Author + information
- Received May 28, 2002
- Revision received July 1, 2002
- Accepted July 18, 2002
- Published online January 1, 2003.
- Maria Ansari, MD*,†,* (, )
- Mark Alexander, PhD†,‡,
- Ali Tutar, MD*,†,§,
- David Bello, MD*,†,§ and
- Barry M Massie, MD, FACC*,†,§
- ↵*Reprint requests and correspondence:
Dr. Barry M. Massie, Cardiology Division (111C), San Francisco VAMC, 4150 Clement Street, San Francisco, California 94121, USA.
Objectives This study examined the outcomes of new-onset heart failure (HF) outpatients managed by cardiologists and primary care (PC) physicians.
Background Several studies have sought differences in outcomes between patients with HF managed by cardiologists and PC physicians, but most focused on inpatients, who often represent later stages of HF, whereas many treatments have their impact by delaying disease progression.
Methods This was a retrospective cohort study of incident HF identified between 1996 and 1997 in a staff model health maintenance organization. Cardiology care was defined as ≥2 visits or ≥25% of total medical outpatient visits to cardiology. Records from a cohort of 403 patients with new-onset outpatient HF were reviewed. The main outcome measure was a combination of death and/or cardiovascular hospitalization at 24 months.
Results Cardiologists’ patients (n = 198) were younger (66 vs. 71 years, p = 0.001), were more likely men (54% vs. 46%, p = 0.01), had coronary artery disease (64% vs. 42%, p = 0.001), and had a low (≤45%) ejection fraction (EF) (66% vs. 44%, p < 0.001) compared with PC physicians’ patients. More cardiologists’ patients received an EF assessment (94% vs. 74%, p < 0.001), angiotensin-converting enzyme inhibitors (83% vs. 68%, p < 0.001), and beta-blockers (38% vs. 22%, p < 0.001). In multivariate proportional hazards modeling that included variables that differed between providers and univariate predictors of outcomes, cardiology care was an independent predictor of a lower risk for the combined outcome (hazard ratio 0.62, confidence interval 0.42 to 0.93, p = 0.02).
Conclusions Cardiology care at this early stage of HF is associated with improved guideline adherence and a reduced risk of the composite outcome of death plus cardiovascular hospitalization.
Despite recent advances in therapy, such as angiotensin-converting enzyme (ACE) inhibitors and beta-blockers, patients with chronic heart failure (HF) continue to experience high mortality and morbidity rates, resulting in five-year survival rates in the community below 50% and approximately one million hospitalizations each year in the U.S. (1–3). Heart failure is one of the few cardiovascular conditions increasing in prevalence, approaching five million patients in the U.S., and is also the second most frequent cardiovascular reason (after hypertension) for outpatient physician visits. Primary care (PC) providers manage the majority of patients with HF, and one goal of some managed care organizations has been to reduce costs by limiting utilization of specialists in treating this common condition.
The rapid evolution in treatment recommendations for HF has led some to question whether specialty care might improve outcomes for these patients. Several studies indicate that cardiologists are more likely to practice in conformity with guideline recommendations, but they also employ more diagnostic tests and procedures than PC practitioners (4–7). In the management of acute myocardial infarction (MI), this combination of greater adherence to guidelines and greater procedural use may be associated with better outcomes (5,8–11), but it is less clear whether this is the case with HF patients (6,12–14).
Prior studies of HF management and outcomes have focused on patients who were hospitalized or referred to specialized HF programs. Such patients are usually the most severely affected, often representing treatment failures, and they experience very high rate of subsequent mortality and morbidity. At this far end of the spectrum, comparisons between PC providers and specialists may not be appropriate. In addition, hospitalized patients with HF are often co-managed, making it difficult to distinguish contributions of general and specialty care. As a result, it may be more relevant and methodologically preferable to examine management practices and outcomes at earlier stages, particularly because the most effective treatments for HF appear to exert their benefit by preventing its progression. Therefore, we designed a retrospective cohort study to examine the characteristics, management, and outcomes of outpatients with new-onset HF managed by PC physicians and cardiologists. This study was undertaken in a staff model health maintenance organization (HMO) and did not involve cardiologists specializing in HF, which enhances its relevance to community practice.
