Temporal Changes in the Racial Gap in Survival After In-Hospital Cardiac Arrest

Key Points

Question  Have recent improvements in survival after in-hospital cardiac arrest been accompanied by a reduction in racial differences in survival?

Finding  In this study of 30 241 black and 81 898 white patients with in-hospital cardiac arrest at 289 hospitals in Get With the Guidelines–Resuscitation, risk-adjusted survival improved over time in both groups and differences in survival narrowed significantly.

Meaning  Racial differences in survival after in-hospital cardiac arrest have decreased substantially at hospitals in a national resuscitation quality improvement program.

Importance  Previous studies have found marked differences in survival after in-hospital cardiac arrest by race. Whether racial differences in survival have narrowed as overall survival has improved remains unknown.

Objectives  To examine whether racial differences in survival after in-hospital cardiac arrest have narrowed over time and if such differences could be explained by acute resuscitation survival, postresuscitation survival, and/or greater temporal improvement in survival at hospitals with higher proportions of black patients.

Design, Setting, and Participants  In this cohort study from Get With the Guidelines–Resuscitation, performed from January 1, 2000, through December 31, 2014, a total of 112 139 patients with in-hospital cardiac arrest who were hospitalized in intensive care units or general inpatient units were studied. Data analysis was performed from April 7, 2015, to May 24, 2017.

Exposure  Race (black or white).

Main Outcomes and Measures  The primary outcome was survival to discharge. Secondary outcomes were acute resuscitation survival and postresuscitation survival. Multivariable hierarchical (2-level) regression models were used to calculate calendar-year rates of survival for black and white patients after adjusting for baseline characteristics.

Results  Among 112 139 patients with in-hospital cardiac arrest, 30 241 (27.0%) were black (mean [SD] age, 61.6 [16.4] years) and 81 898 (73.0%) were white (mean [SD] age, 67.5 [15.2] years). Risk-adjusted survival improved over time in black (11.3% in 2000 and 21.4% in 2014) and white patients (15.8% in 2000 and 23.2% in 2014; P for trend <.001 for both), with greater survival improvement among black patients on an absolute (P for trend = .02) and relative scale (P for interaction = .01). A reduction in survival differences between black and white patients was attributable to elimination of racial differences in acute resuscitation survival (black individuals: 44.7% in 2000 and 64.1% in 2014; white individuals: 47.1% in 2000 and 64.0% in 2014; P for interaction <.001). Compared with hospitals with fewer black patients, hospitals with a higher proportion of black patients with in-hospital cardiac arrest achieved larger survival gains over time.

Conclusions and Relevance  A substantial reduction in racial differences in survival after in-hospital cardiac arrest has occurred that has been largely mediated by elimination of racial differences in acute resuscitation survival and greater survival improvement at hospitals with a higher proportion of black patients. Further understanding of the mechanisms of this improvement could provide novel insights for the elimination of racial differences in survival for other conditions.

Large racial differences in survival exist for in-hospital cardiac arrest despite an overall survival of 20% to 25%.^1-3 For in-hospital cardiac arrests attributable to a ventricular arrhythmia, black patients have 12% lower absolute and 27% lower relative survival compared with white patients.^1,2 Similarly, racial differences in survival have been reported for in-hospital cardiac arrest attributable to asystole or pulseless electrical activity (PEA).³ Although the reasons underlying racial differences in survival remain poorly defined, previous studies^2,4,5 suggest that racial differences may be attributable in part to worse outcomes at hospitals that disproportionately care for black patients.

During the past decade, survival after in-hospital cardiac arrest has improved markedly at hospitals participating in Get With the Guidelines–Resuscitation (GWTG-Resuscitation), a national quality improvement program for in-hospital resuscitation.¹ However, whether improved trends in survival have benefited black and white patients equally has remained unknown. Given the high incidence of in-hospital cardiac arrest in the United States, it is critical to understand whether recent improvements in survival have been accompanied by a reduction in racial differences in survival.

