Patterns of Resuscitation Care and Survival After In-Hospital Cardiac Arrest in Patients With Advanced Cancer

Abstract

Purpose:

Little is known regarding patterns of resuscitation care in patients with advanced cancer who suffer in-hospital cardiac arrest (IHCA).

Methods:

In the Get With The Guidelines – Resuscitation registry, 47,157 adults with IHCA with and without advanced cancer (defined as the presence of metastatic or hematologic malignancy) were identified at 369 hospitals from April 2006 through June 2010. We compared rates of return of spontaneous circulation (ROSC) and survival to discharge between groups using multivariable models. We also compared duration of resuscitation effort and resuscitation quality measures.

Results:

Overall, 6,585 patients with IHCA (14.0%) had advanced cancer. Patients with advanced cancer had lower multivariable-adjusted rates of ROSC (52.3% [95% CI, 49.5% to 55.3%] v 56.6% [95% CI, 53.8% to 59.5%]; P < .001) and survival to discharge (7.4% [95% CI, 6.6% to 8.4%] v 13.4% [95% CI, 12.1% to 14.8%]; P < .001). Among nonsurvivors who died during resuscitation, patients with advanced cancer had better performance on most resuscitation quality measures. Among patients with ROSC, patients with advanced cancer were made Do Not Attempt Resuscitation (DNAR) more frequently within 48 hours (adjusted relative risk, 1.30 [95% CI, 1.24 to 1.37]; P < .001). Adjustment for DNAR status explained some of the immediate effect of advanced cancer on survival; however, survival remained significantly lower in patients with cancer.

Conclusion:

Patients with advanced cancer can expect lower survival rates after IHCA compared with those without advanced cancer, and they are more frequently made DNAR within 48 hours of ROSC. These findings have important implications for discussions of resuscitation care wishes with patients and can better inform end-of-life discussions.

INTRODUCTION

Malignancy is the second leading cause of death in the United States, responsible for 25% of all deaths.¹ Although survival is improving, patients with advanced cancer continue to have high mortality and frequent encounters with the health care system.¹ As outcomes for advanced cancer have improved, so has the aggressiveness of care patients receive at the end of life, including hospital and intensive care admissions.^2-7 Inevitably, some patients with advanced cancer will suffer in-hospital cardiac arrest (IHCA).^8,9

Previous observational data have shown a poor prognosis among patients with advanced cancer who experience IHCA when compared with unselected patients.^10-13 Most of this previous work is limited by small sample sizes and single-center experiences. On the basis of these earlier studies, the extent to which differences in survival are driven by differences in patterns of resuscitation care, including the intensity of efforts, is also unclear. Understanding such issues is critical because it may inform arrest management and postresuscitation care (including goals of care discussion). Ultimately, this information could also help with earlier discussions regarding advanced care planning and the use of cardiopulmonary resuscitation in patients with advanced cancer.

In this context, we assessed patterns of resuscitation care and outcomes in a large national cohort of patients with and without advanced cancer across the United States who suffered IHCA. Our goal was to examine return of spontaneous circulation (ROSC) and overall survival to discharge, as well as the intensity of efforts during and after the IHCA.

METHODS

Data Sources

The Get With The Guidelines – Resuscitation (GWTG-R) registry is a large, multicenter observational registry of patients experiencing IHCA at hospitals throughout the United States. This registry has been described in detail previously in the literature.^14-17 Trained research personnel at each hospital prospectively collect data for patients who experience an IHCA (defined as the absence of a palpable pulse, apnea, and unresponsiveness). Arrest events are identified by collection of cardiac arrest flow sheets, review of paging system logs, and hospital billing charges for resuscitation. The database was also matched with the 2009 American Hospital Association Survey.¹⁸

