Treatment and survival following in-hospital cardiac arrest: does patient ethnicity matter?

Abstract

Aims

Previous research on racial/ethnic disparities in relation to cardiac arrest has mainly focused on black vs. white disparities in the USA. The great majority of these studies concerns out-of-hospital cardiac arrest (OHCA). The current nationwide registry study aims to explore whether there are ethnic differences in treatment and survival following in-hospital cardiac arrest (IHCA), examining possible disparities towards Middle Eastern and African minorities in a European context.

Methods and results

In this retrospective registry study, 24 217 patients from the IHCA part of the Swedish Registry of Cardiopulmonary Resuscitation were included. Data on patient ethnicity were obtained from Statistics Sweden. Regression analysis was performed to assess the impact of ethnicity on cardiopulmonary resuscitation (CPR) delay, CPR duration, survival immediately after CPR, and the medical team’s reported satisfaction with the treatment. Middle Eastern and African patients were not treated significantly different compared to Nordic patients when controlling for hospital, year, age, sex, socioeconomic status, comorbidity, aetiology, and initial heart rhythm. Interestingly, we find that Middle Eastern patients were more likely to survive than Nordic patients (odds ratio = 1.52).

Conclusion

Overall, hospital staff do not appear to treat IHCA patients differently based on their ethnicity. Nevertheless, Middle Eastern patients are more likely to survive IHCA.

Introduction

Sociodemographic disparities in healthcare and outcomes are receiving increasing attention among researchers, policymakers, and society. Research on sudden cardiac arrest (CA) is one domain in which sociodemographic disparities have been explored. The great majority of this research has examined racial disparities (black vs. white) in the USA, focusing on out-of-hospital cardiac arrests (OHCA). A recent meta-analysis of 15 studies shows that black people are less likely to survive an OHCA.¹ Black patients also show worse neurologic outcome upon hospital discharge.² A combination of factors seems to account for these differences, including differences in initial heart rhythm, probability of having a witnessed CA and receiving bystander cardiopulmonary resuscitation (CPR),¹ and post-arrest treatment (e.g. cardiac catheterization).² It has also been suggested that treatment bias originating from medical staff prejudice may be involved,² but this remains unresolved.

Whether there are racial/ethnic disparities in treatment and survival following in-hospital cardiac arrest (IHCA) is less well-explored. IHCA is a common condition associated with a high risk of death.³ There is some IHCA research showing that in the USA, black patients have lower rates of short- and long-term survival.^4,5 However, it is unclear to what extent both IHCA and OHCA racial differences are driven by differences in patient socioeconomic status (SES). After all, racial/ethnic minority groups often score lower on SES indicators, such as income and education,¹ and research has generally shown that higher SES predicts higher survival rates following OHCA, and to some extent even following IHCA.^6–12 While few studies on racial differences have adjusted for SES, one study¹³ finds that racial disparities in OHCA survival disappear when accounting for SES, whereas another¹⁴ finds that they persist.

The current study aims to examine whether there are ethnic differences in cardiac arrest treatment and survival following IHCA, focusing on potential ethnic disparities affecting African and Middle Eastern patients, which to our knowledge, have not been studied in Europe. While doing so, we simultaneously adjust for numerous sociodemographic (age, sex, and SES), clinical (comorbidity, initial heart rhythm, and aetiology), and contextual factors (hospital and year) to increase the certainty that any identified difference can be attributed to patient ethnicity.

Methods

Design

This retrospective registry study was based on merged data from the Swedish Registry of Cardiopulmonary Resuscitation (SRCR) and Statistics Sweden. The study was approved by the Regional Ethical Review Board in Linköping, Sweden (No. 2017/293-31).

The Swedish Register of Cardiopulmonary Resuscitation

The SRCR is a national quality registry that monitors resuscitation practices in Sweden. As of 2018, out of 74 Swedish emergency hospitals, 73 report IHCA data to the registry. Only patients suffering from CA and receiving CPR are reported to the registry. A predefined framework based on the Utstein templates for reporting IHCA is employed.¹⁵ The National Registry Committee conducts random inspections of the data to validate the registry continuously.

At the time of data extraction, the SRCR contained 24 820 registered patients who had received CPR after suffering an IHCA in Sweden from 2005 (start year) to 20 August 2018 (extraction date). The IHCA refers to CA events taking place within the hospital perimeter, without the need for ambulance transport.

