|
| HOME | SUBSCRIPTIONS | CURRENT ISSUE | PAST ISSUES | CARDIOSOURCE | SEARCH | HELP | FEEDBACK |
|
|
|
|||||||||
|
|||||||||||
|
J Am Coll Cardiol, 2005; 45:1654-1664, doi:10.1016/j.jacc.2005.01.050 © 2005 by the American College of Cardiology Foundation |
* University of Hull, Kingston Upon Hull, United Kingdom
Universitätsklinikum, Aachen, Germany
Thoraxcenter, Erasmus MC, Rotterdam, the Netherlands
Manuscript received April 22, 2004; revised manuscript received January 7, 2005, accepted January 11, 2005.
* Reprint requests and correspondence: Prof. John G. F. Cleland, Department of Cardiology, University of Hull, Castle Hill Hospital, Kingston-upon-Hull, United Kingdom (Email: j.g.cleland{at}hull.ac.uk).
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionAppendixSupplementary appendixReferences |
|---|
BACKGROUND: Heart failure is associated with a high rate of hospitalization and poor prognosis. Telemonitoring could help implement and maintain effective therapy and detect worsening heart failure and its cause promptly to prevent medical crises.
METHODS: Patients with a recent admission for heart failure and left ventricular ejection fraction (LVEF) <40% were assigned randomly to HTM, NTS, or UC in a 2:2:1 ratio. HTM consisted of twice-daily patient self-measurement of weight, blood pressure, heart rate, and rhythm with automated devices linked to a cardiology center. The NTS consisted of specialist nurses who were available to patients by telephone. Primary care physicians delivered UC. The primary end point was days dead or hospitalized with NTS versus HTM at 240 days.
RESULTS: Of 426 patients randomly assigned, 48% were aged >70 years, mean LVEF was 25% (SD, 8) and median plasma N-terminal pro-brain natriuretic peptide was 3,070 pg/ml (interquartile range 1,285 to 6,749 pg/ml). During 240 days of follow-up, 19.5%, 15.9%, and 12.7% of days were lost as the result of death or hospitalization for UC, NTS, and HTM, respectively (no significant difference). The number of admissions and mortality were similar among patients randomly assigned to NTS or HTM, but the mean duration of admissions was reduced by 6 days (95% confidence interval 1 to 11) with HTM. Patients randomly assigned to receive UC had higher one-year mortality (45%) than patients assigned to receive NTS (27%) or HTM (29%) (p = 0.032).
CONCLUSIONS: Further investigation and refinement of the application of HTM are warranted because it may be a valuable role for the management of selected patients with heart failure.
| |||||||||||
Effective treatment is available for heart failure due to left ventricular (LV) systolic dysfunction (6). Unfortunately, because of inadequate organization of care, effective pharmacologic treatment often is not given and, when it is, often at doses lower than those shown to be effective (5,7). There is growing evidence that improved organization of heart failure care, as for other malignant diseases, can have a major impact on hospitalization and/or death (8–12). However, relative to the number of patients affected, there is a lack of healthcare staff able to provide expert management for heart failure. Novel methods for the delivery of quality healthcare could increase the effectiveness of management while containing costs and using scarce human resources to maximum effect.
Interest in telemedicine as a way of providing care has been stimulated by the rising costs of hospital treatment, rapid advances in technology, and the wider availability of low-cost, patient-friendly equipment (13,14). Home telemonitoring (HTM) allows the evaluation of patients’ vital signs once or more per day and provides diagnostic information that can be transmitted to health professionals. It has the potential to involve patients more in their own care, assist the titration of medications, improve compliance, and help providers identify early signs of worsening heart failure and its precipitating factors. Home telemonitoring also may assist with care at home or early discharge planning, thereby reducing admissions, hospital days, and rates of mortality. The purpose of the Trans-European Network–Home-Care Management System (TENS-HMS) study was to address some of these issues.
