Association between a Hospital’s Quality Performance for In-Hospital Cardiac Arrest and Common Medical Conditions

Abstract

Background

Public reporting on hospital quality has been widely adopted for common medical conditions. Adding a measure of inpatient survival after cardiac arrest is being considered. It is unknown if this measure would be redundant, given evidence that hospital organization and culture can have hospital-wide effects on quality. Therefore, we sought to evaluate the correlation between inpatient survival after cardiac arrest and 30-day risk-standardized mortality rates for common medical conditions.

Methods and Results

Using data between 2007 and 2010 from a national in-hospital cardiac arrest registry, we calculated risk-standardized in-hospital survival rates for cardiac arrest at each hospital. We obtained risk-standardized 30-day mortality rates for acute myocardial infarction (AMI), heart failure (HF), and pneumonia from Hospital Compare for the same period. The relationship between a hospital’s performance on cardiac arrest and these other medical conditions was assessed using weighted Pearson correlation coefficients. Among 26,270 patients with in-hospital cardiac arrest at 130 hospitals, survival rates varied across hospitals, with a median risk-standardized hospital survival rate of 22.1% and an inter-quartile range (IQR) of 19.7% to 24.2%. There were no significant correlations between a hospital’s outcomes for its cardiac arrest patients and its patients admitted for AMI (correlation of −0.12; P=0.16), HF (correlation of −0.05; P=0.57), or pneumonia (correlation of −0.15, P=0.10).

Conclusions

Hospitals that performed better on publicly reported outcomes for three common medical conditions did not necessarily have better cardiac arrest survival rates. Public reporting on cardiac arrest outcomes could provide new information about hospital quality.

Introduction

Public reporting on mortality has been widely adopted for a few inpatient medical conditions, and there is some evidence that it can improve outcomes when combined with appropriate incentives.¹ In this context, broadening the scope of public reporting to include other conditions such as cardiac arrest is now being considered. Cardiac arrest would be a natural condition to add, as it affects 200,000 hospitalized adults in the U.S. each year,² and only one-fifth of such patients survive to discharge.³ Furthermore, wide variation in risk-standardized in-hospital survival rates⁴ suggests that there is room for improvement. Indeed, hospitals hoping to improve survival after cardiac arrest have begun to implement resuscitation-specific interventions, including identification of errors during resuscitation^{5, 6} and prompt defibrillation with appropriate energy.^{7, 8}

However, given growing evidence that more general aspects of hospital organization and culture (e.g., health information technology adoption), can impact quality of care for multiple conditions,^9-11 it is possible that a publicly reported measure of in-hospital survival after cardiac arrest would be redundant. For example, fair to good agreement exists among the mortality rates of three publicly reported conditions: acute myocardial infarction (AMI), heart failure (HF), and pneumonia.¹² In the case of in-hospital survival after cardiac arrest, it is unknown if a tight correlation exists with publicly reported measures of mortality for AMI, HF, and pneumonia. A strong correlation would suggest that finite resources should not be devoted to expanding public reporting to include cardiac arrest. It would also indicate that improving survival after cardiac arrest depends on identifying hospital factors that plausibly influence multiple disease states. On the other hand, a weak correlation between outcomes for common medical conditions and cardiac arrest would suggest that the quality signal from cardiac arrest is distinct. This would support the inclusion of cardiac arrest in public reports, and the value of resuscitation-specific quality improvement efforts.

