Socio-economic differences in incidence, bystander cardiopulmonary resuscitation and survival from out-of-hospital cardiac arrest: A systematic review Resuscitation

Background: Individuals with a low socioeconomic status (SES) may have a greater mortality rate from out of hospital cardiac arrest (OHCA) than those with a high SES. We explored whether SES disparities in OHCA mortality manifest in the incidence of OHCA, the chance of receiving bystander cardiopulmonaryresuscitation (CPR)orinthechanceofsurvivinganOHCA.Wealso studiedwhethersexandage differencesexistinsuchSESdisparities. Methods: The Medline, Embase and Scopus databases were searched from 01-01-1993 until 31-01-2019. Studies utilising any study design or population were included. Studies were included if the exposure was SES of the OHCA victim or the OHCA location and the outcome was either OHCA incidence, CPR provision and/or survival rate after OHCA. Study selection and quality assessment were conducted by two reviewers independently. Descriptive data and measures of association were extracted, both in the total study population and in subgroups stratified by age and/or sex. This review was carried out following the PRISMA guidelines. Results: Overall 32 studies were included. Twelve studies reported on OHCA incidence, thirteen on bystander CPR provision and fourteen on survival. Some evidence for SES differences was found in each identified stage. In all the studies on incidence, SES was measured over the area of the OHCA victims ’ residence and was consistently associated with OHCA. In studies on bystander CPR, SES of the area in which the OHCA occurred was associated with bystander CPR, while evidence on individual SES was lacking. In studies on OHCA survival, SES of the victim measured at the individual level and SES of the area in which the OHCA occurred were associated, while SES of the victim, measured at the area of residence was not. Studies reporting age and sex differences in the SES trends were scarce. Conclusion: SES disparities in OHCA mortality likely manifest in OHCA incidence, bystander CPR provision and survival rate after OHCA. However, there is a distinct lack of data on SES measured at the individual level and on differences within subgroups, e.g. by sex and age.


Introduction
Out of hospital cardiac arrest (OHCA) is the sudden cessation of effective circulation due to the absence of cardiac pump function, in an out of hospital setting. 1 OHCA has been reported to be responsible for approximately half of all cardiac related deaths and to account for up to 20% of all natural deaths in industrialised countries. 2,3 Marked differences have been shown to exist in OHCA incidence and survival globally. 4 Although great strides have been made in resuscitation in some settings, 5 survival rates remain relatively low. Thus, improvements in both prevention and resuscitation strategies are required to decrease this vast public health burden. Such efforts to reduce OHCA incidence are likely to benefit from identifying high-risk populations.
Socioeconomic status (SES) disparities occur in OHCA mortality (e.g. higher level of income is associated with lower mortality), [6][7][8] but it is unclear from single studies whether SES differences in OHCA mortality manifest mainly in the risk of having an OHCA, in the chance of receiving bystander CPR or in the chance of surviving an OHCA. SES differences in OHCA incidence are conceivable, as established SES gradients in coronary heart disease (CHD) risk factors are likely to translate to OHCA incidence. CHD may lie on the causal path between OHCA and its risk factors or share common risk factors with OHCA. 3,[9][10][11][12][13][14][15] Likewise, the SES of the area in which an OHCA occurs may be associated with (un)favourable resuscitation characteristics such as provision of bystander cardiopulmonary resuscitation (CPR), e.g. due to a lack of knowledge regarding CPR in lower SES areas. 16 Finally, SES of the individual is likely to be associated with OHCA survival outcome, for instance due to the generally higher comorbidity burden among low SES groups. 17 It is unclear which subpopulations may be especially vulnerable. Different individual SES indicators (e.g. income, education level, occupational status or composite measures of deprivation) may differ in their ability to identify vulnerable subgroups with regard to OHCA incidence, bystander CPR provision and survival. Importantly, these indicators of SES, when measured at the individual level, may have differing associations with health outcomes than indicators measured at the area level. 18 Additionally, age and sex differences exist in the socioeconomic gradients for cardiovascular risk factors and health outcomes. 19,20 It is likely that such differences are also found in OHCA. Exploring potential age and sex differences in the socioeconomic gradient of OHCA may serve to further identify groups with high risk of OHCA incidence and/or low chances of surviving an OHCA. Due to the multiplicity in exposure and outcome definitions as well as study populations, patterns of SES differences in OHCA cannot be interpreted from single studies. Thus a systematic review was carried out in order to assess the patterns of SES differences in OHCA.
We carried out a systematic review to explore at which stage, in which populations and which locations disparities in OHCA mortality emerge. We assessed the associations between SES at the level of the individual or that of the area in which OHCAs occurred and (1) the incidence of OHCA, (2) bystander CPR and (3) survival rate of OHCA. Moreover, we compared the consistency across individual SES indicators. Lastly, we investigated whether these SES gradients differ by sex or age group.

