Most of our team members come from public health or nursing backgrounds. We know research has shown that social determinants of health relate greatly impact a person's health. We decided to use a data-driven method for correlating these determinants and hospital readmissions.
What it does
Our machine learning model is trained to flag patients most at risk of hospital readmission based on these social factors: income, education level, zip code, race, and ethnicity.
How I built it
We did a data analysis to select the most important social factors responsible for flagging the patient at risk for readmission. Then, feature selection was carried out using the correlation matrix. We trained our machine learning model on those data points. We then intend to train a ML model on the selected features for flagging the patients. Once the ML model is trained, It could be deployed in the hospitals to identify patients at higher risk for readmission.
Challenges I ran into
Our team members spanned three countries and multiple time zones. We rarely all worked on this project simultaneously. However, we divided up responsibilities and tasks based on our skills and were able to complete this project.
Accomplishments that I'm proud of
We are proud of our application of machine learning to a very relevant health care issue that has long been burdening our healthcare system!
What I learned
We all brought a unique set of skills to this team (computer science, direct patient care experience, public health knowledge). I think we all learned from each other's strengths and gained a greater appreciation for the skills our teammates brought to the table.
What's next for TEAR: Tool for Evaluating & Addressing Readmissions
We used only a couple data points for training our machine learning model. In the future, this model could incorporate more social determinants of health or any other variables of interest to correlate to readmission risk.