two because they see what some would call oh, a downward trend, I could put a regression line on that and call that a downward trend. Well in fact it is getting lower, but the question is did the intervention that we did right here have an effect on the process and its variation? The answer is no. It was getting better, and it continued getting better, but there was no response to the intervention. This unit was up here, it did go downward, but now what’s it doing? It’s heading in the wrong direction. Unit one was here, it shifted and went here after the intervention. It is Unit one that is the only unit that actually can be characterized by static displays of data. It’s not until you lay data out over time that you actually are able to understand the true variation in data. So we ask you to think about, what are the choices you’re making when you look at data? Do you look at data from a static view or from a dynamic view? Hopefully your quality improvement efforts will be using more of a dynamic display of data that plots it over time.