Author + information
- Received December 10, 2017
- Accepted February 22, 2018
- Published online April 30, 2018.
- Khadijah Breathett, MD, MSa,∗ (, )
- Wenhui G. Liu, PhDb,
- Larry A. Allen, MD, MHSc,
- Stacie L. Daugherty, MD, MSPHc,
- Irene V. Blair, PhDd,
- Jacqueline Jones, PhD, RNe,
- Gary K. Grunwald, PhDb,f,
- Marc Moss, MDg,
- Tyree H. Kiser, PharmDg,h,
- Ellen Burnham, MDg,
- R. William Vandivier, MDg,
- Brendan J. Clark, MD, MSg,
- Eldrin F. Lewis, MD, MPHi,
- Sula Mazimba, MD, MPHj,
- Catherine Battaglia, PhD, RNb,k,
- P. Michael Ho, MD, PhDb,c,g and
- Pamela N. Peterson, MD, MSPHc,l
- aDivision of Cardiovascular Medicine, Sarver Heart Center, University of Arizona, Tucson, Arizona
- bVeteran Affairs Eastern Colorado Health Care System, Denver, Colorado
- cDivision of Cardiology, University of Colorado Anschutz Medical Campus, Aurora, Colorado
- dDepartment of Psychology and Neuroscience, University of Colorado, Boulder, Colorado
- eDepartment of Nursing, University of Colorado, Aurora, Colorado
- fDepartment of Biostatistics and Informatics, University of Colorado Anschutz Medical Campus, Aurora, Colorado
- gDivision of Pulmonary Sciences and Critical Care Medicine, Colorado Pulmonary Outcomes Research Group, University of Colorado, Aurora, Colorado
- hDepartment of Clinical Pharmacy, University of Colorado, Aurora, Colorado
- iDivision of Cardiology, Brigham and Women’s Hospital, Boston, Massachusetts
- jDivision of Cardiology, University of Virginia Health System, Charlottesville, Virginia
- kUniversity of Colorado School of Public Health, Denver, Colorado
- lDivision of Cardiology, Denver Health Medical Center, Denver, Colorado
- ↵∗Address for correspondence:
Dr. Khadijah Breathett, University of Arizona College of Medicine, Sarver Heart Center, 1501 North Campbell Avenue, Tucson, Arizona 85724.
Objectives This study sought to determine whether the likelihood of receiving primary intensive care unit (ICU) care by a cardiologist versus a noncardiologist was greater for Caucasians than for African Americans admitted to an ICU for heart failure (HF). The authors further evaluated whether primary ICU care by a cardiologist is associated with higher in-hospital survival, irrespective of race.
Background Increasing data demonstrate an association between better HF outcomes and care by a cardiologist. It is unclear if previously noted racial differences in cardiology care persist in an ICU setting.
Methods Using the Premier database, adult patients admitted to an ICU with a primary discharge diagnosis of HF from 2010 to 2014 were included. Hierarchical logistic regression models were used to determine the association between race and primary ICU care by a cardiologist, adjusting for patient and hospital variables. Cox regression with inverse probability weighting was used to assess the association between cardiology care and in-hospital mortality.
Results Among 104,835 patients (80.3% Caucasians, 19.7% African Americans), Caucasians had higher odds of care by a cardiologist than African Americans (adjusted odds ratio: 1.42; 95% confidence interval: 1.34 to 1.51). Compared with a noncardiologist, primary ICU care by a cardiologist was associated with higher in-hospital survival (adjusted hazard ratio: 1.20, 95% confidence interval: 1.11 to 1.28). The higher likelihood of survival did not differ by patient race (interaction p = 0.32).
Conclusions Among patients admitted to an ICU for HF, African Americans were less likely than Caucasians to receive primary care by a cardiologist. Primary care by a cardiologist was associated with higher survival for both Caucasians and African Americans.
African Americans have the highest risk of heart failure (HF) compared with other races/ethnicities (1) and have the highest overall mortality from HF (2). Yet, compared with Caucasians, African Americans are less likely to receive device therapies (3) and are less often treated with advanced therapies for HF (4). In an observational study of 5 centers from nearly 2 decades ago, African Americans hospitalized with HF were less likely to receive care by a cardiologist than Caucasians (5,6). It is unknown whether these differences in care persist in contemporary settings across the United States.