Identification of study cohort
This was a retrospective cohort study of patients with new-onset HF conducted within the northern California Kaiser Permanente Medical Care Program (KPMCP). The Kaiser system has highly developed and well-validated data management systems, which allow identification of outpatient diagnoses, provider specialty, and patient outcomes. The KPMCP also has a relatively stable membership of insured patients who receive almost all of their care at Kaiser facilities and are not restricted in their access to cardiology specialty care. Specifically, access to specialty care is not limited by financial incentives, policy guidelines, or the need for prior authorization. We identified patients with new-onset HF managed initially in the outpatient setting from July 1996 through August 1997 and evaluated their treatment and outcomes over a two-year period. This study was performed in accordance with a research protocol approved by the institutional review board at KPMCP.
Incident HF cases were identified from the database by an outpatient encounter form with diagnostic codes of HF (International Classification of Disease [ICD]-9 code 428.0), cardiomyopathy (ICD-9 code 425.0), or hypertensive heart disease with HF (ICD-9 402.1, 402.11, 402.91, 404.01, 404.3, 404.11-.15) in individuals enrolled in the KPMCP for at least 12 months. To restrict the cohort to patients with new-onset HF, those with a prior primary or secondary diagnosis of a HF-related diagnosis on a prior hospital discharge were excluded, as were patients admitted within 24 h of their diagnosis. The date of the first encounter was treated as the baseline date for HF diagnosis.
We identified 8,597 potential subjects with HF during the study window who had no prior outpatient visit with an ICD-9 code diagnosis for HF (Fig. 1). Additional patients were excluded because of a prior HF hospital discharge diagnosis (n = 2,941), lack of continuous enrollment in the KPMCP (unless the patient died during the follow-up period) (n = 1,201), and the presence of prespecified conditions thought to dominate subsequent management and outcomes, including end-stage renal disease on hemodialysis, dementia, HIV/AIDS, cirrhosis, or malignancy other than basal cell carcinoma (n = 250). Of the remaining 4,205 potential subjects, a random sample of 529 patients was selected for thorough chart review. Of this initial sample of patients, approximately 80% were followed by PC physicians alone. Therefore, an additional random sample of 165 charts of patients who were seen by any cardiologist was included to enrich the comparison group.
The chart reviews were performed by a professional chart reviewer with a nursing background (n = 619) and by a physician employed part time for this purpose (n = 75). Patients’ charts were reviewed for up to two years before the first HF diagnosis, and for two years after the initial diagnosis to examine the processes of care and clinical outcomes. Of the 694 charts reviewed, 127 patients who had clear evidence of preceding HF (thereby not meeting the inclusion criterion for new-onset HF) and 17 patients with no clinical evidence to support the computer diagnosis of HF using Framingham criteria (15)were excluded. In addition, during chart review, 18 patients who were found to have one or more of the previously listed comorbid conditions that were not listed in the diagnostic database and 85 patients admitted within 24 h of the initial diagnosis were also excluded. Exclusions were also made for missing charts (n = 35) and for lack of patient follow-up beyond the first visit (n = 4). A total of 403 patients remained for analysis.
Comorbid conditions, medication use, and diagnostic test results were extracted from chart review. Coronary artery disease (CAD) was defined as a documented history of prior MI, percutaneous coronary intervention, coronary artery bypass surgery, or a cardiac catheterization showing a ≥70% lesion in any major artery. A medication was considered prescribed if the provider listed the medication as a drug the patient was taking or it was a drug that the patient was refilling at the Kaiser pharmacy. Medication use was evaluated within six months after the diagnosis of HF. Laboratory results were determined through a computerized laboratory database. Deaths were identified from the chart review and by interrogating the state of California death registry. All hospitalizations within the Kaiser system were reviewed. Records of outside hospitalizations, which account for approximately 10% of all hospital admissions for KPMCP members, were available only if documented in the chart.
Before any data analysis, patients were categorized as having cardiology care (CC) on the basis of the number of visits to a cardiologist or the proportion of those visits to the total number of visits to medicine providers using these criteria:
1. More than two office visits to a cardiologist during the follow-up period or before a first major end point (death or cardiovascular hospitalization) if one occurred.
2. For patients with only two visits to a cardiologist, these constituted at least 25% of the total outpatient visits to medicine providers during the follow-up period or before a first major end point.
3. For patients with only one visit to a cardiologist, this visit constituted at least 50% of the total outpatient visits to medicine providers during the follow-up period or before a first major end point.
