Making the Transition From Clinician to Data Scientist: Transferrable Skills

Although I have a non-traditional pharmacy career path that now has me working in a technical role on a data science team at a health system, my training as a pharmacist is enormously useful in my day-to-day work. I love that I am able to use the two things I’m most passionate about professionally, healthcare/pharmacy and programming, to build tools to make healthcare professional's job easier. In this article, I will share two areas that I find that my training in pharmacy has paid dividends, communication skills and healthcare domain knowledge.

Communication Skills

In my (as of this writing) first two years working in health data science, communication has consistently played an outsized role in the success my team has with our tools. Machine learning and Tableau tool development have undoubtedly been important as well but the continued use of our tools by our clinician users has been most impacted by great communication.

The example of great communication that I’ve seen pay the greatest dividends was a project my team started my first month at UnityPoint Health, a tool (originally for Care Management) that we call the Pop Health Toolkit. From its inception, the tool was designed to make finding patients who would benefit from Care Management services easier for the Care Managers. In order to do this, we had meetings (including making a 4 hour round trip to one of our regions) with the healthcare professionals that would be using the tool 6 months later. Our goal in these meetings was to find out what the Care Managers were using to do their work prior, what worked well for them, and what was not working. You can draw parallels to what we were doing with motivational interviewing in the healthcare professional world. We asked open-ended questions and got to the root of what we could do better to make their jobs easier. From these meetings, we identified how we could order patients on our dashboard in a way that would surface the patients who truly needed Care Managers the most. We also replaced several reports that the Care Managers were running manually and separately. This one, combined tool would make stitching together disparate information much easier!

Communication is also crucial within a data science team. In my experience, no major project for our team is implemented by one person. Every team member has a hand in the creation and maintenance of a project. Clearly articulating choices to be made, problems you’re having, and different points of view have been important for my team. If I think about recent modifications I’ve made to some of our code for finding lab values, I had to convert what I thinking into a step-by-step explanation of why it would be more beneficial to change the code and what next stages of the change could look like so my teammate understood the benefits and any drawbacks without having to dig into the database in the same way I did as I was gathering the information.

The communication skills you’ve built as a pharmacist or other healthcare professional will go a long way in making you more successful on a data science team. You’ll be able to better identify what will make your projects more useful for the people using them and will help you work more efficiently with your teammates.

Domain Knowledge / Healthcare Understanding

Understanding the intricacies of healthcare has been enormously beneficial in my day-to-day work. It should come as no surprise to anyone trained as a clinician that healthcare does not operate in the same way that other industries do. In part because of the impact of healthcare (the effect on human life) and unique rules and regulations (HIPAA), healthcare is a daunting system to get used to. My training as a pharmacist helps me adapt to those issues because I was trained and molded in the healthcare system (hopefully someone sees the Dark Knight Rises parallels in that one).

The foundation of our data science tools is the machine learning predictions that identify the likelihood that an adverse health outcome happens to a patient. This is communicated to the healthcare professionals using the tool. To predict these outcomes, we gather data about patients and train machine learning models. Our models take into account which medical conditions each patient has, their lab values, medications, recent healthcare encounters, and much more.

Medical condition interpretation and lab description deciphering strike me as the two most tangible ways my healthcare background has been beneficial. If you looked at a table of labs (like all EHRs will have), you’ll see lab descriptions with things like “sodium”, “potassium”, and “creatinine” in them. You’ll also see hundreds of other codes that cover hospital lab tests, point of care device tests, and other assessments, including the list below:

Sodium
Potassium
Hemoglobin A1c
Hemoglobin
A1c
Point of Care Hemoglobin A1c
Creatinine

Our clinical models frequently contain a feature (predictor) of the last A1c for each patient. As clinicians well know, A1c is used as a measure of diabetes control. A1c is a great example of a lab that requires familiarity with the clinical nature of healthcare. A1c can show up in numerous ways in lab descriptions including “Hemoglobin A1c”, “HgbA1c”, “A1c”, “Point of Care A1c”, etc. There is another lab, “Hemoglobin”, that also shows up in the lab table and is in no way similar to hemoglobin A1c. My healthcare background has prepared me to be able to decipher the difference between hemoglobin A1c and hemoglobin where non-clinicians might struggle with this issue and inappropriately grab the wrong lab codes and values. Another example is the cholesterol labs like HDL and LDL. There are a number of lab descriptions in the lab database that are HDLs, LDLs, or a ratio of HDL along with other cholesterol labs. If you want to grab HDLs only, you have to parse out those HDL:LDL ratios so they don’t show up in the data that your algorithm uses.

My healthcare background also gives me a unique perspective on what can be used to predict the outcome that we are trying to target. This manifests itself in a number of ways. Take an example of trying to predict the risk that a patient will have a myocardial infarction (aka heart attack). Most people could identify that the risk of a heart attack would include a number of heart-related factors like blood pressure, previous heart attacks, etc. As a clinician though, we’re taught about a number of other risk factors for a MI that non-clinicians wouldn’t normally associate like diabetes/A1c, smoking history, drug use, family history, and autoimmune disorders. This extra knowledge allows me to recommend extra data points that may be useful. A more recent example included adding more lab values to our data sets. As we are re-architecting some of our data sets for machine learning, I noticed we were missing some common lab results in our modeling data. We were not using labs like ALT or AST, which can be good predictors of patient health. We also had a data point labeled as “Arterial Blood Gas” but it was only one part of an ABG panel, the pH of the blood. We were missing the other parts of that panel, pCO2 and pO2. Through my education, I had many opportunities to use the “fishbone diagram” for labs (in addition to the diagrams for ABGs, etc) that allowed me to go seek out those diagrams to make sure we had all of the components. I was able to add about 5 or 6 common labs that we weren’t using before.

Wrap Up

In this article, I hope that you saw some of the benefits of making the transition into data science if you’re a healthcare professional and saw examples of the skills that you already have that will make you successful. There are a bunch of skills that you’ll have to master that your healthcare background didn’t prepare you for but having great communication skills and healthcare expertise will make the transition less jarring. You’ll have some unique experiences to share with your team. If you’re not a healthcare professional, no worries, try to find a data science team that either has this type of expertise or really dig into better understanding how healthcare works and learn from those on your team who excel at communication.

-Dalton