Tracking Syphilis Cases in the U.S.

Since 1999 the United States has experienced a dramatic increase in the number of new syphilis cases. The University of Iowa Computational Epidemiology Group has put together 17 years of state-level syphilis data and visualized this data. The visualization is a animated heatmap that shows the case count numbers for all states on a weekly basis.

syphilis_heatmap_01syphilis_heatmap_02

While I think the heatmap does a fairly good job at giving an overview at a given timestamp it is hard to really grasp varieties on different dates. The playback feature helps in this regard but for deeper analysis some sort of direct comparison would be crucial.

Gapminder.com with it’s “traces” mode comes to mind while thinking of possible ways on how to combine timeline based animation and direct comparison over multiple timestamps. Another way would be to introduce a split-screen solution, which I feel would be a inferior solution in terms of usability. Are there any concepts out there that let’s a user compare different states of a heatmap? I would love to hear your thoughts on this.

Share this article

Subscribe for more

Give Feedback

  • http://supervulpes.wordpress.com/ Jesse

    Must. Be. Population. Adjusted.

    It seems to just follow population… almost. Adjusting for pop. would allow the pattern to pop.

  • http://supervulpes.wordpress.com Jesse

    Must. Be. Population. Adjusted.

    It seems to just follow population… almost. Adjusting for pop. would allow the pattern to pop.

  • Pingback: Flight Ready Cases Nvkb61w Keyboard Case 61w W/ Wheels | My Worship Tunes

  • http://www.labescape.com/ Trevor Lohrbeer

    A couple ideas here.

    First, on making two different timestamps comparable, a good approach would be a delta heat map, where you visualize the change between two timestamps. You’d have to use a different color scheme, something like red to white to green. I don’t know how well this would work on a density heat map; I’ve always seen density heat maps with the rainbow color scheme.

    Which brings up the fact that this shouldn’t be a density heat map. The data is representing state-wide data. Density heat maps are best for point-based data (eg: data about cities). Using a density heat map for state data gives the mistaken impression that the syphilis cases are located in specific parts of each state. An area-fill heat map, with the each state filled with the entire color, is a much better approach for this type of data. This would also remove the visual artifact that makes the density dots look like they get bigger and smaller as the color value changes during the animation.

    Finally, the data appears to be cyclical, so ideally you want to have some technique that allows you to remove the cyclical nature. A couple ideas here: an option to aggregate values through the year and show year-by-year comparisons; use small multiples to show all the weeks at once (or aggregates of all the months); use a quantized fill to show multiple values per state at once; or use a different visualization entirely to show time-based trends while maintaining some geographic information (horizon graphs grouped by region come to mind).

    • http://www.artillery.ch/ Wiederkehr

      Thanks for this extensive analysis and the provided ideas. Let’s see if we can have one of the developers / designers of the Computational Epidemiology Group to respond to your suggestions.

  • http://www.labescape.com/ Trevor Lohrbeer

    A couple ideas here.

    First, on making two different timestamps comparable, a good approach would be a delta heat map, where you visualize the change between two timestamps. You’d have to use a different color scheme, something like red to white to green. I don’t know how well this would work on a density heat map; I’ve always seen density heat maps with the rainbow color scheme.

    Which brings up the fact that this shouldn’t be a density heat map. The data is representing state-wide data. Density heat maps are best for point-based data (eg: data about cities). Using a density heat map for state data gives the mistaken impression that the syphilis cases are located in specific parts of each state. An area-fill heat map, with the each state filled with the entire color, is a much better approach for this type of data. This would also remove the visual artifact that makes the density dots look like they get bigger and smaller as the color value changes during the animation.

    Finally, the data appears to be cyclical, so ideally you want to have some technique that allows you to remove the cyclical nature. A couple ideas here: an option to aggregate values through the year and show year-by-year comparisons; use small multiples to show all the weeks at once (or aggregates of all the months); use a quantized fill to show multiple values per state at once; or use a different visualization entirely to show time-based trends while maintaining some geographic information (horizon graphs grouped by region come to mind).

    • http://artillery.ch Benjamin Wiederkehr

      Thanks for this extensive analysis and the provided ideas. Let’s see if we can have one of the developers / designers of the Computational Epidemiology Group to respond to your suggestions.

  • Pingback: The Non-Alarmist Guide to Avoiding Your Environmental Health Risks [Geomedicine] | ReviewBox