Randomized, Controlled Evaluation of Short- and Long-Term Benefits of Heart Failure Disease Management Within a Diverse Provider Network
The SPAN-CHF Trial
Background— Several trials support the usefulness of disease management (DM) for improving clinical outcomes in heart failure (HF). Most of these studies are limited by small sample size; absence of concurrent, randomized controls; limited follow-up; restriction to urban academic centers; and low baseline use of effective medications.
Methods and Results— We performed a prospective, randomized assessment of the effectiveness of HF DM delivered for 90 days across a diverse provider network in a heterogeneous population of 200 patients with high baseline use of approved HF pharmacotherapy. During a 90-day follow-up, patients randomized to DM experienced fewer hospitalizations for HF [primary end point, 0.55±0.15 per patient-year alive versus 1.14±0.22 per patient-year alive in control subjects; relative risk (RR), 0.48, P=0.027]. Intervention patients experienced reductions in hospital days related to a primary diagnosis of HF (4.3±0.4 versus 7.8±0.6 days hospitalized per patient-year; RR, 0.54; P<0.001), cardiovascular hospitalizations (0.81±0.19 versus 1.43±0.24 per patient-year alive; RR, 0.57; P=0.043), and days in hospital per patient-year alive for cardiovascular cause (RR, 0.64; P<0.001). Intervention patients showed a trend toward reduced all-cause hospitalizations and total hospital days. On long-term (mean, 283 days) follow-up, there was substantial attrition of the 3-month gain in outcomes, with sustained significant reduction only in days in hospital for cardiac cause.
Conclusions— In a population with high background use of standard HF therapy, a DM intervention, uniformly delivered across varied clinical sites, produced significant short-term improvement in HF-related clinical outcomes. Longer-term benefit likely requires more active chronic intervention, even among patients who appear clinically stable.
Received November 18, 2002; de novo received February 15, 2004; revision received June 10, 2004; accepted June 17, 2004.
Heart failure (HF) is a major public health problem, affecting ≈5 million in the United States.1,2 It is the most frequent cause of hospitalization within the Medicare population,3 and its frequency is expected to rise in coming years. Mortality rates among patients hospitalized with HF are high.4 The cost of cardiac care in HF patients is growing as the population ages, with the predominant cost driver being hospitalization5; annual direct costs for care of HF patients are estimated to be between $20 and 40 billion.1,3,6–8
Treatments known to improve outcomes in HF are underused. In a recent analysis of Medicare patients, the median rate of initiation of ACE inhibitors in eligible patients across the 50 states was 69%.9
Disease management (DM) has been advocated to increase appropriate medication prescription, to improve compliance, and to reduce hospitalizations. Prior studies show reduced HF readmission rates with nurse-driven DM.10 However, most such studies have not used concurrent randomized control subjects, have been performed within a single center, and/or provided only short-term follow-up. Whether short-term interventions result in sustained improvement in outcomes is unknown. Furthermore, whether the benefits observed from these interventions result from prescription of indicated drugs or from improved compliance and monitoring is unclear.
We performed a prospective, randomized assessment of a 3-month nurse-driven HF DM program, Specialized Primary and Networked Care in HF (SPAN-CHF). We investigated the short- and long-term impacts of this intervention on HF hospital use in patients recently hospitalized for HF who were already well managed with appropriate medications. We explored whether benefits could be (1) realized across a broad provider network, (2) observed within a population in which the rate of prescription of standard treatments was already high, and (3) sustained for a prolonged period.
We enrolled patients at 6 diverse sites (including academic medical centers, community hospitals, and community cardiology practices) in Massachusetts and Rhode Island over 22 months. After giving consent, patients were enrolled during an index HF hospitalization or within 2 weeks of discharge. The investigational review board at each hospital approved the study.
We enrolled 200 patients hospitalized with a primary diagnosis of HF. Patients with HF resulting from ischemic heart disease, dilated cardiomyopathy, valvular heart disease (either surgically treated or deemed inoperable), or hypertensive heart disease were included. Patients were enrolled regardless of ejection fraction or NYHA class. Patients were required to receive primary care through a provider other than an SPAN-CHF investigator. Exclusion criteria included noncardiac debilitating illness such as active malignancy, severe liver disease, severe renal insufficiency (creatinine >3.0 mg/dL), dementia, or obstructive lung disease requiring hospitalization; angina at rest or as the principal cause of activity limitation; myocardial infarction or revascularization procedure during the index hospitalization or within the preceding 30 days; planned revascularization or valvular surgery; or restrictive myopathy, pericardial constriction, or hypertrophic cardiomyopathy.