Patients who did not meet these definitions for CC were classified as having PC if they had at least two visits to internists or family practitioners. Visits to other providers such as nurse practitioners were not counted toward the PC visits.
The prospectively defined primary end point was the composite of death from any cause or a hospitalization for a primary cardiovascular diagnosis, analyzed by time to first event using proportional hazards models, with adjustment for patient characteristics that differed between the provider groups and that showed evidence of univariate prognostic significance. Additional prespecified analyses were all-cause mortality and adherence to HF guidelines developed by the Agency for Health Care Policy and Research (16)and the American College of Cardiology and American Heart Association (17). Guideline recommendations that were evaluated were assessment of left ventricular function and use of ACE inhibitors in patients with ejection fractions (EFs) ≤45% within six months of the diagnosis of HF. Beta-blocker use was examined but was not considered an end point because there were no guideline recommendations for beta-blocker use during the time period evaluated. We also compared the use of lipid-lowering agents and antithrombotic agents in patients with associated coronary or peripheral arterial disease and the use of anticoagulation in patients in atrial fibrillation (AF). Because patient education was not routinely documented in the chart, it was not evaluated as a guideline adherence measure.
All continuous data are expressed as mean value ± SD. Comparisons between care groups were performed using ttests for continuous variables and chi-square analyses for categorical variables. For all comparisons, p < 0.05 was considered statistically significant.
The primary outcome of time to death or first cardiovascular hospitalization was examined by Cox proportional hazards modeling. Age, gender, and the presence of systolic dysfunction (defined as a quantitative EF measurement ≤45% or a qualitative interpretation of “mild,” “moderate,” or “severe” reduction) were prespecified for the proportional hazards analysis. Other variables that were potential univariate predictors of outcome (p < 0.16) were entered stepwise into these models. Statistical analysis was performed with an SAS (Cary, North Carolina) database statistical software package, version 6.12.
The clinical characteristics of patients classified as CC (n = 198) and PC (n = 205) are shown in Table 1. The CC patients were somewhat younger and included a higher proportion of men. They were also more likely to have definite CAD, a history of documented MI, and AF. The CC group had more patients with confirmed systolic dysfunction (63% vs. 32%, p = 0.0012). Comorbid conditions, including diabetes, chronic obstructive pulmonary disease, and peripheral arterial disease, were similarly distributed. The degree of hyponatremia, anemia, and renal insufficiency based on serum laboratory tests within 90 days of the HF diagnosis were not significantly different between the two groups (Table 1).
The recommendation that left ventricular function be assessed in patients with HF was considered fulfilled if an assessment of EF was performed within 24 months before the HF diagnosis or at anytime during the follow-up period. In the CC group, 94% of patients had at least one measurement of EF in the specified time period, whereas only 74% of the PC patients had an EF measurement (p = 0.001).
As shown in Table 2, patients in the CC group were more likely than the PC group to receive ACE inhibitors, beta-blockers, and digoxin within six months of the HF diagnosis. Primary care physicians were more likely to use diuretics (93% vs. 86%, p = 0.26). The use of medications (within six months of the HF diagnosis) for specific comorbid conditions is shown in Table 3. In patients with low EF (≤45%), the CC group had a higher use of ACE inhibitors and beta-blockers within six months of the diagnosis. The use of lipid-modifying therapy (hydroxymethylglutaryl coenzyme A reductase inhibitors, fibric acid derivatives, or nicotinic acid) was 23% in the CC group versus 15% in the PC group (p = 0.027). Specifically, in patients with coronary, cerebrovascular, or peripheral arterial disease, lipid-modifying therapy was higher in the CC group. The use of aspirin was similar between the CC and PC groups in this subset of patients. In patients with AF, warfarin use was higher in the CC group. Finally, patients managed by cardiologists were more likely to have an investigation for ischemia or coronary disease by noninvasive stress testing or cardiac catheterization as a potential cause of their HF (50% vs. 30%, p = 0.001) compared with those managed by PC.
