[MUSIC] Welcome to Module Seven: Data, Variation and Control charts. Why is it so important in quality improvement to examine data and variation and then use control charts? Well, the limitation of variation is really the goal of quality improvement. It's so important for us to have data, to look at that data over time, and to understand where the variation exists, in order for us to implement improvement projects to reduce that variation. A second reason why data and variation are so important, is that when we look at data using control charts or run charts-- data over time-- we're able to see trends. We're able to determine natural variation versus variation that is there because of a special effect or because of our innovation or because something historically has happened. And the third reason for why data is so important in quality improvement is because it provides ways to determine if change is actually an improvement. Without looking at data over time, and looking at variation, it's very difficult to really understand our processes and very difficult to understand if the changes that we're making are really leading to successful improvement. So we have chapter seven, then, to guide us. Chapter seven covers measurement using run charts and statistical process control charts, really to gain this insight into our systems and into our quality improvement projects. The two objectives to this chapter are really to look at data analysis and interpretation using the run charts and statistical process control charts. These are the tools of quality improvement for data. And the second objective is to look at differences between common cause and then special cause variation. What does typical data give you? So, when we look at data using typical ways, we are getting a snapshot in time where we're looking at the average or the median. Also, we look at the range between the pre and the post and then we usually calculate a standard deviation. The limitation to typical data and typical, or usual statistics are that you cannot really look at trends. You really can't look at the whole improvement picture, and it's very difficult to understand if really a change was made. So the key is, in quality improvement that is, to look at measurement over time. The examination of data graphically over time is provided by the use of two of the tools of quality improvement, the control chart and the run chart. Again, this run chart and control chart is specific to looking at data over time and its purpose is to look at trends in your data. Identifying the specific type of variation then using your data in the process will lead us to understanding common cause versus special cause variation. Historically, where did control charts come from? We have to go back to business again and Dr. Shewhart in the Bell Labs. He was the originator of looking at data over time using these methods and most of the original work on statistical process controls was authored by Dr. Shewhart. Deming, also in industry, again, took the control chart and continued the work to demonstrate that it was imperative when you were doing PDSA cycles in the cyclical matter, that it's the only way to capture change. Now I'd like to just look over the difference between common cause variation and special cause. For common cause variation, this is variation that's inherent in every process. So in everything that we do, there's a little bit of variation that occurs. Usually this variation is random, and it's due to regular, natural or ordinary causes. And with common cause variation, it produces processes that are said to be in control or stable. We usually work on processes then, in order to make these processes stable. Now an example of common cause variation would be when you were driving to work each day. And say for example you were interested in reducing your time to getting to work. And so each day you calculated the amount of time it took you, from leaving your garage to getting to work. Now common cause variation would show you that with every day, that value would change. So with each day over time, the amount of time it took you to get to work would just change naturally, in a regular, ordinary way. We would say then your drive to work would be stable or in control. Now let's look at the characteristics of special cause variation. Special cause variation is secondary to irregular or unnatural causes that are not inherent in the process. So when special cause variation is present, the process is usually termed out of control or unstable. We need to then look at what the special cause variation is and address these and remove these from the process in order to reduce the special cause. Going back to the traffic example, say one day there's an accident on your way to work. This accident would be seen as causing your time from leaving to arriving at work to be very much increased. You would identify this amount of time on your run chart as a special cause variation, that on that day, there was this accident that was special cause-- unnatural, not related to your process-- that caused the delay for you getting into work. Now let's look at the specific tools that are used in quality improvement in order to look at your variation and look at your data. And this is the run chart, and as you can see, on the x-axis, or the horizontal axis, this is the chronological advancement of time. So it can be usually seen as days or specifically, the date of each of the data points. On the y axis, then, is the incidence of the phenomenon that you're measuring. In the traffic example, this would be the time you're going to work. So, it would be incremental on the y-axis. Specific in a run chart is the line that is horizontal in the graph itself, and this is the average. In a run chart, the average is seen as the median between the points in the data. For control charts, the specific components of the control chart are similar to those in the run chart. On the y-axis is what you're studying, and on the x axis is time. In the middle on the control chart, then, is the average, or the mean, and the