Aaron Berdofe, MHI – Health Informatics

Data can have a certain beauty to it.

I pulled the current active diagnoses of 1,926 patients from one of my former clients in WV (with permission of course) on March 20, 2010.  After first creating a similarity matrix from the data, I used Multidimensional Scaling to generate plot points of each diagnosis.  Therefore, the distance between each diagnosis’ center in the visualization above represents a relationship of how often one diagnosis occurs within the patient population with another.  The size of each circle represents how many patients the diagnosis occurs on. The number in each circle is the diagnosis code.

WV was recently ranked the unhealthiest state in America.  Unfortunately, the cause behind this largely seems to be due to unhealthy diets and limited physical activity.  My analysis definitely reinforces this.  The top 5 diagnoses were Hypertension (401.9), Hyperlipidemia (272.4), Depression (311), Esophageal Reflux (530.81) and Diabetes Type II (250)*.  Using the visualization, you can easily assume that if a patient comes in who is Hypertensive, there is a high chance that they will have Hyperlipidemia and Diabetes Type II as well.  Not exactly new information for a physician in this type of population or for one who understands the causes behind these disease, but for the rest of us it should be an eye opener.

I’m in the process of coding in some navigational capabilities and text labels in the Processing program I developed to create this so I’ll post it if I get it finished.  It’s a new language for me so don’t hold your breath.

* MS Excel automatically formats numbers when imported so it cuts off “.00” from numbers.

XLSTAT was used for the statistical analysis.