Heart failure (HF) is a leading cause of morbidity and cardiovascular mortality¹ and is a major contributor to increasing healthcare costs,² with the majority of costs attributable to emergent care and hospitalizations.³ In the United States alone, there are >1 million emergency department (ED) visits for HF per year, and the number of such visits has been increasing yearly.⁴ HF will likely have an even greater burden on health care in the future because the prevalence of disease heightens with age.⁵

Editorial see p 1782

Clinical Perspective on p 1814

HF patients are frequently in contact with acute care hospitals, and the point of first contact is often the ED. HF patients who are admitted to hospital from the ED receive rapid medical care; however, nonhospitalized, ambulatory patients may experience varying physician access and care. Although patients who received concurrent care by both a cardiologist and a primary care (PC) physician had the lowest mortality rates after myocardial infarction compared with care by either physician alone,⁶ the impact of shared care in HF patients is a source of debate. Some studies suggest that shared care entails a tradeoff between lower mortality and higher rates of hospitalization,⁷ whereas others found no impact of specialist care.⁸

Often, patients who are discharged home from the ED must navigate their own way through the healthcare system. Even in a universal healthcare environment, there may be variability in the care received by HF patients in the ambulatory, postemergent setting. The pattern of care received is decided largely by the PC provider, who is the gatekeeper of referrals to specialist care. Some HF patients may receive care by a PC physician alone or collaborative care with a cardiologist⁹; others may receive no physician follow-up early after ED discharge. Although outpatient care early after hospital discharge may have a profound effect on outcomes,¹⁰ there is a paucity of data on HF patients who transition to ambulatory care.

We conducted a population-based study of HF patients who visited and were discharged from an ED, examining the type of physician care received within 30 days after discharge. We hypothesized that patients who received early collaborative care by both a cardiologist and a PC physician would have improved outcomes and care.

Methods

Patients

We examined patients who visited any ED for HF in Ontario, Canada, between April 1, 2004, and March 31, 2007. When patients visited an ED multiple times during the study period, the first visit was deemed the inception date. Patients ≥18 years of age were eligible for this study if they were discharged to their residence from the ED without hospital admission and were Ontario residents with a valid health card number.

Data Sources

We used the National Ambulatory Care Reporting System database, which contains information on all ED visits. HF patients were identified by main discharge diagnosis I50 using the International Classification of Diseases coding system. Hospitalizations were identified using the Canadian Institute for Health Information Discharge Abstract Database, and comorbidities were determined by examining all diagnosis codes within 3 years before the index ED visit. Deaths were determined from the Registered Persons Database.

We used the Ontario Health Insurance Plan physician billing and Canadian Institute for Health Information procedure databases (using Canadian Classification of Interventions codes; see Appendix I in the online-only Data Supplement) to determine whether physician visits occurred or diagnostic/interventional cardiac procedures were performed within 100 days after ED discharge. Physician visits and procedures are billed on the date of service, with random audits conducted by the single payer of health services (Ministry of Health and Long-Term Care). Performance of the following procedures was examined: (1) left ventricular ejection fraction testing by echocardiography or radionuclide angiography; (2) stress/exercise testing or nuclear myocardial perfusion imaging; (3) cardiac catheterization; (4) coronary revascularization, including percutaneous coronary intervention or coronary artery bypass graft surgery; and (5) device implantation, including implantable cardioverter-defibrillator or permanent pacemaker. Medication use within 100 days after ED discharge was determined in patients ≥65 years of age using the Ontario Drug Benefit database, which contains information on all prescription medications under the provincial pharmacare program. We determined whether prescriptions were filled for angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, β-adrenoceptor antagonists, digoxin, diuretics (furosemide/loop, thiazide, metolazone), spironolactone, warfarin, or amiodarone.

