By Andrew M. Webster
Editor’s Note: An earlier version of this article was published on the ACTEX Learning blog ( https://blog.actexmadriver.com/)
It was the summer of 2015. I had founded my company, Validate Health, one year earlier. The Society of Actuaries awarded me $10,000 to sponsor a nontraditional internship. My intern, Nate, and I worked with a group of physicians to improve their payment structured under risk contracts. This was Validate Health’s first significant project with a health care provider customer. If we could develop an ongoing relationship then we would be able to raise funds. The stakes were high.
We needed to import claims and medical records, evaluate reserving models and mine the data for patterns of excess medical cost in one month. I was interested in developing a web-based tool to charge monthly subscription fees. All the analytical work was done using R since that was what Nate and I used at the University of Wisconsin. I planned on using the Ruby on Rails web framework to implement the online tool. I was unable to afford an R Shiny license.
The month progressed quickly. We spent three of the four weeks dealing with data problems. By the last week, we were finally evaluating models and pinpointing areas of risk-adjusted high cost. I did not have time to connect Ruby on Rails and R. We had to cut and paste screenshots of R visualizations into a PowerPoint presentation hours before the final meeting.
Then we met with the chief operating officer and the vice president of population health for the physician group. Even though we uncovered reimbursement problems with chronic obstructive pulmonary disease patients and reverse-engineered one of the insurer payment formulas, the meeting did not go well. When we were asked questions about end stage renal disease patients and why reserving factors were high, we were unable to answer on the spot. That was the last meeting with the customer. Later I learned that the group of physicians purchased a program to replace much of the modeling we developed.
By the fall of 2015, my company had exhausted its funding. I went to a tech Meetup to find a temporary tech job where I met a tech entrepreneur who had successfully founded three companies. As I talked with him, he revealed that one of the secrets to his success was using Python for both web development and analytics programming. I had used Python before for various scripting tasks but had never thought of it as a strong web or statistical programming language. If Python was the bridge between web development and analytics then I no longer needed to connect R and Ruby on Rails. As I researched further, I found that other successful Chicago tech companies such as Procured Health, Apervita and SpotHero do their programming exclusively in Python.
From that point on, all our programming has been done in Python as well. By early 2016, Validate Health won the Chicago Python pitch competition and solidified its reputation as a Python-based health InsurTech company. We received funding to attend the annual Python conference and pitch to Google, Yelp and the founder of Python. Using Python enabled us to rapidly develop dozens of analytic applications at low cost. As of late 2017, Validate Health has nearly half a dozen customers. All analytics are automated using Python and customers have control through a web-based interface. We could not have reached this stage if we had continued to use R and Ruby on Rails. Python, and networking with fellow entrepreneurs, saved my company!
If you’d like to hear more about my experience with Python, listen to my webinars on Python for Actuaries Intro: https://www.actexmadriver.com/product.aspx?id=453140807, Advanced: https://www.actexmadriver.com/product.aspx?id=453140950 and Statistical Modeling: https://www.actexmadriver.com/product.aspx?id=453142503
Andrew M. Webster, ASA, MAAA, M.S., is the founder of Validate Health, a member company of the HealthTech incubator MATTER in Chicago, Illinois. He can be reached at andrew.webster@validatehealth.com.