After providing informed consent, patients were equally randomized to intervention or usual care. Randomization lists were generated independently for each hospital (in blocks of 4 patients), stratifying patients first by level of care needed. Class A patients exhibited clinical instability, including those with >1 hospitalization in the preceding 12 months, with active medication adjustment or clinical evidence of right HF; class B included those with only 1 hospitalization in the preceding 12 months and no class A characteristics.
Within 3 days of randomization, the nurse-manager, experienced in HF management, conducted a home visit, meeting with the patient and the respective spouse/partner/caregiver. Depending on patient knowledge and interest, this visit lasted 45 to 90 minutes and focused on dietary and medical compliance, daily weights, self-monitoring, and early reporting of changes in weight or clinical status. At this session, patients and family received a preprinted Patient and Family Handbook. This teaching tool was designed by the SPAN-CHF team and outlined the SPAN-CHF program. Included in the handbook were sections on HF (definition), medications, low-salt diet, importance of daily weight, and clinical signs and symptoms that should prompt a call to the SPAN-CHF nurse or the patient’s primary care physician (PCP). Also included in the handbook were phone numbers for the patient’s SPAN-CHF nurse and the 24-hour SPAN-CHF on-call nurse.
While in the home, the nurse focused on identifying any impediments to and reinforced the importance of compliance to the treatment plan. After the teaching was complete, the nurse performed a cardiovascular examination and a symptom assessment.
Depending on clinical status, the nurse-manager telephoned patients weekly or biweekly, focusing on identifying changes in clinical condition and education reinforcement. At each follow-up phone call, the information discussed during the home visit was reinforced.
Nurse-managers were available by telephone 24 h/d, 7 d/wk. Patients were instructed to report clinical status changes, including any change in weight >2 lb. Nurse-managers received 24-h/d support and weekly management teleconferences with HF physician specialists. Nurse-managers communicated frequently with PCPs, alerting them to changes in patient condition and suggesting regimen changes advised by the HF physician-specialists.
The intervention continued for 90 days after randomization. It was continued for patients with overt clinical instability (class A). After 90 days, stable patients received passive surveillance, during which the nurse-manager remained available for incoming calls but did not place regularly scheduled calls.
Data Collection and Follow-Up
Nonnurse study coordinators, blinded to treatment assignment, performed telephone follow-up in all patients at 3 and 12 months after enrollment to ascertain clinical events. Events were adjudicated by an investigator blinded to treatment group. Medical records were reviewed, and hospitalizations were categorized by primary cause as HF, non-HF but cardiovascular, or noncardiovascular.
End Points and Statistical Analysis
We tested the hypothesis that a uniform DM strategy exerts salutary effects on clinical outcomes in a broad population of HF across a spectrum of practice settings. Our primary end point was hospitalization for HF during the first 90 days after enrollment. Secondary end points were cardiac hospitalizations and all-cause hospitalizations, number of days hospitalized per patient-year of follow-up for HF, cardiac and all-cause hospitalizations, and hospitalizations and hospital days over up to 1 year of follow-up.
Data are presented as both numbers of events and as rates, numbers of events per year of follow-up (±SE). Risk ratios (RRs) were calculated, and statistical comparisons made between treatment groups used χ2 testing. The χ2 and Student t tests were used to examine relationships between baseline characteristics of the 2 groups. The relationship between patient age, the single variable related to treatment, and outcome was assessed with t tests (for events) and Spearman correlations (for days). Multiple logistic regression (for hospital admission), Cox regression (for time to hospital admission), and multiple linear regression (for time spent in hospital) were used to explore the effect that patient age had on relationships between treatment and outcome. Because this testing showed no influence on treatment effects, results are reported as RRs of events and durations per patient-year of observation. Values of P<0.05 were considered statistically significant. All tests were 2 sided and run with SPSS (version 10).