addition in a control chart are the upper control limits and the lower control limits. These are actually statistically determined and are usually about three standard deviations from the mean. When you use a control chart, it's a little bit more accurate for you to understand where the special cause variation is occurring. The control chart, because it is based on statistic is, a more accurate tool for you to use. When you look at run and control charts, the real reason to use them is to distinguish between common cause variation and special cause variation. Again, you're looking at the difference between natural causes, which are common, and the special cause variation, which is something that's occurring that's causing the process to shift. The rules are based on probability, that certain patterns would not occur by chance alone. So special cause is not necessarily positive or negative. You really have to understand what you're studying, and then interpret what the special cause specifically means. It can indicate that a change was effective, or a special cause can also indicate that a change was not effective. Let's look specifically then on what the run chart interpretation rules are. How will you know if the data is reflecting common variation, or if it's reflecting special cause variation? When you're making this determination, you really need about 14 points of data. The rules then, or the interpretation rules for run charts are first, that if you have too few or too many runs in your chart. So you may need to first understand what a run is, before you can understand this interpretation rule. Well a run is 7 points in a row that are either increasing or decreasing. When you're trying to interpret a run chart for looking if there are too few or are too many runs, you need to refer to a run chart table. In the run chart table, it takes into consideration how many points you have in your run chart, and then how many runs are seen as normal, and then it will tell you how many runs are seen as abnormal. In the normal range for runs this indicates that it's common cause variation, that you would expect these certain runs to be occurring over time. However, if there are too many runs in your run chart, then this would indicate special cause variation. That for some reason, there are too many things occurring in that that are really not probable. The second interpretation rule is seven or more consecutive points on one side of the median. So if you look at your run chart, and if you have the median as your middle line, then if you see seven or more consecutive points in a line, on the top of the chart, then this would be seen as some special cause variation. Statistically, it's not probable that you would have these seven consecutive points above the median. This indicate that something is going on in your process. If you look at the traffic example, you may see that there were seven accidents going into work. So that's not a probable situation, that's special cause. And then the last interpretation rule for a run chart is that there's seven consecutive points increase or decreasing no matter where in the run chart. This one is not specific to the median, but just indicates that the likelihood of seven consecutive points increasing or decreasing in time are really a special cause variation. And you need to really try to understand why that might be occurring in your processes. Now lets look at some of the interpretation rules for control charts. A real simple interpretation rule is the first, and that's when there's one or more points outside of the upper or the lower control limit. These are sometimes referred to as astronomical points, where the likelihood of this occurring is very low and therefore, it's special cause variation. You need to be really observant of when these points occur. And think about what's happening at that point in time for you to interpret what could be causing that special cause variation. The astronomical point, or this special cause variation, occurs by chance alone one in 200 times, or 0.5%. So again, all of these rules are statistically based. The second interpretation rule for a control chart is that if there's a run or eight or more successive points on the same sign of the center line. Now this statistically is equivalent to a coin toss coming up with heads eight times in a row. So the probabilities is about one out of 256 times. You may notice that this was also one of the interpretation rules in the run chart. And so, there are similarities between the run and control chart. But as I mentioned in the beginning, the control chart is a more specific, a more accurate way of looking at your data. If you were to see that there were eight or more successive points on the same side of the center line, using this control chart, you would conclude that there's some special cause variation going on here. Again, to interpret this, you would look at that time frame where this is occurring and identify what is happening in the process that may be causing this special cause variation. Another interpretation rule for control charts is seven or more points in a row steadily increasing or decreasing. This is also seen as a trend. Now, each point must be consecutive increasing or decreasing. If there is one aberration to that, then it won't count. Again, this is special cause variation. What's happening in your process that would lead to this increasing or decreasing effect? The next interpretation role for control charts is 14 or more consecutive points alternating up and down, such as a saw tooth pattern. Again, it must adhere to this rule where there are points up and down. If there's any aberration in that pattern, then it does not count as a special cause variation. If there were not 14 points in consecutive points alternating up and down, it would be seen just as common cause variation, that we expect the process to vary over time. However if there's this special sawtooth pattern that you see, you know then that there's a statistical significance in that this would not be occurring by chance. So it's time