Therefore, we used data from a large, national in-hospital cardiac arrest registry and the Centers for Medicare and Medicaid Services (CMS) Hospital Compare reports of hospital-level outcomes to examine the association between a hospital’s survival rate for cardiac arrest and its mortality rate for AMI, HF, and pneumonia. A better understanding of this relationship would help providers assess whether in-hospital mortality after cardiac arrest adds new information on hospital quality, and whether resuscitation-specific interventions around cardiac arrest have the potential to contribute to efforts to decrease in-hospital mortality.¹³

Methods

Data Sources

We conducted our study using 2 data sources. First, we examined hospital resuscitation outcomes using Get With the Guidelines®-Resuscitation, which previously was called the National Registry of Cardiopulmonary Resuscitation. GWTG-Resuscitation is a large, prospective registry of US patients with in-hospital cardiac arrest.¹⁴ Patients qualify for inclusion in the registry if they have a pulseless cardiac arrest, which is defined as the absence of a palpable central pulse, apnea, and unresponsiveness in patients without do-not-resuscitate orders. At participating hospitals, quality improvement personnel use standardized Utstein-style definitions^{15, 16} to collect information on the clinical characteristics and outcomes of consecutive inpatients with cardiac arrest. To enhance data accuracy, research personnel are certified, software supports standardized data collection, and data are routinely re-abstracted for quality assurance purposes.

To obtain 30-day hospital rates of risk-standardized mortality for AMI, HF, and pneumonia, we also used data from the Center for Medicare and Medicaid Services (CMS) on the Hospital Compare website at: www.hospitalcompare.hhs.gov. Hospital outcomes for cardiac arrest, AMI, HF, and pneumonia were then linked by the data coordinating center for GWTG-Resuscitation (University of Pennsylvania) by matching the American Hospital Association identification numbers present in both datasets. Data on hospital characteristics were obtained from the American Hospital Association Annual Survey.

Study Population

Between July 1, 2007 and June 30, 2010, we identified 40,907 patients 18 years of age or older at 368 hospitals within GWTG-Resuscitation with clinical information on an index in-hospital cardiac arrest (Figure 1). Since we were interested in examining the association between risk-standardized hospital survival rates for cardiac arrest and 30-day mortality rates for AMI, HF, and pneumonia, we excluded 148 patients with missing data on survival after cardiac arrest and 273 patients with missing data on location or time of cardiac arrest—critical variables used in deriving risk-standardized hospital survival rates after cardiac arrest. We additionally excluded 99 hospitals representing 8,785 cardiac arrest cases because information on hospital performance for AMI, HF, and pneumonia were not available on Hospital Compare. Nineteen percent of the excluded hospitals were military or Veterans Administration hospitals, and they, on average, had fewer cardiac arrest cases than hospitals included in the study cohort. Other differences between the excluded and included hospitals are summarized in Appendix Table 1. As noted in Figure 1, after additional exclusions, our final study cohort included 26,270 adults from 130 hospitals.

Figure 1.

Study cohort

Study Variables

GWTG-R collects data on multiple patient characteristics, including age, race (white, black, other), sex, initial cardiac arrest rhythm (asystole, pulseless electrical activity, ventricular fibrillation, or pulseless ventricular tachycardia), year of admission, location of arrest (intensive care unit, monitored unit, non-monitored unit, emergency department, other), and time of arrest (night vs. day; weekend vs. weekday). In addition, the registry collects information on the presence or absence of the following conditions within 24 hours of the cardiac arrest: heart failure; myocardial infarction or ischemia; arrhythmia; hypotension; renal, hepatic or respiratory insufficiency; diabetes mellitus; metabolic or electrolyte abnormality; acute central nervous system event (stroke or other); pneumonia; septicemia; major trauma; and malignancy.

Hospital characteristics available from the American Hospital Association data set included teaching status (major, minor, non-teaching), ownership (other non-profit, church, state/local government, and investor), location (urban vs. rural), geographic region (North and Mid-Atlantic, South and Atlantic, North Central, South Central, Mountain/Pacific), and number of beds (<250, 250-499, and 500+), as well as the ratio of full-time-equivalent registered nurses to beds at the hospital.

Study Outcomes

Our outcome of interest was the correlation between a hospital’s risk-standardized survival-to-discharge rate for cardiac arrest and its 30-day risk-standardized mortality rates for AMI, HF, and pneumonia. We used hospital 30-day risk-standardized mortality rates for AMI, HF, and pneumonia for July 2007 to June 2010, as reported on Hospital Compare.^17-19 These mortality rates are adjusted for patient demographics and co-morbid conditions, as has been previously described,^{17, 18} and reflect death from any cause within 30 days of admission.