Multiple studies have demonstrated associations between care by a cardiologist and improved outcomes after an HF hospitalization (7–9). In-patient care by a cardiologist has been associated with a greater likelihood of receiving evidence-based treatments, reduced readmissions, and increased survival (7–10). However, these studies do not isolate outcomes for the highest-risk group: those requiring admission to an intensive care unit (ICU). It is unknown whether the improved outcomes associated with cardiology care in other settings holds true in the ICU setting. Furthermore, racial differences in the ICU are unknown.
Using a national observational database, Premier, we sought to evaluate the relationships between patient race and care by a cardiologist among HF patients admitted to the ICU. We hypothesized that: 1) the likelihood of receiving primary care by a cardiologist versus a noncardiologist would be greater for Caucasians than for African Americans; and 2) primary care by a cardiologist compared with a noncardiologist would be associated with higher in-hospital survival irrespective of race.
The Premier database retrospectively collects administrative data on over 700 U.S. hospitals, accounting for 20% of all in-hospital U.S. discharges (11). Hospital participation in the fee-based Premier database is voluntary and is done to improve hospital quality, outcomes, and efficiency (12). The data used for this analysis were from the subset of patients receiving ICU care at any time during their hospitalization. The database includes patient demographics, comprehensive billing, and procedure data, and is relied upon for quality outcomes research by organizations like the Centers for Medicare and Medicaid Services (11).
Adults 18 years of age or older admitted to any ICU with a primary discharge diagnosis of HF (denoted by discharge International Classification of Diseases-9th Revision-Clinical Modification [ICD-9-CM] 428.x, 414.8, 402, 398.91, 404.11, 404.13, 404.91, 404.93) between 2010 and 2014 were identified (n = 132,428). Race was captured in hospital administrative data (11). Patients with African-American or Caucasian race were included. Other races/ethnicities were excluded due to low representation (Hispanic n = 1,485; other n = 20,677). Patients who received consultative or procedural nonprimary ICU care by a cardiologist were excluded in the primary analysis (n = 5,431) and were included in the sensitivity analysis. The final cohort included 104,835 patients.
ICU admissions for HF occurred in 571 hospitals. Hospitals with <10 admissions to an ICU (n = 29) were excluded to reduce random variation. Hospitals lacking cardiologists were excluded (n = 45). The final hospital cohort included 497 hospitals.
Outcomes of interest
The primary outcome was receipt of primary ICU care by a cardiologist during ICU hospitalization for HF according to race. Primary care by a cardiologist was identified by denotation of both admission care by cardiologist and attending care by a cardiologist in the Premier Healthcare Database. Cardiologist care was defined by billing as cardiovascular disease or cardiac electrophysiology. The secondary outcome was in-hospital survival according to primary ICU care by a cardiologist with stratification by race. Death was defined by an expired status at discharge in the Premier database.
We compared patient and hospital characteristics between Caucasian and African-American patient groups. The following patient-level variables were examined: demographics (age, sex), comorbidities identified by ICD-9-CM codes (atrial arrhythmias, hypertension, diabetes, chronic kidney disease, obesity, end-stage renal disease, ventricular tachycardia, chronic obstructive pulmonary disease, or depression), procedures or cardiovascular events during hospitalization identified by ICD-9-CM or Current Procedural Technology codes (cardiogenic shock, cardiac arrest, cardiopulmonary resuscitation, echocardiography, left ventriculograms, intra-aortic balloon pump, pulmonary artery catheterization, dialysis during admission, or coronary catheterization without percutaneous coronary intervention [catheterization with percutaneous coronary intervention represented a small proportion of the population at 1.3% and was not adjusted for in the model]), and insurance type. Hospital-level variables included type (teaching hospital, urban/rural, geographic location) and racial variation in hospital population (defined by hospitals with both African-American and Caucasian patients vs. those with only 1 patient race). Chi-square tests were used to descriptively compare categorical data, and Wilcoxon nonparametric tests were used to compare continuous data.