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixSupplementary appendixReferences |
|---|
Patients who were ready for or recently discharged after an admission for worsening heart failure were evaluated for inclusion provided their primary care physician agreed. To be included, patients had to have a hospital admission due to or complicated by worsening heart failure lasting >48 h within the last six weeks; to have persisting symptoms of heart failure, a LV ejection fraction <40%, an LV end-diastolic dimension >30 mm/m (height); and to be receiving furosemide at a dose 
40 mg/day or equivalent (e.g., 1 mg of bumetanide or 10 mg of torasemide). In addition to these criteria, patients had to have at least one of the following markers of a further increase in risk: an unplanned cardiovascular admission lasting >48 h within the previous 2 years, an LV ejection fraction <25%, or treatment with furosemide at a dose of 100 mg/day or equivalent. Patients who were younger than 18 years of age; who were deemed unable to comply with home telemonitoring; or who were awaiting revascularization, cardiac resynchronization, or heart transplantation were excluded. The study adhered to local and international guidelines for good clinical practice and was approved by relevant ethical committees for each participating hospital. Written informed consent was obtained from all patients.
Methods. Baseline demographic and social details, clinical history, medication, New York Heart Association (NYHA) functional classification, weight, and physical signs were recorded, and a blood sample was taken for the measurement of hemoglobin, electrolytes, urea, creatinine, and N-terminal pro-brain natriuretic peptide (NT-proBNP; Roche Elecsys proBNP assay, Mannheim, Germany).
Patients were then assigned randomly to receive HTM, nurse telephone support (NTS), or usual care (UC). The main comparison of interest was that between NTS and HTM and, accordingly, twice as many patients were assigned randomly to these groups. The UC group was used as a reference to ascertain whether either HTM or NTS had changed outcome.
After acquiring consent, patients’ baseline data were recorded and sent to an independent statistical group (i.e., Institute for Medical Informatics and Biostatistics, Basel). Random permuted blocks for each center were used to allocate patients to treatment groups. The block size was kept confidential and was varied to avoid investigators predicting which management-group would be the next to be allocated.
All patients were given an individualized written management plan by the investigator that described what pharmacologic treatment they should receive, in what order, and how it should be monitored. All patients required a loop diuretic according to the study entry criteria. The management plan focused on treatment of LV systolic dysfunction with appropriate doses of angiotensin-converting enzyme (ACE) inhibitors and beta-blockers and, if severe symptoms persisted, spironolactone according to regional guidelines (15). Digoxin and anticoagulants were recommended for patients in atrial fibrillation. Patients who could not tolerate or who had contraindications to the aforementioned medication were permitted in the study provided an explanation was given.
For patients assigned randomly to UC, the patient management plan was sent to the patient’s primary care physician, who was asked to implement it. Where the usual organization of care involved nurse specialist titration of drugs, this was allowed. Patients were assessed at a research clinic every four months to assess intervening history, symptoms and signs, renal function, and serum electrolytes. Contact with the research team was discouraged between visits.
Patients assigned randomly to receive NTS also were managed as described for UC except they were contacted by telephone each month by a heart failure specialist nurse to assess their symptoms and current medication. The nurse could proffer advice to the patient at this time and provide feedback to the primary care provider. Patients also were told that they could contact the study nurse by telephone at any time, either directly or by leaving a message on a telephone-answering machine. However, should an out-of-hours emergency arise, they were told to contact their primary care doctor or the ambulance service.
Patients assigned randomly to HTM received instructions on how to use the telemonitoring equipment, and nurse telephone support was offered as for the NTS group. As soon as possible after randomization (median 12 days; upper quartile 24 days), a service engineer visited the patient’s home to install the equipment, which consisted of low-profile, electronic, weighing scales, an automated sphygmomanometer, and a single-lead electrocardiogram using wrist-band electrodes (Fig. 1). Each device contained a short-range radio-transmitter that allowed it to communicate automatically with a hub connected to the patient’s conventional telephone line and, thereby, automatically to a central web server and then via secure Intranet connections to a workstation at each investigator site. Data were encrypted during transmission to ensure patient confidentiality. Patients were asked to make a set of measurements every day before breakfast and before their evening meal, after emptying their bladders, while wearing light clothing, no shoes, and before the next dose of medication. Thus, the patient’s weight, blood pressure, heart rate, and rhythm were monitored twice daily. Values greater than or less than preset limits were notified automatically to the study nurses, who then reviewed the information and took action either directly for any short-term advice or through the primary care physician if long-term changes in therapy were required. Nurses also could scan patient data manually to identify any trends that they considered as requiring action. Study personnel were primarily responsible for implementation of the management plan in patients assigned randomly to HTM. The primary care physician and the investigator were kept informed of all contacts.