Methods
The review was written in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guidelines.

Definitions
SES was defined as a variable describing the relative position in society of an individual or a population, which could include resources and/or skills within the financial, occupational and/or educational domains. SES was defined at the individual level by indicators such as income, highest educational level attained and occupational status. When SES was measured at the individual level, it was referred to in this review by "SESind". In addition, the SES level of the area of residence of the OHCA victims was considered as a proxy for the SES of the individual, henceforth referred to as "SESres". Also, in order to capture the environmental factors that may influence the provision of bystander CPR, SES of the area in which the OHCA occurred was considered, which was referred to in this review as "SESloc". Thus, while SESres and SESloc both measure SES over an area, one characterised the individual based on where the person lives (SESres) and the other the environment where the OHCA took place (SESloc).
OHCA was defined as a sudden cessation of effective circulation due to the absence of cardiac pump function, in an out of hospital setting. Bystander CPR was defined as out-of-hospital CPR provided by bystanders with or without usage of an automated external defibrillator (AED). Survival after OHCA was defined as being alive at the emergency room with return of spontaneous circulation (ROSC), hospitalized with ROSC at any hospital ward (i.e., hospital admission) or being alive at hospital discharge.

Search strategy
We performed a search for studies, worldwide, reporting associations between individual SES or SES of the area in which an OHCA occurs and OHCA incidence, bystander CPR provision and survival after OHCA. A search was performed in the Embase, Medline and Scopus databases from 01-01-1993 databases until 31-01-2019. The lower bound of the year of publication was based on the oldest citation identified 16 in a preliminary search. The search string was developed in collaboration with a medical research librarian (JD), as detailed in Table 1 and Appendix 1. Any study design was permitted as long as it contained original (un)published scientific findings. The search strategy included commonly used indices for SES such as income, education, occupation and unemployment as well as measures of SES such as 'the Townsend deprivation index' and common synonyms for all the identified SES indicators.

Study selection
The results of the search were screened by two reviewers (BN & IO) independently, first by title and abstract, followed by full text review, if not already excluded in the abstract screening. Disagreements between the two reviewers about the inclusion of articles, were resolved by consensus or with arbitration by a third reviewer (IV), if consensus was not reached. Inclusion criteria of the citations are described in Table 1.

Quality assessment
Quality assessment was carried out for each study by two reviewers (BN & IO) independently using the National Institutes of Health (NIH) quality assessment tool for cross-sectional studies, or the appropriate NIH tool for quality assessment for other study designs, as found in Appendix 2. Disagreements between the two reviewers about the quality assessment of the included citations were resolved by consensus or with arbitration by a third reviewer (IV). No studies were excluded on the basis of quality assessment results. However, the quality of the included studies was presented to facilitate interpretation of the results.

Data extraction
The extraction form included identifiers of the citation (including the name of the first author, year of publication and journal), information about the population under study (e.g. city/region, country, ethnic groups, population size and age range), the definitions of exposure and outcome variables used and the results (i.e. risk ratios or odds ratios and confidence intervals). The Supplementary materials, if available, were also checked for relevant results. The extraction table was first piloted in 2 articles 16,21 and checked for inconsistencies between the two reviewers (BN & IO) before proceeding with data extraction from the remaining citations. Information was extracted from published manuscripts. If associations were reported separately across the sexes or in different age groups, this information was also extracted.

Synthesis of results
We gave an overview of the SES gradients by assessing direction, strength and consistency of associations between individual SES indicators and SES of the area in which an OHCA occurred within each outcome measure (OHCA incidence, bystander CPR provision or survival rate after OHCA). We specifically compared consistency across SESind indicators. To be able to understand differences in survival, we also included resuscitation characteristics that may influence the association between SES and survival rate: location of OHCA (residential versus public setting), initial rhythm of the OHCA victim (shockable rhythm versus non-shockable rhythm), whether bystander CPR was performed (yes versus no), witnessed status of the OHCA (yes versus no) and time to defibrillator connection. Moreover, we compared the relative SES differences in incidence and survival across age groups and sex.

