PCORI funded the development of PCORnet®, the National Patient-Centered Clinical Research Network, to make research faster, easier, and less costly to conduct. PCORnet is made up of Partner Networks of healthcare systems, patients and communities, and health plans that harness the power of large amounts of health data.
PCORI supports projects, designed in partnership with federal funding organizations, to improve the quality of data used in clinical research. This project is one of several designed to assess the feasibility of linking patient-level data from research networks with other sources, including disease registries and health plans.
To demonstrate the feasibility of linking the PCORnet database, the Transcatheter Valve Therapy Registry (TVTR), and Medicare claims data; to compare the similarity of PCORnet and Medicare claims data with data in the TVTR; and to create a way to add new data elements into the linked database
||Retrospective descriptive analysis project
|Data Sources and Data Sets
||Records of patients who underwent a transcatheter aortic or mitral valve procedure from 2011 to 2016 at 8 sites
||Review of data from EHRs, the TVTR, and Medicare claims
- A linked database between PCORnet and the TVTR
- Comparison of statistical similarity of PCORnet EHR data, Medicare claims data, and TVTR data
- Data standards to allow for integration of quantitative and qualitative echocardiographic data into the linked PCORnet EHR and TVTR database
The project had three aims:
Aim 1: Create a linked database for PCORnet electronic health record (EHR) data, the TVTR, and Medicare claims data for patients who underwent a transcatheter aortic or mitral valve replacement (TAVR) from 2011 to 2016 at eight sites that are part of the PCORnet network.
Aim 2: Compare differences in baseline characteristics and in-hospital outcomes between patients in the linked database versus patients not in the linked database. Compare the additive value of linking PCORnet EHR or Medicare claims data on in-hospital outcomes and postdischarge data with the TVTR, compared with the TVTR alone.
Aim 3: Establish data standards for quantitative and qualitative echocardiographic data recorded within PCORnet and integrate these data into the linked database.
The project team matched PCORnet EHR data and records from TVTR (including linked Medicare data) to create a combined data set for patients at eight sites in LHSnet, a PCORnet Clinical Data Research Network, between 2011 and 2016. The team used a secure environment to protect patients’ information. The team also obtained a copy of the TVTR registry from 2011 to 2016 linked with Medicare claims data. They linked these records to records in the TVTR using a unique TVTR-specific identifier.
Aim 1: The project team successfully linked TVTR and PCORnet EHR records for 2,809 (96%) of 2,913 patients.
Aim 2: The project team found that baseline characteristics, demographics, and outcomes were generally similar between patients whose PCORnet EHR records did or did not link successfully to PCORNet.
The project team found that combining PCORnet EHR or Medicare claims data with TVTR provided added value across a number of data points compared with the TVTR data alone. For example, for in-hospital outcomes, both PCORnet EHR and Medicare claims data identified additional cases of atrial fibrillation, stroke, and percutaneous coronary intervention not captured in the TVTR data. Both PCORNet and CMS data also identified additional cases of postdischarge events including mortality, stroke, and atrial fibrillation.
Aim 3: All participating LHSNet sites were able to successfully integrate left ventricular ejection fraction (LVEF) data from echocardiographic data sources into the PCORnet Common Data Model (CDM). The project team surveyed the participating sites to come up with a collaborative set of parameters to define LVEF.
The project team found that combining the newly defined LVEF variable in the PCORnet CDM with TVTR alone provided added value at 30 days and one year post-procedure. In addition, a survey administered to participating sites at the completion of the project revealed that the majority of sites had their LVEF data stored either in the EHR or in a separate supplemental storage system with structured data elements that could be extracted through electronic query.
PCORnet’s CDM does not use a unique patient identifier. Therefore, tracking patients is difficult if they get follow-up care at a different facility than where they underwent the TAVR. Also, Medicare claims did not include outpatient data and were only available for patients over the age of 65.
Conclusions and Relevance
The project team was able to link PCORnet EHR data to the TVTR. PCORnet EHR and Medicare claims data showed added value as each data set identified a greater number of fatal and nonfatal outcomes when linked with the TVTR compared with the TVTR alone.
Future projects can use chart reviews to investigate why TVTR showed fewer reported fatal and nonfatal outcomes than PCORnet EHR or CMS Medicare claims data.