Statistical Analyses

We divided hospitals in the GWTG-Resuscitation dataset into quartiles of unadjusted hospital survival rates for cardiac arrest. We then described patient and hospital characteristics in each quartile, comparing continuous variables using one-way analysis of variance, and categorical variables using chi-square or Fisher’s exact test.

Similar to the methodology that CMS uses to calculate risk-standardized mortality for AMI, HF, and pneumonia,^17-19 we estimated risk-standardized cardiac arrest survival rates at hospitals adjusting for the patient characteristics described above. To ensure proper adjustment for clustering of patients within hospitals, 2-level hierarchical logistic regression models were constructed, with survival to discharge as the dependent variable. In these analyses, individual hospitals were modeled as random effects and patient characteristics as fixed effects within each hospital. From the multivariable model, hospital-specific random intercepts were used to calculate predicted survival rates for each hospital. The risk-standardized survival rate was then calculated as follows: (predicted survival rate at a hospital/expected survival rate at the same hospital)*average unadjusted survival rate at all hospitals in our sample. We utilized predicted vs. expected survival rates, as this approach provides likely more accurate risk-standardized estimates for hospitals with low case volumes than an observed vs. expected approach.^20-22 A detailed description of the calculations using a predicted vs. expected approach is provided in Appendix Table 2, which is reproduced from prior work.²³

In our main analyses, we evaluated the correlation between a hospital’s risk-standardized cardiac arrest survival rate and its risk-standardized 30-day mortality rate for AMI, HF, and pneumonia using Pearson correlation coefficients (and in sensitivity analyses, using a Spearman rank correlation). We weighted each hospital by the number of patients with an index cardiac arrest during the study period. We contrasted these primary findings with the correlations between a hospital’s risk-standardized 30-day mortality rate for AMI, HF, and pneumonia.

All statistical analyses on linked data sets obtained from the University of Pennsylvania were conducted at the Mid-America Heart institute using SAS, Version 9.3 (SAS Institute, Cary, NC). All tests for statistical significance were 2-tailed and were evaluated at a significance level of 0.05. The Institutional Review Board at the Mid-America Heart Institute waived the requirement for informed consent as the study used de-identified data.

Results

Patient, event, and hospital characteristics differed when compared across quartiles of unadjusted hospital rates of cardiac arrest survival-to-discharge. Compared with patients treated at hospitals in the lowest quartile of survival, those in the highest quartile of survival were more likely to be white, have an initial rhythm of ventricular fibrillation or pulseless ventricular tachycardia, arrest in a monitored unit or the Emergency Department, and arrest on a weekday (Tables 1 and 2). The pre-existing characteristics of patients differed as well, with those patients treated at hospitals in the highest quartile of survival more likely to have respiratory insufficiency, arrhythmias, hypotension, a prior history of heart failure, incident heart failure, a prior history of myocardial infarction, incident myocardial infarction, metabolic or electrolyte abnormalities, and an acute central nervous system non-stroke event. Compared with hospitals in the lowest quartile of survival, those in the highest quartile of survival were also less likely to be located in the South (Table 3).

Table 1.

Patient characteristics, stratified by quartiles of in-hospital cardiac arrest survival rates

