If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Research Center for Emergency Medicine, Aarhus University Hospital, Aarhus, DenmarkDepartment of Clinical Medicine, Aarhus University, Aarhus, DenmarkDepartment of Anesthesiology and Intensive Care, Randers Regional Hospital, Randers, Denmark
Department of Clinical Medicine, Aarhus University, Aarhus, DenmarkDepartment of Anesthesiology and Intensive Care, Aarhus University Hospital, Aarhus, Denmark
Corresponding author at: Research Center for Emergency Medicine, Aarhus University Hospital, Palle Juul Jensens, Boulevard 99, Bygning J, Plan 1, 8200 Aarhus N, Denmark.
Research Center for Emergency Medicine, Aarhus University Hospital, Aarhus, DenmarkDepartment of Clinical Medicine, Aarhus University, Aarhus, DenmarkDepartment of Anesthesiology and Intensive Care, Aarhus University Hospital, Aarhus, DenmarkPrehospital Emergency Medical Services, Central Denmark Region, Denmark
To investigate how socioeconomic status was associated with the risk of in-hospital cardiac arrest in Denmark.
Methods
We conducted a matched case-control study based on data from nationwide registries in Denmark. A total of 3,449 cases with in-hospital cardiac arrest in 2017 and 2018 were matched at the index time based on age and sex with up to 10 controls from the total Danish population (background controls) and a hospitalized patient population (hospitalized controls), respectively. Household income, household assets, and education were used as measures of socioeconomic status. Conditional logistic regression was used to assess the association between socioeconomic status and the risk of in-hospital cardiac arrest.
Results
Across all analyses of cases and controls, high household income, high household assets, and higher education were associated with decreased odds of in-hospital cardiac arrest. In the analyses of cases and background controls, high household income was associated with 0.45 (95% CI: 0.40, 0.52) times the odds of in-hospital cardiac arrest compared to low household income, which was similar for household assets. Compared to basic education, higher education was associated with 0.50 (95% CI: 0.43, 0.58) times the odds of in-hospital cardiac arrest. The results attenuated marginally after adjustment for comorbidities. Similar albeit attenuated findings were observed in the analyses of cases and hospitalized controls.
Conclusions
In this matched case-control study, high socioeconomic status was associated with lower odds of in-hospital cardiac arrest compared to low socioeconomic status. The findings were consistent across household income, household assets, and education and persisted after adjustment for comorbidities. Strategies are needed to address the socioeconomic inequalities observed in the risk of in-hospital cardiac arrest.
Socio-economic differences in incidence, bystander cardiopulmonary resuscitation and survival from out-of-hospital cardiac arrest: A systematic review.
Socio-economic differences in incidence, bystander cardiopulmonary resuscitation and survival from out-of-hospital cardiac arrest: A systematic review.
In this case-control study, we investigated how SES was associated with risk of IHCA in Denmark.
Methods
Study design, setting, and population
We conducted a case-control study based on nationwide registries and prospectively collected data on IHCA in Denmark. An ethical approval for observational register-based studies is not required in Denmark. Patients and the public were not involved in the development of this study.
Cases were adult patients (≥18 years) with an index IHCA registered in the Danish In-Hospital Cardiac Arrest Registry (DANARREST) from January 1st, 2017 to December 31st, 2018.
Patients with an initial pulse-generating rhythm on the first rhythm analysis and patients with missing data on covariates were excluded.
Controls were retrieved from two separate populations through Statistics Denmark. The first control group (background controls) was retrieved from the total adult population in Denmark. The second control group (hospitalized controls) was retrieved from a population of hospitalized patients in Denmark. Hospitalized patients were defined as patients who had a duration of hospitalization of at least one full day at any hospital department. By using risk-set sampling with replacement, cases were matched based on age and sex with controls who were alive at the date of the index IHCA. Cases were matched with up to 10 controls to optimize efficiency.
Data was obtained from DANARREST, the Danish Civil Registration System, the Income Statistics Register, the Population Education Register, the Employment Classification Module, and the Danish National Patient Registry.
The registry includes data on a number of cardiac arrest characteristics.
Socioeconomic status
SES was defined by individual-level measures of household income, household assets, education, and employment.
Household income was defined as the average annual equivalized disposable household income (Danish Kroner [DKK], 1 EUR = 7.44 DKK) in the three years prior to the index time. The equivalized disposable household income was calculated as the total sum of disposable income of the household divided by the number of household members using the Organisation for Economic Co-operation and Development (OECD) modified scale.