Study Population

We identified 62,931 adult patients (18 years of age or older) experiencing an index IHCA (defined as the first pulseless arrest during a hospitalization) at 390 GWTG-R participating hospitals between April 2006 and June 2010. We restricted the analysis cohort to patients entered in the database after April 2006, because postresuscitation Do Not Attempt Resuscitation (DNAR) data were not collected before this date. To arrive at a more consistent cohort of hospital inpatients, we excluded certain patients. We excluded 12,033 patients who were in perioperative areas (pre-, intra-, or postoperative), other procedural areas (such as the cardiac catheterization laboratory or interventional radiology), emergency departments, or rehabilitation areas, or who were missing event location data. Patients with implantable cardioverter-defibrillators were also excluded (846 patients). Patients with arrest duration of < 2 minutes without ROSC were considered partial or incomplete resuscitations and were excluded (397 patients). Patients with missing event context, sex, or arrest cause data or with missing or implausible event times were also excluded (2,498 patients). Approximately 5% of available records were excluded because of missing data. The final study cohort included 47,157 patients from 369 hospitals.

Independent Variable

The independent variable in the study was advanced cancer, which is defined in the GWTG-R as any solid tissue malignancy with evidence of metastasis, or any blood borne malignancy. This variable represents an active diagnosis of cancer within 24 hours of arrest and was used to identify patients with advanced cancer at the time of IHCA.

Resuscitation Quality

Because knowledge of a patient’s underlying disease process and anticipated prognosis could potentially influence the resuscitation effort, we analyzed for differences in patterns of care during resuscitation.¹⁹ We evaluated duration of resuscitation attempts between patients with and without cancer who did not survive the arrest.²⁰ We evaluated for differences in the proportion of patients who received chest compressions, and whether compressions were started within 2 minutes of becoming pulseless. Among patients with a presenting rhythm of ventricular fibrillation (VF) or ventricular tachycardia (VT), we evaluated for differences in the proportion of patients who received defibrillation within 2 minutes of the start of the event. Among patients with a non-VT or non-VF rhythm, we evaluated for differences in the proportion of patients who received adrenaline (epinephrine) within 5 minutes of becoming pulseless.

Survival End Points

Our end points evaluated (1) survival to hospital discharge (primary end point) and (2) immediate survival of the arrest event with ROSC. Moreover, we hypothesized that early discussions regarding DNAR status may mediate the postarrest intensity of efforts among those achieving ROSC. Therefore, we examined for differences in rates of DNAR within 48 hours after ROSC among patients with and without advanced cancer.²¹

Statistical Analysis

Unadjusted comparisons of patient demographics and clinical presentation were made using t tests and χ² tests, as appropriate. Unadjusted differences in event and/or hospital survival and DNAR status prescription were presented as relative risks (RRs) for survival and were analyzed using χ² tests. Unadjusted differences in duration of resuscitation attempts and resuscitation quality measures were made using t tests and χ² tests, as appropriate.

Differences in outcome between patients with and without advanced cancer were evaluated using Poisson marginal regression models. These models incorporated generalized estimating equations to account for clustering of outcomes in hospitals.²² Poisson regression models were used to directly estimate RR and to avoid the inflation of estimates of standard odds ratios, which are generated using logistic regression in the setting of higher event rates (eg, ROSC).^23,24 Covariate selection was based on previously published and validated mortality models for the ICHA population.²⁵ A P value–based model selection approach was not used, and the same covariates were used in each regression model.

Our models were adjusted for a number of patient-level covariates that have been shown in previous work to be associated with survival,²⁵ including the following patient demographic and clinical covariates: age (as a linear variable); ethnicity; Hispanic ethnicity; illness category (medical [cardiac or noncardiac]; surgical [cardiac, noncardiac, or trauma]); pre-existing conditions before IHCA (myocardial infarction during this admission, hypotension or hypoperfusion, hepatic insufficiency, baseline depression in CNS function, acute stroke, septicemia, renal insufficiency, major trauma, or none); interventions in place at the time of cardiac arrest (invasive airway, ventilator, antiarrhythmic drug, vasopressor or vasodilator drug, chest tube, or arterial line); and cardiac monitoring in place. The pre-existing conditions reflect active diagnoses within 24 hours of the arrest. The models also included IHCA-specific covariates, including event location (intensive care unit, telemetry or step-down area, or general floor); whether the event was witnessed; whether the arrest team was activated using a hospital-wide paging system; the patient’s presenting cardiac arrest rhythm (VF or VT v other), off-hours arrest events (weekend [Saturday or Sunday], or night [11pm to 7am]); and time from admission to arrest event (in days). Finally, our models were adjusted for certain hospital characteristics, including rural versus urban hospital location, membership in a health care system, census division (geographic region), teaching status (on the basis of membership in the Council of Teaching Hospitals), and total number of annual admissions (as a categorical variable). An indicator variable for missing hospital characteristics was also used in the small number of hospitals unable to be matched to the American Hospital Association Survey (17 of 369 hospitals, or 4.6%). We also included the calendar year of admission in the models to account for secular trends.