The registry contains information on prophylactic monitoring (heart rhythm monitoring), CPR delay, treatment during the CA (CPR duration, number of defibrillations, intubation, use of adrenaline/anti-arrhythmic drugs, etc.), survival after CPR, survival to discharge from hospital, 30-day survival, neurologic function (cerebral performance category score) among survivors, and post-arrest treatment (e.g. implantable cardioverter-defibrillator). It further includes basic sociodemographic information about the patient (sex and age), detailed medical data in the form of medical history/comorbidity, initial heart rhythm, likely aetiology of the CA, and information about the situation surrounding the CA (e.g. year and hospital). Finally, the registry includes the medical team’s assessment of the quality of the CA treatment (treatment satisfaction).

Statistics Sweden

The patient’s origin of birth was used as a proxy for ethnicity. These individual-level data were obtained from Statistics Sweden’s Population register (RTB). Origin of birth was categorized at the level of broad geographical regions, such as the Nordics, Africa, and the Middle East. This level of categorization corresponds well with the level at which people hold attitudes and stereotypes of various ethnic groups. It also reflects groups with ethnic, cultural, religious, and lingual similarities, although there are certainly significant differences within these geographical regions. We focused on two ethnic minority groups, namely African and Middle Eastern, which we compared with native Nordic individuals. The reason for this focus is that people from Africa and the Middle East are among the largest ethnic minority groups in Sweden and recent surveys have revealed that Swedish people hold particularly negative attitudes towards them.¹⁶

Information on individual SES was obtained from Statistics Sweden’s Longitudinal integrated database for health insurance and labor market studies (LISA). SES refers to the social standing or class of an individual or social group.¹⁷ We obtained two SES proxies: the highest level of completed education and annual income.

Selection of patients

Patients younger than 40 years of age (n = 603) were excluded from the study, leaving us with a final sample of 24 217 patients. These patients were excluded because our SES proxies are not accurate for younger patients, since many of them have not reached their highest income or level of completed education. Additionally, IHCAs are very rare among young individuals, as is clear from the current data set where patients who are younger than 40 years of age only make up 2.4% of the reported IHCAs. Therefore, these patients are regarded to constitute a selective group with different unobserved initial health.

Variables

Three categories of variables were included in the analysis: outcomes, ethnicity, and control variables. The outcome variables included CPR delay, i.e. the delay from the discovery of the patient suffering CA to the start of CPR, (0 = less than 1 min, 1 = 1 min or longer), CPR duration (minutes), survival immediately after CPR (0 = dead, 1 = alive), and treatment satisfaction (0 = unsatisfactory, 1 = satisfactory).

Ethnicity dummies for each of the following regions were created: Nordic, Middle East, Africa, and Other. Patients from the following regions were included in a category labelled ‘Other’: Western Europe, Southern Europe, Eastern Europe, Asia, North America, South America, and Oceania. The rationale for this is that it is unclear what Swedish people’s attitudes towards people from these regions look like, and hence the basis for expecting disparities is weaker. Relatedly, the category ‘Other’ is a very heterogeneous group with rather few patients from each distinct minority group. Nordic patients were used as a benchmark (i.e. reference category). The rationale for this is that Swedes hold very favourable attitudes towards people from other Nordic countries, and 87% of Swedes report that people from other Nordic countries are similar or very similar to themselves.¹⁶

Age, sex, SES (education and income), comorbidity, hospital, and year of the CA constituted basic control variables. Education was dummy coded into High (1; college/university education) and Low (0; no college/university education). Regarding SES income, a percentile score was created for each patient reflecting the patient’s relative standing in the income distribution. Because the sample contains both retired and working patients, the income variable had to be based on either annual earned income or the retirement pension. The use of a percentile score circumvents the comparison problem that emerges due to earned income generally being higher than retirement pension.

Regarding comorbidity, an index was created. It indicates the number of diagnoses the patient had before the CA. These diagnoses were previous history of heart failure, myocardial infarction, stroke, respiratory insufficiency, diabetes, cancer, and metastatic cancer. The index had a possible range between 0 and 7.

Some regression models also added controls for initial heart rhythm (ventricular fibrillation, ventricular tachycardia, asystole, or pulseless electrical activity) and aetiology of the CA (e.g. myocardial infarction/ischaemia, arrhythmia, heart failure, respiratory insufficiency, and intoxication).

Statistical analysis

The relationship between ethnicity and the dichotomic dependent variables (CPR delay, survival, and treatment satisfaction) was analysed with multiple binary logistic regression analysis. The continuous dependent variable (CPR duration) was analysed with multiple ordinary least squares regressions.

Each regression model also included a set of control variables. All models included our basic control variables: age, sex, SES education, SES income, hospital, and year of the CA. We also computed models including initial heart rhythm and aetiology of the CA. Initial heart rhythm and aetiology are somewhat problematic to use as control variables because they are not strictly predetermined.¹⁸ Initial heart rhythm is mainly assessed after the CA alarm and aetiology is determined post-CA. Essentially, they could to some extent be regarded as outcome variables, because they could be influenced by events around the CA. On the other hand, because initial heart rhythm and aetiology also partly reflect initial health status that is fixed at the time of the CA event, performing separate analyses that also control for these variables would also be appropriate.