We did not account for inflation, since cases and controls were matched on index time. Household assets were defined as the total assets of the household, which included assets in property, monetary institutions, shares, and bonds.
Education was defined as the highest attained educational level according to the International Classification of Education (ISCED) at the index time. The ISCED classifications were grouped into three categories: basic education (ISCED level 0–2 and 9), upper secondary education (ISCED level 3), and higher education (ISCED level 5–8).
Employment was grouped into four categories: retirement, employment, unemployment, and other (i.e., early retirement, sick leave, students, or not otherwise classified). Employment was based on the year prior to the index time.
Statistics
Descriptive statistics were used to describe the population.
The primary SES measures of interest were household income, household assets, and education, while employment was used in subgroup analyses based on retirement. Given that household income and household assets were continuous SES measures, we performed two separate analyses to evaluate whether the association between these measures and risk of IHCA varied contingent on analyses based on categorical versus continuous SES measures. All analyses were performed using the two control groups separately.
Categorical socioeconomic status
To evaluate the association between SES and risk of IHCA, we treated household income, household assets, and education as categorical variables. Household income and household assets were divided into three categories (low, medium, high) based on the tertiles for each combined sample of cases and controls. Low household income and low household assets were chosen as the reference categories. Education was divided into basic education, upper secondary education, and higher education. Basic education was chosen as the reference category.
To assess whether categorical SES was associated with risk of IHCA, we developed separate conditional logistic regression models to yield odds ratios (ORs) with 95% confidence intervals with each SES measure as the independent variable. This was considered the primary analysis. First, we performed a model adjusted for age and sex only.
This model was considered the primary model. Second, comorbidities may act as either confounders or mediators in the association between SES and risk of IHCA (Directed Acyclic Graph in eFigure 1 in Supplemental Materials). Consequently, we then adjusted for comorbidities in model 2 (all comorbidities shown in Table 1). Third, to assess if each SES measure was independently associated with risk of IHCA, we adjusted for the remaining SES measures in model 3. For example, with education as the independent variable, we adjusted for age, sex, comorbidities, household income, and household assets. Due to the risk of collinearity between household income and household assets (Spearman correlation coefficient = 0.65), these were not included in the same model when each of the two measures were the independent variable. Finally, since comorbidities may act as mediators, we then removed comorbidities from the previous model. For example, with household income as the independent variable, we only adjusted for age, sex, and education.
As subgroup analyses to evaluate if the risk of IHCA varied based on sex, all analyses were stratified by sex. Moreover, the association between SES and the risk of IHCA was assessed in the population who were not retired. As a post-hoc analysis of the primary analyses for cases and hospitalized controls, hospital was included as a fixed effect in Model 3.
Analyses to account for missing data are presented in Supplemental Materials.
Continuous socioeconomic status
For this analysis, household income and household assets were treated as continuous variables. Due to the skewness of the data with severe outliers, we ranked the SES measures for each combined sample of cases and controls and presented the variables as percentages from 0 (lowest income/assets) to 100% (highest income/assets). Next, we used separate conditional logistic regression models with linear, quadratic, cubic splines, and restricted cubic splines to yield ORs. These models were adjusted for age and sex. The median of each SES measure was used as the reference. We used three prespecified knots at the 15%, 50%, and 85% percentiles for each SES measure. To evaluate and compare the fit of the models, we used the Quasi-likelihood under Independence Model Criterion (QIC) and reported the model with the best fit.
Without ranking the SES measures, we repeated the separate regression models as described above but excluded outliers below the 1% percentile and above the 99% percentile. Compared to the ranked SES measures, this allowed for a more direct interpretation in exact monetary funds of the association between the SES measures and the risk of IHCA.
Results
Study population
Overall, a total of 4,282 adult patients with IHCA were retrieved from DANARREST, of which 3,449 met all inclusion criteria (eFigure 2). Cases were matched to 34,490 background controls and 34,490 hospitalized controls, respectively. Among background controls, 3,140 (9%) controls were excluded due to missing data on SES. Among hospitalized controls, 926 (3%) controls were excluded due to missing data on SES. Thus, a total of 31,350 background controls and 33,564 hospitalized controls were included in the final population. Table 1 shows the distribution of demographics, socioeconomic status, and comorbidities across cases and controls. eFigure 3 shows the distribution of household income and household assets across cases and controls. Cardiac arrest characteristics of the cases are shown in eTable 2.
