#HHS
Jupyter Notebook JavaScript Python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
api-server
data-loader
jupyter
README.md

README.md

hhs-insights

Vision: Our goverment has wealth of data available on HHS. Our vision is to utilize the available information and make meaningful analysis out of it. For example: By joining the the data sets of doctors and provider information we can get more insight on why a particular area has more frauds. As more efforts are being made to standarize data like doctor ratings we can further leverage and allow people to make a more informed decision about insurance providers.

Goal: Our goal is to allow government to make use of the available data to solve problems on resource management, budgeting and better planning. For example: This can help in population lifestyle equitability, human migration, urban planning and even more.

Problem: Determine availability of providers and quality of doctors. We solve this by combining the datasets and querying for "NPI Deactivation Reason Code" and number of providers through elastic search.

Architecture

Notebooks ==> API server(AWS) ==> ElasticSearch(AWS)

Directory structure

- jupyter/
   - A collecition of ipynb notebooks to render interesting visualizations
- api-server/
   - NodeJS API server that provides a simple API that Jupyter notebooks call into.
- data-loader/
   - scripts to parse and normalize CSV data and store to elasticsearch.
 

API Server

  • Endpoints
  • GET /providers-by-state?speciality=
    • returns an aggregated report of providers grouped by states.
    • can optionally be filtered by speciality, eg speciality=Diabetes
  • GET /providers-by-zip
    • returns an aggregated report of providers grouped by Zip codes.

Interesting Notebooks

  • Fraud occurences
    • Plot a heatmap based on number of provider deactivations due to fraud.
  • Provider distribution by state
    • Plot a heatmap based on raw counts of providers across states.
  • Provider distribution by zip
    • Plot a heatmap based on raw counts of providers across zip code boundaries.

Datasets

  • NPPES provider data from cms.gov. URL
  • ~4.8 million providers data.