Author + information
- Received April 7, 2000
- Revision received July 11, 2000
- Accepted August 18, 2000
- Published online December 1, 2000.
- Andrew D Auerbach, MD, MPH∗,* (, )
- Mary Beth Hamel, MD, MPH†,
- Robert M Califf, MD, FACC‡,
- Roger B Davis, ScD†,
- Neil S Wenger, MD§,
- Norman Desbiens, MD∥,
- Lee Goldman, MD, MPH, FACC∗,
- Humberto Vidaillet, MD, FACC¶,
- Alfred F Connors, MD#,
- Joanne Lynn, MD∗∗,
- Neal V Dawson, MD††,
- Russell S Phillips, MD†,
- for the SUPPORT Investigators
- ↵*Reprint requests and correspondence: Dr. Andrew Auerbach, Department of Medicine, University of California, San Francisco, Department of Medicine-Box 0120, San Francisco, California 94143-0120
The goal of this study was to determine factors associated with receiving cardiologist care among patients with an acute exacerbation of congestive heart failure.
Because cardiologist care for acute cardiovascular illness may improve care, barriers to specialty care could impact patient outcomes.
We studied 1,298 patients hospitalized with acute exacerbation of congestive heart failure who were cared for by cardiologists or generalist physicians. Using multivariable logistic models we determined factors independently associated with attending cardiologist care.
Patients were less likely to receive care from a cardiologist if they were black (adjusted odds ratio [AOR] 0.53, 95% confidence interval [CI] 0.35, 0.80), had an income of less than $11,000 (AOR 0.65, 95% CI 0.45, 0.93) or were older than 80 years of age (AOR 0.23, 95% CI 0.12, 0.46). Patients were more likely to receive cardiologist care if they had college level education (AOR 1.89, 95% CI 1.02, 3.51), a history of myocardial infarction (AOR 1.59, 95% CI 1.17, 2.16), a serum sodium less than 133 on admission (AOR 1.96, 95% CI 1.30, 2.95) or a systolic blood pressure less than 90 on admission (AOR 1.97, 95% CI 1.20, 3.24). Patients who stated a desire for life extending care were also more likely to receive care from a cardiologist (AOR 1.40, 95% CI 1.04, 1.90).
After adjusting for severity of illness and patient preferences for care, patient sociodemographic factors were strongly associated with receiving care from a cardiologist. Future investigations are required to determine whether these associations represent unmeasured preferences for care or inequities in our health care system.
Previous studies have identified nonclinical factors associated with access to general medical care (1–3) and specialized procedures such as cardiac catheterization or cardiac revascularization (4–16). However, few data exist to describe patient factors associated with receiving care from a specialist physician at the time of hospitalization.
Many health care networks have adopted strategies that favor use of generalists, thereby encouraging less resource-intensive patient care (17,18). However, growing evidence suggests that specialty care may produce improved outcomes across a range of diseases, including rheumatoid arthritis (19), critical care medicine (20,21) and cardiovascular disease. Studies of cardiac syndromes suggest worse outcomes when cardiologist care is replaced by less specialized care for patients with unstable angina (22), acute myocardial infarction (MI) (23–25) and congestive heart failure (CHF) (26,27).
As a result, whether a patient receives care from a cardiologist or not may produce substantial differences in patient outcomes. In order to investigate correlates of receiving cardiologist care among patients experiencing an acute exacerbation of CHF, we studied 1,298 patients enrolled in the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT).
SUPPORT was a prospective, multicenter study of decision-making, care and outcomes for seriously ill hospitalized adult patients. Detailed descriptions of the study’s design, sites, patient population, data collection strategies and statistical methods have been published previously (28–31).
We studied patients enrolled in SUPPORT who had a primary diagnosis of acute exacerbation of CHF whose attending physicians were cardiologists or general internists. SUPPORT had two phases: an observation (phase I) and intervention phase (phase II). During the second phase, physicians were randomized to receive information about their patients’ prognoses and preferences for care, and nurse specialists facilitated communication between patients, families and their clinicians. Because the intervention did not affect processes or outcomes for patients with CHF (28), we included patients enrolled during both phases in our analyses.
Five geographically diverse institutions participated in SUPPORT: Beth Israel Hospital, Boston, Massachusetts; UCLA Medical Center, Los Angeles, California; Marshfield Clinic, Marshfield, Wisconsin; Duke Medical Center, Durham, North Carolina and MetroHealth Medical Center, Cleveland, Ohio. The study design was approved by the institutional review board at each site; informed consent was obtained at study entry.