Stage: OHCA incidence
For analyses of OHCA incidence all the included studies reported associations based on the number of events occurring within one year, apart from one study which reported lifetime cumulative incidences. All twelve of the studies that reported on OHCA incidence reported that low SESres was associated with a higher OHCA incidence than high SESres (Fig. 2, Table 3). As all the studies measured SES with an area proxy, it was not possible to compare associations found with OHCA incidence across different indicators of SES (e.g. education level vs. income).
One citation, with sex-specific information, reported evidence for a modifying effect of sex on the association between SESres and OHCA. 38 Pujades-Rodriguez found women with low SESres to have a higher risk of OHCA incidence than high SESres women, while for men this was not the case. Both the papers that stratified results by age found a stronger trend in those aged 0-64 than in the older group. 41,42 For instance, Reinier (2006) reported a weaker association, although in the same direction, between SESres and OHCA incidence in the group aged over 65 compared to the younger age group. Any study design a Any study reporting on mortality from sudden cardiac arrest i.e. the incidence of sudden cardiac death or sudden unexpected death were excluded on the basis of being composed of both the incidence and survival rates of a given population; two aspects we aimed to highlight separately and compare (eTable 2, eTable 3).

Stage: bystander CPR provision
All thirteen studies that reported on bystander CPR provision, measured SESloc, e.g. by household income, mean property value, education, employment or deprivation (Table 4). Studies on individual SES were lacking. Ten of the twelve studies, reporting on CPR provision, found a low SESloc to be associated with a low rate of bystander CPR (Fig. 2). The single paper to investigate AED connection rates by bystanders, found a significant association between a low SESloc and lower AED connection rates, when adjusted for resuscitation characteristics. 22 None of the studies reported associations separately by age or sex groups.

Stage: survival rate after OHCA
Of the fourteen studies that reported on survival rate, five measured SESind, 16 Table 5). Thirteen studies defined the outcome as survival to hospital discharge and one defined the outcome as 30-day survival. 33 The associations between individual SES and survival were mixed, with a clear difference between individual indicators compared to area level indicators (Fig. 2). All five studies that reported results on SESind found a significantly lower survival in the unemployed, 49 those with low property value 16,27,50 and those with low education 51 , but not low occupation level. 51 In contrast, only one of the studies that assessed SESres found a significant difference in OHCA survival rate. 33 For SESloc, all four studies, 21,26,43,45 found a significant trend between higher SES and higher survival. Two of the four studies adjusted for individual level demographic variables, resuscitation characteristics and bystander CPR, factors that may function as mediators in the relationship between SESres and survival rate. 21,49 Four studies 21,24,52,53 reported ROSC on top of survival to hospital discharge (eTable 1). Studies showed consistent results across ROSC and survival to hospital discharge. There was a lack of evidence on differences in the SES gradient by sex and age. Wells et al. 51 found no significant difference between men and women in the association between individual level education or occupation and OHCA survival. None of the other studies made comparisons by sex or age in the association between individual level SES and OHCA survival, nor between SES of the area in which the OHCA occurred and survival rate.

Key findings
In this systematic review, we investigated where the SES differences in mortality from OHCA are likely to stem from by comparing three major stages at which SES differences may arise (OHCA incidence, bystander CPR provision and survival rate after OHCA). We found a relatively consistent trend of a SES influence across the three stages: an increased risk of OHCA incidence, a decreased chance of receiving bystander CPR and a decreased chance of survival after an OHCA in low SES compared to high SES populations and locations. Although SES gradients were found for different indicators in all three stages, sufficient data to compare the strength of associations across stages or across indicators was lacking (Fig. 2). There was minimal evidence that was suggestive of sex and age differences in the SES gradients although it was similarly too sparse to draw conclusions.