Similar models (using the same covariates) were also created to analyze the other study outcomes of resuscitation quality measures, duration of resuscitation attempts (among those without ROSC), and DNAR status within 48 hours after ROSC (among those with ROSC). Duration of resuscitation attempts was analyzed using a linear marginal model (generalized estimating equations type, with normal distribution and identity link function); the remainder of the quality measures and DNAR status within 48 hours after ROSC were analyzed using similar Poisson regression models. Models including DNAR status and cancer status interaction terms were created in a sensitivity analysis to assess for effect modification between cancer and DNAR status.

All analyses considered P < .05 to be statistically significant, and all tests were two sided. Data are reported as mean (95% CI) or median (25th percentile, 75th percentile; interquartile range) where appropriate. Analyses were performed using SAS (Version 9.3; SAS Institute, Cary, NC).

The Institutional Review Board of the Mid-America Heart Institute approved this study and waived the requirement for written informed consent.

RESULTS

Among 47,157 patients suffering IHCA at 369 hospitals, 6,585 (14.0%) had a diagnosis of advanced cancer at the time of cardiac arrest. The median number of patients per hospital was 73 (interquartile range, 27-183). Baseline characteristics of patients by advanced cancer status are listed in Table 1. There were significant differences between the patients with and without advanced cancer across most patient and hospital characteristics. Patients with advanced cancer were younger, had fewer interventions in place at the time of their cardiac arrest, were less likely to be in the intensive care unit at the time of their arrest, were more likely to have pulseless electrical activity as the initial cardiac arrest rhythm, and were more frequently treated at teaching hospitals (Table 1).

Table 1.

Patient Demographics, Pre-existing Conditions, Arrest Event Scenarios, and Hospital Characteristics in the Overall, Cancer, and No Cancer Cohorts

Event Survival

Overall, 29,329 patients (62.2%) achieved ROSC, and 8,431 (17.9%) survived to hospital discharge. There were significant differences in survival between those with and without advanced cancer at both time points, with lower comparative survival. The unadjusted ROSC rate was 57.5% in patients with advanced cancer compared with 63.0% in those without advanced cancer, whereas survival to discharge was 9.6% v 19.2% (P < .001 for both comparisons). The lower survival rates in the patients with advanced cancer compared with those without advanced cancer persisted after multivariable adjustment (Table 2). Patients with advanced cancer had a 7% lower likelihood of achieving ROSC: 52.3% (95% CI, 49.5% to 55.3%) versus 56.6% (95% CI, 53.8% to 59.5%) for an RR of 0.93 (95% CI, 0.90 to 0.95; P < .001). In addition, patients with advanced cancer had a 45% lower likelihood of surviving to hospital discharge: 7.4% (95% CI, 6.6% to 8.4%) versus 13.5% (95% CI, 12.1% to 14.8%) for an RR of 0.55 (95% CI, 0.51 to 0.60; P < .001).

Table 2.