Table 1 lists baseline characteristics. Intervention and control populations were similar in all respects except for a slightly higher mean age in the control group. Because patient age was not related to any of the outcomes examined, it was not necessary to adjust our results for this difference. Forty-two percent of patients were women. Ischemic heart disease was the cause in 61%. Hypertension was present in 69%, and diabetes was present in 48%. Average ejection fraction was 31%, and ≈30% of patients had ejection fraction >40%. Ninety-seven percent were in NYHA class II or III. Baseline medication use was similar in both groups, with ACE inhibitors, angiotensin receptor blockers (ARBs), and β-blockers used frequently. Doses are provided in Table 2. At baseline, 92% were receiving either an ACE inhibitor or ARB. β-Blockers were prescribed in 57% of patients, with a trend toward higher use in the control group. Follow-up averaged 290 days (range, 48 to 366 days) in the intervention group and 277 days (range, 1 to 366 days) in control patients.
Outcomes During Active Intervention
Table 3 lists 90-day outcomes and RRs between the 2 groups. During the initial 90 days, 4 intervention patients (4.1%) and 5 control patients (4.9%) died. During the 90 days of intervention, there were significantly fewer hospitalizations for HF within the intervention group (0.55±0.15 per patient-year alive) than in the control group (1.14±0.22 per patient-year alive; RR =0.48; P=0.027; the Figure). Sixteen percent of intervention patients and 23% of control patients died or were hospitalized for HF (RR, 0.66; P=0.16). There was a significant reduction in hospital days related to HF (4.3±0.4 versus 7.8±0.6 days hospitalized per patient-year alive; RR, 0.54; P<0.001).
The effect of DM on hospitalizations and hospital days resulting from HF during the 90-day intervention period contributed to similar effects, although of lesser magnitude, on cardiovascular hospitalizations (Table 3), with intervention patients hospitalized 0.81±0.19 per patient-year alive and control patients hospitalized 1.43±0.24 per patient-year alive for cardiovascular causes (RR, 0.57; P=0.043). Twenty percent of intervention patients and 27% of control patients died or were hospitalized for cardiovascular cause during the 90-day intervention (RR, 0.72; P=0.21). Days in hospital per patient-year for a cardiovascular cause were substantially reduced in the DM group (RR, 0.64; P<0.001). The impact of the 90-day DM intervention on cardiovascular events and hospital days contributed to a nonsignificant reduction in all-cause hospitalization and hospital days per patient-year alive (RR, 0.92; P=0.34). However, these effects were not statistically significant. For the primary end point, when HF severity at the time of randomization is considered, there was no significant treatment-by-subgroup interaction (RR, 0.48 and 0.51 for intervention patients in class A and B, respectively).
Although patient age did differ between treatment groups (P=0.028), it was not related to any of the outcomes measured. As this predicts and further multivariate analysis showed, the relationships between treatment and outcome measures were not influenced by adjustment for age.
Outcomes During Long-Term Follow-Up
Patient follow-up continued after the 90-day intervention period for up to 1 year from time of randomization (Table 4). Eleven intervention patients (11.3%) and 14 control patients (13.6%) died. Most of the gain achieved during the 90-day intervention was lost during long-term follow-up when analyzed either by cause of hospitalization or by all hospitalizations. Kaplan-Meyer analysis documented that median survival without death or hospitalization was 282 days in the intervention group and 260 days in the control group (P=NS). There was a nonsignificant trend toward residual benefit in the number of hospitalizations for cardiac cause per patient-year alive (RR, 0.79; P=0.13). Ischemic chest pain and arrhythmias accounted for 90% of control and 87% of intervention non-HF cardiac hospitalizations. The intervention group exhibited a trend toward long-term benefits in hospitalization days for HF (RR, 0.87; P=0.07) and for a cardiovascular cause (RR, 0.88; P=0.049). However, there was no benefit in overall hospitalization days during long-term follow up.
HF is the most common discharge diagnosis within the Medicare population. Readmission is frequent, with 1 study demonstrating 44% readmissions during the first 6 months.11 The potential benefits of active clinical management and the underprescription of medications with proven benefit9 suggest a role for DM to improve outcomes for patients with HF. Recent investigations suggest that DM improves quality of care and reduces hospitalization rates.10,12 Most prior trials have ≥1 significant limitations, including small sample sizes; absence of concurrent, randomized controls; limited follow-up; restriction to the urban academic medical center; and low baseline use of medications with proven benefit.
We designed the present investigation as a prospective, randomized, controlled trial to investigate the impact of nurse-driven DM across a broad, varied provider network and postulated that a 3-month intervention would have a short-term benefit on hospital use for HF and that this effect would be sustained for up to 1 year. We found a decrease in hospital use for HF during active intervention. However, discontinuation of the active intervention in this population with a high baseline rate of appropriate medication usage resulted in loss of the initial benefit during long-term follow-up.
