With all the modules that we've just gone through for analysis methods, just want to tie all of those topics into just a case study. So, there is a little bit of deeper understanding of how some of those come into play in an actual case. This was a system, a process project that was several years ago. And as you can see, this is in the A3 format that was described in one of the earlier courses. This is a one-page summary of what went on in a project that took many, many weeks, and had a great deal of effort, but we just have encapsulated into one. So, we'll just kind of walk through the very high level of what went on in this process. This is normothermia. Of course, when you have any patient having normal temperature, you don't want a fever, you don't want them to go into hypothermia as well. But there were just a couple of cases that one of our surgeons found that several of their patients coming into intensive care after their operation, it seemed that the body temperature was lower than expected. So, 36 degrees C, if you get below that, you tend to worry. Now, the physician immediately called together a group, and wanted to see whether this was a common cause where periodically we have patients running into lower body temperatures, or was this a special cause, something that happened just recently, just uniquely, so that we could get some idea of how big the problem was. Is this a global thing that's been happening for a long time, or is this more isolated to situations more recently? If you look at the small control chart on the left hand side, that is where we were characterizing the process. We were just trying to understand exactly what was going on. As you can see, there were a couple of points that were exceeding those control limits that showed us with a little bit of relief. It's that you don't want any patients ever to go outside of those, but at least, it was contained to the most recent set of cases that hasn't been an ongoing problem putting patients at risk that we just had not seen for a long period of time. Then there was a great deal of work that used several of the tools that will go through to analyze and understand the processes and the factors that led to that particular situation. There were patients, neonatal patients showing up in the recovery room after surgery with lower body temperatures. Trying to recover from surgery, and then trying to fight body temperature would put them at, of course, higher risk. Now, on the right-hand side of the A3 shows some of the interventions, and then in the lower right, we will then go through the control chart again for after those interventions have taken place. But let's go through some of the analysis tools that we used on this project. Remember, there are descriptive and inferential statistics. Descriptive are just giving an idea of what is is without trying to look for the interactions, and the causes, and the trends immediately, things like mean, median, and mode. Later on, after you have taken that first snapshot, now you can go into inferential statistics, trying to look at any of the factors, and trends in your process so you can model them into the future to see what future outcomes and behaviors are going to look like. Here is the numerical descriptive statistics, is just the patient temperature. Pure and simple, what is the average temperature of the patients, and how wide is the disbursement of those values. Also can give you the high and the low, the median value as well, but those are just the numbers. Now in descriptive statistics, you can have not only the numerical but the graphical. Here is the patient temperature data, but now in the histogram form. So now, you can see how far away those outliers are. It's easier for the human brain, as most brains, to just see with that picture a better understanding of what the behaviors are like. Still, we're in descriptive statistics, just showing what is is as the baseline. Inferential analysis is now when you can look for those trends. But in this particular case, we were looking at the temperature of the patients is our outcome, the result after their surgery in post-operative. So, between those two columns, the outcomes variables are attributes, that data is variables, because it's time-based, the same way as height, weight. Time is variables. And so, now we are in that left-hand column of tools that we can use. Now, if you look at the inputs and the factors in this particular project, we really didn't have that many attributes, that many categories because they were all, one at a time, very unique type of surgery, but the things that were common that we could do for analysis would be what is the pre-operative temperature? That's a variable's data. What about the age of the patient? Again, variable's data. What about the different values for height and weight in temperature, anything that characterized the patient? All variables data. So, in this particular case, by knowing the type of data that we had for the factors and the outcomes, we really were isolated, or within that variable to variable upper left-hand quadrant of the tools available to us. Here is what the first data chart was, here's the control chart. And in that chart, you've got the center line in the upper lower control limits, and also in there, the dark blue line is the 36 degrees C. So, that's the normal temperature. You didn't want to see any of those patients going below. But if they're fairly close, it might be an actual event, or it could even be a little measurement system problem, it may not be fully accurate. Some of us, myself included, when you weigh yourself on a bathroom scale, you might try a couple of different times trying to look for the best weight. There might be some flexibility, or some slop in that measurement system. So, if your data item's like these temperatures, they might be close, but a little bit out, and there might be further investigation into the measurement system itself. The two red dots on the far right-hand side that are far below not only the blue line, but also those control limits, now that shows that, that was special cause indication that we should investigate more deeply. But that is inferential statistics because the null hypothesis is nothing unique and nothing is going on, but also it has that graphical representation that has the power to go along with the data analysis. Scatter plot is another one of those variables to variables type of tools. In this case, we were looking at if we're concerned with that outcome temperature, what were any of the other temps on that particular patient, on that particular day, with that particular surgery? How much did it change, or was there some sort of correlation? So, if I had the patient's temperature when they began their operation, or when they completed it, or the temperature when they were in the ICU before they even went into surgery, and when they came back to the ICU afterwards, those would still be variable the variable, and could be looked at with the scatter plot. Now, in this case, if you look at that particular grouping, there is null hypothesis of nothing is going on, and even though you might be able to draw some sort of picture out of it, there really isn't a nice linear relationship between those two. The null hypothesis in this case said, there is no statistical correlation between those two temperatures. Again, this is not only inferential, but also the graphical helps in that discussion. And here's one more combination of inferential statistics and graphical. This is when we are looking at the differences between those end-temps, and the post-op-temperature for each patient. So, if I look at the differences, the null hypothesis would be, from the time they finished their surgery and the time they get back into the ICU, it would be zero if not very close. Now, seen graphically and with a statistical analysis, the P-value shows that there is a violation of that null. In this case, the P-value would be very, very low, that would then show that there is something unique, something special going on. By following through each one of these tests in our project, we ended up identifying several items. First of all, there was a measurement system problem, where there wasn't consistent ways to take temperatures. If you think about doing a forehead swipe, or doing a thermometer, or the temperature of the patient when they're in the surgery, those three different methods might provide some sort of variation in temperatures taken. Also, we found that there were transfers and transportation for the patient in very cold situations, and some days of the year. And then if they are not properly warmed, or the bassinet, or the bed that they are transported in, if that is not under power then is not being warmed while transport. So, distance also could be a factor. Each one of those were then taken into account, were addressed by the team. And in the upper right-hand corner, each one of those outcomes from each one of the hypothesis tests, and each of the analyses would then have something that was addressed item by item. If you look at the lower right-hand corner, it has the same control chart with more data added to it. So, showed that after those interventions were put into place, not only did they seem like the right thing to do, but we had data that statistically proved that the process had been fixed, had been improved. We didn't have any more hypothermic patients coming out of the OR. So now, you could use the baseline data to not only characterize it, but then analyze the factors that were necessary to adjust or improve the process. And then those same tools could be used at the end of the project to show that those interventions did, in fact, fulfill the expectations when they were put into place.