Limitations
This systematic review has several limitations. It may have been the case that we missed relevant citations in our search. We included only full papers in the English language, disregarding data published in other languages. However, we noticed that only one article, in German, would have met the inclusion criteria, which suggests that disregarding of this criterion likely would not have changed the conclusions of our review. Additionally, we did not investigate the potential role of publication bias in minimizing the inclusion of negative results in the published literature.
While not related to the methodology applied, our review also suffers from the low number of studies that appeared to be eligible for this systematic review. In particular, a notable lack of studies measuring individual SES, measured at the individual level (SESind; Fig. 2). This constrains the potential inferences that could be made, as SES measured at the individual level usually highlights greater contrasts in health outcomes [54][55][56] and comparisons could have been made between effect sizes of different SES indicators, such as education level and income level. This is a key limitation of the literature, as the comparison between SES indicators may have provided insight into which populations are the most vulnerable. It may also have provided an opportunity to compare findings across national contexts. We also found very few studies that reported results by age or sex groups (Fig. 2). Moreover, we found that several studies had a small sample size, 16,28,29,43,49 which may have reduced the accuracy of the point estimates found in these papers. Finally, the heterogeneity of indicators, outcomes, populations and definitions as well as the  scarcity of papers prevented the performance of statistical analyses, such as meta-analysis, on the pooled results.

Discussion of key findings
Taken together, the evidence from our review suggests that SES influences may occur within each step in the process leading up to SES differences in OHCA mortality. Below we discuss how our findings fit in with previous knowledge on OHCA and SES differences in health and signal some important gaps for further research.
The observation across studies in this review that the incidence of OHCA varies by individual SES is in line with a multitude of previous studies that have shown SES differences in overall cardiovascular disease morbidity, mortality and risk factors. 14,57 These results also are in line with prior studies that have shown a similar SES gradient in sudden cardiac death (eTable 3). It may be the case that the absolute incidence by SES differs across national contexts. However, we observed consistent patterns within the different contexts we observe consistent patterns of relative differences in OHCA incidence across different SES indicators that reflect the relative position of individuals within societal hierarchies. Individual SES differences in OHCA incidence may be mediated indirectly by SES differences in CHD. As 80% of the OHCAs occur in individuals that have CHD, 3 it is likely that CHD lies on the causal pathway between SES and SCA. CHD and OHCA share several common risk factors. 3 SES differences may occur due to behavioural factors (e.g. smoking, physical inactivity), 12,14,15 biomedical factors (e.g. dyslipidaemia, high blood pressure) 9,10 or psychosocial factors (e.g. depression) 11,13 that are both associated with SES and CVD/CHD. Additionally, a more direct pathway linking SES with OHCA incidence may also exist. 58 Both acute and chronic stressors have been linked to SCA, via dysregulation of the autonomic nervous system. Natural disasters, such as earthquakes, a clear source of acute stress, are often followed by a spike in cardiac arrest incidence. 59 Low SES may also be associated with chronic stress 60,61 and both low SES and chronic stress have been associated with sympathetic nervous system overactivity, as assessed by heart rate variability. [62][63][64] In turn, such chronic overactivity in the sympathetic nervous system has been associated with increased risk of cardiovascular disease and cardiovascular mortality and thus possibly also with OHCA. 65 Related to bystander CPR, the majority of studies (10 out of 12) found a consistent association with SESloc. The two papers that did not find a significant difference across SES areas, had a notably small study population 28 (n = 413) or reported older data (1999 as compared to 2008 50 and 2018 22 ). The latter paper may have suffered from a lower general knowledge on CPR at the time of data collection; one to two decades ago, there was generally little CPR training offered to the public. 66 Furthermore, the observation that lower bystander CPR provision was associated with low SESloc is consistent with findings of reduced bystander support in medical emergencies in general in deprived areas. 67,68 This may be due to decreased community cohesion in low SES neighbourhoods, 69 decreased security 70 or simply less knowledge about how to perform CPR and alert the EMS.
It may further be the case that a substantial part of the reported associations between SESloc and bystander CPR provision are actually due to individual SES differences. While the chances of bystanders in an area providing CPR may be the product of factors related to a low SES area, factors associated with the individual OHCA victim may also influence the chances of receiving CPR. Individual demographic variables, such as sex and age, have been suggested to Table 3 (continued) Author, year affect the chances of receiving CPR. 71 Following that logic, one may speculate that factors associated with one's individual SES, may    (continued on next page) influence a bystander to provide CPR or not. However, no studies, as of yet, appear to have investigated the association between individual SES and bystander CPR provision (Fig. 2). Survival after OHCA was found to be consistently associated with a higher SES when measured at the individual level (SESind). The independence from demographic and resuscitation characteristics indicate that differences are likely, at least in part, to be associated with physical resilience, which may be affected by the level of comorbidities. 72,73 In line with this hypothesis, several studies have shown an increasing Charlson comorbidity index to be associated with a decreased chance of survival, though some studies reported no association. [72][73][74] All but one of the studies reporting SESres, observed no association with survival after OHCA. Two of these studies are likely to have suffered from low sample sizes of n = 413 28 and n = 415 49 , while over adjusted statistical models may have hampered the detection of an association in two other studies. 24,27 It may, on the other hand, be the case that individual SES measured at the area level (SESres) generates weaker associations with health outcomes due to the heterogeneity of individuals' SES (SESind) within any given area. 56 Other than serving as a proxy for individual SES, SESres may also capture aspects of the environment in which the OHCA victim resides, which may have an influence on the risk of suffering an OHCA. Individuals living in low SES areas may be more frequently exposed to factors linked to the physical and social environment not captured by SES measured at the individual level. Such factors may include the availability of healthy or fast food outlets, parks and recreational spaces, air pollution levels and neighbourhood safety to name a few. These factors have been reported to be associated with CVD. 75 In contrast to the studies reporting on SESres, we observed an association between OHCA survival and SESloc. Although the reasons are unknown, the differences found in survival rate by SES of the area in which the OHCA has occurred are likely to be, at least in part, due to differences in bystander resuscitation rates. However, the SES differences remained in two studies that adjusted for bystander CPR, suggesting that the disparity in survival rate appears not to be solely dependent on bystander CPR. We speculate that the remaining difference may partly be explained by differences in health-care related factors. Potential reasons for SES differences in these healthcare related factors may be a number of system factors, including the distribution of ambulances in the city, neighbourhood structures that influence Emergency Medical Services response time (including road design and traffic patterns), and proximity of high socioeconomic status neighbourhoods to hospital. 76 These factors were not assessed in the studies included in this review. Other factors, not included in this review include potential SES differences in in-hospital care. This choice was made given the focus on an out-of-hospital population. While, SES differences in survival to discharge may be mediated by both pre-as well as in-hospital factors, other reported (pre-hospital) indicators of survival showed consistent results. Thus, it is likely that SES differences reported in survival to discharge was not solely influenced by in-hospital factors.