	Quartiles of In-Hospital Cardiac Arrest Survival Rates
	Quartile 1 (Low) N=36 (n=6,544)	Quartile 2 N=31 (n=6,538)	Quartile 3 N=24 (n=6,509)	Quartile 4 (High) N=39 (n=6,679)	P-value*
In-Hospital Cardiac Arrest Survival, %	5.4 to <18.1	18.1 to <21.7	21.7 to <25.6	25.6 to <35.6
	Patient Characteristics
Demographics
Age
Mean ± SD	65 ± 16.6	66 ± 15.7	66 ± 16.0	65 ± 15.9	< 0.001
Median (IQR)	67 (54.0, 78.0)	68 (56.0, 78.0)	68 (56.0, 79.0)	67 (55.0, 78.0)
Male	3,751 (57.3%)	3,736 (57.1%)	3,814 (58.6%)	3,992 (59.8%)	0.007
Race
White	3,881 (59.3%)	4,744 (72.6%)	4,849 (74.5%)	5,060 (75.8%)	< 0.001
Black	1,915 (29.3%)	1,534 (23.5%)	1,201 (18.5%)	886 (13.3%)
Other	202 (3.1%)	134 (2.0%)	163 (2.5%)	196 (2.9%)
Pre-Existing Conditions
Respiratory insufficiency	2,122 (32.4%)	2,723 (41.6%)	3,115 (47.9%)	2,806 (42.0%)	< 0.001
Renal insufficiency	2,052 (31.4%)	2,229 (34.1%)	2,285 (35.1%)	2,216 (33.2%)	< 0.001
Arrhythmia	1,433 (21.9%)	2,008 (30.7%)	2,120 (32.6%)	2,156 (32.3%)	< 0.001
Diabetes mellitus	1,854 (28.3%)	2,155 (33.0%)	2,013 (30.9%)	2,046 (30.6%)	< 0.001
Hypotension	1,129 (17.3%)	1,833 (28.0%)	1,898 (29.3%)	2,140 (32.9%)	< 0.001
Heart failure this admission	929 (14.2%)	1,184 (18.1%)	1,225 (18.8%)	1,123 (16.8%)	< 0.001
Heart failure prior to admission	1,165 (17.8%)	1,317 (20.1%)	1,390 (21.4%)	1,273 (19.1%)	< 0.001
MI this admission	772 (11.8%)	1,026 (15.7%)	1,344 (20.6%)	1,230 (18.4%)	< 0.001
MI prior to admission	645 (9.9%)	801 (12.3%)	1,449 (22.3%)	1,029 (15.4%)	< 0.001
Metabolic/electrolyte abnormality	711 (10.9%)	1,042 (15.9%)	846 (13.0%)	1,102 (16.5%)	< 0.001
Septicemia	1,149 (17.6%)	1,054 (16.1%)	1,118 (17.2%)	1,148 (17.2%)	0.15
Pneumonia	764 (11.7%)	1,005 (15.4%)	809 (12.4%)	897 (13.4%)	< 0.001
Metastatic/hematologic malignancy	836 (12.8%)	800 (12.2%)	748 (11.5%)	788 (11.8%)	0.13
Baseline depression in CNS function	624 (9.5%)	1,096 (16.8%)	661 (10.2%)	649 (9.7%)	< 0.001
Hepatic insufficiency	485 (7.4%)	455 (7.0%)	462 (7.1%)	561 (8.4%)	0.007
Acute CNS non-stroke event	273 (4.2%)	539 (8.2%)	476 (7.3%)	561 (8.4%)	< 0.001
Acute stroke	294 (4.5%)	218 (3.3%)	240 (3.7%)	216 (3.2%)	< 0.001
Major trauma	340 (5.2%)	292 (4.5%)	262 (4.0%)	330 (4.9%)	0.008
Interventions in Place †
Assisted/mechanical ventilation	2,411 (36.8%)	1,914 (29.3%)	2,375 (36.5%)	2,277 (34.1%)	< 0.001
IV vasoactive agents	2,043 (31.2%)	1,825 (27.9%)	2,035 (31.3%)	1,970 (29.5%)	< 0.001
IV antiarrhythmics	367 (5.6%)	367 (5.6%)	468 (7.2%)	498 (7.5%)	< 0.001
Pulmonary artery catheter	84 (1.3%)	120 (1.8%)	241 (3.7%)	221 (3.3%)	< 0.001
Dialysis/extracorporeal filtration	267 (4.1%)	209 (3.2%)	228 (3.5%)	237 (3.5%)	0.05
Intra-aortic balloon pump	69 (1.1%)	78 (1.2%)	156 (2.4%)	143 (2.1%)	< 0.001
Internal cardiac defribrillator	113 (1.7%)	116 (1.8%)	104 (1.6%)	113 (1.7%)	0.88