Accordingly, we examined trends in survival after in-hospital cardiac arrest among black and white patients in GWTG-Resuscitation and assessed whether racial differences have narrowed over time. To explain the underlying mechanism of survival improvement, we evaluated differences between black and white patients in acute resuscitation survival (ie, return of spontaneous circulation [ROSC]), postresuscitation survival (ie, survival to discharge in patients with ROSC), and resuscitation quality measures. Finally, we examined whether hospitals with a higher proportion of black patients with in-hospital cardiac arrest had the largest increase in survival.

The design of the GWTG-Resuscitation registry has been previously described in detail.⁶ The registry includes consecutive patients with in-hospital cardiac arrest (defined as the absence of a palpable central pulse, apnea, and unresponsiveness in patients without a do-not-resuscitate [DNR] order) who received cardiopulmonary resuscitation (CPR). Patients are screened and enrolled by trained personnel at participating hospitals. Data on cardiac arrests are collected using standardized Utstein-style definitions to allow uniform reporting of clinical variables across sites.^7-9 To optimize capture of cases, various methods are used: centralized collection of cardiac arrest flow sheets, hospital paging-system logs, routine checks of code carts, pharmacy tracer drug records, and hospital billing charges for resuscitation medications. Rigorous training of research personnel for operational definition adherence before data acceptance, a periodic reabstraction process, and use of standardized software are used for ensuring data completeness and accuracy in the registry. The institutional review board at the University of Iowa approved the study and waived the requirement for informed consent.

From January 1, 2000, through December 31, 2014, a total of 189 782 patients 18 years or older with an index in-hospital cardiac arrest were identified (eFigure 1 in the Supplement). Because our study focused on racial differences between black and white patients, we excluded 18 876 patients of other races or patients for whom race information was unavailable. We also restricted our cohort to cardiac arrests that occurred in an inpatient ward or intensive care unit and excluded 34 148 cardiac arrests that occurred in operating rooms, procedure suites, or emergency departments because of the unique circumstances associated with cardiac arrest at such locations. In addition, 601 patients with missing data on survival or calendar year of arrest were excluded. Finally, to obtain statistically reliable estimates of temporal trends in survival for black and white patients, we restricted our cohort to hospitals with at least 3 calendar years of data submission, a mean annual case volume of at least 10 cardiac arrest cases, and at least 10 black and 10 white patients. Thus, our final study population comprised 112 139 patients at 289 hospitals. Data analysis was performed from April 7, 2015, to May 24, 2017.

Our primary exposure variable was black or white race as reported by patients or their family members. Patient variables included demographic characteristics (age, sex), medical conditions before or at the time of cardiac arrest (congestive heart failure [current or prior]; myocardial infarction [current or prior]; diabetes; renal, hepatic, or respiratory insufficiency; baseline central nervous system depression; acute stroke; pneumonia; hypotension; sepsis; major trauma; metastatic cancer; or metabolic or electrolyte abnormality), cardiac arrest characteristics (initial rhythm [ventricular fibrillation, pulseless ventricular tachycardia, asystole, or PEA], the use of a hospital-wide cardiopulmonary arrest alert, time of cardiac arrest [daytime: 7 am-10:59 pm, nighttime: 11 pm-6:59 am, weekend: Saturday or Sunday], cardiac arrest location [not monitored, telemetry, or intensive care unit]), and interventions in place before the arrest (use of intravenous vasopressors, antiarrhythmic drugs, mechanical ventilation, and dialysis). Data regarding hospital characteristics were obtained from the American Hospital Association data and included geographic census region (Northeast, Midwest, South, or West), type of hospital (major teaching: hospitals with a fellowship program; minor teaching: hospitals with only a residency program; or nonteaching: hospitals without a residency or fellowship program), ownership (nonprofit or profit), location (urban or rural), hospital bed size (<100, 100–499, or ≥500 beds), and availability of cardiac catheterization laboratory.

