OBJECTIVES: This study was conducted as part of the IMI BigData@Heart consortium to identify and evaluate quality and accessibility of real world, patient-level data sources incorporating clinical information about patients with heart failure (HF) across 120 countries.

METHODS: A literature review targeting primary research publications from non-interventional studies for adult patients with HF was conducted in Medline and EMBASE from 1 January 2010 to 13 March 2018. Identified literature was reviewed and a list of unique data sources was derived. Information for each data source was recorded in a database abstraction tool. Data sources were examined and information was extracted on data types (electronic health records, omics, clinical data etc), study design, patient numbers, clinical characteristics collected (both HF-specific and general), follow-up and potential for access and linkage.

RESULTS: From an initial sample of 11,889 identified publications, 1,326 unique HF data sources were identified. Of these, 322 data sources (39 international and 283 single country) covering 74 countries were selected for detailed review. A total of 74% were annotated as HF disease-specific and 26% as generic and the most common study design was disease registry (44%). The five countries with the highest number of data sources were the US (61), Germany (19), Italy and Spain (16 each), and Canada (12). The most-commonly recorded clinical features were patient demographics (94%), hospitalizations (81%), comorbidities (77%), mortality (75%), and left ventricular ejection fraction (73%); the least complete data features collected were drug codes (10%), dates of procedures and prescriptions (7%), caregiver involvement (6%) and use of procedure codes (6%). A total of 5% of data sources reported to have collected -omics data.

CONCLUSIONS: This study created a comprehensive collection of real-world data sources for HF that represents a rich resource for researchers. The curated dataset is available as an open-source resource at http://dx.doi.org/10.6084/m9.figshare.6281291.v1.