Patients with CHF were included in SUPPORT if they were admitted to the hospital or transferred to the intensive care unit (ICU) with the primary diagnosis of an acute exacerbation of CHF. One of the following specific criteria were also required for inclusion: 1) history of severe CHF at baseline, New York Heart Association class III or IV and medications before admission that included two or more representatives from the diuretic, vasodilator or angiotensin-converting enzyme inhibitor drug classes or 2) New York Heart Association class IV CHF, systolic blood pressure 100 mm Hg or less or hypotension precluding use of medications listed or 3) chart documentation of CHF and left ventricular ejection fraction (LVEF) less than 20%.
Patients were excluded from SUPPORT if their CHF was due to valvular disease, restrictive cardiac disease, pericardial disease, iatrogenic fluid overload or renal failure. Patients were excluded if they were pregnant, not English speaking, nonresident foreign nationals, less than 18 years of age, transferred from another hospital and not to an ICU, had acquired immunodeficiency syndrome, had an estimated length of stay of less than 72 h or died or were discharged within 48 h of study entry. For this analysis eligible patients cared for by physicians who were not cardiologists or general internists were also excluded.
The specialty of the primary attending physician for each patient was recorded at admission by administrative offices at each site. Interviewers administered a physician questionnaire two to six days after SUPPORT study entry. In the phase I questionnaire physicians were asked how long they had cared for the patient and if they planned to care for the patient after hospitalization. This information was not collected in phase II.
Patient data were collected by chart abstraction and by interviews with patients and their surrogates (defined as the person who would make decisions for the patient if the patient was unable to do so). Data collected from charts included age, gender, insurance type, site of patient enrollment, presence of comorbid illnesses, medical history (e.g., MI or ventricular tachycardia), whether the patient was enrolled in SUPPORT while in an ICU and the acute physiology score (APS). The APS is the physiology-based component of Acute Physiology and Chronic Health Evaluation (APACHE) III and includes laboratory measurements, vital signs and Glasgow coma score. The APS has been shown to predict in-hospital mortality, with higher scores indicating increased risk (32).
Patient demographics, number of dependencies in activities of daily living, preferences for cardiopulmonary resuscitation and preferences for life-extending care in the event of terminal illness were compiled during interviews of patients or their surrogates between three and six days after study entry. Patient preference data for life-extending care were unavailable for 298 patients, respectively. In these cases we assumed the patient would desire more aggressive care—as would be done in clinical practice. Demographics including race, education and income were also obtained directly from the patient or their surrogate. When information regarding patient race, education or income was unavailable, data were imputed using methods described previously (33,34).
In addition to data collected during SUPPORT we abstracted charts of a random sample of generalist patients for the presence of specialty consultation and changes in attending specialty (such as transfer to a specialty service) during hospitalization. Generalist patients’ charts were available for abstraction at four of five SUPPORT study sites; 98 (17.7% of generalist patients) charts were abstracted.
We used descriptive statistics to characterize patients in the study. For bivariable comparisons we employed the Fisher exact test or the Wilcoxon rank-sum test.
Using multivariable logistic regression models with the specialty of the attending physician as the dependent variable, we assessed the independent effects of patient sociodemographic, preference and severity of illness measures upon the likelihood of being assigned to cardiologist care. Independent variables were initially chosen based on statistical significance in bivariable analyses as well as possible biologic or system-based association with receiving specialty care. Predictors were then included or excluded from models based on their independent statistical significance within the model, observed confounding effects or their ability to maintain face validity. Specific variables included were patient age, gender, race, insurance type, education, income, history of MI, ventricular tachycardia or ventricular fibrillation, whether the patient suffered an MI during the study hospitalization and whether an abnormal serum sodium, serum albumin or systolic blood pressure was measured in the first three study days. Models also contained adjustments for number of patient comorbidities, acute physiology score (APS) on day 1, number of dependencies with activities of daily living, presence of dementia as a comorbidity, whether or not the patient was in an ICU at study entry and site of patient enrollment. We used the c-statistic to evaluate the rank order discrimination of the model. Models included the entire cohort, performed with a c-statistic of 0.88 indicating excellent discrimination, and had a nonsignificant Hosmer-Lemeshow test, suggesting adequate calibration. Due to small sample size, patient data from the chart abstraction process were not analyzed using multivariable methods. All statistical analyses were performed using SAS 6.12 or 7.0 for Windows (SAS Institute, Cary, North Carolina).