Household income and household assets
In the analyses of cases and background controls, high household income was associated with 0.45 (95% CI: 0.40, 0.52) times the odds of IHCA compared to low household income, which attenuated marginally after incremental adjustment for comorbidities and education (Fig. 1, eTable 3). These findings were similar in men and women (eTable 3). Among individuals who were not in retirement, high household income was associated with 0.21 (95% CI: 0.17, 0.28) times the odds of IHCA compared to low household income (eTable 3). These findings attenuated marginally after incremental adjustment for comorbidities and education. Similar findings were observed in the analyses evaluating household assets (Fig. 1, eTable 3).
Fig. 1Association between categorical socioeconomic status and risk of in-hospital cardiac arrest comparing cases and matched background controls. Analyses are presented as odds ratios (ORs) with 95% confidence intervals. Thus, an OR > 1 indicates an association with increased odds of in-hospital cardiac arrest compared to low SES/basic education. Conversely, an OR < 1 indicates an association with decreased odds of in-hospital cardiac arrest compared to low SES/basic education. Model 1: Age and sex. Model 2: Age, sex, and comorbidities. Model 3: Age, sex, comorbidities, and remaining socioeconomic status. Model 4: Age, sex, and remaining socioeconomic status.
Similar albeit attenuated findings were observed in the analyses of cases and hospitalized controls. High household income was associated with 0.75 (95% CI: 0.66, 0.85) times the odds of IHCA compared to low household income, which persisted after incremental adjustment for comorbidities and education (Fig. 2, eTable 4). These findings were similar in men and women (eTable 4). High household assets were associated with 0.80 (95% CI: 0.71, 0.91) times the odds of IHCA compared to low household assets, which persisted after incremental adjustment for comorbidities and education (Fig. 2, eTable 4). However, women had a weaker association, which disappeared after incremental adjustment for comorbidities and education (eTable 4). Among individuals who were not in retirement, high household income was associated with 0.48 (95% CI: 0.38, 0.62) times the odds of IHCA compared to low household income (eTable 4). These findings attenuated marginally after incremental adjustment for comorbidities and education. Similar findings among individuals who were not in retirement were observed in analyses of household assets (eTable 4). The post-hoc analysis in which hospital was included in Model 3 showed results similar to the primary analysis (eTable 4).
Fig. 2Association between categorical socioeconomic status and risk of in-hospital cardiac arrest comparing cases and matched hospitalized controls. Analyses are presented as odds ratios (ORs) with 95% confidence intervals. Thus, an OR > 1 indicates an association with increased odds of in-hospital cardiac arrest compared to low SES/basic education. Conversely, an OR < 1 indicates an association with decreased odds of in-hospital cardiac arrest compared to low SES/basic education. Model 1: Age and sex. Model 2: Age, sex, and comorbidities. Model 3: Age, sex, comorbidities, and remaining socioeconomic status. Model 4: Age, sex, and remaining socioeconomic status.
Across all analyses of continuous household income and household assets for cases and controls, higher household income and household assets were generally associated with a decreased odds of IHCA compared to the median household income and the median household assets, respectively (Fig. 3, eFigure 4). Conversely, lower household income and household assets were generally associated with a higher odds of IHCA compared to the median household income and the median household assets, respectively. Compared to the findings from the analyses of cases and background controls, the findings from the analyses of cases and hospitalized controls were slightly attenuated (Fig. 3, eFigure 4).
Fig. 3Association between continuous socioeconomic status and risk of in-hospital cardiac arrest comparing cases and matched controls. Analyses are presented as odds ratios (ORs) with 95% confidence intervals. Thus, an OR > 1 indicates an association with increased odds of in-hospital cardiac arrest compared to the median SES. Conversely, an OR < 1 indicates an association with decreased odds of in-hospital cardiac arrest compared to the median SES. Household income (blue) and household assets (red) are presented as percentages. All models are adjusted for age and sex. Three prespecified knots were used at the 15%, 50%, and 85%-percentiles for each SES measure. The following models are reported based on the best model fit as determined by the QIC. (A) Household income for cases and background controls: Linear spline model. (B) Household income for cases and hospitalized controls: Linear spline model. (C) Household assets for cases and background controls: Quadratic spline model. (D) Household assets for cases and hospitalized controls: Restricted cubic spline model.
