Video Transcript: Run Charts (Part 2)

Bob Lloyd, PhD, Executive Director Performance Improvement, Institute for Healthcare Improvement

Now that we have the basic components of a run chart, let’s think about how we’d use this. Let me tell you a personal story. When I first started working in the quality improvement area years ago, I was living in Chicago, and I commuted 35 miles from my home to the hospital where I worked. So, I would be hung up in traffic for what seemed to be an inordinate amount of time. So, what did I do? I said, “Let’s start plotting my commute time.”

So, I’m going to do it by day, Monday, Tuesday, Wednesday, etc. ― not counting weekends, just workdays. On the y axis, I’m going to plot the number of minutes it takes me to commute, zero to… ? I put an upper range of 120 minutes to give myself some time in Chicago; there could be some hang-ups ― traffic, weather, etc.

So I start plotting. Then I realized that I needed to make two charts: one for my morning commute, and then one for the evening commute. At first I thought they were probably the same, but then I realized after I plotted the data that they were quite different. So let’s look at the morning commute.

I start plotting my data by Monday, Tuesday, and I plot the chart, and it’s going along, and we’ve got some variation, and I put my center line, which, as we know, is the median (x with a tilde above it). And lo and behold, I found that my average ― center line, the median time ― was about 48 minutes. It was about 35 miles I had to travel.

Now, what dawned on me after I did this chart was basically that it was random variation. It was stable, it was predictable; some days it would be up, some days it would be down, but overall it was up and down, and there were no violations of any of the rules (not a shift, not a trend, too few variations, no astronomical data points ― just up and down, back-and-forth). And what quickly realized is that, while there was about an average of 48, that I was actually spending a lot less time commuting than I thought. It was actually more of a mental issue: I’m stuck in traffic, and you think time is going very slowly. Not until you plot it though do you actually see what’s happening.

Then I stated to think about how I could actually improve this process. So what did I do? I started coming up with my Model for Improvement, and started thinking about what changes could I make? So here I have the y axis and my x axis and my minutes and my days along the x axis.

So now I came up with this brilliant idea, I’ve got my average just bouncing along here at about 48 minutes, and right here is where I decided to make a change. I thought, after looking at a map, that I could take a different route, and this looked like a less congested route than following the major freeways in Chicago. So I extend my center line, which is my average of about 48 minutes, the median, and now I’m going to see what happens to my commute after I’ve made my change. I’ve actually instituted a PDSA test right here.

I thought this was a great idea. I plot the data, and what do I discover? That, actually, now I have a shift in my process because I had more than six data points above the median, and it was actually now taking me longer, and if I computed the average, the center line, for this new commute time, it actually took me about 55 minutes on the average. It was a failed PDSA.

So what do I do? I don’t do that route anymore. I go back to my baseline, and say “Are there any other changes I can make?” Extend my center line, which is my baseline data, and now I’m going to see, “Can I come up with new ideas that may, in fact, give me what I was looking for ― a lower commute time?” and I would know that, again, by the rules.

I did discover some interesting things, that you run into special events. So I’m commuting (and, again, looking at the AM commute, which is up and down, back-and-Forth), and on the average about 48 minutes, but then one day — wonk! — I had this spike. It literally took me two-and-a-half hours to get home. And this turned out to be a rapid snowstorm that hit on Valentine’s Day, and Chicago was just stuck. It was one of those where you call home and say, “I have no idea when I’m going to get home”; we had a special cause, a snowstorm.

On the other hand, there was a time I had to go in on a holiday to have a special meeting, and, instead of my regular time, what we saw was a drop. When I annotated — which is a key thing to do, you annotate your charts, put a note on it — so here I wrote “snowstorm,” and here I wrote “holiday” because I went in on a holiday when there was far less traffic on the road. Now these are not normal parts of the commute. I saw on average, my commute was still about 48 minutes, but I was able to annotate and demonstrate why these were anomalies and not part of the normal variation.

So, pick something! Whether you run, cycle, golf; commute time; apply it to a work process. Start to collect some data, plot it on a chart over time, and start applying these simple run chart rules to understand the variation that lives in your process.