Physician Type

Physician specialty was determined from the Ontario Physician Human Resource Data Centre database.¹¹ Physician specialty was defined by the “functional specialty” determined, via telephone survey, by the type of care constituting the majority of the physician's practice. Therefore, internal medicine specialists who provided primarily cardiology care were also considered cardiac specialists. Patients who visited a PC physician alone were compared with those who received collaborative care by PC and cardiac specialist or no physician visit (NoMD) within 30 days after ED discharge. Although there is no time specified for a physician visit after ED discharge,¹² a 30-day post-hospital discharge visit is a clinical benchmark.¹³,¹⁴

Outcomes

The coprimary outcomes were death and composite of death and hospitalizations or repeat ED visits for HF within 1-year follow-up. The composite outcome was selected because of the importance of repeated hospital visits on prognosis¹⁵ and healthcare burden. Secondary outcomes were the composite of death and all-cause hospitalizations or ED visits at 1 year.

Statistical Analysis

Continuous and categorical variables were compared across models of physician care with ANOVA and the χ² test, respectively. We intended to correlate ambulatory treatments and diagnostic investigations occurring within the first 100 days after ED discharge in an outpatient setting, with similar minimum 100-day observation periods for each patient; therefore, we excluded patients who died, were admitted to hospital, or visited an ED for HF during this period. We conducted a propensity-matched analysis to determine the effect of physician care on outcomes by estimating propensity scores for collaborative versus PC, NoMD versus PC, and other 2-way comparisons.¹⁶ The variables used to derive the propensity score included the following covariates based on published HF-specific mortality¹⁷ and morbidity¹⁸ risk-adjustment models: age, sex, myocardial infarction, prior HF hospitalization, coronary artery bypass graft surgery, percutaneous coronary intervention procedure, implantable cardioverter-defibrillator, permanent pacemaker, diabetes mellitus, hypertension, cancer, unstable angina, cardiopulmonary-respiratory failure and shock, pneumonia, protein calorie malnutrition, trauma, major psychiatric disorders, arrhythmias, decubitus/chronic skin ulcer, anemia, and peripheral vascular disease, cerebrovascular disease, respiratory disease, rheumatologic disease, peptic ulcer disease, neurological disease, renal disease, atherosclerotic disease, valvular disease, other heart disease, gastrointestinal disease, or severe hematologic disease. Patients were matched by use of a greedy matching algorithm on the logit of the propensity score with a caliper width of 0.2 of the SD of the logit of the propensity score. Balances in baseline covariates were evaluated by calculating standardized differences.

Time to event analysis for the propensity matched samples was performed with a Cox proportional-hazards model that stratified on matched pairs. All patients were followed for 1 year after ED discharge, with the last follow-up date of March 31, 2008. Unadjusted Kaplan–Meier and adjusted survival curves were plotted.¹⁹ Statistical significance was indicated by a 2-tailed value of P<0.05. All analyses were performed with SAS version 9.1.3 (SAS Institute Inc, Cary, NC).

Results

Study Cohort

After identification of unique ED visits (Figure 1), 10 599 patients in the following comparison groups were identified: collaborative care (n=1478), PC (n=6596), cardiology (n=535), or NoMD (n=1990). Validation studies compared with clinical Framingham criteria have found high predictive value of HF diagnosis codes.²⁰

• Download figure
• Download PowerPoint

Physician Specialty

In total, 9173 PC and 548 cardiac specialists were evaluated. To validate physician specialty, we classified each diagnostic code billed during cardiac specialist or PC visits as cardiovascular or noncardiovascular (Appendix II in the online-only Data Supplement). Of 10 122 450 fee codes billed by cardiac specialists, 9 219 571 (91.1%) were cardiovascular. In contrast, of 227 843 789 fee codes billed by PCs, only 29 100 210 (12.8%) were related to cardiovascular disease. Median time to first physician visit was 5 days (25th and 75th percentiles, 2 and 12 days) for PC, 3 days (25th and 75th percentiles, 1 and 7 days) for collaborative care, and 9 days (25th and 75th percentiles, 4 and 18 days) for cardiology alone. Median time to cardiac specialist visit for collaborative care was 12 days (25th and 75th percentiles, 5 and 20 days); 1072 patients (72.5%) visited a PC physician before the cardiac specialist.

