Untangling sex disparities in cardiac arrest survival: navigating the complex interplay of predictors

This editorial refers to ‘Underlying reasons for sex difference in survival following out-of-hospital cardiac arrest: a mediation analysis’ by C. Miedel et al., https://doi.org/10.1093/europace/euae126.

Much effort has been put into improving survival following out-of-hospital cardiac arrest (OHCA), including training the population in cardiopulmonary resuscitation (CPR), optimizing dispatcher-recognition of OHCA and dispatch-assisted CPR, early defibrillation, and advanced in-hospital care.¹ Following these initiatives, survival has improved in many communities, but mostly among selected patient groups such as those with witnessed cardiac arrest, an initial shockable heart rhythm, of male sex, those who received bystander CPR, or early defibrillation.² Female patients are often older, have more comorbidities, are more likely to live alone, and thus have an unwitnessed arrest with a non-shockable heart rhythm, as first-recorded rhythm.^3–5 It has also been reported that women are less likely to receive bystander CPR, early defibrillation, and advanced in-hospital care.^4,6 Thus, not surprisingly, the greatest improvement in OHCA survival has been observed among male patients.^3–6 However, results have been conflicting as to which factors are the most significant drivers of this survival disparity. The most important remaining questions are how much of the disparity in observed survival can be attributed to modifiable factors such as timely recognition of cardiac arrest, bystander CPR, or use of automated external defibrillator (AED) and how much is due to unmodifiable factors such as age, unwitnessed status, or prior comorbidities.

Since it is not possible to conduct a randomized controlled trial to understand the effect of sex on survival and how different patient and cardiac arrest characteristics may affect survival differently, according to sex, it is only possible to try to understand these relationships through registry studies using epidemiological methods. Sweden has a long-standing tradition of high-quality OHCA data collection and initiatives to improve pre-hospital and in-hospital care that have been associated with improved OHCA survival.⁷

In Europace, Miedel and colleagues report ‘Underlying reasons for sex difference in survival following out-of-hospital cardiac arrest – a mediation analysis’.⁸ The authors have investigated 30-day survival after OHCA according to sex, and the underlying covariates that may explain potential differences in survival. The study included 43 293 non-EMS witnessed OHCAs from the well-established, high-quality Swedish Registry for Cardiopulmonary Resuscitation, from 2010–2020. Out-of-hospital cardiac arrests were linked to the National Patient Registry and Statistics Sweden for access to comorbidity and socioeconomic factors, respectively. A modified Poisson Regression Analysis and multivariable analyses were used to explain differences in 30-day survival according to sex. The importance of mediators was investigated using a mediation analysis. The main results were that women were less likely to survive 30 days following OHCA mainly due to lower proportion of shockable heart rhythm, older age, and lower income.

There are interesting methodological aspects of this study. The investigators provided analyses where a range of covariates were treated as confounders and then an analysis where the importance of the same variables was examined, but this time, as mediators. To understand the difference between the two approaches, it is necessary to consider and establish the time from which survival is examined. If time zero is defined as the time of cardiac arrest, then sex and age can be considered confounders since they are established at the time of cardiac arrest. But other important variables including bystander CPR, initial shockable heart rhythm, and defibrillation occur later than time zero, and should then be defined as mediators (or intermediates). So, any analysis that has the time of cardiac arrest as time zero needs to examine most variables as mediators. Another relevant possibility is to define time zero as when the ambulance arrived, and initial heart rhythm was assessed. In this scenario, bystander CPR, prior defibrillation, and initial heart rhythm have been established, and thus they can be viewed as confounders. Therefore, depending on the perspective of the study i.e. the definition of time zero, both approaches are relevant, but their differences should be appreciated. Establishing time zero at the time of ambulance arrival is highly relevant for an analysis focused on patient treatment since all information is available at this time, whereas an analysis focusing on cardiac arrest itself as time zero is more relevant if the main interest of the analysis is the basic characteristics of the patient who suffered the cardiac arrest.

For the current analyses, we learn most about the basic characteristics of women from the mediation analysis and most about treatment perspectives from the confounder analysis. Both are important, but there is a caveat. Confounder analysis is generally well understood. For the important variable initial shockable rhythm, we can present survival for the two sexes with and without an initial shockable rhythm and these four proportions are easy to interpret. It is more difficult with the mediation analysis. We learn from the calculations that nearly half of the lower survival among females was due to a non-shockable heart rhythm. But what does this actually mean? In many cases, it is the proportion mediated on a scale depending on the regression method, and this is not easy to translate to a clinical message. Obviously, the presence of an initial shockable heart rhythm is an important mediator, but the quantitative message is not easy to appreciate. Fortunately, in the current study, the message does not change when important covariates are considered either confounders or mediators, but the example also shows that setting time zero such that a confounder analysis is relevant provides the results that are most easily understood.

The first important message from this simultaneous conduction of analyses addressing mediation and confounding is that statistical measures that are most easily clinically interpreted should be the preferred option when choosing statistical analyses for a clinical audience.

The second most important message from this study is that, in Sweden, the lower survival among female patients with cardiac arrest seems to be driven by a lower proportion of shockable rhythm, higher age, and lower income, compared with male patients. The first two factors are not easily targeted by sex-specific interventions and thus not likely modifiable. Whether a higher income among female patients would increase their chances of survival remains unknown. However, higher income has been associated with higher survival across several patient groups, for both men and women. Importantly, the main findings from this study are not likely generalizable to communities where female patients are less likely to receive bystander CPR and/or defibrillation prior to EMS arrival. For this reason, analyses such as the present study are important to tease out modifiable factors across different communities striving to improve outcomes for patients that have suffered an OHCA. Future studies should also include data from the first link in the chain of survival, recognition of cardiac arrest, as sex disparities in recognition of OHCA may play an important role and can be targeted through community interventions.