PCORnet Blog

Want to Use Medicare Claims Data to Support Your Study? PCORnet’s CMS Linkage Team Shares How It Can Be Done

The Centers for Medicare and Medicaid Services (CMS) collect over two billion data points per year through the payment of Medicare health claims, and until now, this wealth of information has been largely inaccessible to researchers. Always seeking out new pathways to better insights, PCORnet, the National Patient-Centered Clinical Research Network, launched its CMS Linkage Pilot Project to answer this challenge.

The pilot team developed a process for using Medicare claims data to supplement PCORnet data in pragmatic clinical trials, such as the ADAPTABLE study comparing the effectiveness of different daily aspirin dosing for heart attack and stroke prevention. In a newly released white paper, the project team describes the processes and data flows used successfully in the pilot, as well as lessons learned and recommendations.

“As researchers, we have always been excited about the potential of Medicare data to fill in critical information gaps found in electronic medical records, but technological challenges have hindered our past efforts,” said Brad Hammill, Associate Professor of Population Health Sciences at DCRI.

“By pulling this data into PCORnet’s innovative architecture, we were able to rapidly extract important insights. For example, we found evidence of heart attacks that occurred when patients were traveling out-of-state in the Medicare data. These types of events are not usually captured in a patient’s primary electronic health record and are useful in giving us a more complete picture of the patient experience.”

For those looking to learn from the CMS Linkage Pilot’s success, the team has a few tips:

  • Allow ample time for preparing and processing the CMS Data Use Agreement. Processing times of six months are not uncommon with complex Data Use Agreements involving multiple institutions. In the case of the pilot team, those institutions were two PCORnet Clinical Data Research Networks, Mid-South and the Greater Plains Collaborative, plus PCORnet’s Coordinating Center at the Duke Clinical Research Institute (DCRI).
  • Social security numbers are most reliable for identifying patients within Medicare data. If those aren’t available, try using the Medicare beneficiary identification number.
  • Use the CMS Data Linkage project team’s tools to transform Medicare data to the PCORnet Common Data Model format. By doing so, you can easily analyze both the Medicare data and PCORnet data using the same analytic programs. 

Much more about the CMS Linkage Pilot Project is detailed in the newly published white paper which is available on the PCORnet Commons. If you have questions on the pilot, you can send those to [email protected]