Abbreviations and definitions: N is number of hospitals, n is number of patients; SD is standard deviation; IQR is interquartile range; MI is myocardial infarction; CNS is central nervous system; IV is intravenous

Note: Numbers do not sum to sample total for every row, because some characteristics had missing data.

^*Continuous variables compared using one-way analysis of variance. Categorical variables compared using chi-square or Fisher’s exact test.

^†Interventions in Place refers to interventions already in place when need for chest compressions and/or defibrillation was first recognized.

Table 2.

Event characteristics, stratified by quartiles of in-hospital cardiac arrest survival rates

	Quartiles of In-Hospital Cardiac Arrest Survival Rates
	Quartile 1 (Low) N=36 (n=6,544)	Quartile 2 N=31 (n=6,538)	Quartile 3 N=24 (n=6,509)	Quartile 4 (High) N=39 (n=6,679)	P-value*
In-Hospital Cardiac Arrest Survival, %	5.4 to <18.1	18.1 to <21.7	21.7 to <25.6	25.6 to <35.6
	Event Characteristics
Admission year
2007	2,135 (32.6%)	2,152 (32.9%)	1,990 (30.6%)	2,224 (33.3%)	0.008
2008	2,202 (33.6%)	2,255 (34.5%)	2,333 (35.8%)	2,311 (34.6%)
2009	2,207 (33.7%)	2,131 (32.6%)	2,186 (33.6%)	2,144 (32.1%)
Arrest Time
Day	3,108 (47.5%)	3,333 (51.0%)	3,340 (51.3%)	3,425 (51.3%)	< 0.001
Night	1,267 (19.4%)	1,232 (18.8%)	1,195 (18.4%)	1,214 (18.2%)
Weekend	2,169 (33.1%)	1,973 (30.2%)	1,974 (30.3%)	2,040 (30.5%)
Location
ICU	3,427 (52.4%)	3,041 (46.5%)	3,166 (48.6%)	3,036 (45.5%)	< 0.001
Monitored	905 (13.8%)	1,072 (16.4%)	932 (14.3%)	1,219 (18.3%)
Non-monitored	1,197 (18.3%)	1,306 (20.0%)	1,173 (18.0%)	926 (13.9%)
ER	548 (8.4%)	583 (8.9%)	641 (9.8%)	825 (12.4%)
Procedural	339 (5.2%)	400 (6.1%)	495 (7.6%)	549 (8.2%)
Other	128 (2.0%)	136 (2.1%)	102 (1.6%)	124 (1.9%)
Initial Rhythm
Asystole	2,348 (35.9%)	2,207 (33.8%)	2,065 (31.7%)	1,948 (29.2%)	< 0.001
PEA	3,212 (49.1%)	3,167 (48.4%)	3,061 (47.0%)	3,217 (48.2%)
VF	579 (8.8%)	680 (10.4%)	843 (13.0%)	911 (13.6%)
PVT	405 (6.2%)	484 (7.4%)	540 (8.3%)	603 (9.0%)

Abbreviations and definitions: N is number of hospitals, n is number of patients; SD is standard deviation; IQR is interquartile range; ICU is intensive care unit; PEA is pulseless electrical activity; VF is ventricular fibrillation; PVT is pulseless ventricular tachycardia

Note: Numbers do not sum to sample total for every row, because some characteristics had missing data.

^*Continuous variables compared using one-way analysis of variance. Categorical variables compared using chi-square or Fisher’s exact test.

Table 3.

