We conducted a multicenter prospective observational study of patients from 64 participating hospitals using the non-traumatic nationwide OHCA registry data.14 The present study evaluated the epidemiology and outcomes of patients that suffered OHCA with a presumed medical etiology in Korea and is registered with ClinicalTrials.gov, number NCT03222999. To enhance the effectiveness in research, several research committees have been organized. In addition, the Data Safety and Monitoring Board Committee was organized to provide data quality control.14

Data were obtained from the KoCARC-pro registry database in Korea for cardiac arrest events that occurred between October 1, 2015 and June 30, 2017. Patients older than 18 years with OHCA who were transported to participating hospitals via emergency medical service (EMS) with resuscitation efforts initiated and whose arrest was identified as being of a presumed medical etiology by emergency physicians in the ED were included. Patients with obvious signs of death (e.g., rigor mortis or dependent lividity), terminal illnesses documented in their medical records, those under hospice care, pregnant women, and those with pre-documented ‘Do Not Resuscitate’ cards were excluded. Additionally, those with non-cardiac etiology arrest, those with missing hospital outcomes, patients with a missing first recorded rhythm and cases of trauma, poisoning, electrocution, primary respiratory arrest, drowning, asphyxia, hanging, and other injuries were also excluded.14

The following information was extracted from the KoCARC database: demographics, including gender, age, home address, and location of event; CPR-related characteristics such as presence of a witness, CPR by a bystander, initial ECG rhythms obtained pre-hospital arrival or in the ED, and provision of CPR by EMS or in the ED (epinephrine, advanced airway, antiarrhythmic, extracorporeal membrane oxygenation [ECMO]); CPR-related time variables such as basic life support (BLS) interval (time from collapse to initiation of chest compression attempt at the scene), defibrillation interval (time from collapse to initiation of electrical shock), and pre-hospital interval (time from collapse to arrival in an ED); and clinical outcomes, including presence of ROSC, ED outcomes (admitted, died, or transferred), hospital outcomes (discharged alive, died, or transferred), neurological outcomes at the time of discharge, and whether post cardiac arrest treatment (i.e., target temperature management [TTM] or percutaneous coronary intervention [PCI]) was provided.

Patients with a first documented ECG rhythm in the field or in the ED were divided accordingly into 3 rhythm categories: initial shockable (IS), converted to shockable (CS), and never shockable (NS).4 We extracted patients with OHCA meeting the criteria for each of the TOR rules and calculated the sensitivities, specificities, positive predictive values (PPV), and negative predictive values (NPV) and their respective 95% confidence intervals (CI) to identify OHCA patients with a risk of mortality and poor neurologic outcomes at hospital discharge.1115

The 6 combinations of TOR criteria were as follows (Table 1):

Survival to discharge was defined as final discharge to home or transfer to another facility after admission to the hospital. Neurological status was quantified using Cerebral Performance Category (CPC) scores, which are based on a 5-point scale in which scores of 1 (good recovery) and 2 (moderate disability) are defined as favorable neurologic outcomes.

The primary outcome measures were based on the group proportions of survival to hospital discharge and of survival with a favorable neurologic outcome in the 3 ECG rhythm categories. The secondary outcome measure was validation of previously researched TOR rules.

Data were analyzed using IBM SPSS Statistics version 23.0 (IBM corp., Armonk, NY, USA) and MedCalc 17.4.4 version (MedCalc Software, Mariakerke, Belgium). Descriptive statistics are presented as medians with interquartile ranges (interquartile range [IQR]: 25th and 75th percentiles) and categorical variables are presented as counts and percentages.

The significance of differences between the 3 groups (IS, NS, and CS) was tested using the Kruskal-Wallis test or analysis of variance with a post-hoc analysis (correcting for multiple comparisons with the Bonferroni method) for continuous variables, and the χ² test or Fisher's exact test for categorical variables.

Before conducting the multivariable analysis, we conducted a co-linearity test of variables associated with OHCA events and time-related parameters. In this study, the tolerance of each variable was larger than 0.1 (range, 0.72–0.98), while the variance inflation factor was less than 10 (range, 1.02–1.40). This demonstrates that the variables introduced into the model generally did not exhibit multi-collinearity. We used logistic regression for the multivariable analyses, and the outcome of interest was survival to hospital discharge. To determine the logistic model calibration, we calculated the Hosmer-Lemeshow goodness of fit. Odds ratios (OR) greater than 1 were indicative of a beneficial effect on survival. ORs and 95% CIs were derived for all covariates. The characteristics of the adjusted OR are described using Forest plots. P < 0.05 was considered statistically significant.

The study was approved by the Institutional Review Board of Kyungpook National University Hospital (KNUH 2015-11-013), which waived the requirement of informed consent. Approval for this study was obtained from the institutional review boards of each of the 64 participating hospitals.

Of 4,757 patients registered in the KoCARC-pro registry database during the 21 months of registration, 4,608 were analyzed after excluding 6 patients with in-hospital cardiac arrest and 143 patients younger than 18 years. Among the 4,608 patients, 65.1% were men, and 59.5% had a witnessed OHCA (Fig. 1 and Supplementary Table 1). Furthermore, 46.7% received CPR from a bystander. A total of 13.3% of the patients had ROSC prior to hospital arrival. The percentages of patients who underwent PCI, TTM, or ECMO were 11.7%, 9.3%, and 2.9%, respectively.