In the sensitivity analyses to account for missing data, the results were comparable to the primary analyses (eFigure 5–6).
Education
In the analyses of cases and background controls, higher education was associated with 0.50 (95% CI: 0.43, 0.58) times the odds of IHCA compared to basic education only, which attenuated marginally after incremental adjustment for comorbidities, household income, and household assets (Fig. 1, eTable 3). These findings were similar in men and women (eTable 3). Among individuals who were not in retirement, higher education was associated with 0.25 (95% CI: 0.19, 0.33) times the odds of IHCA compared to basic education only (eTable 3). These findings attenuated marginally after incremental adjustment for comorbidities, household income, and household assets.
Similar albeit attenuated findings were observed in the analyses of cases and hospitalized controls. Higher education was associated with 0.70 (95% CI: 0.60, 0.81) times the odds of IHCA compared to basic education only, which persisted after incremental adjustment for comorbidities, household income, and household assets (Fig. 2, eTable 4). These findings were similar in men and women (eTable 4). Among individuals who were not in retirement, higher education was associated with 0.47 (95% CI: 0.35, 0.62) times the odds of IHCA compared to basic education only (eTable 4). These findings attenuated marginally after incremental adjustment for comorbidities, household income, and household assets. The post-hoc analysis in which hospital was included in Model 3 showed results similar to the primary analysis (eTable 4). However, the effect size was slightly larger for higher education.
In the sensitivity analyses to account for missing data, the results were comparable to the primary analyses (eFigure 5–6).
Discussion
In this matched case-control study, we systematically evaluated the association between SES and the risk of IHCA in Denmark. We found that high SES was associated with decreased odds of IHCA compared to low SES, which persisted after adjustment for comorbidities. These findings were consistent across all SES measures, control groups, men and women, and among individuals who were not retired.
The preceding literature on the association between SES and the risk of IHCA is sparse and has several limitations. First, a study by Song et al. found that lower area-level household income based on residential zip codes was associated with a higher risk of intraoperative IHCA.
Second, a study by Merchant et al. found no clear association between area-level median household income based on hospital zip codes and the risk of IHCA,
although a high income hospital area was associated with 0.84 (95%: 0.71, 1.00) times the risk of IHCA in the adjusted analyses compared to a low income hospital area. Both studies reported income on an area-level only and did not report on any individual-level SES measures. Ecological measures of SES may not accurately correspond to individual-level measures, which may potentially underestimate true effects.
Influence of individual- and area-level measures of socioeconomic status on obesity, unhealthy eating, and physical inactivity in Canadian adolescents.
In comparison, we present the findings of the first study to systematically evaluate the association between individual-level SES and the risk of IHCA. We found that high household income was associated with a substantially lower risk of IHCA compared to low household income, which was consistent across adjusted analyses, sexes, control groups, and among non-retired individuals only. Our findings were consistent across multiple individual-level SES measures, including household assets, and education. Our SES measures have several advantages. First, compared to individual income, equivalized disposable household income may be a more useful SES measure because it adjusts for family size and associated costs of living.
Thus, it may reflect what an individual can afford. Second, as individuals age, the relative significance of household assets may increase compared to household income.
Finally, we were able to evaluate the association with household income and household assets as continuous measures. This allowed us to assess the association between SES and the risk of IHCA across the continuum of SES, thereby providing further evidence of a socioeconomic gradient in the risk of IHCA.
SES may be associated with the risk of IHCA through several dynamic and multifactorial pathways.
Educational attainment and mean leukocyte telomere length in women in the European Prospective Investigation into Cancer (EPIC)-Norfolk population study.
Educational attainment but not measures of current socioeconomic circumstances are associated with leukocyte telomere length in healthy older men and women.
, which may then influence the distribution of SES in the population at risk of IHCA. However, since higher SES may potentially be associated with increased risk of do-not-resuscitate orders,
it is unclear how the distribution of do-not-resuscitate orders may have influenced our results. In our study, we found that the association between high SES and lower risk of IHCA persisted after adjustment for comorbidities. Thus, SES may potentially impact the risk of IHCA through pathways which are not directly related to the increased burden of comorbidities that were included in this study. This may include the utilization of health care services. Low SES may be associated with increased end-of-life acute hospital-based care,
which may potentially increase the risk of IHCA. While previous studies have found that healthcare providers may exhibit negative biases against lower SES,
Relationship between alcohol-attributable disease and socioeconomic status, and the role of alcohol consumption in this relationship: a systematic review and meta-analysis.
