Visualizing Medical Data
I love my friends. When they see a graph or a fun data set, sometimes they’ll share it with me as blog-spiration.
This is how I became the proud owner of some real-life blood pressure data, as my friend has been requested by their doctor to record this information several times a day. As with any new dataset, I tried out several types of graphs to see which type provided the most useful information. I looked for outliers in the data so that we could investigate and explain them. I also looked for ways to categorize the information in the data set that might help me tease out patterns that are hidden within overall trends.
In this multiple line plot, I have both Diastolic and Systolic blood pressure readings going back as far as October 2011. There was a point that spikes above 160 Systolic. To investigate, I right clicked on the graph and opened up the Brushing tool, which allowed me to click on any point in the graph and label it. Brushing the point allowed me to quickly find out that the high reading was on February 11. If I wanted to be nosy, I could ask my friend if there was something stressful going on that day or if the reading was due to error or equipment malfunction.
We could also see if time of day had an impact on blood pressure readings. I took the raw data and broke it down by hour of the day using STATISTICA Extract, Transform, and & Load, which is typically used to aggregate and combine much larger and complicated data sources, but well-suited for handling time-stamped data of any size. This time, I added heart rate data on the right Y axis of the multiple line plot.
Each point represents an average of all readings from each one-hour time block in the day. For example, the first point is an average of all readings ever taken between midnight and 1am since October. We can see the slight upward trend through the morning hours, the respite around noon – perhaps during or in anticipation of a lunch break – and the continued upward trend through the evening hours until around 8pm.
Finally, my friend was asked to start tracking the blood pressure data taken from each arm. We can look at the distribution of data from each arm using a Categorized Histogram.
The additional statistics added to the top of each graph allow us to confirm what our eyes gather from the histograms – that the means are higher from the right side readings and the variability is higher on the left side readings. Once again, I used the Brushing tool on the graphs to turn “off” the one outlier point from the left side. The graph and corresponding statistics updated immediately to remove the effect of the outlier point.
The ability to easily track such important health information coupled with the growing ability to analyze it, monitor it, and suggest interventions when necessary even from afar….Living in the future is pretty exciting, isn’t it?