In the results section, regression analyses including only basic control variables are reported in Panel A, whereas analyses also including heart rhythm and aetiology are reported in Panel B.

In all regressions, we included all observations for which the dependent and the ethnicity variables are non-missing. We used robust standard errors, and the level of statistical significance was set at P < 0.05. The analysis was performed in Stata 16.¹⁹

Results

Descriptive statistics

Most patients had no CPR delay ≥1 min (59.1%) and the CPR duration was on average 16.2 min (SD = 14.8). The survival rate immediately after CPR was 51.6%, and 71.8% of the healthcare professionals reported being satisfied with the treatment provided. Additional descriptive statistics can be found in Table 1.

Regression analysis

Results for multiple logistic regression analysis and multiple ordinary least squares regression analysis are reported in Table 2. The results reveal significant ethnic differences with respect to both CPR delay and survival after CPR, but not for CPR duration and treatment satisfaction, when basic control variables were included in the regression models (Table 2, Panel A). Middle Eastern patients are less likely to receive delayed CPR [odds ratio (OR) = 0.71] and are more likely to be alive (OR = 1.60) after resuscitation compared with Nordic patients. African patients or the ‘Other’ category do not differ significantly from Nordic patients with respect to any of the outcomes in this regression model (or in the regression models reported below, and hence they are not discussed further in the Results section).

When initial heart rhythm and aetiology of the CA are added as controls (Table 2, Panel B), the difference in CPR delay between Nordic and Middle Eastern patients is no longer statistically significant. Middle Eastern patients still have significantly higher odds of surviving than Nordic patients (OR = 1.52) with the odds of surviving being ∼50% higher for Middle Eastern patients.

A posthoc analysis of the data shows that Middle Eastern patients were more likely to be heart rhythm monitored before the CA than Nordic patients even when adjusting for our basic control variables (t = 3.16; P = 0.002). Because heart rhythm monitoring is also correlated with significantly less CPR delay (rho = −0.213), shorter CPR duration (rho = −0.163), higher survival (rho = 0.239), and more treatment satisfaction (rho = 0.095), we followed up on its potential role as a mediator for the higher survival probability for Middle Eastern patients. Middle Eastern patients were no longer significantly more likely to be monitored when adjusting for aetiology, suggesting that the reason for the difference in monitoring was medical. To be even more cautious, we added heart rhythm monitoring as an additional control variable in the model, which in all other respects was identical to those reported in Table 2 (Panel B), and the results remain virtually unchanged. Thus, more prophylactic heart rhythm monitoring does not seem to explain why Middle Eastern patients are more likely to be alive after CPR.

Discussion

Overall, we find that African and Middle Eastern patients are treated relatively similarly to Nordic patients when suffering an IHCA in Sweden. Although Middle Eastern patients were somewhat less likely to have delayed CPR, this difference was no longer statistically reliable when aetiology and heart rhythm were adjusted for. One difference that remained statistically robust across our different models was the substantially higher likelihood of survival among Middle Eastern patients.

These findings were unexpected given that African and Middle Eastern individuals are subject to prejudice and face discrimination in Sweden, for example, in the labor market.^20–23 It is possible that the right to equal healthcare, stipulated in the Swedish Health and Medical Services Act, and the steadily growing public debate on this topic in Sweden have had an impact on healthcare professionals. When treating patients from ethnic minority groups, they might become mindful of their behaviour and exert effort into not treating patients differently based on ethnicity. When people are held accountable for their decisions, they are also less likely to act on their bias.²⁴ Because medical staff treating IHCA often work in teams where one’s behaviour is visible to others, and therefore accountability should be high, they might be more likely to monitor their own behaviour, especially when ethnicity is concerned. Any consciously held negative attitudes and associated emotions felt when treating the ethnic minority groups may have been successfully overcome.^25,26 Another possibility is that due to the seriousness of the studied IHCA setting, ethnic attitudes may play a weaker role compared to the labor market^22,23 which does not involve life or death decisions. This might explain why we find no ethnic treatment differences after adjusting for medical factors in this study. Because these explanations are admittedly speculative, an important task for future research is to examine ethnic attitudes among IHCA medical staff, and how such staff thinks and feel when they make treatment decisions about IHCA patients with different ethnic backgrounds.