To explore our findings we performed multiple secondary analyses (not presented). Analyses limited to patients with complete preference (n = 1,000) and LVEF (n = 772) data were not different from those presented. We explored the effect of site of care using site-stratified and site-limited analyses as well as analyses including adjustment for care in a transplantation center or stratified by admission to a heart transplantation center. These analyses did not suggest confounding by site-specific factors. Finally, we examined generalized estimating equations allowing for clustering of data by site of care and noted no difference in our findings related to site-specific grouping of effects.
Compared with generalists, cardiologists were older (45.7 vs. 40.3 years) and more likely to be men (91.4% vs. 80.9%). For the 448 patients (61% of phase I cohort, n = 735) with physicians who indicated whether they would provide care after discharge, similar proportions of cardiologists and generalists stated they would be providing care after discharge.
Patient characteristics (Table 1)
Of the 1,298 patients who met eligibility criteria, 555 were cared for by general internists, and 743 were cared for by cardiologists. Cardiologists’ patients were younger, more likely to be men, less likely to be black and more likely to have private insurance. Cardiologists’ patients were also more likely to have more than 12 years of education and have an annual income greater than $11,000.
The median APS score was lower among cardiologists’ patients, and serum albumin was higher, but they had lower systolic blood pressure and lower serum sodium at admission.
Specialty consultation among generalist patients
Of the generalist patient charts reviewed, 40.8% (n = 40) received cardiology consultation. No patients were transferred to the care of a cardiologist during hospitalization although one was transferred to another hospital for cardiac transplantation.
Patients with an annual income greater than $11,000 received cardiologist consultation (50.0% vs. 22.6%, p < 0.05) more often. No other factors were significantly associated with consultation. Patients who received consultation were more likely to be less than 60 years old (21% vs. 9%, p = 0.55), less likely to be black (13% vs. 19%, p = 0.40), more likely to have a history of MI (45% vs. 31%, p = 0.16) and more likely to want life-extending care (58% vs. 41%, p = 0.12).
Factors associated with care by attending cardiologists (Fig. 1)
In multivariable models patients were more likely to receive cardiologist care if they had 16 or more years of education. Patients were less likely to receive care from a cardiologist if they were black, insured by a combination of Medicare and Medicaid, uninsured or insured by Medicaid alone. Patients were also less likely to receive specialty care if they earned less than $11,000 annually. Patients with three or more comorbidities and those 70 years of age or older were also less likely to be cared for by a cardiologist.
Patients who had a history of MI were more likely to be cared for by a cardiologist as were patients with serum sodium less than 133, systolic blood pressure less than 90 mm Hg or a preference for life-extending care. A history of ventricular fibrillation or ventricular tachycardia was not significantly associated with receiving cardiologist care.
Results of subset analyses (not presented) among the phase I patients whose physicians provided information about their role in care after discharge (n = 448) were similar to those based upon the whole cohort. Whether a patient’s physician had cared for them for more than 1 month was not associated with the likelihood of receiving cardiologist care (AOR 0.81, 95% CI 0.45, 1.46).
In this cohort of adults hospitalized with an acute exacerbation of CHF, having a cardiologist attending was independently associated with younger age, higher education and income, better insurance and white race. These associations were significant even after adjustment for patient severity of illness, medical history, preferences for care and whether their physician provided care beyond hospitalization.
Socioeconomic factors and differences in care
Our results are consistent with studies describing factors associated with differences in the quality and amount of medical care patients receive. Lower socioeconomic status is associated with reduced access to services such as breast examinations, mammograms, Pap smears or childhood immunizations (35). Associations between end-stage renal disease and lower socioeconomic status are thought to result from reduced access to appropriate antihypertensive therapy (36,37). Patients who are poor, uninsured or who have completed less education are less likely to have a regular medical provider (1,2). Despite a higher burden of illness black patients use medical services less frequently and more often forgo care for economic reasons (38). Even after seeking care patients in vulnerable populations are more likely to experience adverse events (39). Care may be modified by age, with studies observing that elderly patients receive fewer hospital resources and less aggressive care (40).