Of the 4,608 OHCA patients, 389 cases who had unmatched medical records or whose pre-hospital time variables had not been recorded were excluded. Overall 4,219 patients were enrolled into the survival analysis. According to the patients' confirmed ECG rhythms and any subsequent rhythm changes, participants were divided into the IS (n = 821), CS (n = 498), and NS rhythm groups (n = 2,900) (Fig. 1 and Table 2). The IS group was relatively younger with a higher proportion of men, public events, witnessed cardiac arrests, bystander CPR, and pre-hospital defibrillation rates (Table 2). The CS and NS groups had similar rates of pre-hospital witnesses, bystander CPR, and degrees of acidosis in hospital. The CS group had the lowest pre-hospital ROSC prior to ED arrival and survival to discharge (5% and 4.2%, respectively) among the 3 groups, and, similar to the IS group, favorable neurologic outcomes to discharge was low.

The total survival to discharge rate was 11.7%, and survival with good neurologic outcomes was 7.9%. In the IS group, these rates were 35.3% and 30.2%, respectively, showing that those in the IS group had better neurologic outcomes than those in the NS group which had a total survival to discharge rate of 5.7% and survival with good neurologic outcome rate of 1.9% (crude OR, 9.0 [95% CI, 7.28–11.12] and 22.39 [95% CI, 16.49–30.39], respectively, Table 3). In the CS group, rates of total survival to discharge and survival with good neurologic outcomes (4.2% and 2.0%, respectively) were not significantly different than those of the NS group (crude OR, 0.73 [95% CI, 0.47–1.15] and 1.06 [0.54–2.09]).

Multivariate analysis identified ECG rhythms, age (< 65 years), place of cardiac arrest, presence of a witness, and ROSC before ED arrival as factors related to survival to discharge (Fig. 2A). Factors related to discharge with good neurologic outcomes were ECG rhythm, age, presence of a witness, and ROSC before ED arrival (Fig. 2B). Although gender and bystander CPR were significant in the univariate analysis, they were not significantly associated with survival outcomes in the multivariate analysis (Table 3).

We validated the TOR criteria against the 4,608 adult cardiac arrest cases in this study. When the traditional TOR-BLS and TOR-ALS rules, which do not include ECG rhythms, were applied to our patient pool, the numbers of patients who met TOR criteria but survived to discharge were 118 (2.6%) and 17 (0.4%), respectively (Table 4). Additionally, the number of patients who were discharged with a CPC score of 1 or 2 were 27 (0.6%) and 4 (0.1%), respectively (Table 5). When the Goto's TOR rule, which considers TOR for non-shockable rhythms in the field with unwitnessed events and no pre-hospital ROSC, was applied, the number of patients who met TOR criteria but survived to discharge was 27 (0.6%), and the number of patients who were discharged with CPC scores of 1 or 2 was 4 (0.1%). However, all patients who met the SOS-KANTO's TOR rule (asystole in the field and in the ED), new TOR model 1 (asystole in the field), or new TOR model 2 (asystole in the ED) criteria, had poor neurologic outcomes of CPC 3 or higher, with none of the patients showing favorable neurologic state with CPC 1 or 2 (specificity = 1.000). Regarding their effectiveness at detecting poor outcomes based on the area under the curve (AUC), the international TOR-BLS, Goto's TOR rule, new TOR model 1, and new TOR model 2 criteria (which include the absence of pre-hospital ROSC) had higher AUCs than the SOS-KANTO criteria, which does not include the absent pre-hospital ROSC condition (Tables 4 and 5). Overall, the new TOR model 2 criteria were the most effective at predicting in-hospital mortality and unfavorable neurologic outcomes, with AUCs of 0.853 and 0.911, respectively. Once TOR was established based on these criteria, none of the patients met criteria for CPC 1 or 2 (all specificity and PPV = 1.000).

This study had a few limitations. First, as with other multicenter observational studies, the integrity, effectiveness, and ascertainment bias of the data are potential limitations. We strove to reduce the potential biases in this multicenter observational study by implementation of the quality control committee in KoCARC. Second, facility or regional differences in EMS resources, CPR quality, and post-cardiac arrest care may affect survival outcomes; however this study did not analyze differences in inter-hospital variabilities. These variables may have potentially introduced confounding effects. Third, the specificity of the criteria for in-hospital mortality was not 1.0. If ECMO-CPR is performed or a new CPR-device is applied according to technological developments, the TOR rules could be changed.22 Withdrawal or termination guidelines for ECMO-CPR have not been studied previously, and the volume of ECMO cases in this study was as low as 2.9%, and thus could not be analyzed separately. TOR rules in patients with ECMO criteria and socio-ethical issues should be discussed and a consensus process will be needed. Therefore, externally validating the new TOR criteria obtained in this study using KoCARC data collected in a particular period in the future through a prospective study would be necessary to apply the criteria to patients with a low possibility of inpatient or other clinical situation recoveries.

This study found that patients with non-shockable rhythms on the pre-hospital initial ECGs that were converted to shockable rhythms still had poor survival outcomes, comparable to those of the NS group. Furthermore, survival outcomes and TOR predictions varied widely depending on the type of ECG rhythms and time of ECG documentation. In the future, types of ECG rhythms as well as time of ECG documentation should be considered in treatment algorithms and prognostication among OHCA patients.