Our intervention was designed to enhance dietary and medication compliance and to facilitate communication between patients and caregivers. Nurse-managers acted as facilitators between patients and PCPs. HF physician-specialists acted as consultants, primarily to nurse-managers and occasionally to PCPs and cardiologists, when so requested. HF specialists did not routinely see the patients.
Primary End-Point Analysis
Our intervention was associated with a statistically significant 52% reduction in the number of HF hospitalizations per patient-year at 90 days. This salutary effect was not maintained on long-term follow-up.
Secondary End-Point Analysis
At 90 days, HF DM was associated with significant reductions in cardiac hospitalizations and a trend toward reduction in all-cause hospitalizations. Similar trends were present in the percentage of patients admitted for cardiac and all causes over the first 90 days after enrollment. Similar to HF hospitalizations, there was a long-term loss of effect on cardiac and all-cause hospitalizations.
The intervention led to significant reductions in days hospitalized because of HF and cardiac causes, with a trend toward a reduction in all-cause hospitalizations. Again, the impact was attenuated on long-term follow-up. The impact on hospital days resulting from cardiac causes was nominally statistically significant, and there was a trend toward reduced hospital days related to HF, but there was no effect on all-cause days hospitalized.
In contrast to other patients, within this well-treated population, there was less opportunity to improve the maintenance medical regimen. Therefore, benefits observed during the first 90 days are unlikely to have stemmed from improved natural history of disease but rather from aggressive fluid management and providing an alternative to emergency department care. The attrition in early beneficial outcomes on long-term follow-up suggests that longer-term benefit in such a population requires more active chronic intervention, even among patients who appear clinically stable.
Comparison With Other Trials
Rich and colleagues13 were the first to provide definitive evidence for the effectiveness of DM in improving clinical outcomes in HF patients. In their single-center study of high-risk patients, they reported a reduction in rehospitalization 90 days after discharge in the DM group. As was seen in our 90-day results, the principal effect of their intervention was to reduce HF hospitalizations. Unlike ours, the population studied by Rich et al was composed largely of elderly, predominantly black women with systolic hypertension and a low baseline rate of therapy with ACE inhibitors and β-blockers. Improved prescription rates for appropriate medications, particularly improved blood pressure treatment, in the intervention group may have altered the natural history in these patients, resulting in the reported reductions in HF hospitalizations on long-term follow-up.
Kasper et al14 tested HF DM at 2 academic medical centers in a population limited to patients judged to be at increased risk of readmission after HF hospitalization. There was a trend toward a reduction in death or HF hospitalization over 6 months in the group randomized to HF DM. That trial, unlike ours, used a complex algorithm to manage patients and provided meals and transportation for medical services to patients who had a financial need. Stewart et al15 performed a randomized trial of multidisciplinary HF DM in 200 patients at a tertiary care center in Australia. Intervention patients experienced a 40% reduction in the primary end point—death or unplanned readmission over 6 months—but had more elective days in hospital. In contrast to the present study, repeated home visits were permitted, and all patients randomized to DM were mandated treatment by a cardiologist. In a single-center DM trial limited to elderly patients, Cline et al16 used an intervention stressing education, medication compliance, and diuretic self-adjustment. The intervention group in this study showed an improvement in time to readmission with trends toward improvement in other parameters (hospitalizations per patient, percent population readmitted, days hospitalized). Like our trial, this study documented attenuation of early clinical benefit during longer-term follow-up. Unlike ours, this study was limited to elderly patients, enrollees in the DM group were followed up in a hospital clinic, and nurse-managers were not available 24 h/d.
Krumholz and coworkers,17 in a single-center randomized study, tested a 1-year educational intervention without medical management in 88 patients with HF. In intervention patients, there was a significant reduction in the primary end point of death or hospital readmission. This study confirmed results of an earlier trial18 that showed that a brief educational intervention, despite a lack of effect on hospitalizations, led to a significant improvement in self-care behavior. In contrast to our study, the intervention group reported by Krumholz et al showed increasing benefit for intervention patients on long-term follow-up.