Hospital characteristics, stratified by quintiles of in-hospital cardiac arrest survival rates

	Quartiles of In-Hospital Cardiac Arrest Survival Rates
	Quartile 1 (Low) N=36 (n=6,544)	Quartile 2 N=31 (n=6,538)	Quartile 3 N=24 (n=6,509)	Quartile 4 (High) N=39 (n=6,679)	P-value*
In-Hospital Cardiac Arrest Survival, %	5.4 to <18.1	18.1 to <21.7	21.7 to <25.6	25.6 to <35.6
	Hospital Characteristics
Teaching Status
Major teaching	6 (16.7%)	5 (16.1%)	9 (37.5%)	10 (25.6%)	0.21
Minor teaching	14 (38.9%)	7 (22.6%)	8 (33.3%)	10 (25.6%)
Non-teaching	16 (44.4%)	19 (61.3%)	7 (29.2%)	19 (48.7%)
Ownership
Other Non-Profit	17 (47.2%)	15 (48.4%)	12 (50.0%)	26 (66.7%)	0.13
Church	2 (5.6%)	6 (19.4%)	6 (25.0%)	5 (12.8%)
State/Local Government	7 (19.4%)	4 (12.9%)	5 (20.8%)	4 (10.3%)
Investor	10 (27.8%)	6 (19.4%)	1 (4.2%)	4 (10.3%)
Region
North and Mid-Atlantic	5 (13.9%)	3 (9.7%)	3 (12.5%)	7 (17.9%)	0.03
South Atlantic	14 (38.9%)	13 (41.9%)	9 (37.5%)	4 (10.3%)
North Central	6 (16.7%)	8 (25.8%)	8 (33.3%)	15 (38.5%)
South Central	8 (22.2%)	7 (22.6%)	4 (16.7%)	6 (15.4%)
Mountain/Pacific	3 (8.3%)	0 (0.0%)	0 (0.0%)	7 (17.9%)
Urban	34 (94.4%)	26 (83.9%)	21 (87.5%)	37 (94.9%)	0.35
Full-time equivalent nurse ratio
<1	2 (5.6%)	4 (12.9%)	0 (0.0%)	1 (2.6%)	0.08
1 to <1.5	15 (41.7%)	11 (35.5%)	6 (25.0%)	7 (17.9%)
1.5 to <2	11 (30.6%)	11 (35.5%)	13 (54.2%)	13 (33.3%)
2 to <2.5	7 (19.4%)	5 (16.1%)	3 (12.5%)	11 (28.2%)
2.5 to <3	1 (2.8%)	0 (0.0%)	1 (4.2%)	4 (10.3%)
3+	0 (0.0%)	0 (0.0%)	1 (4.2%)	3 (7.7%)
Number of beds
<250	16 (44.4%)	8 (25.8%)	8 (33.3%)	17 (43.6%)	0.43
250 to 499	13 (36.1%)	17 (54.8%)	8 (33.3%)	14 (35.9%)
500+	7 (19.4%)	6 (19.4%)	8 (33.35)	8 (20.5%)

Abbreviations and definitions: N is number of hospitals, n is number of patients; SD is standard deviation; IQR is interquartile range

Note: Numbers do not sum to sample total for every row, because some characteristics had missing data.

Hospital Variation in Outcomes for Cardiac Arrest, AMI, HF, and Pneumonia

After risk-standardization of hospital rates of cardiac arrest survival-to-discharge, we found that the median risk-standardized cardiac arrest survival rate among the 130 hospitals in our study was 22.0%, but this rate varied across hospitals (inter-quartile range [IQR]: 19.6% to 24.0%; range, 12.8% to 33.6%) (Figure 2). To contextualize this variation, there was also site-level variation in hospital outcomes for common medical conditions. The median 30-day mortality rate (i.e., death from any cause within 30 days of admission) for AMI was 15.3% (IQR: 14.4% to 16.6%; range, 12.4% to 19.6%), for CHF was 11.2% (IQR: 10.3% to 12.3%; range, 8.6% to 15.5%), and for pneumonia was 11.7% (IQR: 10.7% to 13.0%; range, 8.4% to 18.3%) (Appendix Figures 1A to 1C).

