Data Science: to Pursue or not Pursue | My Experience taking Stanford’s Data for Medicine Class
Every now and then, something happens in life that dramatically alters your trajectory on where you want to go. My experience taking Stanford University School of Medicine’s Data Science for Medicine course with professor Nigam Shah is one of them.
TRANSITION TO HEALTH TECH
Rewinding for a second to provide some context, I’m a pharmacist by training and have worked in hospitals and health systems for over a decade. However, I made the switch to health tech at Verily about 7.5 months ago. As with any job transition, especially one that is in an arguably different industry, it’s likely that you’ll need to acquire and hone new skill sets to be proficient. (Here’s a list of resources I used if you’re curious). I began doing that during the pandemic in 2020 and really ramped up in late 2021 when I realized that I actually wanted to take a leap of faith from a stable, informatics job I had at Mayo Clinic. The culmination of my learning experiences over those 1.5 years led me to strongly favor developing my technical skills as I perceive the value of clinicians who can code to be quite high. I expand on that hypothesis in a previous article: Why All Clinicians Should Learn to Code.
THE INTERSECTION OF EXPERIENCE AND IMPACT
Personally, I wasn’t just looking to jump to a new industry just for the kicks, I was also frustrated by the momentum of healthcare as a whole. The actors that govern our healthcare system are complex, matrixed, and full of red tape. Rightfully so to ensure patient safeguards, however, I also think there are plenty of opportunities to streamline our progress towards the quadruple aim of healthcare. Further, I also think it’s important that a strong foundation of all facets of the healthcare system (e.g. healthcare, policy, tech) is critical to solving the issues we’ll face in the years to come. If anything, the pandemic has definitely underpinned how intertwined all of these components are. Think of federal and state initiatives in shifting resources and usage of artificial intelligence by Moderna to develop one of the first covid-19 vaccines. The big question is:
What exactly is the ideal path to make such an impact? Especially for someone that has primarily clinical experience?
CAREER PATH AMBIGUITY
I’ve been trying to answer that question for myself over the last few months I’ve been at Verily as daily interactions with product managers, software engineers, and data scientists have become the norm. Given my former role at Mayo Clinic as an informatics pharmacist, which was most similar to a product manager, I had initially thought I’d develop my technical skills for a few years before jumping into a formal product role. My rationale, which isn’t limited to healthcare, is that I feel managers and leaders should have a solid foundation in the areas they are leading. How can one lead a team of engineers, UX designers, or marketers if they have never done it themselves? I imagine this notion was borne out of my strong belief that the most talented informatics pharmacist are those that have actually practiced pharmacy. Your experience in that role (e.g. pharmacist, software engineer) is what gives you a unique insight into how to tackle the problems you’ll face as an informatics pharmacist, product manager, and leader.
THE GOOD & THE BAD OF DATA SCIENCE
…okay, I think I’ve veered off track enough already. So how does this all relate to the data science course I just took? (Btw, I finished my last lecture yesterday so I’m celebrating the fact that I’ll have personal time again on evenings and weekends). The answer: It has provided clarity as to what data scientists do and a viable career path that aligns nicely with my experience and goals. The problem: it’s really hard lol….
Since I’ve rambled pretty much this entire article, let me provide some highlights in the aftermath of passing (thankfully) this class.
- It’s really hard. I already mentioned this, but it’s worth noting again. I also think this opinion is partly due to my personal lack of knowledge in programming and mathematics, which, are prerequisites for this class. As a graduate-level course, it’s definitely more about learning how to apply existing knowledge vs. just learning new concepts to regurgitate. I should also note that I didn’t come into this class with zero knowledge.
- It optimizes both clinical and technical domains. I explored software engineering quite a bit in the last few months and even mapped out my 5-year plan to pursue an MS in Computer Science at Stanford University. However, the more I learned about data science and software engineering, the more confident I felt in data science being the best intersection of my skills. For anyone reading this, keep in mind this is subject and relative to your own goals and interests.
- The perceived incremental impact is high. Relative to any work I’ve previously done, I think the opportunity within data science to contribute significantly and meaningfully to healthcare is high. I could design alerts to remind every clinician to provide adequate care coordination for patients that are being discharged or I could develop predictive models to tailor resources and interventions toward patients that have preventable readmissions.
- I learned frameworks. I’m not new to data science and have been exploring it for a few years. In fact, I made a video with my colleague Samuel Ubanyionwu, PharmD, BCPS in 2019, about how we should be thinking about upskilling our fellow pharmacists in a world of artificial intelligence. However, despite dabbling in various courses and exploratory learning, I never really had a framework to approach building models. That is, not all models (e.g. predicting sepsis) may need to be built. The utility of models should also be evaluated — among other things like fairness. If the ramifications of a false negative vs. a false positive are that stark, perhaps we should reconsider building it at the expense of more carbon emissions. I now have one thanks to this class.
- I learned…. a lot of R. I did a few courses on Coursera back in 2020, but it pales in comparison to what I’ve learned over the last 10 weeks. I can’t say that I’m an expert in it, but I can definitely do some damage. The most fascinating, yet frustrating thing about the class was homework 4: feature engineering. Learning how to manipulate multiple tables (e.g. conditions, notes, labs, medications) from the electronic health record (EHR) and transform them into patient-feature matrixes was both challenging and insightful. The former was because I had to constantly Google how and why an R function should be utilized, but the latter was because the homework included free reign to engineer your own features of interest. It definitely piqued my curiosity as I had a lot of questions about the data set given my domain knowledge.
NEXT STEPS
To be honest, I’m not entirely sure. In fact, I don’t imagine being certain about a career path is common for most either. So, if you’re also pondering this question, know that you’re not alone. What I do know though is that data science is high on my list to explore further. I’ll continue to do so through Stanford University School of Medicine’s Biomedical Informatics Graduate Program and potentially continue down the MS Computer Science path. I have the next 5 years of it planned out anyways.