Patient Characteristics

Patients receiving collaborative care had higher burden of coronary heart, arrhythmic, and valvular disease than other care groups (Table 1). There were no systematic differences in noncardiac comorbidities between physician types. In the 6 months before the ED visit, 94.4% (PC), 95.9% (collaborative), 79.8% (cardiology), and 61.8% (NoMD) of patients visited a PC physician; 96.8% (PC), 98.2% (collaborative), 95.0% (cardiology), and 78.2% (NoMD) were previously seen by any physician.

Outcomes

Survival curves for death are shown in Figure 2 and composite outcomes are shown in Figures 3 and 4 for 30-day event-free survivors. Adjusted survival curves are shown in Figures I through III in the online-only Data Supplement. Compared with PC (1-year mortality, 10.4%), crude death rates were significantly reduced with collaborative care (7.2%; P<0.001 versus PC), nonsignificantly lower with cardiology care (9.5%; P=0.526 versus PC), and higher for NoMD care (15.5%; P<0.001 versus PC). The combined 1-year rates of death, ED visits, or hospitalizations for HF were 18.7% (collaborative; P=0.004 versus PC), 18.9% (cardiology; P=0.077 versus PC), 27.3% (NoMD; P<0.001 versus PC), and 22.2% (PC). All-cause ED visits or hospitalizations and death were also lower with collaborative (57.5%; P<0.001) and cardiology (57.2%; P=0.007) care and higher with NoMD (66.2%; P=0.007) compared with PC (63.1%) at 1 year.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Propensity-Matched Analyses Compared With Collaborative Care

In a 2:1 propensity-matched analysis, we compared 2871 PC (mean age, 71.2±12.2 years) with 1477 collaborative care (71.7±11.1 years) patients. After propensity matching, standardized differences were ≤0.10, suggesting balance in covariate distribution between groups (Figure IVand Table I in the online-only Data Supplement). Collaborative care was associated with significant mortality reduction, with a hazard ratio of 0.79 (95% confidence interval [CI], 0.63 to 1.00; P=0.045), and all-cause hospitalization, ED visit, or death, with a hazard ratio of 0.86 (95% CI, 0.78 to 0.94) compared with PC (P=0.001). There was a trend toward a reduction in the composite of hospitalization or ED visit for HF or death, with a hazard ratio of 0.87 (95% CI, 0.75 to 1.01; P=0.06). In an exploratory propensity-matched subanalysis, we compared 1015 patients receiving collaborative with 526 receiving cardiologist care (Table II in the online-only Data Supplement). There was a trend to increased mortality with cardiologist care alone, with a hazard ratio of 1.41 (95% CI, 0.98 to 2.03; P=0.067). However, there were no significant differences in the combined end point of death or repeat ED visit or hospitalization for HF (hazard ratio, 1.06; 95% CI, 0.82 to 1.36).

• Download figure
• Download PowerPoint

Propensity-Matched Analysis of Primary Versus NoMD Care

In a 2:1 propensity-matched analysis comparing 3807 PC (age, 77.0±11.5 years) with 1979 NoMD (77.2±12.8 years) patients, baseline covariates were balanced between groups (Figure IV and Table III in the online-only Data Supplement). Compared with the NoMD group, patients receiving PC had reduced mortality risk, with a hazard ratio of 0.75 (95% CI, 0.64 to 0.87; P<0.001). Similarly, PC was associated with reduction in the composite outcome of HF hospitalization, ED visit, or death, with a hazard ratio of 0.88 (95% CI, 0.79 to 0.98; P=0.024). PC was not associated with a reduction in the composite of all-cause hospitalization, ED visit, or death, with a hazard ratio of 0.97 (95% CI, 0.89 to 1.04; P=0.367).

Diagnostic and Revascularization Procedures

Patients who received collaborative care had the highest rates of left ventricular ejection fraction evaluation, noninvasive testing for ischemia detection, and cardiac catheterization (Table 2). Those with NoMD care had the lowest rates of all diagnostic tests and procedures. Analysis of the timing of diagnostic and therapeutic interventions demonstrated that the majority of cardiac diagnostic tests and therapeutic interventions were performed after the cardiac specialist visit (Figure 5).

