# Beyond simple charts: Design of visualizations for big health data
Ola and Sedig present four visual designs specifically for conveying large healthcare datasets . The authors describe the characteristics of healthcare datasets as high volume, low veracity, great variety, and high velocity.
Following the pattern-based framework (see the appendix for the 14 patterns) by Sedig and Parsons , the authors propose 4 visual design categories: demography visualization, chronology visualization, geography visualization and overview visualization.
The authors use a subset of the Global Burden of Diseases dataset , which contains over 12 million records.
The authors blend patterns (Stack•Track•Token•Group•Link•List•Coordinate) to create a design based on demography visualization, as shown in Figure 1. The sunburst layout encodes age group, cause, risk and location in four layers. The coordinate axes at the center show relationships between cause, risk and location, as shown in Figure 2.
Figure 3 shows the design for the chronology visualization (Fusion•Coordinate•Token•Hierarchy•Cell•Link•Group), it facilitates the exploration from a temporal perspective.
Figure 4 shows the design for the geography visualization (Branch•Token•Coordinate•List•Group•Spectrum•Area•Cell), it facilitates the exploration from a temporal perspective. The chord diagram at the upper left shows the global cause-risk relationships from a risk-centric point of view, as opposed to the cause-centric point of view at the upper right. A choropleth map is shown in between.
The stadium-like layout at the bottom of Figure 4 shows disease labels for the selected region, cause-risk relationships visualized in a matrix is connected to a disease distribution matrix, via a sankey diagram.
Figure 5 shows the design for the overview visualization (Branch•Token•Group•Cell•Hierarchy•Stack). The design allows the user to explore age-cause-risk-location relationships and serves as an entry point that links to other visualizations described above. The main components are treemaps linked via a sankey diagram.
In the related work section, the authors cite EHR Vis systems LifeLine [^LifeLine], LifeLine2 [^LifeLine2], [^Midgaard]. Analysis of time-oriented data is inspired by Aigner et al.  and Wang et al. 
- Data source: Global Burden of Diseases
- Size: over 12 million records used
- Spatial dimensionality: 2D
- Temporal dimensionality: static
- Type: multivariate
- Libraries used: D3.js
- Chord diagram
- Information Visualization of EHR
- Geospatial Visualization of EHR
- Area: used to map data items onto visualizations in such a way that their boundary, shape, region, and/or area are encoded.
- Branch: used to map data items onto visualizations and organize them in a branched and/or subdivided fashion.
- Cell: used to map data items onto visualizations and organize them by segmenting, compartmentalizing, or containing them within cell-like structures.
- Coordinate: used to map data items onto visualizations and organize them with respect to a frame of reference.
- Cycle: used to map data items onto visualizations and organize them in a circular, wheel-like, rotational, spiral, and/or cyclical fashion.
- Fusion: used to map multiple data items onto a single visualization in a continuous fashion, such that the items are integrated and fused together.
- Group: used to map data items onto visualizations and organize them by congregating them close to each other.
- Hierarchy: used to map data items onto visualizations and organize them in a hierarchical, multi-level, pyramid-like fashion, where higher levels are superior to or contain and encompass lower level items.
- Link: used to map data items onto visualizations and organize them by connecting them together using paths, routes, lines, or other similar structures.
- List: used to map data items onto visualizations and organize them by placing in a sequential, successive fashion.
- Spectrum: used to map data items onto visualizations and organize them in a spectral fashion. Often instantiated using multiple saturation or luminance values of a particular hue, or using multiple hues or textures.
- Stack: used to map data items onto visualizations and organize them by placing on top of one another in a piled or stacked fashion; visualizations are often placed on top of one another such that they are touching or are very close together.
- Token: used to map one or more data items onto a visualization that can be regarded as a unit, whether in atomic form or composite form made of discrete parts.
- Track: used to map data items onto visualizations and organize them in a lane-, stripe-, and/or track-like fashion.
Ola, O., & Sedig, K. (2016). Beyond simple charts: Design of visualizations for big health data. Online Journal of Public Health Informatics, 8(3). https://doi.org/10.5210/ojphi.v8i3.7100 ↩︎
Sedig, K., & Parsons, P. (2016). Design of Visualizations for Human-Information Interaction: A Pattern-Based Framework. Synthesis Lectures on Visualization. https://doi.org/10.2200/s00685ed1v01y201512vis005 ↩︎ ↩︎
Institute for Health Metrics and Evaluation (IHME) University of Washington. (2015). GBD Compare. Retrieved from https://vizhub.healthdata.org/gbd-compare/ ↩︎
Aigner, W., Miksch, S., Muller, W., Schumann, H., & Tominski, C. (2008). Visual Methods for Analyzing Time-Oriented Data. IEEE Transactions on Visualization and Computer Graphics, 14(1), 47–60. https://doi.org/10.1109/TVCG.2007.70415. ↩︎
Wang, T. D., Plaisant, C., Quinn, A. J., Stanchak, R., Murphy, S., & Shneiderman, B. (2008). Aligning temporal data by sentinel events. Proceeding of the Twenty-Sixth Annual CHI Conference on Human Factors in Computing Systems - CHI ’08, 457. https://doi.org/10.1145/1357054.1357129 ↩︎