In this video, we will cover a real life example of a visualisation. This graph or this visualisation shows the performance of one quality measure at a large health system. You can see that there are different colors, different types of graphs, and different graphs that are representing different attributions and filters, on the right. We will dissect each piece within this visualization, to explain how they all fit together into one. Before I get started, I wanted to share with you that we've noticed when audiences or the users look at the dashboards or these visualizations, they start from the top left and up at the bottom right. So you'll see the direction here, the flow of your essentially your site, from the top left, going to the right, and it comes down to the bottom left and goes back to the right. So we should try to use this flow to design our visualizations. Now, let's take a look at each piece individually. Use of colors and labels. So this visualization was on the left side of the entire visualization, and we can notice that there are titles and column headings for each region, as well as the year that's representing the data, and we use conditional formatting to color the cells. In this case the jreens and the reds indicate whether the metric, the 10 metrics that are listed on the left, are meeting the target goals. If it's above, it's green, or if it's below, it's red. So we can use conditional formatting to easily visualize, what metric is doing well or not within a given region. This lists all the metrics, the 10 metrics in one page, which is easy to see rather than having 10 different pages, and it shows the criteria for green and red colors as we just discussed, and it displays actual results in addition to color, so you can see the labels within each color to look at the percent score compared to the target. The next piece is the trend graph. Let's talk about the difference between a trend graph and a cross sectional graph. In this graph, what we're seeing is performance over time, and we can see that for example, the first metric at the top, decreased and then it's coming back up. So that tells us, "Oh, this is what's happening over time." However, when we look at the previous visualization, where we have the number, the percentage, as well as the color coded based on whether it's meeting target or not, is actually a cross sectional data point, where we're looking at 2017 and 2018 data to say, overall throughout those years, this metric is performing above or below for each of the regions. So that's the difference between cross-sectional versus trending. Both are quite helpful, depending on the need. And it's useful to track performance over time, because you want to know if you're making progress over time, and many times, we also use the trend graphs to compare pre and post intervention periods. In healthcare, we do a lot of interventions to make improvements, and it's very important to know whether we are making progress based on the intervention or not. So trends are quite helpful, as well as cross sections for those studies. The next piece is ranking, you'll see that, these two visualizations are essentially doing the same thing but with different attributions. Attributions can be things like the facility, so within a region, there could be multiple facilities like hospitals or it can also be providers within a hospital. So here, we list facility 1 and facility 2, and then on the second visualization we say, provider 1 and provider 2. So these two attributions are telling us the rank of each facility and provider based on the metric. And we can see that the poor performance are listed at the top if it were possible to scroll down, we would see a lot of the high performers, both on the providers, as well as the facilities, that are doing well on the specific metric. So, let's talk about sample size. We see that there is an equal sum number on both graphs. Sample size is quite important to derive important findings that could be used for decision making. If we are talking about very low sample sizes, low ends, it will be difficult to make any decisions because they only represent a few cases. And in many visualizations, you can use 50 or 100 as your threshold and to make sure that you are excluding any attributes that have small sample sizes. And just to emphasize, data accuracy is a must at this level of attribution. When you're looking at for example, from the regional level that we saw from this slide. Let me drill down to the facility, as well as to provide our level data, it's very important to have accurate data. Many times the data might be accurate at a facility or a regional level but once you drill down to the specific providers or nursing units, the data accuracy becomes less reliable. So it's important to keep an eye on that. Now, the benefit of having a list or the rank of providers or facilities like the visualization that you see here, is that it creates the Hawthorne effect, which means if you shared this with the providers or the nursing units or the facilities, and let them know that this graph is transparently shared with leadership or other managers throughout the organization, the Hawthorne effect creates the impact of, oh, somebody is looking at our metrics, so we should probably do better. It's very similar to how you used to get report cards in high school, I Imagine yourself not getting those report cards, it would be much more difficult or you wouldn't be as incentivized to do better in school because nobody's seeing your grades. That's what we call Hawthorne effect. And drilling down to this level of detail is critical, if you want to create that effect. And it can be an intervention on its own. So we talked about interventions in the previous slide, where we try to improve quality or performance, whatever it may be, Hawthorne effect can be a standalone intervention where you will see improvements, by just sharing the data and compared to other sites or facilities or providers. Let's talk about filters. On the very right of the entire visualization, we had some filters. Filters create the ability to drill down or isolate certain attriibutions based on the user's needs. In this case, different data ranges for viewing the cross-sectional or trending data is possible by changing the date range filter towards the middle part of the screenshot, and there can be as many filters as you want in a visualization. However, usability should be considered, because you don't want somebody to open your product or your visualization and have 100 different filters that they can change, and that actually creates more confusion. So it's important to have a balance between a number of filters and usability. Lastly, drilled down visualization. So we talked about the ranking of facilities and providers. Many times, it's not just those types of attributions where we're just drilling down based on the groups of attributions like region, to facility, to provider, or to nursing unit. Those are important but there are some, I mean this is very frequent to actually where we have a lot of process measures that contribute to that outcome measures. So in this slide, we're looking at outcome measures, the green and the red is based on the outcome of a certain performance metric related to quality improvement or whatever it may be. But in this slide, we're talking about process measures that contribute to the outcomes. Because, if we want to make improvements, we have to know where to start. So if process measures that contribute positively or negatively to the outcomes are defined, then we need to track them. We can use both cross-sectional and trending data. In this case, we're just looking at cross-sectional data for the process measures. But by looking at these process measures, we can tell what measure is having the biggest impact on the outcome or we can also tell which process measure is doing well or not, and how should we intervene to correct that. So it's important to have these drill downs, for process measures. Some other notes related to this video is, that you should know it takes time to develop good visualizations, and you need to consider design as well as capabilities of the tool. Many times, the audience or the customers would demand, I want to create a visualization that does X Y Z, but from a technical perspective, in terms of the program or the tool that you're using, that might not be possible. So you have to communicate with your customer to make sure that expectations are met with the reality of the tools that you're using. And data artist is a new field in this data visualization industry in my opinion. Data scientists are quite common and they're well known for their ability to manipulate data and extract insight through fundamentally the analysis tools that you heard about earlier, but data artists, is somebody who knows how to design visualizations for maximum consumption and understanding by the audience. So that is a niche but growing field. And lastly, when you're creating visualizations keep everything simple, that is more important than complex. So the last bullet point is simple is greater or better than complex. So keep that in mind as you're creating your visualizations.