Primary outcome: Receipt of primary ICU care by a cardiologist
Hierarchical logistic random effect models, separating between- and within-hospital effects, were used to determine the association between primary care by a cardiologist and patient race. Due to large hospital variation in the racial composition of patients, the association between race and care by a cardiologist may come from both patient and site levels. A between-within model separates the patient- and site-level race effects by including the hospital race proportion in the model. Because we were interested in the patient-level race effect, we adjusted for the hospital-level race effect as a confounder. We also adjusted for all other patient- and hospital-level variables listed in the previous text.
We addressed extreme racial variation in hospital population by performing additional analyses using only hospitals that had both Caucasians and African Americans. In addition, analyses were performed among subgroups of age (<65 years vs. ≥65 years), sex, and hospital type (rural vs. urban), with an interaction term between race and each subgroup because of known racial differences in HF presentation in these predetermined subgroups (1,13).
Sensitivity analyses that are described further in the Online Appendix include: 1) exclusion of patients with end stage renal disease and patients requiring dialysis during admission (n = 18,133); 2) exclusion of patients who experienced a major cardiac event (n = 7,102); and 3) inclusion of patients who received consultative or procedural care by a nonprimary cardiologist during ICU admission (n = 3,933).
We also performed a secondary analysis examining racial differences in receipt of care by a cardiologist by time to evaluate if the timing of the implementation of the Affordable Care Act Medicaid Expansion was associated with differences in care. Specifically, we included a binary covariate indicating the time before or after January 2014 (implementation of the Affordable Care Act Medicaid Expansion) and evaluated the interaction between time and race in the original model.
Secondary outcome: In-hospital survival
Cox models with inverse probability weighting were used to determine the association between care by a cardiologist and in-hospital survival and were stratified by race. The final Cox model was weighted with the stabilized inverse probability of censoring for each patient at different time points to adjust for possible informative censoring of in-hospital all-cause mortality by hospital discharge (14). Procedures or cardiovascular events during the hospitalization were treated as time-dependent variables in estimating the censoring weights. The denominator model for stabilized weights included race, cardiologist, time-dependent variables, and the remaining baseline variables used in the logistic model for primary care by a cardiologist. The numerator model included the same baseline variables. The final Cox weighted model included race, care by a cardiologist, interaction of race and cardiologist, and the baseline variables. A robust sandwich estimator was used to estimate the standard error of the log hazard ratios adjusting for clustering of data by hospitals. R package ipw (R Foundation for Statistical Computing, Vienna, Austria) was used for the inverse probability weighting analyses, and SAS version 9.4 (SAS Institute, Cary, North Carolina) was used for all other analyses.
Baseline characteristics of the study population by race
From January 2010 through December 2014, 104,835 patients with a primary discharge diagnosis of HF were admitted to an ICU. Of these, 84,182 (80.3%) were Caucasian, and 20,653 (19.7%) were African American (Table 1). Approximately one-half were men (51.9% Caucasian men, 49.2% African-American men). Caucasian patients were on average 11 years older than African-American patients (mean age: 73.1 years Caucasians, 62.3 years African Americans). Over one-third of patients had systolic HF (34.5% Caucasians, 36.5% African Americans), less than one-third had diastolic HF (30.7% Caucasians, 27.4% African Americans), and over one-third had an unknown HF type (34.8% Caucasians, 36.1% African Americans). The majority had some form of health care insurance. African Americans had lower proportions of Medicare (Caucasians 79%, African Americans 61.4%) and higher proportions of Medicaid (Caucasians 5.4%, African Americans 15.5%). There were modestly higher proportions of Medicaid after 2014 in both racial groups (before 2014: Caucasians 5.2%, African Americans 15.2%; after 2014: Caucasians 6.3%, African Americans 16.7%). There were racial differences in comorbidities, with more atrial arrhythmias, chronic obstructive pulmonary disease, and depression among Caucasians, and more diabetes, chronic kidney disease, end-stage renal disease, and obesity among African Americans.
