You're looking at data of the Agency for Healthcare Research and Quality. Specifically, you're looking at data on disparities within 'priority populations', which are groups with unique healthcare needs or issues that require special attention. This visualization has been created for the Visualizing.org Health Disparity Visualization Challenge.
For each group the data is collected into a single set. This will result in a question that can be answered for each individual table: "how does this group as a whole perform compared to the other groups?". For instance, if 'less than highschool' scores low on every other element (race, ethnicity, income level, etc.), then the average of this group will be low, and thus one can conclude that this group as a whole performs badly. Visually this is shown by ordering the groups and elements of the groups by average, where the worst performing group or element of a group is shown on top.
The groups and elements within a group requiring most attention are always on top. All elements of a group are ordered from 'most attention required' on top, and 'least attention required' at the bottom. The elements are sorted by the white lines, which represent the average for this element. If you check the 'show grouped' checkbox, the groups will also be ordered themselves by 'most attention required' on top, and 'least attention required' at the bottom. The groups themselves are sorted by the black lines, which is the group average and again showing the 'most attention required group' on top. The thick bars represent the standard deviation: near 70% of all values for this element fall within this range. The thin bars represent the range for this element: from minimum value to maximum value.
The source tables used for the visualizations can be found here.
Some of the challenges I ran into were:
This visualization has been created for the Visualizing.org Health Disparity Visualization Challenge, and is developed using Protovis and jQuery. The source tables of the NHQR have quite some missing data, and this visualization does not resolve this problem.