Hgraph – Your health in one picture
hgraph.org
hgraph.org
- The use of a circular arrangement is questionable. Judging the distance of items from the center along radial axes is not something humans are good at. And drawing lines between the points adds uninformative visual noise. Better to just have a little bar chart with consistent axes.
- All the 'data' is presented without scales, numeric values, or trends over time. The one overall numeric value is presented without any context. We need a scale, a min and max, and a population average for this to be remotely meaningful. My first assumption is that this is like school grades in the US, where 85 is perhaps normal and 60 is abysmal. Is that right? Will that make sense to those from countries without grade inflation?
- The directionality is all messed up. Sleep is bad when it's low, that makes sense. Nutrition is bad when it's too high? But 'nutrition' is good, right? It's not clear what 'conditions' and 'environment' are, or how they could be too low.
- I would guess that the relative importance of these pieces of information differs by individual, gender, race, etc. Someone with diagnosed cancer should be marked as such rather than just being high on 'conditions.' And individuals with specific diagnosed ailments or family histories should have specific measures listed more prominently as important predictors.
- Some information that isn't good or bad but just descriptive belongs on here, like age, occupation, and, for women, period/pregnancy/menopause status.
There are a few dozen ways to present and graph data, that humans can quickly lock onto. The circular arrangement may or may not be the penultimate method to visualize health data, but it does afford a few good byproducts like seeing outliers quickly, identifying potential patterns (if you're looking across people), etc.
The high-level hGraph with ~18 metrics (we're pretty sure this isn't The Set) is the galactic view, just to give you an instant sense of your health. Once you dive into the data, the level of fidelity increases. Think of the Eames 10x10 flick of seeing the universe and then zooming in to get to the atomic levels that reveal trends, history, etc.
Directionality is how you construct the model (and then interpret it). You can eat too much. You can live in a hazardous environment (with too little X and too much Y). You're right that demographics will have a severe impact on metrics. That bias should be represented in the viz/algorithm.
hGraph needs some TLC from clinicians, policy-makers, and we the patients in order to get it to the next level.
Check out the screenshots of the health detail displays. If you have questions, fire away.
-Juhan, juhan@mit.edu, hGraph co-author
- I would recommend being very careful with the wording. One can have too much "food" but "nutrition" has positive connotations and no lay person would speak of "too much nutrition" or "too much environment".
- Please read the experimental material on information display. People study these things and polar display is very poor for picking outliers and recognizing patterns relative to almost any other option, including just printing numbers next to labels with no graph.
- For building "interactive" data displays, please read Bret Victor's "Magic Ink", specifically the section on the evil of requiring interaction to display information. [1] One of the problems of the circular layout is that 18 is about as many items as can fit, thus you need interactions to see more data. With most other displays of this data, you could easily fit far more information into one, interaction-free display, which could even be printed out for use in a wider variety of contexts.
- You could also use the clearer display to show actual numerical labels or trends over time. (Surely someone with high but declining blood pressure should appear different than someone with high and increasing blood pressure.)
- Finally, I am very curious about the decision to include the same set of metrics for everyone and to display them as a single graph. Some of these metrics belong together on one scale, some belong on separate scales, some belong only on people with specific conditions, etc.
- Are you planning to do a controlled test of this against other high-level patient summaries? Obviously my guess is that this graph and its interactions would fare worse than a one-sentence textual description provided by a past doctor. "Patient is a well-dressed and intelligent elderly man, with a successful history of controlling his diabetes through diet and exercise, but still showing signs of hypertension" seems to me much better than a graph with conditions: red that requires zooming in to see the word "diabetes" and LDL right on the edge that requires zooming in to see a declining trend over time.
[1] http://worrydream.com/MagicInk/#interactivity_considered_har...
(And yes, I did try it. Didn't turn out great for me, but I have a variety of fun medical problems I wouldn't expect to see covered here.)
I know most people have more than one thing that they need to do with regards to their health. Health is interconnected and negative habits (poor sleep) lead to other negative habits (poor diet). But that works on the flip side too. You sleep more and you walk more so you eat better, etc. By improving one thing you're likely to improve the rest.
Side note: We have an analog mechanism for daily behavior change, due out in Spring 2014 ... the Health Axioms: http://www.healthaxioms.com.
I see these circle/star plots sometimes but don't see the advantage over bar charts in conveying information or meaning. Am I missing something that a circle of dots adds that a line of bars or dots doesn't?
This looks awful. Why would want to do a polar plot for a time series?
Just so you can rotate it with your finger to go back in Time?
You can't really make out the rate of change if it's not linear.
Sleep and excercise are easily forgotten or given up when 'free time' gets harder to achieve, while both have so many benefits (or rather, none of the disadvantages that too little of both have).
Looks like it's a family with a baby in the household?
I believe hGraph is just taking a stab at presenting this data in a different way, one that could be more useful to patients that shudder at the thought of having to look at a spreadsheet.
If good explanations are shown along with the factors and questionable factors – such as weight – have enough context, this could be more than just a neat way to show the health of someone.
How is this thing making a determination on healthy weight without taking into account other factors like height, age, build, etc?
Thanks for your feedback!
when i go to [0] and change the default 3 drinks a week to 0 drinks a week my score drops from 100 to 90, that is a full 10%. Why? Acetaldehyde and acetic acid, both metabolites of ethanol are rather potent toxins. I would expect a better rating from keeping toxins out of my body, not a worse one. Then i changed the number of cigarettes from 0.5 a day to 0 a day. Nothing changed. Not smoking surely is much better from a health perspective than smoking 3-4 cigarettes each week.
I simply don't get it.
I'm more alarmed by the way the score drops if you weigh less than 258 lbs... especially with a 32.5 inch waistline.
Edit: there are some neat sliders on that page you can move the "optimum" values and green ranges for acceptable scores on the graphs on the left hand side to develop a saner scoring system.
You can manipulate what is "healthy" (min/max) in the chart by editing the end points of the green zone on the left hand side.
On the right hand side you can input your values for each metric to see it plotted on the hGraph circle + get a score.
Feel free to submit what you think is healthy in the bottom right.
Thanks for the feedback!
Eric
If you'd like to play with hGraph check out a tool we made during development.
We created this tool for physicians to give us feedback on the engine that drives hGraph. You can define what is considered "healthy" and rank each metric's importance in your overall health.
Put in your numbers and see how healthy you are.
Eric
Great design by the way!