Medication Use

The highest rates of diuretic use were observed in those who received collaborative care, suggesting that these patients had more symptomatic or severe HF (Table 3). Use of evidence-based therapies was also highest with collaborative care, including those with high-risk characteristics such as diabetes mellitus, ischemic heart disease, arrhythmias, and renal dysfunction (Table 3). Use of nonsteroidal antiinflammatory drugs and cyclooxygenase-2 inhibitors was lower with collaborative care. Among those receiving collaborative care, most medication changes, specifically initiation of new drugs or dose increases, were performed after a PC visit (Figure 6 and Figure V in the online-only Data Supplement).

• Download figure
• Download PowerPoint

Sensitivity Analyses

An analysis that included patients who had events within the first 31 to 100 days after ED discharge was similar to the primary results, although inclusion of these patients yielded an older patient cohort. Analysis of 1642 collaborative care patients (age, 77.5±12.6 years) matched to 3210 PC patients (77.3±11.3 years) demonstrated improved outcomes with collaborative care, with hazard ratios of 0.84 (95% CI, 0.70 to 1.01; P=0.057) for death and 0.88 (95% CI, 0.78 to 0.99; P=0.039) for the composite of death and hospitalizations or ED visits for HF. In a propensity-matched analysis of 4466 PC (age, 77.3±11.3 years) versus 2321 NoMD (77.5±12.6 years) patients, NoMD care was associated with increased risk, with hazard ratios of 1.35 (95% CI, 1.20 to 1.52; P<0.001) for death and 1.09 (95% CI, 0.99 to 1.19; P=0.068) for the composite of death and hospitalizations or ED visits for HF. Results were also similar when socioeconomic status (eg, income quintile) and rural versus urban residence were included in the propensity-matched models (Table IV in the online-only Data Supplement).

Discussion

Patients who are discharged home from the ED are vulnerable to recurrent morbid events and death but are often left to navigate the healthcare system with variable degrees of medical guidance. In this study of patients discharged home from the ED without hospital admission, we examined those who received early collaborative care provided by a PC physician and a cardiac specialist. We found that patients who received collaborative care within 30 days of discharge had lower rates of death, recurrent emergency visits, and hospitalizations. A sizable proportion of patients did not visit any physician within 30 days after ED discharge, and they experienced greater death and higher risk of the composite of HF-specific outcomes and mortality. Patients who received early collaborative care were more likely to undergo important diagnostic tests, including cardiac functional assessment and evaluation for myocardial ischemia, were more likely to undergo coronary revascularization, and received more evidence-based drug therapies than PC patients alone.

Analyses similar to ours have not been conducted in many other jurisdictions.²¹ Many prior studies have not examined concurrent or shared care in the ambulatory setting but instead have focused on care provided by cardiologists versus generalists.²² In 1 prior study, comanagement by a cardiologist improved care quality among those recently discharged from hospital, whereas mailed treatment recommendations for HF and myocardial infarction had no effect.²³ We found that the majority of patients received HF care provided by their PC physician without concurrent cardiologist care, which is concerning given the increasing complexity of management and the frequent need for cardiac diagnostic or therapeutic tests. Many patients did not visit any physician early after ED discharge, which may attest to the lack of understanding by both physicians and patients of the importance of timely discharge transitional care. The gaps in care experienced by patients without an early physician visit are magnified because they are not afforded the benefit of rapid care that is provided to those who are hospitalized.²⁴

Some studies have found contrasting results. In a study of patients discharged after hospitalization, shared care was found to increase hospitalizations but decrease mortality compared with PC.⁷ Conversely, in a community-based study of 10 hospitals, cardiology care decreased hospitalizations but did not affect mortality among HF patients discharged after hospitalization.²⁵ Our study provides new insights because we examined a wide range of important outcomes, including ED visits and hospitalizations.¹⁵ Second, we examined patients with acute HF who were not admitted to hospital, were considered well enough to be discharged, and were not as ill as the hospitalized patients who have been the subject of prior studies. Furthermore, the study cohort was comprised of those who survived the initial month(s) after discharge, when increased risk of adverse events occurs,^26,27 and thus could be considered to represent a milder spectrum of HF than hospitalized patients.