The primary outcome was survival to hospital discharge. Secondary outcomes included acute resuscitation survival (defined as ROSC for at least 20 minutes) and postresuscitation survival (defined as survival to hospital discharge among patients who achieved ROSC). Given their overall association with survival, we also evaluated the following resuscitation quality measures as secondary outcomes: (1) device confirmation of correct endotracheal tube placement, (2) a monitored or witnessed cardiac arrest event, (3) time to first chest compression of 1 minute or less, (4) time to first defibrillation of 2 minutes or less for ventricular tachycardia or ventricular fibrillation, and (5) administration of epinephrine or vasopressin for PEA or asystole within 5 minutes.^10-13 In addition to the 5 individual measures, we derived an overall composite measure of defect-free care, defined as the proportion of patients who received all interventions for which they were eligible.

Given the large sample size, we compared baseline characteristics between black and white patients by using standardized differences. An absolute standard difference value greater than 10% was considered to be a statistically meaningful difference.¹⁴

Next, we calculated unadjusted rates of survival to discharge among black and white patients for each calendar year and then examined temporal trends by race over time. To evaluate the effect of patient characteristics on the above temporal trends, we constructed 2-level multivariable hierarchical regression models that account for clustering of patients within hospitals and therefore avoided overestimation of statistical associations.¹⁵ Moreover, because hospitals may begin and end participation in GWTG-Resuscitation at different years, such models account for clustering of observations within each year. In these models, hospital site and admission calendar year were included as random effects, whereas race, race × calendar-year interaction term, and patient characteristics described in the Study Variables subsection of the Methods section were included as fixed effects. In the aforementioned model, we specified a temporal correlation structure and assumed that the effect of race and the race-time interaction did not vary by hospital site. We used the model coefficients to calculate risk-adjusted rate of survival among black and white patients for each calendar year and then subtracted the risk-adjusted survival rate among black individuals from that among white individuals for each year to obtain annual absolute survival differences between black and white patients.

We conducted additional analyses to identify factors that explain the underlying mechanism of survival improvement. First, to determine whether reduction in survival difference between black and white patients over time was attributable to improvement in acute resuscitation survival, postresuscitation survival, or both, we constructed additional hierarchical multivariable logistic regression models for each outcome by using the same approach as described in the previous subsection. Second, we examined temporal trends in rates of defect-free care and the individual resuscitation quality measures between black and white patients to determine the extent to which improvement in survival was attributable to a temporal improvement in these measures. Furthermore, we repeated our analyses of racial differences in acute resuscitation survival after additionally adjusting for defect-free care as a patient-level categorical variable to determine whether temporal changes in adherence to resuscitation guidelines explained the survival difference by race.

On the basis of prior work² showing that lower survival rates among black patients were largely attributable to the concentration of black patients at hospitals with lower survival rates after in-hospital cardiac arrest, we stratified hospitals according to the proportion of black patients at each hospital. Hospitals at which the proportion of black patients was below the hospital median were classified as hospitals with a lower proportion of black patients, and hospitals above the median were classified as hospitals with a higher proportion of black patients. Hospital characteristics were compared between hospitals with lower and higher proportions of black patients. To determine whether narrowing of survival differences was more marked in one hospital group compared with the other, we examined temporal trends in risk-adjusted rates of survival to discharge at hospitals with lower and higher proportions of black patients, in the overall population, and separately for black and white patients at these hospitals. Risk-adjusted survival rates were derived using the aforementioned hierarchical models and included a calendar year by hospital group interaction term.

All analyses were performed with SAS statistical software, version 9.2 (SAS Institute Inc). All tests were 2 sided, and P < .05 was considered to be statistically significant.

Among 112 139 patients with in-hospital cardiac arrest, 30 241 (27.0%) were black (mean [SD] age, 61.6 [16.4] years) and 81 898 (73.0%) were white (mean [SD] age, 67.5 [15.2] years). Differences in baseline characteristics by race are summarized in Table 1. Compared with white patients, black patients were younger and less likely to have an initial rhythm of ventricular fibrillation or pulseless ventricular tachycardia and current or prior myocardial infarction but were more likely to have renal insufficiency, diabetes, baseline central nervous system depression, and septicemia and to have received dialysis before the arrest. Temporal trends in baseline characteristics by race are summarized in eTable 1 in the Supplement.

