The latest generation of Visual Data Analytics software finally improved the user experience for non-techies and a people who do not want to become experts in how to use the tool. The paradigm-change caused by better usability is that now more decision-makers, front-line operators, and big thinkers from around the organization have the Visual Data Analytics software at their fingertips. Curiosities can be quickly measured. New ideas can be tested and legitimized, -or discarded. All without the time and mediation previously required to have reports made by someone else.

The downside of this democratizing of data is small in comparison the upside, but worth being aware of. One common issue is created by by using the wrong viz (or chart type) for the intended purposes. With some luck, the poor choice of viz didn’t hurt anything. However, it is likely your viz is a less effective. The worst case scenario is that the sloppy viz skews the facts and results in the wrong conclusion. The easiest way to avoid using the wrong chart type is to determine the correct chart type before hand.
The Purpose of Data Visualization (The Why)
Data visualizations are often created to serve one of three purposes: Discovery, Communication, and Understanding[mfn]Keim, D. A., Mansmann, F., Schneidewind, J., & Ziegler, H. (2006). Challenges in visual data analysis. IEEE International Conference on Information Visualization, 9-16. doi: 10.1109/IV.2006.31[/mfn]. This is the Why of your viz. Knowing which of these you are working towards will help you stay on track.
Discovery
Confirmatory analysis is when you test a hypothesis you have a hunch about.Exploratory data analyses may identify things you didn’t know.
Communication
A well designed visualization is the most efficient and effective manner to communicate complicated information
Understanding
Understanding real relationships and events through through visual cues and implicit information.
Choosing the Best Viz Type (The What)
Each chart type has strengths and limitations, so it is important to choose the chart type that gives you best chance to answer your question (Discovery), for example. The ‘right’ chart depends on both the problem at hand, and the data available. Let’s say you are developing a viz to help a manager schedule nurses. The purpose(communication) is to show her reference patient volume as well as forecasted patient volume. The problem at hand , patient volume over time , suggests you should choose a Line Graph. However, Line Graphs need Date/Time data. If you don’t have this data, a bar graph would be a fine runner up. If you get curious and think that Day of Week might be a mediator of patient volume(Discovery), a Heat Map organized by week might be a good chart to start with.
Bar Chart
Considerations
- Great for comparing data across categories
- Categories should be mutually exclusive
- Shows highs and lows well
- Stacked bars show proportion information within a category
Example Qs:
- How does Hospital A perform compared to peers?
- Which age group is our highest and lowest for readmission?
Line Graph
Considerations
- Shows trend over time
- An area chart is a line graph that show proportional contribution.
Example Qs
- How has this patient’s risk score been trending?
- Which category of personnel costs are driving the overall changes?
- Has the number of rejected claims changed in two years?
- How many patients of this type will we see next month?
Pie Chart
Considerations
- Shows relative proportions
- Hard to compare across pie charts
- Pies are good additions to viz’s like maps
- Limit pie wedges(6-ish)
Example Qs
- What is our payer mix?
- What is are patient demographics?
- Describe our revenue and expenditures.
Maps
Considerations
- Any kind of location data
- Embed symbols, pie charts, etc.
- Combine point data with area data
- Overlay population data
Example Qs
- Where are our patients coming from?
- Which patients are near us but seeing other providers?
- What areas are not referring patients to us?
Scatterplot
Considerations
- Exploring relationships between numeric data
- Add categories (color) and a third numeric dimension (size) for more information
- Add a trend line to clarify
Example Qs
- Is there a link between hours on shift and number of medical errors? If so, is it different for different shifts?
- Is duration of surgery linked to LOS? Is that because of HAIs?
Heat Map
Considerations
- Shows measures between two categorical ‘axis’ to highlight trends
- Order (or organize) categories for clarity
- Use square size to indicate additional information
Example Qs
- Are there certain days of week, month, or other calendar trends?
- How are all of our departments (or providers)doing across the 10 item preparedness checklist.
Box and Whisker Plot
Considerations
- Shows distributions of data
- Easy to spot outliers or skewed data
Example Qs
- What procedures are disproportionately driving our average LOS?
- Which services have the most variability in quality?
- Which facilities are the most consistent?