Sex and age differences
The association between individual SES and OHCA incidence was stronger in women than in men. 19,20 This finding is in line with previous findings on sex differences in social inequalities in cardiovascular disease. 20 While in men, SES gradients in total mortality are steeper, women exhibit a greater SES disparity in cardiovascular disease mortality, morbidity and risk factors (e.g. smoking) than men. 20 Two Table 4 (continued) Author    studies that showed SES differences in sudden cardiac death stratified the results by sex and found contradicting results. One 7 found a significant modifying effect by sex with a stronger gradient in men while the other 6 did not find a significant interaction with sex although the SES gradients appeared stronger in women than in men. Age differences in the association between individual SES and OHCA incidence were found in two studies. 41,42 Both were limited in only stratifying OHCA cases aged below and above 65 years of age. The higher age group was found to have narrower SES differences in OHCA incidence than the lower age group, 41,42 implying that while absolute differences in incidence may still be larger in high SES versus low SES groups, the relative SES differences are attenuated at a higher age. This finding is in accordance with findings of several studies that have found converging SES differences in health above the age of around 60. This is attributed to a ceiling effect at higher ages due to the morbidity in high SES groups catching up at later ages to their low SES counterparts at lower ages. 77

Conclusion
We found a consistent trend of an increased risk of OHCA incidence and a decreased chance of receiving bystander CPR and surviving an OHCA in low SES compared to high SES populations and locations. This suggests that SES disparities in OHCA mortality likely manifest at each of these stages relevant to OHCA. Despite the clear SES differences, this review has highlighted several gaps in the literature. One such gap is caused by the lack of detailed information in which specific populations SES disparities in OHCA occur, e.g. differences between men and women and by age groups, and a distinct lack of data on SES measured at the individual level. Furthermore, research into the factors mediating SES disparities in OHCA incidence, bystander CPR and survival rate after OHCA is warranted. This could inform targets such as risk factor prevention in socioeconomically deprived individuals and targeted AED placement in socioeconomically deprived areas.

Conflicts of interest
None.