Medicare Drug Spending Cost Analytics Dashboard
We created a dashboard that displays the cost of drugs over years 2011 - 2015 under Medicare. This tool was created to show insights on sudden changes in cost through the use of charts. Some of the preprocessing was done using Hadoop to display knowledge of Cloud Computing Applications (aggregating generic drug data).
This tool would be useful for anyone conducting research on the increasing of Medicare drug spending (Part D) and how it related to the annual part D premium increases published by the Social Security Administration.
Our results were meant to be neutral, meaning the user is free to make his or her own conclusion as to why the spending and consumer breakdown trends are increasing or decreasing and determine the best steps to address these trends.
While this information is primarily about Medicare drug spending, this information could be useful to anyone involved; shareholders and stakeholders alike.
Pre-processing Data and Dashboard Setup
The data we acquired from data4democracy (https://data.world/data4democracy/drug-spending) was mostly clean and ready to use. However, we wanted to subset data based on the represented generic and non-generic brands. Hadoop was used for this step.
Technologies and Tools
Example of API response
Hadoop MapReduce — Medicare Part D spending list for each medication details the generic name of the branded medicine name, the year, and total spending on the medication over the years (between 2011-15). Created a list of of cumulative spending and a list of top 20 purchased generics each year over the years for each generic drug.
AWS — We used AWS technologies like load balancer, clusters, EBS which runs the docker container inside it and make available the analysis of data through the node.js application.
Docker — We used a Docker image with node.js to run the hosted application which was available externally through load balancer URL.
D3.js — For representing the data analyzed we are using D3.js to generate the dynamic graphs and showing the information which can be easily consumed by the users.
REST — We are using npm package express.js to create the REST API and make the readily data available to client as JSON as needed which can be further used for showing graphs etc. Additionally, data were filtered and processed using various npm packages like csvtojson, underscore.js.
Discussion and Future Improvement
By making it easier to aggregate and visualize spending information, the web interface allows any user to retrieve data to aid their research in drug prices. The information can be used to argue the need to make a targeted effort towards prevention of diseases where the user sees an upward spending trend.
Results can be further improved by allowing the user to drill deeper into the aggregate data, though this is beyond the original scope of the project. At its current state, the web interface provides plenty of information to the user.