During the study period, black patients had lower unadjusted rates of survival to discharge compared with white patients (15.4% vs 19.9%; P < .001). Nonetheless, survival improved over time in both racial groups (Figure 1). After model adjustment, risk-adjusted survival improved from 15.8% in 2000 to 23.2% in 2014 among white patients and from 11.3% in 2000 to 21.4% in 2014 among black patients (P for trend <.001 for both races) (Table 2). However, the magnitude of survival improvement was greater among black patients compared with white patients on an absolute scale (survival difference between black and white patients decreased from 4.5% to 1.8% during the 15-year period; P for trend = .02) (Table 2) and a relative scale (compared with baseline, survival over time improved by 89.4% among black patients and by 46.8% among white patients; P for interaction <.001) (Table 2).

Unadjusted rates of acute resuscitation survival improved over time in both racial groups (Figure 2). After risk adjustment, acute resuscitation survival increased from 47.1% in 2000 to 64.0% in 2014 among white patients and from 44.7% in 2000 to 64.1% in 2014 among black patients (P for trend <.001 for both); by 2014, racial differences in acute resuscitation survival were eliminated (survival difference between black and white patients: 2.3% in 2000 to –0.1% in 2014; P for trend = .002) (Table 2). In contrast, although risk-adjusted postresuscitation survival also improved for both patient groups (Table 2 and Figure 2), the absolute differences in postresuscitation survival between black and white patients persisted during the study period (survival difference between black and white patients: 1.6% in 2000 to 2.7% in 2014; P for trend = .85).

We found that the rate of defect-free care improved over time among white individuals (73.3% in 2000 and 79.6% in 2014) and black individuals (68.9% in 2000 and 79.5% in 2014; P for trend <.001 for both groups). However, the magnitude of improvement over time was similar in both groups (P for interaction = .26) (eTable 2 and eFigure 2 in the Supplement). When examined separately, temporal trends in performance on individual quality measures varied. Of importance, racial differences in performance on individual measures, if any, were small and did not change over time (eTable 2 in the Supplement). Moreover, after additional adjustment for defect-free care, differences in acute resuscitation survival by race that were noted earlier in the study (2000-2010) became larger (eTable 3 in the Supplement).

Overall, the median proportion of black patients in the 289 study hospitals was 23.1% (interquartile range [IQR], 13.0%-38.1%; total range, 1.4%-94.2%). When categorized as hospitals with a lower proportion of black patients (below median) and a higher proportion of black patients (above median), the median proportion of black patients was 12.9% at hospitals with a lower proportion of black patients (IQR, 8.0%-17.5%) and 38.1% at hospitals with a higher proportion of black patients (IQR, 29.5%-52.2%). Hospital characteristics were similar between the 2 hospital groups except that private ownership was more common among hospitals with a higher proportion of black patients (eTable 4 in the Supplement). As shown in Figure 3, overall risk-adjusted survival at hospitals with a higher proportion of black patients was 11.9% in 2000 compared with 16.7% at hospitals with a lower proportion of black patients. By 2014, risk-adjusted survival had improved to 21.2% at hospitals with a higher proportion of black patients and 23.4% at hospitals with a lower proportion of black patients. Although survival improved over time in both hospital groups (P for trend <.001), the magnitude of improvement was much larger at hospitals with a higher proportion of black patients (P for interaction = .01). When survival rates were examined separately for black and white patients in both hospital groups (eFigure 3A and B in the Supplement), survival improvement over time was consistent with our main findings.

Our study has 3 principal findings. First, rates of survival to discharge after in-hospital cardiac arrest have improved over time among black and white patients, and racial differences in survival have narrowed. Second, greater improvement in survival among black individuals compared with white individuals was largely mediated by elimination of racial differences in acute resuscitation survival. Third, hospitals with a higher proportion of black patients with in-hospital cardiac arrest had lower baseline survival but the largest increases in survival over time. Collectively, our findings highlight a marked narrowing in racial differences in survival after in-hospital cardiac arrest.

