Guest Post: Hear from Lown’s data science intern, Maya

As a rising junior Computer Science student at the University of Massachusetts Amherst, I was excited to join the Lown Institutes data science team as a summer intern. I looked forward to applying what I have learned in the classroom to real-world data sets and problems, as well as the opportunity to learn more about the American health care system. Over the past three months, I have completed projects using statistics and programming skills. Through analyzing the Lown Institute’s Hospital Index of Social Responsibility, I learned about the characteristics that make a high quality hospital. I examined data on financial assistance policies and extraordinary debt collection actions, medical overuse, and the impact of ownership type on safety net hospitals’ clinical outcomes.  

Here is a summary of the key projects I completed during my internship. I added more detail of these on the Data Team’s Substack, in case you are interested in the technical details.

• Created an app for data analysis requests using a Large Language Model (LLM). I created my first Shiny application and used a publicly available code package, querychat, that allows users to “chat” with their data using natural language processing. I built this for the Avoiding Overuse data, so now users can ask questions like what is the average overuse rate of spinal fusion in hospitals in the Northeast? and have the LLM provide the answer using the data. 

• Completed data requests. Not everything can be answered using an LLM, and so I also practiced my own analysis skills. Journalists, hospitals and public health researchers ask the Lown Institute specific questions about its datasets and index, and I had the opportunity to dig into the data and help with these. This included which hospitals are consistently performing high on the Hospital Index of Social Responsibility over the past five years, and a comparison of financial assistance policies across select healthcare systems. 

• A deeper analysis into the association between ownership type and patient outcomes in safety-net hospitals. We wanted to see if there was a difference between safety-net hospitals (those with the highest share of low income patients) that were nonprofit, versus those that were for-profit or publicly owned. To do this, I coded several different models and created many visualizations. Overall, there were significant differences between publicly owned hospitals and nonprofit/for-profit hospitals, although there may be confounding due to factors like hospital size. 

Working at the Lown Institute this summer was a wonderful opportunity to gain real-world experience. I am grateful to the team for helping structure projects that allowed me to use my existing skills and to progressively develop new skills that I will carry forward in future work.