The improved outcomes were likely mediated in part by improved processes of care resulting from knowledge of the importance of pathophysiology, targeted diagnostic testing, and pharmacotherapy for HF among cardiac specialists.²⁸ Collaborative care resulted in higher rates of cardiac diagnostic testing that guides treatment decisions (eg, cardiac functional assessment, ischemia evaluation) and interventions. Collaborative care resulted in increased use of medications that are beneficial in subgroups with coronary heart disease, diabetes mellitus, and arrhythmias, regardless of the underlying left ventricular systolic function. Medication use was also higher in those with increased mortality risk (eg, renal dysfunction), in whom greater potential benefits of treatment may be realized.²⁹ Patients receiving collaborative care also had higher rates of use of loop diuretics, metolazone, and amiodarone, which indicate increased baseline degree of sickness than other care groups. The synergistic contribution of PC to improved outcomes may have been related to timely physician visits, partnership with cardiologists in increasing HF medications to target doses, and provision of care for noncardiac comorbidities.³⁰

Our study has several implications. First, patients discharged from the ED are at risk of death and substantial morbidity, and they should be considered to have had a serious event that requires rapid follow-up. Second, after “dischargeable acute HF,” early physician involvement using a collaborative approach may lead to improved care quality and a substantial reduction in mortality and morbidity in the post-ED context. Finally, HF patients who are discharged from the ED should be provided with a follow-up option that includes cardiac specialist involvement in conjunction with PC or a comparable variant of care.

Our study had limitations that merit comment. We did not evaluate internal medicine generalists or other subspecialists as comparators. However, we did include internists in our analysis if they were functionally cardiac specialists on the basis of their practice patterns. We did not specifically examine the volume of HF patients treated because prior analyses suggested no volume-outcome relationship.³¹ Because we did not have clinical data at baseline, it is possible that there may have been other clinical factors that could not be adjusted for in our between-group analyses. For example, we did not have information on left ventricular function; however, many of the diagnostic investigations, therapeutic interventions, and drug therapy subgroups considered were relevant regardless of the underlying ejection fraction. We did not have information on drug doses, but prior studies suggest that higher doses were attained despite lower blood pressures³² and renal insufficiency³³ among those who received cardiology care. Finally, to create a comparison sample that could be examined for early physician visits and subsequent care, our primary analyses excluded events that occurred within 100 days to allow for equivalent durations of time for the ascertainment of changes in drug therapy, diagnostic tests, and therapeutic interventions that would occur after the physician visit. Although we conducted sensitivity analyses including these patients and found consistent results, examination of very early events and care (within the first 30 days) was beyond the scope of this study.

Conclusions

Among HF patients discharged from the ED, the risk of death and morbidity of recurrent hospital visits was reduced in those who received care within 30 days after discharge that was shared by a PC physician and a cardiac specialist. Among those who survived the first 30 days after ED discharge, early collaborative care resulted in improved use of diagnostic tests, interventions, and cardiovascular medications compared with PC alone, likely mediating an improvement of subsequent outcomes. Importantly, a large proportion of HF patients discharged after emergency care did not receive any physician follow-up, and these individuals were at the greatest risk of death and adverse events. Substantial opportunities to improve the care and outcomes of the large number of HF patients after ED discharge exist.

Sources of Funding

The Institute for Clinical Evaluative Sciences is supported in part by a grant from the Ontario Ministry of Health and Long Term Care. The opinions, results, and conclusions are those of the authors, and no endorsement by the Ontario Ministry of Health and Long Term Care or by the Institute for Clinical Evaluative Sciences is intended or should be inferred. This research was supported by operating grant MOP 86718 and a clinician-scientist award (Dr Lee) from the Canadian Institutes of Health Research, a Career Investigator Award from the Heart and Stroke Foundation of Ontario (Dr Austin), and a Canada Research Chair in Health Services Research (Dr Tu).

Disclosures

None.

Footnotes