The numbers of primary events (deaths and cardiovascular hospitalizations) are shown in Table 4. There was no significant difference between the numbers of either deaths or hospitalizations. However, when proportional hazards models were constructed that adjusted for the patient characteristics that differed between the groups and other variables that had prognostic significance, CC was associated with a reduction in the primary outcome of time to death or cardiovascular hospitalization (hazard ratio (HR) 0.65, 95% confidence interval [CI] 0.45 to 0.96, p = 0.03). Ejection fraction ≤45% (HR 1.77, 95% CI 1.2 to 2.6, p = 0.0052) was the most significant predictor and the only other independent predictor of outcome in the multivariate analysis (Table 5). There was a significant interaction between specialty care and low ejection fraction (p = 0.01) suggesting that the beneficial effect of cardiology care was driven by a reduction of events in the group of patients with low ejection fraction. This study was not powered to look at the effect of CC on cardiovascular hospitalizations and death separately, but when analyzed in multivariate analysis there is a trend toward reduced cardiovascular hospitalizations (HR 0.66, 95% CI 0.4 to 1.0, p = 0.05), but not deaths in the CC group.
This study demonstrates several important differences between the characteristics, management, and outcomes of outpatients with new-onset HF treated by cardiologists and those treated by PC physicians. First, patients managed by cardiologists differ significantly from those managed by PC physicians; specifically, they tend to be younger, more often men, and to have more severe cardiac dysfunction and more associated cardiac diseases. The second finding is that guideline recommendations were more frequently fulfilled in the CC patients. Third, and most notably, when the results are adjusted for the significant differences between patients managed by the two groups of providers, CC was an independent predictor of a reduced hazard for the composite outcome of death or hospital admission for a cardiovascular diagnosis.
Relationship of present findings to prior studies
It is expected that patients receiving CC have more severe or more complex heart disease, but the differences in gender and age may reflect referral bias or a lower propensity to seek out specialty care among women and older patients. It should be noted that patients with dementia, terminal illnesses, and nursing home residence were specifically excluded from this cohort in order to eliminate patients in whom guideline recommendations may not have the same priority and in whom the outpatient physician may have less control over patient management. Because these patients are less likely to receive specialty care, these exclusions also serve to limit the differences between the two provider groups. Nonetheless, the differences observed in the present study are consistent with prior studies in patients with HF and coronary disease (10,12–14,18,19).
Prior studies have also shown that cardiologists adhere more closely to guidelines than PC physicians, both in HF and coronary disease patients (4,5,20–22). In the present study, more CC patients had measurements of left ventricular function and more had assessments for myocardial ischemia. A higher proportion of CC patients were treated with ACE inhibitors, and this was specifically the case in those with documented low EF. Although at the time these patients developed HF, the substantial benefit of beta-blockers in reducing mortality and morbidity in HF patients was only becoming apparent and the use of beta-blockers was not widely recommended, it is noteworthy that a higher proportion of CC patients were receiving a beta-blocker, despite their lower EFs. Finally, cardiologists were more likely to use recommended therapies for related coexisting conditions, such as aspirin and lipid-modifying agents for CAD, peripheral arterial, and cerebrovascular disease, and warfarin for AF. It should be noted that although guideline-recommended treatment rates by cardiologists were relatively higher in this study, the absolute rates were still not ideal. Other models of care such as HF case management programs may have resulted in even higher rates of guideline adherence.
In the case of post-MI patients, improved guideline adherence has been associated with improved outcomes in-hospital and during one-year follow-up in several studies (9,10). In one such study, multivariate analyses suggested that the greater use of recommended medications and of early reperfusion strategies explained some, but not all, of this improvement in outcome (22).
Improved outcomes secondary to specialty care have not necessarily been the case with HF patients (5,7,14,19). A number of methodologic challenges may explain the variability in the previously reported HF study results, which have generally involved hospitalized patients. In the inpatient setting, specialty consultation and collaborative treatment are common and are often undocumented (“curbside consultations”) and, therefore, it is often difficult or even impossible to determine which providers are most responsible for a patient’s management. A focus on hospitalized patients may be more appropriate for acute conditions such as MI or unstable angina, but provides a very circumscribed window into the management and outcomes of patients with illnesses such as chronic HF, where mortality during a given admission contributes relatively little to the long-term outcome and where the management and treating physicians often differ markedly from the in-patient to the postdischarge period.
