Speaker 1: In this video, you will learn how to use Minitab in the Healthcare module to make data-driven decisions that improve patient wait times. Let's start by talking about some common key performance indicators in the healthcare industry. By reducing wait times, analyzing costs and utilization, increasing patient satisfaction, and improving patient safety, the patient experience will become a competitive advantage for your organization. How do we do this? Let's analyze patient wait times in an emergency department. As we begin investigating a process, we will use visualizations to obtain a baseline of where the process currently stands. Select Visualize Wait Time to see the tools that are suited for this baseline visualization. If you are curious about how to set up your data for this analysis or how to run this analysis, click the info link to see help information for patient wait times. Visualizations summarize data and help us better understand key characteristics. For example, what were the lowest and highest wait times? In what ranges do most of the wait times fall? To find out, let's select Visualize the Distribution of Wait Time.

Speaker 2: Specify the wait time column from your Minitab worksheet and click OK.

Speaker 1: A histogram displays a wealth of information, including many things we might not notice looking at the raw data. This graph shows that the data are right skewed with some extremely high wait times on the right side. We also see that most patients experience wait times between 85 and 100 minutes. Now let's visualize the relationship between patient wait time and other variables by choosing Visualize Relationships with Wait Time. We'll look at the relationship between average patient wait time and two categorical variables, the type of issue that brought the patient to the hospital and the facility they visited, by selecting Visualize Multiple Categorical Variables by Average Wait Time. In graph variables, we'll enter wait time, in row variables, we'll enter the facility column, and in column variables, we'll enter the condition column. A heat map lets us investigate the relationship between categorical variables and the average wait time. Here, the highest wait times correspond to the dark red rectangles. Notice that the longest wait times correspond to patients that came to the facility with respiratory issues, particularly at certain facilities. Visualizations are a great way to start investigating data, but they can only take you so far. Predictive analytic techniques allow you to consider many more variables at the same time. Using predictive analytics, you are likely to uncover insights that visualizations might miss. Let's return to the Healthcare Chooser and run some predictive analytics using predict wait time. Because this is a larger data set of observational data, it is likely the relationship between wait time and several other variables or predictors will be complex, so we will use complex relationships with multiple predictors to predict wait time. The response is wait time. The continuous variables are the numeric variables, number of physicians and number of patients in the department. The categorical predictors are the attributes, or qualitative variables, facility, condition, diagnosis code, and private insurance.

Speaker 2: Relative variable importance values range from 0% to 100%.

Speaker 1: The most important variable always has a relative importance of 100%. Notice that here the most important predictor of wait time is the patient's actual diagnosis. This is something we hadn't discovered with our visualizations. We can use the predict button to obtain predictions for a newly arriving patient's wait time. We can enter individual values, add the number of physicians, the number of patients, the facility, their condition, their diagnosis code, and whether or not they had private insurance. The resulting prediction tells us that a person or patient with that demographic could expect a 78.8 minute wait in the emergency department. To view how CART regression models work, let's interact with the CART tree. A regression tree splits the data into high and low values of the mean wait time based on business rules obtained from the predictors. Here, the overall mean wait time is 105 minutes, but for those with a specific classification of diagnosis, the mean time was quite a bit higher, specifically 247 minutes. We can also find combinations that lead to the highest wait times. Here, the worst wait times all correspond to patients with a specific diagnosis that were seen at a handful of facilities. Using this predictive model, we determined that patients with a specific diagnosis remained in the emergency room the longest. Now that we know where to look, we can dig deeper into the process associated with this type of diagnosis to find opportunities for improvement. Once we implement the process improvement, we can verify its success by determining whether the reduction in wait times for patients with this diagnosis was statistically significant. Back in the chooser, we'll return to patient wait times, choose decrease wait time, and demonstrate that average wait time decreases after a process improvement. Now, we'll enter a column of wait times and the column of sample IDs, which represent whether the time is associated with before or after the process improvement. In options, we will choose the less than hypothesis to test whether there was a decrease in wait times after the process improvement. Because the p-value here is less than .05, we can conclude that our improvement efforts were a success. Our journey from visualizations to predictive analytics resulted in a statistically significant reduction in patient wait times. Thanks for joining me today. I hope this short overview gives you some ideas for how analytics can improve your patient's experience and give you a competitive advantage.