The CVD Surveillance dataset is comprised of patients hospitalized for one of the following conditions between January 1, 1999 and December 31, 2009:
- Acute Coronary Syndrome (ACS) defined as acute myocardial infarction or unstable angina.
- Heart Failure (HF).
- Stroke including ischemic, hemorrhagic or transient ischemic attack.
The CVD Surveillance study utilizes patient data from 15 participating sites across the US. The dataset was compiled using standardized programs that convert local source data extracts into a common dataset structure with the same mapped variables and corresponding definitions. Kaiser Permanente Northern California (KPNC) serves as the data-coordinating center and manages all requests regarding the data.
795,964 total patients were identified; with 287,808 (36.2%) having a diagnosis of ACS, 492,341 (61.9%) having a diagnosis of HF and 208,558 (26.2%) having a diagnosis of ischemic of hemorrhagic stroke.
Table 1. CVD Surveillance Dataset Member Counts by Site
Number of Members||
Percent of Dataset|
|Kaiser Permanente Northern California||246,362||31.0%|
|Kaiser Permanente Southern California||219,144||27.5%|
|Kaiser Permanente Northwest||53,419||6.7%|
|Kaiser Permanente Hawaii||53,317||6.7%|
|Kaiser Permanente Colorado||44,895||5.6%|
|Kaiser Permanente Georgia||41,800||5.3%|
|Group Health Cooperative||39,180||4.9%|
|Harvard Pilgrim Health Care||19,041||2.4%|
|Fallon Community Health Care||17,770||2.2%|
|Henry Ford Health System||16,332||2.1%|
|Marshfield Clinic Research Foundation||9,058||1.1%|
|Geisinger Health System||1,490||0.2%|
|Scott & White||1,124||0.1%|
Figure 1. Distribution of Age and Gender for Subjects in the CVD Surveillance Study
All member data was extracted using the Virtual Data Warehouse (VDW). The VDW is a standardized resource comprised of: (1) computerized datasets stored behind separate security firewalls, including variables with identical names, formats, and specifications; (2) informatics tools that facilitate storage, retrieval, processing, and managing the datasets; and (3) detailed data documentation. All data are linked by a unique identifier so that data elements can be rolled up to the individual person-level. We used the VDW to characterize each sites cohort of patients with regard to relevant exposure and outcome variables. KPNC merged all site-specific VDW data into multisite datasets. However, some variables were not available at all sites and not all sites contributed data across the entire study period.
Matlock DD, Groeneveld PW, Sidney S, Shetterly S, Goodrich G, Glenn K, Xu S, Yang L, Farmer SA, Reynolds K, Cassidy-Bushrow AE, Lieu T, Boudreau DM, Greenlee RT, Tom J, Vupputuri S, Adams KF, Smith DH, Gunter MJ, Go AS, Magid DJ. Geographic variation in cardiovascular procedure use among Medicare fee-for-service vs Medicare Advantage beneficiaries.
JAMA 2013; 310:155-62.