CARES study of 26,895 pediatric cardiac arrests builds survival risk scores from field data
Only 11.9% of children in the 2013-2023 registry data survived to discharge, and the authors say simple decision-tree scores sorted patients into low, moderate and high survival groups.
Key takeaways
- A study published November 10, 2026 in The American Journal of Emergency Medicine built survival risk scores from 26,895 pediatric out-of-hospital cardiac arrests in the CARES registry, 2013 to 2023.
- In that data, 27.8% of children survived to hospital admission and 11.9% survived to hospital discharge.
- The authors built three separate scores, for infants, for preschool and school-age children, and for adolescents, and found decision-tree models the most useful because they were simpler.
- The key predictors were gender, arrest location, witness status, cause of arrest, bystander CPR, prehospital AED use and first rhythm.
- For crews: the scores were validated only on a held-out 30% of the same registry, and the record does not describe any field use or external validation.
A study published November 10, 2026 in The American Journal of Emergency Medicine reports new risk scores for predicting which children survive an out-of-hospital cardiac arrest, built from prehospital variables that responders already record [1]. The authors, Mawani, McNally, Knight, Shen and Ebell, drew on the Cardiac Arrest Registry to Enhance Survival, known as CARES, for the years 2013 through 2023 [1].
The analysis included 26,895 pediatric patients, according to the PubMed record of the paper [1]. Of those, 27.8% survived to hospital admission and 11.9% survived to hospital discharge [1].
The authors wrote that pediatric out-of-hospital cardiac arrest "is associated with poor survival and neurological outcomes" and that their goal was to build and internally validate scores for survival to discharge [1]. The record available in the packet is the abstract. The full paper, its tables and the scores themselves are not included.
What the records show
The team used two statistical approaches, logistic regression and classification and regression trees, often shortened to CART [1]. They built the scores on 70% of the registry data and tested them on the remaining 30% [1]. Model performance was judged on discrimination and predictive accuracy [1].
Rather than one score for all children, the authors produced three, split by age: one for infants, one for preschool and school-age children, and one for adolescents [1]. The abstract does not give the age cutoffs for those groups.
The variables that mattered most were gender, arrest location, witness status, the etiology of the arrest, bystander CPR, use of a prehospital AED, and first rhythm [1]. All of these are fields that EMS agencies participating in CARES already report.
Both modeling approaches performed in a range the authors describe as good [1]. The area under the receiver operating characteristic curve, a standard measure where higher values mean better separation of survivors from non-survivors, was:
| Age group | Logistic regression | CART |
|---|---|---|
| Infants | 0.81 | 0.78 |
| Preschool and school-age | 0.77 | 0.76 |
| Adolescents | 0.82 | 0.82 |
Despite the slightly higher numbers for logistic regression in two age groups, the authors preferred the CART models [1]. They wrote that the tree models "are simpler to follow" and sorted more patients into the low and high survival categories while using fewer predictors [1].
Each score places a patient into a low, moderate or high survival category. The abstract reports the average survival probability for each category by age group [1]:
- Infants: 3% in the low group, 13% in the moderate group, 33% in the high group.
- Preschool and school-age children: 4%, 11% and 29%.
- Adolescents: 4%, 13% and 48%.
The spread is widest among adolescents, where the high survival group approaches one in two and the low group sits near one in twenty-five, per the figures in the abstract [1].
Why it matters for EMS
CARES is populated largely by EMS agencies, and the predictor list in this study is made up of information that is known on scene or within minutes of patient contact. Witness status, bystander CPR, whether a public AED was applied, and the first rhythm are all documented in the prehospital record. The authors describe the inputs as "easily identified prehospital variables" [1].
The authors frame the scores as tools for research and for monitoring the quality of medical services, and say they "can potentially support clinical decision-making" [1]. The abstract does not describe any specific decision the scores are meant to inform, such as transport, termination of resuscitation or destination choice, and it does not report testing in any of those settings. EMS clinicians follow local protocols and medical direction.
For systems that benchmark their pediatric outcomes, the baseline figures are the most immediately usable piece of the record. A survival-to-discharge rate of 11.9% across a decade of national registry data gives agencies a reference point for their own pediatric arrest numbers [1].
What the records do not tell us
The packet contains only the PubMed abstract. Several details that would be needed to assess or apply the scores are absent from that record:
- The scores themselves. The abstract does not list the decision-tree branches, the thresholds or the point values for any of the three age groups.
- Age definitions. The record does not state where the infant, preschool and school-age, and adolescent groups begin and end.
- External validation. Validation was internal, using 30% of the same registry [1]. The record does not report testing on data from outside CARES or from outside the United States.
- Neurological outcome. The outcome modeled was survival to discharge [1]. The abstract does not report whether the scores predict neurological status at discharge.
- Which CARES agencies and years contributed. The abstract gives the overall span of 2013 to 2023 but no breakdown by region or by participating agency.
- Calibration and confidence intervals. The abstract reports point estimates for discrimination and average survival by group, with no intervals.
- Funding and conflicts of interest. The record does not include this information.
The paper carries the DOI 10.1016/j.ajem.2026.10.040 and is listed as a journal article [1]. The record does not indicate whether it is in final published form or an early online version.
What to watch
The authors state the scores can facilitate research and quality monitoring [1]. The record does not name any planned external validation, implementation study or guideline review, so there is no next document or date to point to from the packet.
Questions readers ask
What percentage of children survive out-of-hospital cardiac arrest?
In the CARES registry data from 2013 to 2023 analyzed in this study, 27.8% of 26,895 pediatric patients survived to hospital admission and 11.9% survived to hospital discharge.
What factors predicted survival in pediatric cardiac arrest in the CARES study?
The authors identified gender, arrest location, witness status, etiology of arrest, bystander CPR, use of a prehospital AED, and first rhythm as the most important predictors.
How accurate were the pediatric OHCA risk scores?
Area under the curve values were 0.81, 0.77 and 0.82 for logistic regression and 0.78, 0.76 and 0.82 for CART across the infant, preschool and school-age, and adolescent groups. Validation was internal, on 30% of the same registry data.
Are the CARES pediatric cardiac arrest risk scores used by EMS in the field?
The abstract says the scores can facilitate research, monitor quality of medical services and can potentially support clinical decision-making. It does not report any field implementation or external validation.
Sources
- Mawani M et al. Predicting survival in pediatric out-of-hospital cardiac arrest: Development and internal validation of risk scores. Am J Emerg Med, 2026 (PubMed) — The American journal of emergency medicine (2026-11-10). pubmed.ncbi.nlm.nih.gov/43002825/
This is news, not medical advice. EMS clinicians: follow your local protocols and medical direction.


