Termination-of-resuscitation rule for emergency department physicians treating out-of-hospital cardiac arrest patients: an observational cohort study

The present study was a nationwide population-based observational study in all adult patients ages 18 years and older for whom resuscitation had been attempted after OHCA in Japan from 1 January 2005 to 31 December 2009. Cardiac arrest was defined as the cessation of cardiac mechanical activities confirmed by the absence of signs of circulation [17]. The cause of arrest was presumed to be of cardiac origin unless evidence suggested external causes (trauma, hanging, drowning, drug overdose and asphyxia), respiratory diseases, cerebrovascular diseases, malignant tumors or any other noncardiac causes. Attribution of noncardiac or cardiac etiology was made by the physicians in charge in collaboration with the EMS personnel. This study was approved by the Ethical Committee of Kanazawa University. The requirement for written informed consent was waived.

Japan has approximately 127 million residents in an area of 378,000 km², approximately two-thirds of which is uninhabited mountainous terrain [18]. Details on the Japanese EMS system have been described elsewhere [18, 19]. Briefly, municipal governments provide EMS through approximately 800 fire stations with dispatch centers. The Fire and Disaster Management Agency (FDMA) of Japan supervises only the EMS system nationwide, and individual EMS systems are operated by each local fire station. Generally, an ambulance crew consists of three EMS staff members, including at least one emergency lifesaving technician (ELST). ELSTs are allowed to use several resuscitation methods, including semiautomated external defibrillators, insertion of an adjunct airway, insertion of a peripheral intravenous line and administration of Ringer lactate solution. Since July 2004, only specially trained ELSTs have been permitted to insert a tracheal tube, and since April 2006, they have been permitted to administer intravenous epinephrine in the field under online physician instruction. Since October 2006, all EMS providers have performed cardiopulmonary resuscitation (CPR) according to the Japanese CPR guidelines [17], which are based on the 2005 AHA Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care [20]. As EMS personnel in Japan are legally prohibited from terminating resuscitation in the field, most OHCA patients undergo CPR by EMS providers and are transported to hospitals, except in cases where fatality is certain [18]. The length of the on-scene effort by EMS personnel is not predetermined before transport is initiated. Epinephrine use is implemented according to the FDMA resuscitation guidelines for ELSTs [17, 21]. The permitted dosage of epinephrine is 1 mg per attempt, and repeated doses may be administered under the physician’s instruction.

The FDMA launched a prospective population-based observational study of all OHCA victims who received EMS in Japan [18] since January 2005. EMS personnel at each center, using an Utstein-style template, recorded data for OHCA victims with the cooperation of the physician in charge [22]. All data were transferred and stored in a nationwide database developed by the FDMA for public use. For this study, we analyzed this anonymous database with the permission of the FDMA. The main items included in the database are patient sex, age, etiology of arrest (presumed cardiac origin or not), bystander witness status, bystander CPR with or without automated external defibrillator use, initial identified cardiac rhythm, bystander category (that is, if there was a bystander, then whether the bystander was a layperson or EMS personnel), whether epinephrine was administered, whether advanced airway management (including endotracheal tube, laryngeal mask airway and esophageal-tracheal tube) were used, whether ROSC was achieved before arrival at the hospital, time of the emergency call, time of vehicle arrival at the scene, time of ROSC, time of vehicle arrival at the hospital, one-month survival and neurological outcome at one month after cardiac arrest. Neurological outcome was defined on the basis of the Cerebral Performance Category (CPC) scale: category 1, good cerebral performance; category 2, moderate cerebral disability; category 3, severe cerebral disability; category 4, coma or vegetative state; and category 5, death [22]. This CPC categorization was determined by the physicians in charge. The call-to-response time was calculated as the time from the emergency call to the time of vehicle arrival at the scene. The call-to-hospital arrival time was calculated as the time from the emergency call to the time of vehicle arrival at the hospital.

The main outcome measures were specificity, positive predictive value (PPV) and area under the receiver operating characteristic (ROC) curve for the newly developed TOR rule for emergency department physicians. Secondary outcome measures were those related to the BLS TOR rule.

Kolmogorov–Smirnov and Lilliefors tests were performed to evaluate the distributions of continuous variables, and we found that all continuous variables were not normally distributed (all P < 0.01). Therefore, the Wilcoxon and Kruskal–Wallis tests for continuous variables and the χ² test for categorical variables were performed to compare the basic characteristics of patients between the development and validation groups. Multivariate logistic regression analyses including 11 variables were performed to assess the association between prehospital variables and one-month death or unfavorable neurological outcome in the development group. Continuous variables are expressed as medians (interquartile range (IQR)) [1st to 3rd quartiles], and categorical variables are expressed as percentages. As an estimate of effect size and variability, we report odds ratios (ORs) with 95% confidence intervals (CIs). All statistical analyses were performed with the JMP version 10 statistical discovery software package (SAS Institute Inc, Cary, NC, USA). All tests were two-tailed, and P < 0.05 was considered statistically significant.