However, the above explanations do not speak to why we find that Middle Eastern patients are more likely to survive IHCA. It is possible that this finding is explained by the healthy migrant effect, which is a frequently observed phenomenon whereby non-Western immigrants to western countries constitute a select and stronger subgroup of their native country.^27,28 The healthy migrant effect, however, typically fades away for each generation living in a western country and much research shows that immigrant children ultimately develop a higher risk of cardiovascular disease and death than their peers originating in the current country of residence.²⁹ Our data suggest that it is unlikely that the Middle Eastern survival advantage is explained by better cardiac arrest care for this group of immigrants. CPR duration was not significantly associated with ethnicity, and ethnic differences in CPR delay became non-significant when CA aetiology and initial heart rhythm were adjusted for. Furthermore, the Middle Eastern survival advantage persisted despite adjustment for heart rhythm monitoring. Finally, the medical staff did not report higher treatment satisfaction for patients of Middle Eastern origin.

With regards to African patients, we did not observe anything suggesting a healthy migrant effect. Immigrants from Africa are on average younger than the Swedish general population. They also tend to originate from North and Eastern Africa, many of whom have experienced decades of hardship and destabilized societies, which may have hampered their abilities to lead a healthy lifestyle, despite being able to migrate to distant countries.

Clinical implications

In healthcare, it is important to monitor quality aspects with respect to the treatment in different clinical scenarios. Such a process may force healthcare providers to reflect on their work and enable the potential for improvement. One aspect of quality is to not treat patients differently based on their ethnic group membership unless medically justified. Our results suggest that Middle Eastern or African patients are not receiving negative discriminatory treatment when suffering IHCA in Sweden. Hence, there seems to be no urgent need for Swedish hospitals to allocate resources on having their staff undergo diversity training targeted at changing attitudes towards these two ethnic minority groups. However, there might be healthcare settings where ethnic discrimination occurs, and which warrant diversity training.

Limitations

The sample size for the African group is rather small, which may have prevented us from detecting small group differences in treatment and survival for these patients. Moreover, the current research is based on Swedish data and they may not generalize to other countries. In fact, Sweden is considered to be at the forefront of equality,³⁰ and it is possible that the two studied ethnic minorities fare worse during IHCA treatment in other parts of the world, especially considering that black vs. white disparities have been found in the USA.^4,5

Another limitation of the study is that we did not adjust for the specific ward in which the CA took place. However, whether the patient’s heart rhythm was monitored before the CA, which we adjusted for in our final analysis, could be regarded as a proxy for the type of ward. In intensive care units, most patients receive heart rhythm monitoring, whereas, for general wards, most patients do not. In addition, we did not have access to data on whether patients were transported and admitted to an intensive care unit during the CA.

The comorbidity index used in the current research has not undergone any validation, which is a limitation. However, the same index has been used in one previous study¹² where it was negatively associated with socioeconomic status, as would be expected.³¹ Moreover, as expected, a closer look at the current data set shows that our index is positively correlated with age (rho = 0.11, P < 0.0001) and mortality (rho = 0.06, P = 0.0001). Together these findings offer some predictive validity in favour of our comorbidity index. It is nevertheless desirable that comorbidity measures undergo extensive validation before use when feasible, although lack of such validation seems to be a general weakness in the IHCA (and OHCA) literature.

Relatedly, as with any other registry study, we could not control for all potential ethnic health differences before the IHCA, as discussed above in the context of the healthy migrant effect.^27,28 It is possible that some unobserved initial health variable(s) explain the higher survival among Middle Eastern patients.

Conclusions

We conclude that when patients suffer a sudden cardiac arrest in Swedish hospitals, they appear to receive equal treatment regardless of ethnic background. Yet, the odds of surviving IHCA are ∼50% higher for Middle Eastern patients relative to Nordic patients.

Code availability

Stata do-files (code) will be shared by the corresponding author upon request.

Ethics approval

This research has been conducted according to the principles of Helsinki and was approved by the Regional Ethical Review Board in Linköping, Sweden (No. 2017/293-31).

Consent to participate

Informed consent is not applicable since the data come from various registries upheld by Swedish authorities. Personal identifiers had been removed for all patients appearing in the registries before the researchers were given access to the data.

Funding

This research was supported by the Swedish Research Council for Health, Working Life and Welfare (FORTE) (grant no. 2018-00256 to J.A.).

Conflicts of interest: None declared.

Data availability

The data that support the findings of this study are available from The Swedish Register of Cardiopulmonary Resuscitation and Statistics Sweden, but restrictions apply to the availability of these data, which were used under license for this study, and so are not publicly available. Researchers interested in using the data may contact The Swedish Register of Cardiopulmonary Resuscitation, email: [email protected], and Statistics Sweden, email: [email protected].

References