As a part of my training, I have enrolled in a master’s course at the local university, UTSA. The program is the Master of Data Analytics or MSDA. I have completed three semesters of course work that have included learning statistical methods using primarily R, python and SAS. I wasn’t quite sure what I was getting myself into at the start of the program, but it has proved quite valuable for my CI training. I am involved in some very interesting predictive modeling projects on the periphery and I can easily follow along. These courses have allowed me to see how I can be a bridge between statisticians or engineers with the clinical side.
So what encompasses data analytics? Data analytics is broad and includes different types of data analytics including descriptive, diagnostic, predictive and prescriptive.
- Descriptive – this type of analysis describes what has happened and is great at showing past performance
- Diagnostic – this type of analysis tries to explain why something happened, such as a root cause analysis
- Predictive – this is a sexy part of data analytics that includes predictive modeling and artificial intelligence
- Prescriptive – this type of analysis tries to answer the question "what should be done" by deciding the best course of action in a given situation
Data analysis is large part of clinical informatics and it encompasses many things. Predictive modeling and artificial intelligence is, in my opinion, the sexy part of data analytics. There have been many models created to improve various areas of medicine that can improve quality and reducing cost.
The other important aspect of data analytics is that it often involves “big data”. Medical records can certainly be viewed as big data. The electronic medical record holds a wealth of information from patient’s demographic information such as age or gender, their medical history, prescription history, vitals, etc. They also contain many different types of data. The physician will often write “free text” notes while vitals are often stored in discrete numerical fields. Knowing how to handle all this data to arrive at something meaningful is important.
But I would also argue that knowing what information is important and makes sense is also important when analyzing the data. There are many crazy correlations that exist that don’t necessarily translate to causation. Spurious Correlations website is a great example of this. For a medical example, if we used random information available to us in the medical record, we may find a correlation between patients that stubbed their toe and urinary tract infections. That example should sound crazy to just about everyone. However, less obvious findings may be missed without the proper medical knowledge. This is where I see physicians trained in Clinical Informatics filling in that gap.
doctordatalyze.com 