for you again to take the data, look at this time frame and identify what's happening in the process during this time frame that would cause this process to have a special cause. It's going to be giving you cues to understand what's happening in the process, so that you can have a better impact on effecting the process. The reason for the statistical process control charts, and the run charts, is because in-quality improvement. It's important for us to look at data over time. To really emphasis this point, I would just like to just demonstrate the difference between static data, and then this dynamic data. And as you can see on the left, this is an example of static data. Where you're looking at the data average before your intervention and then in the middle is your intervention, and then after your intervention. Unfortunately in static data interpretation, you can conclude things that sometimes aren't exactly the truth. For example, if we were to ask whether this intervention was effective or not, looking at the static data, we could run a T-test or some traditional statistics and determine that there is significance in that there is a reduction over time between pre and post. But if we look at the dynamic data then, in the table or the figure next to it on the right, we can see that prior to this intervention even being implemented. There was already a decline in the incidence. And that it may not actually have been the intervention that caused this change but that there probably was a trend down, or seven or more consecutive points in the line, to indicate that there was some type of special cause variation that occurred even before the intervention was implemented. As a good quality improvement expert what you would do with that data, in the dynamic form, is you'd look at that time period where that special cause variation occurred prior to the intervention being implemented. And you would try to understand what was occurring there, so that you could extend whatever was happening in the process there, into the future. This is why it's so important to keep track of data over time. And why the tools of quality improvement for data variation are so integral to our quality improvement projects. Here's another example of looking at data in a dynamic fashion. And again, you can see that the chart can be annotated, or you can also demonstrate different interventions that are being implemented at different times. This is another advantage using dynamic data, charts, and the control charts, and the run charts, because you can actually identify which intervention is having a bigger effect. Again, this would be the static way of looking at the data, and it would be really impossible to understand which intervention was having an effect when you're looking at data in the pre and post mode. You've understood now that the interpretation rules are a way to evaluate the effectiveness of the change or changes that you're making. And maintaining the old process control limits and the center line, really don't recalculate with new points being initiated in this control chart. Basically if we look at this specific control chart and we understand that there was one change that was made, and we see that there is now a special cause variation, that this intervention was effective and is causing a reduction and an improvement in our piece of work. We may want to think about creating another type of control chart where we can split the control chart so that we can look at the process before the change and then after the change. By using a different control chart, we will be able to look at special cause variation after the change was made. Here's an example of the control chart that has been split, where we look at the process prior to the intervention and then after the intervention. After the intervention, the control limits are recalculated and the center line is recalculated because there's enough data points after the intervention was implemented in order to examine the new process. Now note that in the split control chart, that the new process has both a lower mean and less process variation and this indicates that there's really a tighter control in the process and this is a really good sign. The advantage of splitting the control chart then is that we can then institute a new intervention using the new control chart upper and lower control limits. And we are able to then identify if we can even cause more special cause variation, more improvement in our process in order to really get the best results we can get. Going back to then the prior slide, this is our process with the intervention being implemented where we're looking at the data over time. And when we do not split the control chart, it's really hard to see where the intervention then is becoming the standard process. Again, when we split the control chart, we're able to find, and identify, and observe, how this new process now becomes standard practice and also, you can see that the variation becomes reduced. You're able to then also identify where new interventions will have an impact. I know data is a hard subject for many front line health care professionals, and it's usually something we really just avoid. But the use of control charts and run charts is essential for all of your quality improvement efforts. It helps you to become a leader. It helps you to communicate the processes that you are trying to improve. It helps you to show your administration that what you're doing is having an effect. In this chapter, we looked at data analysis and interpretation using these two tools, run chart and control charts. And we tried to emphasize what the difference is between common cause and special cause variation. Now we're going to move to application of using run charts and control charts. The application is not a video, but the application will move you through specific case studies and then will have you create run charts and control charts so that you can gain competency and confidence in using these very helpful and important tools for your quality improvement work. [MUSIC]
