Epidemiology and outcomes of in-hospital cardiac arrest in critically ill children across hospitals of varied center volume: A multi-center analysis

Objective

To describe epidemiology and outcomes associated with cardiac arrest among critically ill children across hospitals of varying center volumes.

Methods

Patients <18 years of age in the Virtual PICU Systems (VPS, LLC) Database (2009–2013) were included. Patients with both cardiac and non-cardiac diagnoses were included. Data on demographics, patient diagnosis, cardiac arrest, severity of illness and outcomes were collected. Hierarchical cluster analysis was performed to categorize all the participating centers into low, low-medium, high-medium, and high volume groups using the center volume characteristics (annual hospital discharges per center, annual extracorporeal membrane oxygenation per center, and annual mechanical ventilators per center). Multivariable models were used to evaluate association of center volume with incidence of cardiac arrest, and mortality after cardiac arrest, adjusting for patient and center characteristics.

Results

Of 329,982 patients (108 centers), 2.2% (n = 7390) patients had cardiac arrest with an associated mortality of 35% (n = 2586). In multivariable models controlling for patient and center characteristics, center volume was not associated with either the incidence of cardiac arrest (OR: 1.00; 95% CI: 0.95–1.06; p = 0.98), or mortality in those with cardiac arrest (OR: 0.93; 95% CI: 0.82–1.06; p = 0.27). These associations were similar across cardiac and non-cardiac disease categories. Furthermore, we demonstrated that there was no correlation between incidence of cardiac arrest and mortality in those with cardiac arrest across different study hospitals in adjusted models.

Conclusions

Both incidence of cardiac arrest, and mortality in those with cardiac arrest vary substantially across hospitals. However, center volume is not associated with either of these outcomes, after adjusting for patient and center characteristics.