Improving Healthcare Cost Transparency Using AI

By law, beginning in mid-2022, health insurers are required to publish “Machine Readable Files” (MRF) which convey in-network rates for covered items and services.  The goal behind the legislation driving the regulation is to increase transparency in healthcare pricing to consumers. 

In practice, actual utilization of the MRFs is complicated by:  

  • Minor differences in the way the files are presented or structured.  
  • The sheer volume of the data  
  • Specialized technology required to access and process the MRF files  

As a result of these complications, most MRF data today remains largely inaccessible to the public.    

The burgeoning availability of powerful AI toolsets holds promise for making MRF data more generally available.  This project proposal involves work using consumer-grade AI tools to access, analyze, and summarize published MRF data from health insurers.    

If successful, the net outcome of the project would be an increase in cost transparency as envisioned by the original legislation.  

Objectives

Aggregate and leverage MRF and AI together for the purpose of improving visibility to rates for covered items and services to the public.  

The project involves creation of several supporting components:  

  • A code solution written in a common language (e.g. Python) for creating usefully specific AI prompts from common language used to describe healthcare services.  
  • A code solution written in a common language (e.g. Python) for leveraging consumer grade AI APIs for the purpose of accessing and retrieving published MRF data.  
  • A code solution written in a common language (e.g. Python) for aggregating and analyzing MRF pricing data obtained from multiple sources by CPT (service) code.  
  • A code solution written in a common language (e.g. Python) for visually representing pricing variance around queried healthcare services.  

Motivations

The main objective is to gain knowledge and experience in utilizing Artificial Intelligence technologies, and large language models, etc. along with consumer grade APIs.  The outcome is to develop and apply a machine learning model to derive valued and accurate data from machine readable files to support quality health services and outcomes for consumers. 

Qualifications

Minimum Qualifications:
  • Some experience with APIs and JSON or XML 
  • Some experience with Python

Preferred Qualifications:
  • Some experience with AI/ML Python libraries such as pandas 
  • Familiarity with LLMs


Details

Project Partner:

Andrew Diestel

NDA/IPA:

No Agreement Required

Number Groups:

1

Project Status:

Accepting Applicants

Keywords:
PythonMachine Learning (ML)New Product or GameConsultancy
Card Image Capstone