Racial differences in clinical outcomes between black and white patients have been described for several medical conditions, including in-hospital cardiac arrest, and remain a primary driver of lower life expectancy among black patients.^2-5,16,17 However, few studies have examined whether racial differences in clinical outcomes have changed over time. In a study¹⁸ using data from the National Health Interview Survey, differences in breast and colorectal cancer screening rates between black and white individuals persisted during a 10-year period from 2000 to 2010. In contrast, a study¹⁹ from the GWTG–Heart Failure program found that differences in the use of implantable cardiac defibrillators for primary prevention that were present during 2005 were eliminated by 2010. Similar studies^20,21 from the GWTG programs in stroke and coronary artery disease also found a temporal reduction in racial differences with regard to provision of guideline-recommended treatments. However, those studies primarily focused on process-of-care measures and have not linked these changes to narrowing of racial differences in clinical outcomes. In contrast, our study found a significant reduction in racial differences in survival after in-hospital cardiac arrest, primarily through elimination of differences in acute resuscitation, and highlighted that increases in survival were greatest in hospitals that disproportionately care for black patients.

Previous studies^2,4,5 have also highlighted the role of hospital quality in explaining racial differences. An earlier GWTG-Resuscitation study² found that poor survival among black patients with in-hospital cardiac arrest owing to ventricular arrhythmias was largely attributable to the concentration of black patients in poor-quality hospitals. In fact, the hospital site at which patients had their cardiac arrest accounted for 40% of the racial difference in survival in that study. At the start of our study, survival was 4.8% lower at hospitals with a higher proportion of black patients compared with hospitals with a lower proportion of black patients. Although survival improved at both hospital groups over time, survival increases were larger at hospitals with a higher proportion of black patients than at hospitals with a lower proportion of black patients, suggesting that reductions in racial differences over time were mediated, at least in part, by greater improvement in survival outcomes at hospitals that disproportionately serve black patients with cardiac arrest.

We also found that, although racial gaps in acute resuscitation survival closed during the study period, racial differences in postresuscitation survival persisted between black and white patients. Clinical practice guidelines in recent years have emphasized important resuscitation quality measures.²² We found that a temporal improvement in survival in both racial groups was accompanied by a modest improvement in provision of defect-free care. However, differences by race in provision of defect-free care and individual quality measures were small and unlikely to account for the temporal reduction in survival differences between black and white patients. It is possible that strategies to improve CPR quality, audiovisual feedback, routine mock codes, and team debriefing, which are currently unmeasured in GWTG-Resuscitation, explain the aforementioned trends in survival by race.^22-24

There were notable differences between black and white patients in our study. Although black patients were younger and had a lower prevalence of cardiovascular comorbidities, the prevalence of most noncardiovascular comorbidities (eg, diabetes, septicemia, and pneumonia) was higher. Therefore, black patients were more likely than white patients to have a cardiac arrest attributable to asystole or PEA, rhythms that are associated with worse survival. These findings suggest that illness severity was higher and expected survival was lower in black individuals compared with white individuals. Second, prior studies^25,26 suggest that black patients are less likely to adopt DNR status compared with white patients. Moreover, a recent study²⁷ of more than 300 000 patients hospitalized with heart failure found that, although rates of DNR have increased modestly (10.4% in 2007 to 11.3% in 2010) over time, the increase was largely confined to white patients. Rates of DNR among black patients remained unchanged in that study. Therefore, the reduction in racial differences in survival over time noted in our study is unlikely to be a result of greater adoption of DNR status by black patients.

Strengths of our study include the use of a large and contemporary database with a standardized definition of in-hospital cardiac arrest, detailed information on important variables, and representation from a large number of hospitals. Of importance, we examined survival differences between black and white patients not only on a relative scale but also on an absolute scale. Given that black patients had lower survival at baseline, a greater relative improvement over time alone would be less meaningful if it were not accompanied by a reduction in absolute difference in survival. Thus, our study could serve as a paradigm for future studies examining appreciable changes in racial differences over time for other conditions.

