Blue Cross Blue Shield Smart Watch Program
Which patients would benefit the most from a smart watch?
January 2021 - May 2021
Chronic Condition Count by Gender
During the Spring semester of 2021, Blue Cross Blue Shield Rhode Island decided to give my Business Analytics Consulting class a project using its own customer medical data set. The premise of the project is to utilize the data set, which contains thousands of patients' medical records, to figure out the top 200 patients that would benefit the most from a free smartwatch.
For the detailed project, please visit: BCBS and Smartwatch
Dataset
BCBSRI provided our team with multiple data sets that include patients’ medical records, the age of their accounts, and communications between patients and BCBS. Since we had to work with thousands of raw data on different spreadsheets, our team had to join the data set using patients’ IDs and then clean it. Many variables were removed and/or joined in order to create the best data set with no missing values and multicollinearity problems.
Theory and Analysis
n order to solve the problem that BCBS presented, our team first decided to make the assumption that the smartwatch that patients will be receiving will be Apple Watch, which was the best product with the most health-related feature at that time. Our team also decided to use the risk score variable to determine the top 200 patients. The higher the risk score, the more likely that the patient will benefit from getting a smartwatch
Results
Looking at the final 200 members, more than 50% have 3-4 chronic conditions and the majority fall in the age range of 60-80 years old. 82% of the final group also suffer from a heart-related disease that can cause significant risk for Afib. We believe that this result is similar to the list of members who already received a smartwatch, continuing the trend that was set by the previous batch of smartwatch recipients. On top of that, all of the 200 members have a response rate higher than 60%, proving that they will communicate with Blue Cross Blue Shield to talk about the program
What I Learned
Technical Tools
I learned how to use SPSS to test for correlation and multicollinearity. On top of that, I also learn how to create a regression model and use it to predict and test for prediction accuracy.
For more details on this project, please visit BCBS and Smartwatch.
Risk Score by Gender
Chronic Diseases by Age and Gender
Using SPSS to figure out the most significant variables, our team created a regression model to predict the Health Risk score. We then use the prediction accuracy method of Mean Absolute Error and found out that the average error, or difference, between our predicted health risk using our regression model and the provided risk score is only around 0.37. The top 200 patients with the highest predicted health risk are the 200 patients we are looking for
Significant variables in Our Regression Model
Overview
In this project, I was tasked to create a master data set by joining different data sets and cleaning it by removing all NAs and unused variables and joining different variables. I was also in charge of the exploratory analysis and part of the project in which I used Tableau to create data visualizations. Finally, I utilized SPSS to test for correlation and multicollinearity between variables in order to help create the final predicting model.