How an EDW Enables the Best Healthcare Visualizations
There’s a lot of truth to the adage that a picture is worth a thousand words. In the world of healthcare analytics, we would say that a good graph is worth a thousand data points. Visualizations allow the human brain to consume data very quickly. A visualization can encapsulate thousands of records in such a way that they can instantly be understood.
Previously, we discussed front end business intelligence (BI) tools and their strengths and weaknesses. We mentioned that the core strength of these tools is their ability to visualize data to represent a snapshot of information captured at a particular point in time. What’s more, they create charts and graphs that can be broken into multiple data stratification.
Now, we’ll share a few powerful types of visualizations a BI tool—running on a healthcare enterprise data warehouse (EDW)—can create and how they can help an organization measure their quality improvement efforts.
Statistical Process Control Charts for Healthcare
Statistical process control charts are very meaningful for tracking performance in a sophisticated, statistically sound way. They are one of the best tools for showing process improvement. A control chart is essentially a line graph that shows a trend over time—but it also includes control limits. These control limits help you determine whether the variation you see among your data points is normal, acceptable variation or whether you should be concerned about it.
Here’s an example of how a control chart works. Let’s say your goal is to increase patient satisfaction by reducing the time it takes to admit a patient. First, of course, you have to determine how you’re going to streamline your process to decrease admit time. Then, you set up a control chart to track whether your intervention is effectively reducing this time.
The principal elements of your control chart would be:
- The vertical axis, which represents the individual values you observe—in this case, the time it takes to admit a patient
- The horizontal axis, which represents time—in this case, the months since you began your intervention
- The centerline, which indicates your mean performance level. This mean is calculated from the baseline data. In our example, it would be the mean time it took to admit a patient before your process-improvement intervention.
- An upper control limit, which is typically three standard deviations above the mean
- A lower control limit, typically three standard deviations below the mean
Since you want to reduce the time it takes to admit a patient, you hope all of your data points moving forward will fall below the mean. Some, however, will inevitably fall above the mean.
What the control limits show you is whether you need to be concerned about the data that falls above the mean, and whether you have a generally stable and predictable process. If the data falls above the mean but still below the upper control limit, you can usually ignore it as an example of the kind of acceptable variation that occurs in any process. If data falls above the upper control limit, you know you need to identify and address the problem that occurred in that instance. If the data falls below the lower control limit, it might represent a case where you want to repeat the action throughout the process given the time saved. Regardless, it’s worth studying to find out why.
Can you imagine trying to make sense of all that data without a visualization? Even more, can you imagine trying to get your admissions staff to make sense of all that data without a visualization? You want your staff to be actively engaged in looking at the data to measure their own progress, and visualizations help them digest the information.
If your intervention is successful, the control chart will display a downward trend over time across multiple data samples. Whether or not this time reduction improves patient satisfaction is another measure altogether—and would require tapping into a different data source and using a different visualization. The possibility for new visualizations never ends!
Scenario Analysis Visualizations for Healthcare
Another example of a powerful visualization tool is one that helps you perform scenario analysis. This means that you not only can see a metric, but you can use a variety of filters to choose what levels of information you see. For example, your visualization might feature age filters that help you determine how your intervention drives different outcomes for patients in different age groups. It’s important to note that scenario analysis is more than just standard filtering—it’s granular filtering that can adjust different variables in your data.
Here is an example of a straightforward scenario analysis visualization we built for a hypothetical client seeking to reduce the rate of unnecessary C-sections in their hospital. In this simulation, the client wanted to be able to see how different cervical dilation values, as one example of granular filtering available, correlated to the C-section rate. The visualization tool allowed this client to filter the data based on the dilation value—from 1 to 10—to see how the C-section rate changed for different dilation values. The client was also able to filter on gestational age, induction method or whether the newborn was “Marked as Delivered” as some additional criteria used in the scenario analysis. One OBGYN at the hospital is using this tool to analyze how much opportunity a mother is given to have a natural labor before the C-section occurs. From that point, the OBGYN can delve into the data to answer the question: If we didn’t let the mother progress to vaginal birth, why not?
Empowering Clinicians and Staff with Healthcare Visualizations
As we mentioned earlier, one of the greatest opportunities of having these visualizations is getting them into the hands of frontline staff to engage them with the data and empower them to do their jobs better. Our clients, nurses, doctors and program heads are running these visualizations and using them to make decisions. To be widely effective, analysis capability should be rolled out to the masses, not just to the analysts assigned to compile a report for those who are making the clinical decisions. Empower the people making the decisions to see the data firsthand, do scenario analysis and come up with new answers—and even new questions.
Here is a very simple example of how a large health system successfully made a visualization tool available to frontline staff to help drive improvement. The health system was trying to reduce its 30- and 90-day readmission rates for heart failure patients. One of the interventions they put in place was making sure each patient had a follow-up appointment scheduled within 48 to 72 hours before being discharged from the hospital. One of the hospital’s first steps was getting a baseline of how often individual staff members were scheduling these appointments. Most fell below the 50 percent range, and some were lower than 20 percent. They also discovered that some of these low numbers resulted from staff documenting the appointment data in text fields rather than entering it as discrete data.
To help drive improvement, the hospital set up a simple incentive program: whoever had the best stats at the end of the contest would receive a small prize. Importantly, leaders made the improvement data available from a workgroup-developed dashboard which was then reviewed at a weekly meeting so they could track their own statistics. Over the course of that intervention, the hospital saw significant improvement from everyone involved—people whose percentages had been down in the 20s and 30s rose all the way up to the 70s and 80s. A bit of healthy competition and the ability to visualize their progress drove their behavior.
This is just a small example of the power of running visualization tools on top of a healthcare enterprise data warehouse (EDW).
Read about how business intelligence tools drive smarter decision-making. If you don’t have a visualization tool yet, learn about how to evaluate a business intelligence tool before you buy.
What sort of visualization tools have you implemented at your organization? Are they running on top of a robust EDW? What success stories can you share with us?