Cardiac Arrest Registry to Enhance Survival (CARES)

Emory University Woodruff Health Sciences Center

Includes all non-traumatic out-of-hospital cardiac arrests (OHCAs) where the patient receives resuscitative efforts (CPR and/or defibrillation) by 911 responders, or is defibrillated by a bystander prior to the arrival of 911 responders, that occur in the CARES catchment area (covering 49% of the US population).

Unit of Observation:
Personally Identifiable Information Available for Linking:
Alabama, United States of America
Years Available:
5% of project budget if externally funded; Free if internally funded
Frequency of Updates:
Annually with an approximately four month lag (e.g., CY 2018 are expected in April 2019)

All non-traumatic out-of-hospital cardiac arrests (OHCAs) where the patient receives resuscitative efforts (CPR and/or defibrillation) by 911 responders, or is defibrillated by a bystander prior to the arrival of 911 responders.


Researchers may request de-identified CARES data by state or nationally. To request data, researchers must submit a project proposal to the CARES Data Sharing Committee (see Appendix B of the CARES Data Sharing User Guide). To inquire more about the national data set, researchers should contact Rabab Al-Araji, the CARES Epidemiologist. To inquire about state-specific data, researchers should contact the respective CARES State Coordinator. See the CARES contact page for contact information. 

If the proposal is approved, researchers must secure a Non-Disclosure Agreement (NDA) and Institutional Review Board (IRB) approval before receiving the requested de-identified data set. 

Abstracts must be submitted within three months of the date that the data set was provided, and draft manuscripts must be submitted within nine months of the date the data set was provided. All abstracts and manuscripts must be reviewed by the CARES Data Sharing Committee prior to submission. Manuscripts must include the words “and the CARES Surveillance Group” in the authorship line following the individual authors, and the "Acknowledgement" section of all manuscripts should reference the CARES participating sites.

Timeline for Access

CARES will provide feedback on research proposals within four weeks of submission. If the Data Sharing Committee approves the request, CARES staff will then provide the requested de-identified data after the researcher signs the NDA form, gains IRB approval, and attends a data review webinar.

Lag Time

Files are updated annually, with updates usually available in early April (e.g. CARES 2018 data becomes available in April 2019).


If the project is funded from external sources, CARES charges 5% of the project award amount for data access. If the project is funded internally (by the researcher’s institution), there is no fee to access CARES data. See Appendix F of the CARES Data Sharing User Guide for details.


All data is de-identified, with fields that could identify a patient, EMS agency, or hospital removed from the data set. CARES includes a randomly generated ID for EMS agencies and/or hospitals in the data set to allow researchers to adjust for those in a multivariate analysis. Appendix E of the CARES Data Sharing User Guide includes the data set’s structure.

Identifiers Available for Linking

  • While direct identifiers are not available, researchers have used probabilistic matching techniques to link CARES patient-level data with other data sets such as Medicare inpatient data.
  • Researchers used variables such as date of arrest/hospital admission, patient age and gender, de-identified admitting hospital, and diagnosis and procedure codes in this probabilistic matching.

Linking to Outside Data Sources

This data set includes 911 call center data, EMS data, and hospital data. It is also linked yearly to a number of census-tract level variables including: median household income, median age, race, average household size, unemployment rate, educational attainment, and urbanicity (the degree to which an area is urban). CARES can assist with linking outside datasets provided by the researcher to CARES data, but will strip the dataset of all identifiers before returning it to the researcher.

Data Contents

Partial List of Variables

Demographics (age, sex, race/ethnicity), cardiac arrest-specific data (location, type of arrest, witness status, presumed etiology), resuscitation-specific data, neighborhood-specific data (using Centrus Desktop Geocoder, linked to: median household income, median age, race, unemployment rate, educational attainment, etc.). See appendix E of the CARES Data Sharing User Guide for a full list of variables.

J-PAL Randomized Evaluations Using this Data Set


Other Research Using this Data Set

See a complete list of publications using the CARES data set here.

For an example of research linking CARES data to outside data, see this study linking it to Medicare data using probabilistic matching techniques:
Chan, Paul S. et al. 2016. “Long-Term Outcomes Among Elderly Survivors of Out-of-Hospital Cardiac Arrest.” Journal of the American Heart Association, 5. doi:10.1161/JAHA.115.002924.

Last reviewed