Primary outcome: Primary ICU care by cardiologist for Caucasians compared with African Americans
The unadjusted proportion of patients with a primary discharge diagnosis of HF who were admitted to an ICU by a cardiologist was 15.0% in Caucasian patients and 14.8% in African-American patients. In adjusted analyses, Caucasians were more likely than African Americans to receive primary ICU care by a cardiologist to the ICU (odds ratio [OR]: 1.42; 95% confidence interval [CI]: 1.34 to 1.51) (Figure 1). Similar patterns were observed when hospitals were restricted to those that had both Caucasian and African-American patients with HF in the same hospital (OR: 1.43; 95% CI: 1.35 to 1.52; 455 hospitals and n = 101,830) (Figure 1).
Patients receiving primary ICU care by cardiologists compared with noncardiologists were slightly younger (69.1 years vs. 71.5 years) (Table 2). A higher proportion of patients receiving primary ICU care by a cardiologist were men (59.7% vs. 49.9%) and were admitted to urban hospitals (91.1% vs. 84.6%). Among subgroups, Caucasian patients were more likely to be admitted by a cardiologist compared with African-American patients after adjustment for age and hospital rurality (p ≤0.001). Differences in age groups and rurality did not alter the relationship (interaction of race with age: p = 0.27; race with rurality: p = 0.95) (Figure 1). However, the relationship between race and care by a cardiologist did differ by sex. Caucasian women were more likely to be admitted by a cardiologist compared with African-American women (OR: 1.31; 95% CI: 1.21 to 1.42), but the difference was greater among Caucasian than African-American men (OR: 1.51; 95% CI: 1.40 to 1.64; interaction of race with sex: p = 0.002) (Figure 1).
In sensitivity analyses, results were consistent when patients with end-stage renal disease and patients requiring dialysis during admission were excluded (Figure 2). The odds of receipt of care by a cardiologist were higher for Caucasians versus African Americans (OR: 1.49; 95% CI: 1.41 to 1.59). Similarly, after excluding patients with major cardiac events (Online Table 1), Caucasians were more likely to receive care by a cardiologist than African Americans (OR: 1.45; 95% CI: 1.36 to 1.55) (Figure 2). Findings were consistent when patients who received consultative or procedural nonprimary ICU care by a cardiologist were included. The odds of care by a cardiologist (primary or nonprimary) was higher among Caucasians than African Americans (OR: 1.45; 95% CI: 1.38 to 1.53) (Figure 2).
Before January 2014, the odds of care by a cardiologist was higher for Caucasians than African Americans (OR: 1.58; 95% CI: 1.48 to 1.68) (Figure 3). After the January 2014 implementation of the Affordable Care Act Medicaid Expansion, the odds of care by a cardiologist remained higher for Caucasians than African Americans (OR: 1.39; 95% CI: 1.23 to 1.57). However, after January 2014, the racial difference was reduced (interaction with time: p = 0.05).
Secondary outcome: In-hospital survival with primary ICU care by cardiologist compared with noncardiologist
The unadjusted all-cause in-hospital mortality was 7.13% among patients receiving primary ICU care by cardiologists and 8.23% among patients receiving primary ICU care by noncardiologists. The unadjusted all-cause in-hospital mortality by race was 6.21% among African Americans and 8.52% among Caucasians. After adjusting for patient- and hospital-level variables, primary care by a cardiologist was associated with better in-hospital survival compared with a noncardiologist (hazard ratio: 1.20; 95% CI: 1.11 to 1.28) (Figure 4). The interaction of race with primary care by a cardiologist was not statistically significant (p = 0.32), demonstrating that primary care by a cardiologist is associated with improved survival irrespective of Caucasian or African-American race.
In a large national sample of HF ICU hospitalizations, African Americans were less likely than Caucasians to receive primary ICU care by a cardiologist. Primary ICU care by a cardiologist compared with a noncardiologist was associated with higher in-hospital survival irrespective of race. Our findings suggest that care by a cardiologist may improve in-hospital survival among African Americans admitted to ICUs with HF.
Contemporary associations between race and care by a cardiologist exhibit a persistent but declining racial disparity in Caucasian and African-American patients. Previous studies from over 2 decades ago demonstrated nearly 90% increased likelihood of a HF patient receiving cardiology care if the individual was Caucasian instead of African American (5,6). Our findings revealed more than a 40% increased likelihood of Caucasians receiving care. Changes in health care policy may contribute to these findings, because African Americans are more disproportionately underinsured than Caucasians (15). Our time analysis displayed a modest reduction in racial differences in receipt of cardiology care after January 2014, when the Affordable Care Act Medicaid Expansion was implemented. This study does not demonstrate causal improvement in receipt of cardiology care. However, increasing access to insurance may have a role in reducing racial disparities, as suggested in other studies (4,16).