Unadjusted Survival Rates, Unadjusted Relative Risk for Survival by Cancer Status, Multivariate-Adjusted Survival by Cancer Status, and DNAR Status for Patients With and Without Cancer

Resuscitation Quality

To understand whether survival differences were potentially a result of less aggressive resuscitations among those with advanced cancer, we examined the duration of resuscitation among those who died during acute resuscitation as a surrogate measure of resuscitation intensity. A total of 17,828 patients (37.8%) did not survive the initial resuscitation (ie, did not achieve ROSC). Mean duration of resuscitation among nonsurvivors was 22.5 minutes (95% CI, 21.7 minutes to 23.3 minutes) in patients with advanced cancer compared with 24.2 minutes (95% CI, 23.9 minutes to 24.6 minutes) in those without advanced cancer (P < .001). After adjustment for potential confounders, there continued to be a significant difference: 22.5 minutes (95% CI, 21.5 minutes to 23.5 minutes) in patients with advanced cancer and 24.1 minutes (95% CI, 23.5 minutes to 24.7 minutes) in patients without advanced cancer (P < .001; Table 3). Performance on resuscitation quality measures was high in both groups of patients (those with and without advanced cancer), although patients with advanced cancer had slightly higher performance on the measures. Table 3 presents unadjusted and multivariate-adjusted performance on the quality measures. Missing information on the performance measures made up a small proportion and was well balanced between patients with and without advanced cancer.

Table 3.

Resuscitation Quality Measures for Patients With and Without Cancer

DNAR Status

Among the 29,329 patients (62.2%) who achieved ROSC after initial resuscitation, 13,077 (44.6%) had a DNAR order placed at some time before discharge, with 8,331 (28.4%) made DNAR within 48 hours after the event, 4,005 (13.7%) made DNAR > 48 hours after the event, and 741 (2.5%) made DNAR with undetermined timing. There was an association between the presence of advanced cancer and rates of DNAR status. The rate of DNAR status at any time after ROSC and before discharge was 55.6% in patients with advanced cancer compared with 43.0% in patients without advanced cancer (RR, 1.29 [95% CI, 1.25 to 1.34]; P < .001). Differences in DNAR rates between the groups were most pronounced within the first 48 hours, with rates of 37.3% in patients with advanced cancer compared with 27.1% in patients without advanced cancer (RR, 1.38 [95% CI, 1.31 to 1.44]; P < .001). After the first 48 hours, DNAR rates in the two groups were similar, with rates of 14.6% and 13.5% in patients with and without advanced cancer, respectively (RR,1.08 [95% CI, 1.00 to 1.18]; P = .06). The difference in overall DNAR status rates between groups persisted after adjustment, and this was entirely driven by DNAR status change in the first 48 hours after the event (Table 2).

Finally, the difference in survival to discharge (among survivors of the initial arrest) between patients with and without advanced cancer was attenuated after adjustment for DNAR status, rising from an RR of 0.61 (95% CI, 0.57to 0.65; P < .001) to 0.74 (95% CI, 0.67 to 0.80; P < .001).

DISCUSSION

Using a large national registry of IHCA, we observed that patients with advanced cancer made up a sizeable portion of those undergoing resuscitation care. We also found that survival rates after cardiac arrest were lower for patients with advanced cancer compared with other patients who suffer a cardiac arrest, independent of other patient and hospital factors. We noted that patients with cancer were made Do Not Resuscitate (DNR) more frequently within the first 48 hours after the arrest, although this does not fully explain the differences in survival to discharge between patients with and without cancer. We found few clinically significant differences in resuscitation quality between patients with and without cancer. Collectively, our findings provide much-needed insights into resuscitation patterns, DNAR decision making, and survival outcomes among patients with advanced cancer who suffer an IHCA. Our findings reassuringly support the notion that patients with advanced cancer receive cardiac arrest care consistent with their documented code status order.

Although previous research has demonstrated poor survival in patients with advanced cancer, to our knowledge, no previous study has evaluated cardiac arrest survival in a large nationwide cohort of patients. Although we were unable to determine survival rates for specific cancer diagnoses, we can say that patients with advanced cancer (ie, metastatic disease or hematologic malignancy) can expect, on average, a survival rate to hospital discharge of < 10%. This is a survival rate that is one half of that of patients without advanced cancer, independent of other risk factors. Although the survival rate is low, it is not so low as to be considered futile, as has been suggested by some single-center studies.^11,26-28 This is important information to include in a discussion of DNAR status among patients being admitted to the hospital with a cancer diagnosis. Goals of care ideally would be discussed with patients before the need for hospitalization, especially in the context of advanced malignancies lacking effective long-term treatment options.