An area of interest may be the potential reverse causation between SES and comorbidities (i.e., comorbidities leading to low SES). In this study, it is unclear if comorbidities are confounders or mediators for the association between SES and the risk of in-hospital cardiac arrest. However, evidence from multiple studies, including longitudinal studies, suggest that reverse causation may not account for the strong association between SES and comorbidities.
In our study, we found that higher education was associated with lower odds of in-hospital cardiac arrest. Since education cannot be reduced once it is attained, confounding from comorbidities may not be able to explain this association. Thus, comorbidities may likely act as mediators of the association between socioeconomic status and the risk of in-hospital cardiac arrest. To evaluate if our results were confounded or mediated by comorbidities, we incrementally adjusted for comorbidities in a separate model. We found that the results persisted in these analyses. Consequently, SES may be associated with the risk of in-hospital cardiac arrest through pathways that are not directly related to the burden of comorbidities. Another area of interest is whether each SES measure is independently associated with the risk of in-hospital cardiac arrest. To assess this, we further adjusted our models for the remaining SES measures. We found that the results persisted but were slightly attenuated, particularly in the analyses of cases and hospitalized controls. This implies that individual measures of SES may be independently associated with the risk of in-hospital cardiac arrest.
This study provides the initial insights into the association between SES and the risk of IHCA. To address the socioeconomic inequalities observed in this study, strategies and initiatives on several levels are necessary. First, additional research needs to inform on the complex mechanisms in which SES may influence the risk of IHCA. Second, clinical improvement initiatives must address the inequalities that affect vulnerable populations. This may include strategies that aim to identify patient populations at risk of IHCA, training of healthcare professionals on the diverse effects of SES on health, and precision health with targeted interventions. In turn, this may yield more effective treatments.
Thus, health systems may need to advance public health programs that address the upstream socioeconomic conditions that may influence the risk of IHCA.
Limitations
The findings of our study need to be interpreted in the context of the register-based observational matched case-control study design. First, we adjusted for multiple confounders, however residual or unmeasured confounding may still exist. Second, there was missing data for SES. However, the sensitivity analyses to account for missing data confirmed our results. Third, the distribution of DNR orders across SES may have influenced the distribution of SES among patients with IHCA. It is unclear how this may have affected our results. Fourth, in the analyses of cases and hospitalized controls, collider bias may potentially have influenced the results.
However, we believe such bias is likely small and towards the null. Finally, the study was conducted in the setting of a universal healthcare system in a welfare state. Consequently, the observed socioeconomic inequalities are not necessarily generalizable to other healthcare systems and welfare models.
Conclusions
In this matched case-control study, high SES was associated with lower odds of IHCA compared to low SES. The findings were consistent across household income, household assets, and education and persisted after adjustment for comorbidities. Strategies are needed to address the socioeconomic inequalities observed in the risk of IHCA.
CRediT authorship contribution statement
Nikola Stankovic: Methodology, Software, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Mathias J. Holmberg: Methodology, Software, Validation, Data curation, Writing – review & editing, Supervision, Project administration. Asger Granfeldt: Conceptualization, Methodology, Writing – review & editing, Supervision, Project administration. Lars W. Andersen: Conceptualization, Methodology, Validation, Resources, Writing – review & editing, Supervision, Project administration, Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This study was supported by the Karen Elise Jensen’s Foundation and Helsefonden.
Appendix A. Supplementary material
The following are the Supplementary data to this article:
Socio-economic differences in incidence, bystander cardiopulmonary resuscitation and survival from out-of-hospital cardiac arrest: A systematic review.
Influence of individual- and area-level measures of socioeconomic status on obesity, unhealthy eating, and physical inactivity in Canadian adolescents.
Educational attainment and mean leukocyte telomere length in women in the European Prospective Investigation into Cancer (EPIC)-Norfolk population study.
Educational attainment but not measures of current socioeconomic circumstances are associated with leukocyte telomere length in healthy older men and women.
Relationship between alcohol-attributable disease and socioeconomic status, and the role of alcohol consumption in this relationship: a systematic review and meta-analysis.
00165-4&id=gr1.jpg){kind=link}
00165-4&id=gr2.jpg){kind=link}
00165-4&id=gr3.jpg){kind=link}