Our findings are consistent with prior studies demonstrating that care by a cardiologist is associated with increased survival among patients admitted with HF (7,9). In a decade of follow-up at a large academic center, 60-day mortality was reduced among patients with HF cared for by a cardiologist compared with noncardiologist, but in-hospital mortality was not significantly different between the 2 groups (7). In the Get With The Guidelines Database from 2005 through 2008, an increase in care by hospitalists compared with cardiologists for patients admitted with HF was associated with a modest increase in 30-day mortality (9). In addition, care by a cardiologist compared with noncardiologist has been associated with an increased likelihood to receive evidence-based medication upon discharge and reduced readmission rates (7–10).
African Americans have higher cumulative rates of death from HF than Caucasians, but lower in-hospital mortality (1,17). This does not mean that in-hospital care by a cardiologist is associated with worse outcomes in African Americans, but rather the contrary. African Americans have a higher risk of progression of left ventricular systolic dysfunction than Caucasians (18), and thus have a greater need for a specialist’s care to help reduce progression of disease and provide options for end-stage care. The observed racial differences between in-hospital and long-term mortality rates are thought to be because African Americans have a lower severity of HF than Caucasians at the time of admission (17). Systolic dysfunction was more common than diastolic dysfunction for both Caucasians and African Americans, and the HF phenotype was missing for approximately one-third of patients from both racial groups, which is similar to most large databases (19). However, measures of HF severity, such as New York Heart Association functional classification and INTERMACS profiles (classification of HF severity and limitations), were not available. Adjustment for procedures and events during hospitalization may have accounted in part for any differences in HF severity.
A myriad of factors may contribute to the racial differences in receipt of care. Compared with Caucasians, African Americans have higher proportions of hypertension (1), more comorbidities (1), and higher levels of underinsurance (15). Many of these findings were observed in this dataset. In particular, African Americans had higher rates of hypertension, diabetes, and end-stage renal disease. However, differences remained after adjusting for these factors and after removing high-risk groups. Unmeasured factors such as racial bias may be an underlying factor (20). Given the associations with time in this study and a U.S. health care policy in flux, negative bias toward patients with Medicaid should be further examined, particularly among African Americans who have higher proportions of Medicaid. Finally, variability in the identification of HF may vary by race and requires additional study.
In addition, the overall proportion of patients receiving care by a cardiologist was small. This may be related to growing intensivist/critical care and hospital medicine specialties (21,22). The prevalence of HF in the United States is expected to approach 8 million over the next few years (23), and we acknowledge that cardiologists cannot be expected to manage all of these patients. A collaboration between hospitalists, intensivists, and cardiologists will be necessary (24). However, better survival for critically ill patients requiring ICU stay for HF has been associated with care by cardiologists, and this should be considered irrespective of patient race.
First, this dataset did not include physician details such as: the source of admission, type of physician caring for the patient prior to ICU admission, and unbilled informal consultations with cardiology. In addition, this dataset did not include in-patient or discharge medications or cause of death. Second, the ICU classification of open versus closed unit was not available for this analysis. This may be associated with the proportion of patients cared for by a cardiologist, but should not be associated with racial differences because only hospitals with cardiologists were included. Third, approximately one-third of the patients were missing HF phenotype classification. For that reason, we did not adjust for risk based upon HF phenotype. Fourth, race was reported by administrative data rather than patient self-report alone. It is possible that patients of unidentified races could contribute to the results disparately. However, this contemporary dataset reports <0.01% missing data (11). Finally, as with any observational study, there are possible unmeasured confounders.
Among a contemporary cohort of patients admitted to the ICU with a diagnosis of HF, African Americans were less likely than Caucasians to be admitted by a cardiologist. Admission by a cardiologist compared with a noncardiologist was associated with better in-hospital survival irrespective of race. Future research should identify strategies to reduce racial differences in receipt of care by a cardiologist and improve quality of care and outcomes for patients who receive care from noncardiologists.