Stewart and Horowitz19 recently analyzed a group of 297 patients, the vast majority admitted for HF, pooled from 2 separate trials assessing the impact of HF DM in patients ≥55 years old with compromised LV systolic function from a single center in Australia. Over ≈4 years, they documented significant reductions in the composite of unplanned readmission or death and improved total and event-free survival in DM patients.19
In contrast to these other trials,13–19 we enrolled patients across a broad network of providers that ranged from academic medical centers to community practice sites. Furthermore, we did not limit our intervention to patients deemed to be at high risk for readmission on the basis of clinical characteristics other than a recent hospitalization for HF, which itself is associated with ≈40% 6-month rehospitalization rate.11 Our findings over the first 90 days demonstrate that a DM protocol may be applied successfully to improve short-term clinical outcomes across diverse clinical environments.
The active treatment group underwent intensive intervention for the initial 90 days after randomization. If clinically stable, patients were passively followed up. It appears likely that there are 2 mechanisms by which a DM program may influence clinical outcomes. The first is altering the natural history of disease by improving the rate of prescription of medications that have demonstrated benefit in randomized clinical HF trials. When such a mechanism is operative, it is likely that improved clinical outcomes may be observed beyond the time of active intervention. In the present study, enrolled patients had a high background use of ACE inhibitors, ARBs, and β-blockers. Consequently, in contrast to prior trials in which patients had relatively low baseline prescription rates for medications of proven benefit in HF,13–19 our DM program did not have a substantial opportunity to modify the natural history of the underlying disease process on the basis of increased use of evidence-based therapies.
In the absence of such an effect, as when appropriate medications are already being prescribed at a high rate, the impact of the intervention is likely to occur solely through aggressive fluid management and providing an alternative to an emergency department visit. Our data and those from others15 suggest that the beneficial effects related to this second mechanism are dependent on continued active management and that these benefits may be lost on longer-term follow-up after discontinuation of the active intervention. Our trial supports previously published data20 that suggest that by facilitating access to medical care, DM may lead to increased long-term medical resource use, eventuating in increased hospitalization, including that for noncardiac diagnoses. As cited previously,20 enhanced communication and access to PCPs in intervention patients (via SPAN-CHF nurses) may have led to investigation of noncardiac illnesses that were previously undetected or unreported. This is a potential limitation of all DM programs that needs to be considered, especially in evaluations of the cost-effectiveness of these programs. Conversely, other trials14,17 suggested that early benefits of HF DM are magnified when that intervention is continued for longer periods of time.
We have extended findings of prior randomized, controlled trials by demonstrating the clinical benefits of a 90-day nurse-driven HF DM intervention delivered uniformly across a diverse provider system in a population with an initial high background use of ACE inhibitors, ARBs, and β-blockers. We documented significant 90-day reductions in hospital use resulting from HF and cardiac causes, with significant attrition in these improved outcomes on longer-term follow-up. Improved long-term benefit in a population with adequate baseline medical treatment likely requires more active chronic intervention, even among patients who appear clinically stable.
This study was funded in part by grants from the Fannie E. Rippel Foundation and the Hewlett-Packard Corporation.
1999 Heart and Stroke Statistical Update. Dallas, Tex: American Heart Association; 1999. Publication 75231–4596.
National Heart, Lung, and Blood Institute. Morbidity and Mortality Chartbook on Cardiovascular, Lung, and Blood Diseases, 1996. Washington, DC; US Department of Health and Human Services; 1996.
O’Connell JB, Bristow MR. Economic impact of heart failure in the United States: time for a different approach. J Heart Lung Transplant. 1993; 13: S107–S112.
MacIntyre K, Capewell S, Stewart S, et al. Evidence of improving prognosis in heart failure. Circulation. 2000; 102: 1126–1131.
National Inpatient Sample, Release 5, 1996 Data. Washington, DC: Agency for Health Care Policy and Research; 1999. Publication PB 99–500480.
Konstam MA, Kimmelstiel CD. Economics of heart failure. In: Balady GJ, Pina IL, eds. Exercise and Heart Failure. Armonk, NY: Futura Publishing Co, Inc; 1997: 19–28.
Cline CMJ, Israelsson BYA, Willenheimer RB, et al. Cost effective management programme for heart failure reduces hospitalisation. Heart. 1998; 80: 442–446.
Jaarsma T, Halfens R, Huijer Abu-Saad H, et al. Effects of education and support on self-care and resource utilization in patients with heart failure. Eur Heart J. 1999; 20: 673–682.
Stewart S, Horowitz JD. Home-based intervention in congestive heart failure. Circulation. 2002; 105: 2861–2866.