National Medical Price Repository

posted October 8, 2025

Prices for medical services in the U.S. are primarily set by private insurance companies who negotiate with healthcare providers. Until recently, these negotiated rates were largely considered trade secrets. In July 2022, however, the Transparency in Coverage regulation went into effect, requiring insurers to provide machine-readable files that show the negotiated prices for all the services they cover. The availability of these data has created an opportunity for economic research into healthcare costs.

A team of researchers has joined with MSI on a project to gather the data and create a National Medical Price Repository that could be a resource for researchers and others searching for answers to important questions about healthcare spending and competition. The research team includes:

• Professor Jean Abraham, School of Public Health, University of Minnesota
• Professor Steve Parente, Minnesota Carlson, University of Minnesota
• Assistant Professor Conor Ryan, Department of Economics, Pennsylvania State University

The researchers decided to partner with MSI for this project in order to solve some problems associated with extracting and processing the data provided by the insurers:

• The amount of data is huge. The companies post monthly datasets online that contain hundreds to thousands of files, which have different formats from company to company.
• The datasets contain “zombie” or “ghost” codes that describe services that are impossible (for example, the cost of a surgical procedure performed by a psychiatrist). These unrealistic codes must be dealt with.
• The data are formatted in a hierarchical format called JSON, which is a format not used by most researchers in health economics.

MSI application developer Reid Wagner worked with the research team to create an iterative workflow to process the files. In this process, Python and Bash, as well as some manual steps, are used to download the data. The data are then parsed and condensed using Rust and Python. The results are sent to the researchers who can aggregate and analyze them. The process for a single insurer is shown in the image below. This process would be repeated for each of the other insurance firms.

The researchers are analyzing 70 “shoppable services” for this study: Lab Testing, Evaluation and Management, Surgery, Imaging, and Obstetrics. Some preliminary results are that prices for individual procedures vary by an order of magnitude, that there are similar variations across zip codes and within zip codes, and that larger insurers pay higher prices to physicians. The study is ongoing with new data provided by the insurers.