Figure 2.

Distribution of risk-standardized hospital rates of in-hospital cardiac arrest survival

Relationship between Hospital Outcomes for Cardiac Arrest, AMI, HF, and Pneumonia

In unadjusted analyses, there were no significant associations between a hospital’s cardiac arrest survival rate and its 30-day risk-standardized mortality rate for AMI (correlation of −0.11, P=0.13), HF (correlation of 0.03, P=0.46), or pneumonia (correlation of −0.12, P=0.66). After risk-standardization of hospital cardiac arrest survival rates, we found weak but statistically insignificant negative correlations between a hospital’s cardiac arrest survival rate and its mortality rates for AMI (correlation of −0.12, P=0.16), HF (correlation of −0.05, P=0.57), and pneumonia (correlation of −0.15, P=0.10) (Figures 3A to 3C). In sensitivity analyses, results were very similar using the Spearman correlation coefficient.

Figure 3A. Correlation between hospitals’ risk-standardized cardiac arrest survival rates and 30-day risk-standardized acute myocardial infarction (AMI) mortality rates.

The size of the circle represents the number of patients at each hospital, with the smallest hospital having 34 patients and the largest hospital having 922 patients.

Figure 3C. Correlation between hospitals’ risk-standardized cardiac arrest survival rates and 30-day risk-standardized pneumonia mortality rates.

The size of the circle represents the number of patients at each hospital, with the smallest hospital having 34 patients and the largest hospital having 922 patients.

In contrast, 30-day risk-standardized mortality rates for AMI, HF, and pneumonia, were all positively correlated (for AMI and HF, correlation of 0.40; for AMI and pneumonia, correlation of 0.37; and for HF and pneumonia, correlation of 0.49), and these correlations were statistically significant (all p-values < 0.001).

Discussion

Given wide variation in hospital survival rates for cardiac arrest, we examined whether hospitals that perform well for common medical conditions also excel in outcomes for cardiac arrest. We found no significant correlations between a hospital’s publicly reported 30-day mortality rates for AMI, HF, or pneumonia, and its in-hospital survival rate for cardiac arrest. These findings suggest that the quality signal from cardiac arrest is distinct from that conveyed by the other measures. Because of this, it is likely that public reporting on in-hospital cardiac arrest would add to existing information on hospital quality. Our findings are also consistent with value in resuscitation-specific interventions to improve in-hospital cardiac arrest survival.

In clinical areas other than resuscitation, there is growing interest in identifying how the organizational and cultural traits of hospitals – such as nurse staffing, leadership, safety culture, and implementation of health information technology – may impact a broad range of patient outcomes.^{9-11, 24-26} For example, Bradley et al. found that certain aspects of a hospital’s organizational environment (e.g., openness to creative problem-solving) were associated with lower 30-day risk-standardized mortality rates for AMI.²⁷ Others have demonstrated an association between health information technology and hospital-wide quality.^9-11 One might postulate that a hospital’s cultural and organizational environment could also impact the quality of care for resuscitation.

However, most prior work in cardiac arrest has sought to describe and improve hospital processes of care that apply only to resuscitation.^6-8 Empiric evidence to support such a focused approach has been lacking. It is quite plausible that high-performing hospitals for some diseases (e.g., HF, AMI, and pneumonia) are not the same hospitals that excel in the care of cardiac arrest patients, thereby necessitating different types of quality improvement interventions. For example, standing orders and electronic reminders may be helpful for AMI, HF, and pneumonia (all of which have long lead times between diagnosis and discharge), but of little use for cardiac arrest, where prompt recognition and rapid response times (e.g., defibrillation and cardiopulmonary resuscitation) are associated with improved survival.^{8, 28}