|
Outcome measures. The primary outcome was days lost because of death or hospitalization in acute medical/surgical beds for any reason during 450 days, and the primary comparison of interest was between patients assigned randomly to NTS and HTM. The assumption was that the primary effect of HTM, compared with NTS, would be to reduce bed-days occupancy. Because this outcome is determined partly by the duration of follow-up and by mortality, a fixed duration of follow-up was incorporated into the definition. After an interim analysis (see "Interim analysis"), the duration of follow-up was reduced to 240 days. All-cause mortality, symptoms, and optimization of medication were secondary outcomes. Investigators were asked to classify hospitalizations as due to heart failure, other cardiovascular, or noncardiovascular. Deaths were classified as sudden, due to circulatory failure, or due to other causes.
Sample size. We assumed that HTM would not alter survival, but we included a survival component in the primary end point because the number of days alive is a major determinant of days in hospital. We estimated that mortality would be 2% per month for the first 3 months, 1.5% per month for the next 3 months, and 1% per month thereafter (4). We assumed that the NTS groups would spend an average of 10% of days-alive (41 days) in hospital (4) and that this would be reduced to 6% (25 days) by HTM. In other words, days lost to death or hospitalization would be reduced from 80 days to 63 days during 450 days follow-up, representing 17.8% and 14.0% of days exposure, respectively. We also expected a highly skewed distribution in the primary outcome, with approximately 25% of patients in the control group surviving until the end of the study without any hospital admission. Using a Wilcoxon (Mann-Whitney) rank-sum test, a study with 145 patients in each of the two arms of primary interest was calculated to have 80% power to show significance at p = 0.01 using a two-sided test. A total of 195 patients per arm would provide 90% power. The power calculations appeared robust to a 10% to 20% rate of discontinuation of HTM. Accordingly, we planned to assign 200 patients randomly to HTM, 200 to NTS, and 100 to UC.
Statistical analysis. Analyses were conducted by intention-to-treat. Baseline characteristics were expressed as mean and standard deviation or median and interquartile range. Treatment groups were compared using a two-sample Wilcoxon test. Outcome measures were expressed as differences between means with 95% confidence intervals, calculated using the Scheffe’s multiple comparison procedure. Categorical variables were analyzed using the chi-square test. Continuous outcome variables were not normally distributed and, therefore, the Kruskal-Wallis test was used.
For the survival distribution and statistical comparison, the Kaplan-Meier estimation method with 95% confidence level for survivor function was used. An exploratory multivariate Cox regression analysis was performed that included the following covariates: assigned group, age, NT proBNP, body mass index, systolic and diastolic blood pressure, hemoglobin, sodium, urea, creatinine, NYHA functional classification, loop and potassium-sparing diuretics, ACE inhibitors, and beta-blockers. Covariates were entered into the model if the p value was <0.1. The final model was analyzed using the selected covariate and the variable treatment group and applying Cox regression with stepwise forward selection (entry p value 0.05).
Predefined subgroup analyses included age (lower two terciles vs. upper tercile), gender, etiology of heart failure (ischemic heart disease vs. other), LV ejection fraction (lower two terciles vs. upper), dose of loop diuretic (40 mg vs. >40 mg/day), and plasma concentration of NT-proBNP (lower tercile vs. upper two terciles). The statistical analysis was conducted using the SAS System version 8.2 (SAS Institute, Cary, North Carolina).
