Descriptive statistics are reported as mean (SD), median (25%, 75%), or number (percent) as indicated. Comparisons of continuous variables between two groups were made with t-tests; ANOVA was used for comparisons across three or more groups. Likelihood ratio chi-square analyses were employed when comparing discrete variables across groups. A univariable logistic regression model was used to calculate the odds ratios of survival across age groups and effect of initial cardiac rhythm. Multiple logistic regression was used to model the relationship between survival and potential predictors of outcome. Potential predictors derived from the adult literature and a priori expert opinion included age, witnessed arrest, bystander CPR, EMS scene time (<10 minutes versus ≥ 10 minutes), airway management, and attempts at vascular access.
An analysis of scene time among the pediatric age groups was performed for time <10 minutes. This interval was chosen a priori because the authors considered an EMS scene time <10 minutes as a “scoop and run” approach. Ten minutes was estimated to be the minimum time for EMS providers to arrive at the patient’s side, assess the patient, provide initial resuscitation efforts, and transfer the patient to the transporting vehicle. Post hoc analyses of scene time among the combined pediatric age groups versus adult scene time were performed.
The incidence rates were calculated per 100,000 person-years for the 12-month period March 1, 2006 to February 28, 2007 for both pediatric and adult populations to avoid bias introduced by seasonal variation. Because one site had incomplete data, only 10 sites were included in the incidence calculations. For each site, the age category and sex specific rates were calculated; these rates were standardized by age and sex to the North American population of the US 2000 census and Canadian 2001 census (21.4 million persons within the 10 ROC sites). The site rates, weighted by the site population, were then averaged to obtain overall rates. Additionally to counteract the possibility of incomplete ascertainment at each site, a “hot deck” multiple imputation scheme was used. For a given month, if an agency reported incomplete capture of cases, or if the number of submitted cases was substantially fewer than expected, based on the average agency rate over March to August 2006 (P<0.005), we assumed that ascertainment was not complete for that month at that agency. This was primarily an issue at the start-up of reporting for a few agencies. The number of cases for an agency during such a month was then imputed using a Poisson model adjusted for calendar month and agency within each site. Arrest characteristics such as age and sex for the imputed cases were determined by randomly sampling a case from the agency in question. This imputation was repeated 10 times; incidence rates were averaged over repetitions.12 (link),16 ,17 (link) Imputation was used for 3.5% of the pediatric cases and 5.1% of the adult cases.
These data were collected as part of an observational study that met the requirements for minimal risk research in the US and Canada and was approved by 74 Institutional Review Boards and 34 Research Ethics Boards. The authors had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.
An analysis of scene time among the pediatric age groups was performed for time <10 minutes. This interval was chosen a priori because the authors considered an EMS scene time <10 minutes as a “scoop and run” approach. Ten minutes was estimated to be the minimum time for EMS providers to arrive at the patient’s side, assess the patient, provide initial resuscitation efforts, and transfer the patient to the transporting vehicle. Post hoc analyses of scene time among the combined pediatric age groups versus adult scene time were performed.
The incidence rates were calculated per 100,000 person-years for the 12-month period March 1, 2006 to February 28, 2007 for both pediatric and adult populations to avoid bias introduced by seasonal variation. Because one site had incomplete data, only 10 sites were included in the incidence calculations. For each site, the age category and sex specific rates were calculated; these rates were standardized by age and sex to the North American population of the US 2000 census and Canadian 2001 census (21.4 million persons within the 10 ROC sites). The site rates, weighted by the site population, were then averaged to obtain overall rates. Additionally to counteract the possibility of incomplete ascertainment at each site, a “hot deck” multiple imputation scheme was used. For a given month, if an agency reported incomplete capture of cases, or if the number of submitted cases was substantially fewer than expected, based on the average agency rate over March to August 2006 (P<0.005), we assumed that ascertainment was not complete for that month at that agency. This was primarily an issue at the start-up of reporting for a few agencies. The number of cases for an agency during such a month was then imputed using a Poisson model adjusted for calendar month and agency within each site. Arrest characteristics such as age and sex for the imputed cases were determined by randomly sampling a case from the agency in question. This imputation was repeated 10 times; incidence rates were averaged over repetitions.12 (link),16 ,17 (link) Imputation was used for 3.5% of the pediatric cases and 5.1% of the adult cases.
These data were collected as part of an observational study that met the requirements for minimal risk research in the US and Canada and was approved by 74 Institutional Review Boards and 34 Research Ethics Boards. The authors had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.