Performance on the quality measures we evaluated was slightly higher in patients with advanced cancer. Although there was a modest, statistically significant difference in duration of resuscitation attempts between patients with and without advanced cancer, this difference is likely clinically insignificant. The major difference in care that we found was that patients with cancer had a higher rate of conversion to DNAR status within the first 48 hours after the arrest. Although we can only speculate as to the factors leading to this observation, conversion to DNR status likely results from an informed discussion of poor prognosis. However, the higher rate of early postarrest DNAR status may influence survival in patients by not allowing for appropriate resuscitation and sufficient time to achieve meaningful recovery. Notably, patients with malignancy also arrested more frequently in general medical wards, raising the possibility that there may have been delayed recognition of illness severity, leading to differing outcomes.

Our study has several limitations. First, our data source does not have granular data on the type of advanced cancer or the approaches and individual treatments that were used for the malignancy. Therefore, the capture of patients with advanced cancer may have been inaccurate and may have led to some degree of misclassification bias. Second, the GWTG-R enrolls only patients who are not DNAR at the time of their IHCA; therefore, our data do not include information on patients who were DNAR at the time of IHCA and thus did not receive resuscitation care. As a result, the patients with and without advanced cancer in this analysis may reflect a less sick population than the totality of patients with cancer in the hospital. This may be the primary factor affecting the higher survival rate seen in our study relative to previous research. Third, there were no data about whether DNAR was discussed before the IHCA event, or on why patient preferences for DNAR status changed after the event among those who achieved ROSC. Fourth, approximately 5% of eligible records were excluded because of missing data, possibly introducing selection bias. Fifth, we did not have access to the overall denominator of patients with or without cancer admitted to study hospitals; therefore, we could not report on whether patients with malignancy experience IHCA or are made DNAR more frequently. We note that, nationally, cancer diagnoses account for 4.3% of all hospital admissions; therefore, patients with cancer seem to make up a disproportionate share of patients undergoing resuscitation.²⁹

Despite these limitations, our study highlights a number of areas for further research. Future studies that include data about patients’ underlying cancer diagnoses and previous treatments will be important to determine whether certain populations with cancer have better survival after IHCA do than others. These data will be crucial to inform patients when discussing resuscitation care wishes, especially in the context of advanced malignancy and the possibility of ongoing cancer-directed therapy. Although it may be appropriate to attempt resuscitation in some patients with advanced cancer, the expectation of long-term survival for many patients with cancer may be small, and resuscitation may be less appropriate without a reasonable expectation of meaningful survival. This will require early and frank discussions around goals of care between providers and patients and their families when a patient is hospitalized.

In conclusion, patients with advanced cancer can expect lower survival after IHCA compared with those without advanced cancer, they have modestly shorter duration of resuscitation attempts, and they are more frequently made DNAR within 48 hours of ROSC. These findings have important implications for discussions of resuscitation care wishes with patients and can better inform end-of-life discussions.

AUTHOR CONTRIBUTIONS

Conception and design: Jeffrey T. Bruckel, Sandra L. Wong, Steven M. Bradley, Brahmajee K. Nallamothu

Collection and assembly of data: Jeffrey T. Bruckel, Sandra L. Wong

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Patterns of Resuscitation Care and Survival After In-Hospital Cardiac Arrest in Patients With Advanced Cancer

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/journal/jop/site/misc/ifc.xhtml.

Jeffrey T. Bruckel

Stock or Other Ownership: AvantGarde Health

Sandra L. Wong

No relationship to disclose

Paul S. Chan

Research Funding: National Heart, Lung, and Blood Institute

Steven M. Bradley

No relationship to disclose

Brahmajee K. Nallamothu

Leadership: United Healthcare Scientific Advisory Board