COMPETENCY IN MEDICAL KNOWLEDGE: Among patients admitted to an ICU for HF, African Americans are less likely to receive care by a cardiologist compared with Caucasians. Further, care by a cardiologist is associated with better in-hospital survival irrespective of race. Prior work has shown that African Americans have lower in-hospital mortality but higher overall mortality attributed to HF. Thus, future studies should evaluate whether equitable care by a cardiologist translates to a reduction in overall HF mortality rates for African Americans.
TRANSLATIONAL OUTLOOK: Interventions that reduce racial differences in receipt of care and improve quality of care by noncardiologists during ICU admissions for HF are indicated. Implementation research objectives should include reaching heterogeneous racial groups in both sexes.
Dr. Breathett has received support from the American Heart Association (AHA) Strategically Focused Research Network (#16SFRN29640000); the National Institutes of Health (NIH) L60 MD010857; the NIH/NCATS Colorado Clinical and Translational Sciences Institute (ULI TR001082); the University of Colorado, Department of Medicine, Health Services Research Development Grant Award; and the University of Arizona Health Sciences, Strategic Priorities Faculty Initiative Grant. Dr. Allen has received grant funding from NIH (K23 HL105896), the Patient Centered Outcomes Research Institute (CDR-1310-06998), and the AHA (#16SFRN29640000); and has served as consultant for Janssen, Amgen, Boston Scientific, and Novartis. Dr. Daugherty has received grant funding from the National Heart, Lung, and Blood Institute (NHLBI) (K08 HL103776 and RO1 HL133343) and the AHA (#2515963). Dr. Blair has received grant funding from the NHLBI (RO1 HL133343) and the AHA (15SFDRN24180024). Dr. Vandivier has received grant funding from the Flight Attendant Medical Research Institute (CIA092054 and 150001F). Dr. Clark has received grant funding from the NIH (K23 AA021814). Dr. Ho has served as a consultant for the AHA and Janssen Inc. Dr. Peterson has received grant funding from the AHA. Contents are the authors’ sole responsibility and do not necessarily represent official NIH views. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- heart failure
- intensive care unit
- International Classification of Diseases-9th Revision-Clinical Modification
- odds ratio
- Received December 10, 2017.
- Accepted February 22, 2018.
- 2018 American College of Cardiology Foundation
- Benjamin E.J.,
- Blaha M.J.,
- Chiuve S.E.,
- et al.
- ↵National Center for Health Statistics. Recent Trends in Heart Failure-related Mortality: United States, 2000–2014. NCHS Data Brief No. 231, December 2015. Available at: http://www.cdc.gov/nchs/products/databriefs/db231.htm. Accessed May 1, 2016.
- Breathett K.,
- Allen L.A.,
- Helmkamp L.,
- et al.
- Auerbach A.D.,
- Hamel M.B.,
- Califf R.M.,
- et al.
- Selim A.M.,
- Mazurek J.A.,
- Iqbal M.,
- Wang D.,
- Negassa A.,
- Zolty R.
- Uthamalingam S.,
- Kandala J.,
- Selvaraj V.,
- et al.
- Kociol R.D.,
- Hammill B.G.,
- Fonarow G.C.,
- et al.
- ↵Premier, Inc. Premier Applied Sciences supports health care transformation through generation of real-world evidence. Available at: https://www.premierinc.com/transforming-healthcare/healthcare-performance-improvement/premier-applied-sciences/. Accessed August 15, 2016.
- ↵Premier, Inc. Healthcare alliance: who it benefits. Available at: https://www.premierinc.com/healthcare-alliance/who-it-benefits/. Accessed October 26, 2017.
- Hernán M.A.,
- Brumback B.,
- Robins J.M.
- McMorrow S.,
- Long S.K.,
- Kenney G.M.,
- Anderson N.
- Davidson B.L.
- Morrow D.A.,
- Fang J.C.,
- Fintel D.J.,
- et al.
- Heidenreich P.A.,
- Trogdon J.G.,
- Khavjou O.A.,
- et al.
- Collins S.P.,
- Pang P.S.,
- Fonarow G.C.,
- Yancy C.W.,
- Bonow R.O.,
- Gheorghiade M.