Similar patterns have been described for the care of cardiovascular illnesses. Older patients are less likely to undergo cardiac testing or to receive thrombolytics (4,8,41). Racial differences in the appropriate use of coronary angiography have been noted in community settings (10–14,16,42) and in equal-access systems (43,44). Black patients are less likely to receive beta-adrenergic blocking agents by the third postinfarction day (45). Payor source affects use of invasive procedures and revascularization, with uninsured patients undergoing such procedures less frequently (46–49). Finally, patient educational level and preferences for care are important in determining whether patients undergo cardiac catheterization (50).
Barriers to specialty referral
Differences in care due to sociodemographic factors propagate across prehospital, intrahospital and posthospital phases of care (51). This study examines a single point in the process where differences can accrue: assignment to specialty care at the time of hospitalization. We hypothesized that the choice of physician at admission was driven by longitudinal patient-doctor relationships, yet our analyses did not suggest this effect. We also hypothesized that patient’s preferences for care would be important. However, adjusting for patient’s preference did not eliminate the independent effect of socioeconomic measures, suggesting the existence of other biases. Possible biases include physicians’ perceptions of patient prognosis, specific referral relationships between providers or subtleties of patients’ prior history that affect referral at the time of hospitalization. For example, it is possible that the older patients of generalists had undergone cardiovascular workups long before SUPPORT and were not considered for further evaluation. It is also possible that referral to cardiovascular specialty care was solely biased by patient demographics such as race, as has been seen in previous studies (52).
Factors associated with likelihood of specialty care may have an important impact on care quality. Referral to cardiologist care is associated with more frequent use of efficacious therapies and may improve outcomes of patients with cardiovascular illnesses (53). Although variable in absolute magnitude of benefit, specialty care of acute MI has been observed to impart a survival advantage (23–25,54). For patients with CHF, specialists have better knowledge of medications known to improve outcomes (55), use these medications more frequently (56) and are more likely to dose them appropriately (57). When generalists continue to provide care, cardiology consultation results in more frequent use of therapies associated with improved survival (58).
The SUPPORT patient population was a severely ill, hospitalized population with a poor short-term prognosis, and our results may not be generalizable to other patients or sites. SUPPORT excluded non-English speaking patients and those who died or were discharged within 72 h of admission, further limiting generalizability. We do not have data describing patient preference for specific caregiver specialties or interventions, nor did interviewers collect data about physicians’ preferences for referral. Misclassification bias among the generalist cohort due to specialty consultation or transfer to a cardiologist’s service after admission is also possible. Examination of generalist patients’ charts showed that transfer to specialty care was a rare event, however. These data suggest that, although consultation was common, similar nonclinical factors were associated with receiving cardiology consultation. A final shortcoming regards the lack of LVEF information for the entire cohort. Subset analyses among patients with available ejection fraction information were not different from those presented.
In this population of seriously ill patients hospitalized with CHF, having a cardiologist attending was associated with patient sociodemographic factors even after adjusting for severity of illness, longitudinal care by the physician and preferences for care. Further studies are required to investigate whether associations in receiving specialty care among patients of vulnerable sociodemographic groups reflect systematic biases, differences in patients’ preferences or clinical criteria not included in our analyses.
The authors would like to thank Jane Soukup, MS, for her assistance in assembling the database used for this paper.
☆ Supported by the Robert Wood Johnson Foundation.
Presented in abstract form at the annual meeting of the Society of General Internal Medicine, Washington DC, May 1998.
- Acute Physiology and Chronic Health Evaluation
- Acute Physiology Score
- congestive heart failure
- intensive care unit
- left ventricular ejection fraction
- myocardial infarction
- Study To Understand Prognoses And Preferences For Outcomes And Risks Of Treatments
- Received April 7, 2000.
- Revision received July 11, 2000.
- Accepted August 18, 2000.
- American College of Cardiology
- Gurwitz J.H,
- Osganian V,
- Goldberg R.J,
- Chen Z.Y,
- Gore J.M,
- Alpert J.S
- Schreiber T.L,
- Elkhatib A,
- Grines C.L,
- O’Neill W.W
- Casale P.N,
- Jones J.L,
- Wolf F.E,
- Pei Y,
- Eby L.M
- Reis S.E,
- Holubkov R,
- Edmundowicz D,
- et al.
- Edep M.E,
- Shah N.B,
- Tateo I.M,
- Massie B.M