Our findings should be interpreted in light of the following limitations. First, participation in GWTG-Resuscitation is voluntary; therefore, our results may not be generalizable to nonparticipating hospitals. Second, information on race was based on self-report, raising the possibility for misclassification of race. However, a previous study²⁸ found that the magnitude of misclassification of race by self-report compared with race reported on death certificate and medical record is small. Third, although we adjusted for a number of potential confounders in our risk-adjusted analyses, the chance of residual confounding attributable to unmeasured variables remains. Fourth, we lacked information on important acute resuscitation (eg, quality of CPR, frequency of interruptions), postresuscitation variables (eg, quality of nursing care, multidisciplinary teams), and hospital-specific quality improvement initiatives (eg, routine mock codes), thus limiting our ability to examine their association with racial differences in survival. Fifth, an increase in the use of DNR orders may account for the observed trends in our study. However, that seems unlikely given the temporal increase in illness severity, particularly among black patients, and previous reports^25,29,30 that suggest lower use of DNR and comfort-directed care and higher use of life-prolonging measures among black patients compared with white patients. Finally, because our study focused on differences between white and black patients, the findings are not applicable to other races.

We found a significant reduction in racial differences in survival after in-hospital cardiac arrest, largely mediated by elimination of racial differences in acute resuscitation and greater survival improvement at hospitals with a higher proportion of black patients. Further understanding the mechanisms of that improvement could provide novel insights for the elimination of racial differences in survival for other conditions.

Corresponding Author: Saket Girotra, MD, SM, Division of Cardiovascular Diseases, Department of Internal Medicine, University of Iowa Carver College of Medicine, 200 Hawkins Dr, Ste 4427, Roy Carver Pavilion, Iowa City, IA (saket-girotra@uiowa.edu).

Accepted for Publication: June 2, 2017.

Published Online: August 9, 2017. doi:10.1001/jamacardio.2017.2403

Author Contributions: Drs Joseph and Girotra had full access all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Joseph, Chan, Girotra.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Joseph, Girotra.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Zhou, Jones, Vaughan-Sarrazin, Girotra.

Obtained funding: Girotra.

Administrative, technical, or material support: Girotra.

Study supervision: Girotra.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Chan reported serving as a consultant for the American Heart Association. No other disclosures were reported.

Funding/Support: This study was funded by grants K08HL122527 (Dr Girotra) and R01HL123980 (Dr Chan) from the National Heart, Lung, and Blood Institute, National Institutes of Health.

Role of the Funder/Sponsor: Get With the Guidelines–Resuscitation is sponsored by the American Heart Association, which had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. The manuscript was reviewed and approved by the Get With the Guidelines–Resuscitation research and publications committee before journal submission.

Group Information: In addition to the study authors Paul S. Chan, MD, MSc, Steven M. Bradley, MD, MPH, and Saket Girotra, MD, SM, the American Heart Association Get With the Guidelines–Resuscitation Investigators include Dana P. Edelson, MD, MS, and Matthew M. Churpek, MD, MPH, PhD (University of Chicago, Chicago, Illinois), Benjamin Abella, MD, MPhil (University of Pennsylvania, Philadelphia), Monique L. Anderson, MD (Duke University School of Medicine, Durham, North Carolina), Romergryko Geocadin, MD (Johns Hopkins University School of Medicine, Baltimore, Maryland), Zachary D. Goldberger, MD, MS (University of Washington School of Medicine, Seattle), Patricia K. Howard, PhD, RN (University of Kentucky Healthcare, Lexington), Michael C. Kurz, MD (University of Alabama School of Medicine, Birmingham), Vincent N. Mosesso Jr, MD, (University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania), Boulos Nassar, MD, MPH (University of Iowa Carver College of Medicine, Iowa City), Joseph P. Ornato, MD, and Mary Ann Peberdy, MD (Virginia Commonwealth University Medical Center, Richmond), and Sarah M. Perman, MD, MSCE (University of Colorado School of Medicine, Aurora).