Interim analysis. An interim analysis was conducted by the independent statistical group (Institute for Medical Informatics and Biostatistics) after 426 patients had been recruited. They requested that recruitment of patients should stop and the trial brought to a close because of a large difference in mortality rates between the UC care group and those assigned randomly to NTS or HTM and because it was unlikely that the primary end point would be reached. After examining the mortality data, the steering committee agreed to close the study to recruitment, inform the investigators of the mortality difference, but continue follow-up until October 2002 to allow an evaluation of days lost to death or hospitalization over the course of 240 days in almost all patients. This point became the revised primary outcome measure.
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixSupplementary appendixReferences |
|---|
|
|
|
|
|
|
|
|
|
Data on patient contacts other than hospitalization were collected by monthly review in the NTS and HTM groups but only every four months in the UC group (Table 8). This difference may have led to under-reporting of contacts in the latter group. The NTS and HTM groups were associated with a substantial and similar increase in reported patient-contacts, including emergency room, home and office visits, and telephone contacts, compared with UC. The overall number of contacts also was significantly greater with NTS than HTM during 450 days of follow-up. There was a substantial substitution of home and office visits with telephone contacts with HTM compared with NTS.
|
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixSupplementary appendixReferences |
|---|
Most patients in this trial reported few or no symptoms at the time of assignment despite severe cardiac dysfunction, which may reflect the ability of intensive therapy to control symptoms. Alternatively, the assessment of symptoms in the aftermath of an episode of severe decompensation may be unreliable, either because of the patient’s perception that symptoms have improved or because the patient has not yet tried to return to his or her usual activities. Mortality and recurrent hospitalization rates were high, indicating that a simple assessment of symptoms at discharge is not an adequate guide to prognosis. The severity of LV systolic dysfunction, high rates of comorbidity, significant renal dysfunction, and low arterial pressure all predicted a poor prognosis (16). In addition, the median plasma concentration of NT-proBNP was markedly elevated and similar to that reported in studies of severe persistent heart failure (17). This study not only emphasizes that patients with severe cardiac dysfunction and a poor prognosis may have only intermittently severe symptoms but also shows that outcome can be modified if a high standard of care is provided.
Days lost to death or hospitalization, the primary outcome measure in this study, is a relatively novel but highly relevant outcome in clinical and health economic terms. Kaplan-Meier (time-to-first-event) curves are an accurate method for displaying the effects of treatment on mortality but may be misleading for nonfatal events (18). In "time-to-first-event" analyses for a composite of fatal and nonfatal events, a minor event has the same value as death, and a patient who has a single, short, early admission is ascribed a worse outcome than one who has multiple later, long-term admissions. Hospitalization, especially when brief, may reflect early detection of problems and good care rather than an adverse outcome. Days lost to death or hospitalization, assuming all patients are followed for a similar time, combines mortality with a more sensible measure of morbidity; the number of days spent in hospital.
There are several reasons why increased monitoring by nurses or telemonitoring might improve outcome in this population. Better organization of care and patient support might increase the likelihood of patients being initiated and maintained on appropriate treatment for heart failure. This hypothesis was confirmed for HTM but not for NTS. Earlier detection of cardiovascular and noncardiovascular problems by well-organized care also may have led to more prompt and effective therapy for a variety of problems, both cardiovascular and noncardiovascular.
Two recent reports support the notion that telemonitoring may improve mortality in patients with heart failure. The Weight Monitoring in Heart Failure (WHARF) trial reported a 10.4% absolute and 56.2% relative reduction in mortality during an average follow-up of 169 days in 280 patients with a HTM system for symptoms and weight compared with UC (14). In one of the largest trials of disease management for heart failure, a 5% absolute and 38% relative reduction in mortality was reported with NTS compared with UC among patients with heart failure due to LV systolic dysfunction but not those without systolic dysfunction (12). The fact that benefit occurred only in the group of patients in whom pharmacologic interventions have been shown to reduced mortality provides supportive evidence that this response was rather specific to the management program and not just a general response to improved care. However, neither study showed a reduction in hospitalizations or other healthcare use. Perhaps it is time to move away from always imputing an adverse outcome to healthcare contacts or hospitalizations in trials and service audits. If such events improve how patients feel and reduce their mortality then, provided they are affordable, they should be welcomed, especially when they prevent rather than respond to crises.