There is very little information about differences in outcomes among HF outpatients managed by cardiologists and by PC providers, and most reports have focused on transplant programs and specialized HF clinics. The present study was designed to minimize several of these confounding factors. Using an HMO population allowed ascertainment of the providers and accurate definition of the provider groups. In addition, the patients in the CC and PC groups were more similar than in most studies, most likely because they were derived from the same HMO pool and access to specialty care was not restricted. Some conditions often differentially distributed between patients followed by cardiology and PC physicians, such as dementia and certain severe comorbidities, were excluded in advance. The focus on patients with new-onset HF was a unique feature of the study. This approach was selected because the identity of the physicians managing the patients’ HF could be ascertained, whereas in patients with chronic HF, especially those who have been hospitalized in the past, it is less clear who may have evaluated the patients and initiated medications. Furthermore, the initial management of HF patients may play a critical role in their subsequent natural history and prognosis, so that the impact of interventions such as treatment with angiotensin-converting inhibitors and beta-blockers and differences in the management of coronary disease may have greater impact on this population.
This was a prospectively designed study, but the cohorts were identified and the data were gathered retrospectively. As noted earlier, some patients were excluded to limit the differences between CC and PC patients, but significant differences remained, including older age and more women in the PC group and more systolic dysfunction, coronary disease, and AF in the CC group. Although adjustments were performed in the multivariable analyses, there may have been other unmeasured confounders, which could account for differences in outcomes. In addition, differences in medication use were determined from crude utilization rates based on chart review and did not account for evidence of prior drug intolerances or contraindications.
As is the case in clinical practice, many of the patients in this cohort were seen by both PC providers and cardiologists. In the PC group, 18% of patients saw a cardiologist at least once during the follow-up period. The categorization of such patients into CC or PC groups is of necessity somewhat arbitrary. The criteria employed in this study were designed to assure at least moderate participation by a cardiologist in the CC group, but at the same time to allow patients to be classified as having CC even if they had only a small number of visits to cardiologists, as requiring a larger number or percent of visits would have resulted in preferential assignment to PC for patients who had early outcomes. Importantly, the criteria for this classification were developed before any of the analyses were performed.
Lastly, this study involved new-onset HF patients in a community outpatient cohort and, therefore, the results may not be generalizable to other practice settings or patients later in their disease course.
Conclusions and implications
This study suggests that participation by cardiologists in the care of patients with recent-onset HF managed in the outpatient setting is associated with reduced mortality and morbidity, as demonstrated by a significant reduction in the composite of death and cardiovascular hospitalization. This is likely to be due at least in part to better adherence to guideline recommendations, particularly the greater use of medications such as ACE inhibitors and beta-blockers, which have been shown to reduce these adverse outcomes. Additional research is required to determine whether these results can be extended to chronic HF in the ambulatory care setting or to the postdischarge phase of patients hospitalized with HF.
We thank Ms. Irene Tekawa for her assistance in programming support.
☆ This work was supported by a grant-in-aid from the California Affiliate of the American Heart Association (Dr. Alexander, Principal Investigator; Dr. Massie, Co-Investigator), by a Health Services Research and Development Grant from the Department of Veterans Affairs (Dr. Massie, Principal Investigator; Dr. Ansari, Co-Investigator), and by American College of Cardiology-Merck and NRSA awards (Dr. Ansari).
- angiotensin-converting enzyme
- atrial fibrillation
- coronary artery disease
- cardiology care
- confidence interval
- ejection fraction
- heart failure
- health maintenance organization
- hazard ratio
- International Classification of Disease
- Kaiser Permanente Medical Care Program
- myocardial infarction
- primary care
- Received May 28, 2002.
- Revision received July 1, 2002.
- Accepted July 18, 2002.
- American College of Cardiology Foundation
- American Heart Association
- Edep M.E.,
- Shah N.B.,
- Tateo I.M.,
- Massie B.M.
- Fonarow G.C.,
- Stevenson L.W.,
- Walden J.A.,
- et al.
- Nash I.S.,
- Nash D.B.,
- Fuster V.
- Casale P.N.,
- Jones J.L.,
- Wolf F.E.,
- Pei Y.,
- Eby L.M.
- Reis S.E.,
- Holubkov R.,
- Edmundowicz D.,
- et al.
- Konstam M.D.K.,
- Baker D.
- ↵Guidelines for the Evaluation and Management of Heart Failure. Report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on Evaluation and Management of Heart Failure). J Am Coll Cardiol 1995;26:1376–98
- Auerbach A.D.,
- Hamel M.B.,
- Davis R.B.,
- et al.