In our data, lack of correlation between high-performing hospitals for HF, AMI, and pneumonia, and those for cardiac arrest indicate that general efforts to improve quality for common medical conditions do not -- in and of themselves -- translate into high quality for cardiac arrest. At a minimum, our results suggest that wide variation in 30-day mortality rates for AMI, HF, and pneumonia among hospitals is not due to differences in achieving survival to discharge after cardiac arrest. Our findings could further be interpreted as supporting the need to pursue resuscitation-specific efforts to improve survival for patients with in-hospital cardiac arrest. Such interventions may include those targeted at preventing cardiac arrests (e.g., hospital monitoring, rapid response teams, remote intensive care unit monitoring), improving acute resuscitation care (e.g., times to defibrillation and vasopressors, high-quality chest compressions with minimal interruptions), and/or optimizing post-resuscitation survival, as well as strengthening resuscitation systems of care (e.g., simulations of and debriefing after cardiac arrest). While leadership and culture no doubt lay the foundation for quality improvement efforts across a variety of domains, our results are consistent with the hypothesis that these factors are necessary but not sufficient for improved cardiac arrest survival. Future studies are needed to determine which resuscitation-specific interventions have the greatest impact on cardiac arrest survival, and how each of these interventions interacts with broader determinants of hospital quality.

Our results are in keeping with prior work that has found that hospitals’ performance on a variety of quality measures is not always tightly correlated. For example, the correlation between hospital performance on process measures for AMI, HF, and pneumonia is modest.^{12, 29} Others have found that hospital performance on process measures is weakly correlated with outcome measures.^30-32 In contrast, a more limited number of studies have found that hospital mortality rates across a variety of surgical procedures may be related.³³ Similar to prior work,¹² we also found that 30-day risk-standardized mortality rates for AMI, HF, and pneumonia were correlated among hospitals. To date, however, few studies have described the association between hospital performance on quality outcomes in different clinical domains (e.g., resuscitation and medicine). While we found that outcomes for AMI, HF, and pneumonia are not tightly correlated with outcomes for cardiac resuscitation, additional research will be needed to identify what set of quality measures for which conditions provides the best picture of overall hospital quality at a reasonable cost.

Our study has several important limitations. First, we used data collected at hospitals enrolled in GWTG-Resuscitation, so our findings may not be generalizable. However, GWTG-Resuscitation includes acute care hospitals located throughout the nation, and these hospitals care for a variety of patients and have diverse structural characteristics. Second, we were only able to examine survival or mortality for a few inpatient conditions, but these are very common conditions with a long history of public reporting. Furthermore, our outcome of survival is arguably of greatest importance to patients. Third, we were limited to use of registry data in our adjustments for severity of illness. Moreover, we did not examine functional or neurological status at discharge. Related to this, we did not have information about how hospitals might vary in their aggressiveness around end-of-life care. Hospital culture regarding palliative care could be an unmeasured confounder. For example, it is possible that hospitals with higher quality CHF, AMI, and pneumonia care were also better at resuscitation, but our results did not reflect this because of unmeasured hospital-level differences in severity of illness and aggressiveness around end-of-life care. Finally, we measured associations and cannot make conclusions about causality. For example, it is possible that the lack of a correlation in mortality for common inpatient medical conditions and cardiac arrest survival reflects decades of public reporting on process and outcome measures for AMI, HF, and pneumonia, and relative neglect of other conditions such as cardiac arrest.

In summary, we found that a hospital’s 30-day mortality rates for HF, AMI, and pneumonia were not correlated with its in-hospital survival rate for in-hospital cardiac arrest. This supports current efforts to make in-hospital cardiac arrest a publicly reported measure, as it would likely yield new information about hospital quality. It further supports the need for resuscitation-specific interventions to improve in-hospital cardiac arrest survival rates.

Figure 3B. Correlation between hospitals’ risk-standardized cardiac arrest survival rates and 30-day risk-standardized heart failure (HF) mortality rates.

The size of the circle represents the number of patients at each hospital, with the smallest hospital having 34 patients and the largest hospital having 922 patients.