Patients assigned randomly to NTS spent the highest proportion of days in hospital, which appears to reflect, on the one hand, the high mortality rate in the UC group, because death prevents hospitalization and, on the other hand, a reduction in hospital days with HTM. The reasons for the reduction in days spent in hospital by HTM appear complex. Patients were more likely to be hospitalized with heart failure in this group, perhaps reflecting the early detection of deterioration and possibly some false alarms. However, hospital stays were shorter for patients assigned to HTM, which may reflect improved planning of both admissions and discharges so that effective care was delivered quickly by appropriately trained staff. In addition, there may have been a greater willingness to discharge a patient knowing that monitoring and titration of therapy could be facilitated at home. In this context, it could be argued that many of these hospitalizations were appropriate, contributed to the continued well-being of the patient, and that HTM was encouraging a more flexible and dynamic use of a greater spectrum of resources appropriate to patients’ needs.
Improved access to care, either by nurses or by telemonitoring, appeared to lead to an increase in patient contacts, which is not surprising. Home visits are potentially an expensive method of delivering health care because the health care professional may spend considerable time traveling to and from the patients home, although visiting the patient in his or her home environment may help ensure that the patient is complying with medication and coping with his or her situation (8). Office visits make fewer demands on the healthcare professional’s time but more demands on the patient, including time and travel costs. By comparison, telephone calls are an inexpensive way of making healthcare contacts. In the context of the TEN-HMS study, substituting a large proportion of face-to-face contacts with telephone calls does not appear to have an adverse effect on the uptake of therapy, symptoms, hospitalization, survival, or patient satisfaction when supported by telemonitoring.
Although many patients were elderly, their acceptance and ability to cope with the HTM technology was high. Few patients asked for the equipment to be removed or failed to comply with daily measurements. Good or very good satisfaction with HTM was reported by 96% of patients.
This trial is one of the first substantial, prospective, randomized ones for HTM for patients with heart failure. The results are sufficiently encouraging to warrant both service development and further research. Improved selection of patients and tailoring the duration of HTM to the patient’s needs could enhance the benefits and lower the costs of therapy further. There was little practical experience of HTM for heart failure when this study was planned, and nurses and medical staff had to learn how to use the technology as the study progressed. Staff training programs based on this study also could improve the effectiveness of HTM. Finally, improvements in devices, communication, and data processing for decision support all are likely to increase the potential of HTM to benefit patients.
|
Appendix |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixSupplementary appendixReferences |
|---|
|
Supplementary appendix |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixSupplementary appendixReferences |
|---|
|
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixSupplementary appendixReferences |
|---|
Related Article
This article has been cited by other articles:
|
|
 |
|
  F. Halimi, J. Clementy, P. Attuel, X. Dessenne, W. Amara, and on behalf of the OEDIPE trial Investigators Optimized post-operative surveillance of permanent pacemakers by home monitoring: the OEDIPE trial Europace, September 4, 2008; (2008) eun250v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J Morguet, P. Kuhnelt, A. Kallel, U. Rauch, and H.-P. Schultheiss Utilization of telemedicine by heart disease patients following hospitalization J Telemed Telecare, June 1, 2008; 14(4): 178 - 181. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J G F Cleland, N Freemantle, J-C Daubert, W D Toff, F Leisch, and L Tavazzi Long-term effect of cardiac resynchronisation in patients reporting mild symptoms of heart failure: a report from the CARE-HF study Heart, March 1, 2008; 94(3): 278 - 283. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Jaarsma, M. H. L. van der Wal, I. Lesman-Leegte, M.-L. Luttik, J. Hogenhuis, N. J. Veeger, R. Sanderman, A. W. Hoes, W. H. van Gilst, D. J. A. Lok, et al. Effect of Moderate or Intensive Disease Management Program on Outcome in Patients With Heart Failure: Coordinating Study Evaluating Outcomes of Advising and Counseling in Heart Failure (COACH) Arch Intern Med, February 11, 2008; 168(3): 316 - 324. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Ritzema, I. C. Melton, A. M. Richards, I. G. Crozier, C. Frampton, R. N. Doughty, J. Whiting, S. Kar, N. Eigler, H. Krum, et al. Direct Left Atrial Pressure Monitoring in Ambulatory Heart Failure Patients: Initial Experience With a New Permanent Implantable Device Circulation, December 18, 2007; 116(25): 2952 - 2959. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. D. Solomon, J. Dobson, S. Pocock, H. Skali, J. J.V. McMurray, C. B. Granger, S. Yusuf, K. Swedberg, J. B. Young, E. L. Michelson, et al. Influence of Nonfatal Hospitalization for Heart Failure on Subsequent Mortality in Patients With Chronic Heart Failure Circulation, September 25, 2007; 116(13): 1482 - 1487. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. W F B.-A. de la Porte, D. J A Lok, D. J van Veldhuisen, J. van Wijngaarden, J. H Cornel, N. P A Zuithoff, E. Badings, and A. W Hoes Added value of a physician-and-nurse-directed heart failure clinic: results from the Deventer-Alkmaar heart failure study Heart, July 1, 2007; 93(7): 819 - 825. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Holland, I. Brooksby, E. Lenaghan, K. Ashton, L. Hay, R. Smith, L. Shepstone, A. Lipp, C. Daly, A. Howe, et al. Effectiveness of visits from community pharmacists for patients with heart failure: HeartMed randomised controlled trial BMJ, May 26, 2007; 334(7603): 1098 - 1098. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. O Grancelli and D. C Ferrante Telephone interventions for disease management in heart failure BMJ, May 5, 2007; 334(7600): 910 - 911. [Full Text] [PDF]
|
|
|
|
 |
|
 G. Pare, M. Jaana, and C. Sicotte Systematic Review of Home Telemonitoring for Chronic Diseases: The Evidence Base J. Am. Med. Inform. Assoc., May 1, 2007; 14(3): 269 - 277. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M Gary Nicholls, A Mark Richards, and Christchurch Cardioendocrine Research Group Disease monitoring of patients with chronic heart failure Heart, April 1, 2007; 93(4): 519 - 523. [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Rozenman, R. S. Schwartz, H. Shah, and K. H. Parikh Wireless Acoustic Communication With a Miniature Pressure Sensor in the Pulmonary Artery for Disease Surveillance and Therapy of Patients With Congestive Heart Failure J. Am. Coll. Cardiol., February 20, 2007; 49(7): 784 - 789. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W.H. W. Tang and G. S. Francis The Year in Heart Failure J. Am. Coll. Cardiol., December 19, 2006; 48(12): 2575 - 2583. [Full Text] [PDF]
|
|
|
|
 |
|
 S. Myers, R. W. Grant, N. E. Lugn, B. Holbert, and J. C. Kvedar Impact of Home-Based Monitoring on the Care of Patients with Congestive Heart Failure Home Health Care Management Practice, October 1, 2006; 18(6): 444 - 451. [Abstract] [PDF]
|
|
|
|
|
|
J. G.F. Cleland Reply J. Am. Coll. Cardiol., August 15, 2006; 48(4): 851 - 852. [Full Text] [PDF]
|
|
|
|
|
|
P. S. Chan, B. K. Nallamothu, and R. A. Hayward Keeping Apples and Oranges Separate: Reassessing Clinical Trials That Use Composite End Points as Their Primary Outcome J. Am. Coll. Cardiol., August 15, 2006; 48(4): 850 - 850. [Full Text] [PDF]
|
|
|
|
|
|
F. Koehler and S. D. Anker Noninvasive Home Telemonitoring: The Trans-European Network-Home-Care Management System J. Am. Coll. Cardiol., August 15, 2006; 48(4): 850 - 851. [Full Text] [PDF]
|
|
|
|
|
|
W.H. W. Tang and G. S. Francis The Year in Heart Failure J. Am. Coll. Cardiol., December 6, 2005; 46(11): 2125 - 2133. [Full Text] [PDF]
|
|
|
|
|
|
E. P. Havranek Improving the Outcomes of Heart Failure Care: Putting Technology Second J. Am. Coll. Cardiol., May 17, 2005; 45(10): 1665 - 1666. [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||