The results of multivariate logistic regression analyses, including 11 variables to determine the factors associated with one-month death and CPC categories 3 to 5, are shown in Table 2. No prehospital ROSC had the highest adjusted OR for one-month death (adjusted OR, 25.8; 95% CI, 24.7 to 26.9) followed by unshockable initial rhythm (adjusted OR, 2.76; 95% CI, 2.54 to 3.01) and unwitnessed by bystanders (adjusted OR, 2.18; 95% CI, 2.09 to 2.28). We selected three prehospital variables that had adjusted ORs greater than 2.0 for predicting one-month death as predictors for one-month outcomes. Therefore, these three variables were incorporated into the new TOR rule, meaning that TOR was newly defined as appropriate in cases fulfilling all three criteria: no prehospital ROSC, unshockable initial rhythm and unwitnessed cardiac arrest by bystanders. No prehospital ROSC was also strongly associated with one-month CPC categories 3 to 5 with an adjusted OR of 38.4 (95% CI, 35.9 to 41.0).

Table 3 shows the results of the analysis of derivation (development group) and validation (validation group) data sets of the new TOR rule for predicting one-month death. The rates of patients who fulfilled all three criteria in the development and validation groups were 57.1% and 57.3%, respectively, both of which would be considered futile. The new TOR rule showed a specificity of 0.875 (95% CI, 0.870 to 0.881), PPV of 0.992 (95% CI, 0.991 to 0.992) and area under the ROC curve of 0.851 (95% CI, 0.850 to 0.852) in the development group. The specificity, PPV and area under the ROC curve for the validation group were 0.903 (95% CI, 0.894 to 0.911), 0.993 (95% CI, 0.992 to 0.993) and 0.874 (95% CI, 0.872 to 0.876), respectively. Table 4 shows the results of the analysis of derivation and validation data sets of the new TOR rule for predicting a one-month unfavorable neurological outcome. The new TOR rule showed a specificity of 0.939 (95% CI, 0.933 to 0.945), PPV of 0.998 (95% CI, 0.998 to 0.999) and area under the ROC curve of 0.922 (95% CI, 0.921 to 0.923) in the development group. The specificity, PPV and area under the ROC curve for the validation group were 0.966 (95% CI, 0.958 to 0.973), 0.999 (95% CI, 0.999 to 0.999) and 0.942 (95% CI, 0.941 to 0.944), respectively.

Table 5 shows the results of analysis of the 2009 validation data set (N = 105,030) of the BLS TOR rule for predicting one-month death. The rate of patients who fulfilled all three criteria of the BLS rule in the validation group was 81.6%, which would be considered futile. The BLS TOR rule showed a specificity of 0.779 (95% CI, 0.767 to 0.791), PPV of 0.988 (95% CI, 0.988 to 0.989) and area under the ROC curve of 0.811 (95% CI, 0.809 to 0.813). Table 6 shows the results of the analysis of the validation data set of the BLS TOR for predicting one-month unfavorable neurological outcomes. The BLS TOR rule showed a specificity of 0.928 (95% CI, 0.917 to 0.938), PPV of 0.998 (95% CI, 0.997 to 0.998) and area under the ROC curve of 0.880 (95% CI, 0.871 to 0.889).

The potential limitations of the current analysis are as follows. First, we did not evaluate detailed in-hospital interventions. We assumed that OHCA patients received standard ALS according to the Japanese CPR guidelines [17], which were based on the 2005 AHA guidelines [20]. Second, there may be unmeasured confounding factors that could have influenced outcomes. However, the use of uniform data collection on the basis of Utstein-style guidelines for reporting cardiac arrest, the large sample size and a population-based design were intended to minimize these potential sources of bias. Third, the relevance of our results to other communities with different emergency care protocols and different OHCA causes remains unknown. Although TOR outside the hospital is not allowed in some Asian countries [31‐33], EMS systems in those countries are different from the Japanese EMS system. Therefore, studies in other countries may be required to validate our results. Fourth, emergency department physicians must pay attention to the use of the new TOR rule for patients with special conditions such as accidental hypothermia. Fifth, emergency department physicians should decide whether further medication is needed, even if patients meet the newly derived TOR rule. Finally, no measure of interrater reliability is obtainable from the data set to guide the determination of the rate of misclassification.