In this issue, Cha and colleagues present a retrospective study using a national database including 337 Korean Emergency Departments (ED). They investigate whether average cardiopulmonary resuscitation (CPR) duration in each ED (defined as the length of time CPR was conducted in patients who did not achieve return of spontaneous circulation (ROSC)) was associated with hospital survival. They included adult patients with presumed cardiac etiology of arrest from 2006 to 2010. Of the 107,736 patients identified, 69,455 had a presumed cardiac etiology of arrest and did not achieve ROSC prior to ED arrival. Almost half of those received no CPR in the ED, and after some additional exclusions, 30,691 patients were included in the analysis.

In patients who did not achieve ROSC, longer resuscitation times were associated with several factors generally linked to better outcomes, including initial shockable rhythm, witnessed arrest, and shorter transport time. Average ED CPR duration ranged from 11 to 45?min, and there was a slight positive correlation overall between CPR duration and survival. They then separated patients into 3 groups based on average duration of CPR in the ED. These groups were <20?min, 20–29?min and ?30?min. The three groups were very similar in most baseline characteristics recorded, including percentage of shockable rhythms, response interval and transport time. Using the 20–29?min group as a reference range, patients at an ED in the <20?min group had an adjusted OR for survival to discharge of 0.40 (95% CI 0.26–0.62), while those at an ED in the ?30?min group had an adjusted OR of 1.07 (95% CI 0.88–1.31). The authors conclude that average ED CPR duration of <20?min is associated with decreased survival. They refrain from drawing firm conclusions about the overall correlation due to the weakness of that relationship (Spearman’s coefficient 0.14).

This study includes some interesting findings, although with several limitations. The number of patients in the group with average CPR <20?min was comparatively very small (712, compared to approximately 15,000 in each of the other groups). Rates of bystander CPR also trended lower in this group, and it is possible that there were unidentified characteristics of the hospital (e.g., community vs. academic) or patient population (e.g., high prevalence of cancer or not) at these few centers that had an effect on outcome. The lack of short and long-term neurologic outcome is also a limitation, as it is not clear if the higher survival rate reflects survival with good neurologic outcome or if most of these patients were left with severe neurologic injury. The substantial higher mortality in the <20?min group however, as well as the overall correlation between CPR duration and survival, remains intriguing.

The authors chose to look at outcomes stratified by average ED duration of CPR in patients who did not achieve ROSC in order to minimize the bias of patients with good prognostic signs being resuscitated for longer. This theoretical bias is borne out in their data showing that longer resuscitation times at the individual level are associated with good prognostic factors such as younger age and shockable rhythm. By looking instead at how long ED personnel attempt resuscitation before declaring a death when efforts are not successful, they are attempting to get at the effect of average hospital practices rather than patient characteristics. This approach has been used before, notably with similar results. In 2012, Goldberger and colleagues utilized the Get With The Guidelines-Resuscitation database, which includes data on in-hospital cardiac arrest at 435 hospitals in the United States. They found that survival was higher in hospitals in the highest quartile of median CPR time in non-survivors (25?min), and lowest in hospitals in the lowest quartile (16?min). The authors of the current paper point out that Emergency Medical Services are different between countries (with only Basic Life Support being available in Korea, where this study was done), and that therefore caution should be used in applying these findings to other countries. The fact that similar results were obtained in their out-of-hospital population in Korea and an in-hospital population in the Unites States however suggests that this may be a more generalizable finding.

After the paper by Goldberger and colleagues, other investigators used the Get-With-The-Guidelines-Resuscitation database to look for patient and hospital factors associated with longer resuscitation efforts. They found that younger age and female gender were associated with longer resuscitation attempts in those without ROSC. This differs from the current study, which showed that men were more likely to have longer resuscitation attempts. Khan et al. also found that good prognostic signs such as shockable rhythm and witnessed arrest were associated with longer CPR. On the hospital level, being at a teaching hospital and in an urban setting were both associated with longer CPR efforts. The authors of the current paper reflect on a similar possibility in their discussion, suggesting that a possible reason for better outcomes at hospitals with longer average CPR duration could be that these are hospitals where more interventions are available, such as extracorporeal membrane support.

The majority of papers examining duration of resuscitation and outcome after cardiac arrest have focused on duration of resuscitation for individual patients who do obtain ROSC, not averages in non-survivors for institutions. This body of literature has generally shown that longer duration of arrest is associated with worse outcome, but even here there is some suggestion that resuscitation attempts may not be futile beyond 20 or even 30?min. In a study of pediatric in-hospital arrest in the Get-With-The-Guidelines-Resuscitation database, Matos and colleagues confirmed that longer downtime is associated with worse survival and worse neurologic outcome. However, they also found survival rates of approximately 16% even in patients who received CPR for >35?min, with 60% of those having a good neurologic outcome. These rates varied by patient category (better in surgical patients and worse in trauma patients). Other patient populations may be much less likely to recover after prolonged resuscitation, as illustrated in an observational study of pediatric drowning victims, where survival rates if CPR >30?min was required were essentially zero. In a prospective observational study of adult patients who obtained ROSC after out-of-hospital arrest and were treated with therapeutic hypothermia, longer downtime was associated with worse prognosis, but even in those with downtime >20?min, 23% had a good neurologic outcome. Clearly, these studies are each looking at different patient groups, and longer downtime for an individual remains a poor prognostic indicator. However, the body of literature examining this question from both the patient and the hospital perspective, including the work presented in this issue, appears to suggest that prolonged CPR and resuscitation may not be futile in the appropriate context. The authors of the current paper suggest that a randomized trial of CPR duration could help shed light on whether longer CPR duration would actually improve outcomes. Although this might be the ideal way to answer this question, this would be a difficult trial to design, both in terms of blinding and medical ethics. Prolonged CPR is almost certainly not a practice that should be employed indiscriminately, as there are cases where a favorable outcome is highly unlikely. Based on the relationship we see in this paper as well as the others discussed here however, identifying patients who might benefit from prolonged resuscitation should be a priority for future research.

Conflict of interest statement