- Robyn A Clark, scholar1,
- Sally C Inglis, scholar2,
- Finlay A McAlister, associate professor3,
- John G F Cleland, professor4,
- Simon Stewart, professor5
- 1Division of Health Sciences, University of South Australia, Adelaide, Australia
- 2Faculty of Health Sciences, University of Queensland, Brisbane, Australia
- 3Division of General Internal Medicine, University of Alberta, Edmonton, Canada
- 4Academic Cardiology, University of Hull, Hull
- 5Preventative Cardiology Unit, Baker Heart Research Institute, Prahran, Melbourne, Vic 3004, Australia
- Correspondence to: S Stewart firstname.lastname@example.org
- Accepted 19 February 2007
Objective To determine whether remote monitoring (structured telephone support or telemonitoring) without regular clinic or home visits improves outcomes for patients with chronic heart failure.
Data sources 15 electronic databases, hand searches of previous studies, and contact with authors and experts.
Data extraction Two investigators independently screened the results.
Review methods Published randomised controlled trials comparing remote monitoring programmes with usual care in patients with chronic heart failure managed within the community.
Results 14 randomised controlled trials (4264 patients) of remote monitoring met the inclusion criteria: four evaluated telemonitoring, nine evaluated structured telephone support, and one evaluated both. Remote monitoring programmes reduced the rates of admission to hospital for chronic heart failure by 21% (95% confidence interval 11% to 31%) and all cause mortality by 20% (8% to 31%); of the six trials evaluating health related quality of life three reported significant benefits with remote monitoring, and of the four studies examining healthcare costs with structured telephone support three reported reduced cost and one no effect.
Conclusion Programmes for chronic heart failure that include remote monitoring have a positive effect on clinical outcomes in community dwelling patients with chronic heart failure.
Chronic heart failure is a common diagnosis, carries a poor prognosis, and affected patients are major consumers of healthcare resources.1 As the prevalence of chronic heart failure is increasing this situation will deteriorate unless new management strategies are developed.2 The effectiveness of multidisciplinary non-pharmacological approaches for improving outcomes in patients with chronic heart failure has been well established in over 30 randomised trials.3 4 5 6 7 As most of these trials have tested multifaceted approaches, however, it has been difficult to identify the incremental benefits of the components of each intervention.6 Nevertheless, it is clear that within most populations access to these programmes is limited as a result of barriers related to funding or geography.8 As a result interest is increasing in remote monitoring models for delivering care, which incorporate information communication technology either as telemonitoring (transfer of physiological data such as blood pressure, weight, electrocardiographic details, and oxygen saturation through telephone or digital cable from home to healthcare provider) or as regular structured telephone contacts between patients and healthcare providers, which may or may not include the transfer of physiological data.6
Earlier reviews of multidisciplinary programmes for chronic heart failure have been unable to make definitive conclusions about the value of remote monitoring strategies given the paucity of relevant studies and patient numbers at the time of these analyses.5 6 However, several studies with relatively large numbers of patients have since been published, permitting a more detailed analysis. We evaluated the effect of remote monitoring strategies in patients with chronic heart failure and whether the effect differed by the type of technology used for the communication of information.
We updated two earlier systematic reviews that dealt with telemonitoring5 6 by searching 15 electronic databases using search methods recommended by the Cochrane Heart Review Group.9 All randomised trials evaluating remote monitoring programmes published between 1 January 2002 and 6 May 2006 were included. Databases searched included the Cochrane library and the Cochrane CENTRAL register of controlled trials, Medline (1 January 2002 to 6 May 2006), Embase, CINAHL (1 January 2002 to 6 May 2006), AMED, ISI web of knowledge, HSTAT, Ingenta, Zectoc, LILACS, and science citation index expanded (to search forward to detect studies citing the original reviews), DARE, national research register, Psych Info, and web of science. We also hand searched the reference lists in 21 published systematic reviews of disease management programmes in chronic heart failure,3 4 5 6 7 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 149 review articles on telephone support programmes in chronic disease, and those studies identified in our electronic searches that met the inclusion criteria. Unpublished conference proceedings were reviewed and published abstracts were included if the authors replied to our request and sufficient details and outcomes of studies were retrieved. Finally, we communicated with the principal investigators of the identified trials and with national and international experts in the specialty to identify any studies we had potentially missed. We did not restrict study inclusion by language but did limit our review to only randomised controlled trials.
We applied the highly sensitive search strategy from the Cochrane Collaboration.9 Keywords for searches of the database included heart failure (exp), cardiac failure (exp), telemedicine (exp), telecare (exp), telemonitoring (exp), teleconsultation (exp), teleconference (exp), telecommunications (exp), case management (exp), comprehensive health care (exp), disease management (exp), health services research (exp), home care services (exp), clinical protocols (exp), patient care planning (exp), nurse led clinics and special clinics (exp), randomised controlled trial(s), controlled clinical trial, random allocation, double blind method, single blind method, clinical trial(s), research design, comparative study, follow-up study, and prospective study.
Search strategies were written for each database and double checked by the second reviewer, under the direction and supervision of a medical librarian.
Types of interventions
Remote monitoring programmes started by a health professional (medical, nursing, social work, pharmacists) for patients with chronic heart failure living at home were eligible for inclusion if the monitoring was carried out at least once in the first month after hospital discharge, was targeted towards the patient (that is, the patient had to be the person on the telephone), was structured (as opposed to offering telephone follow-up on an “as needed” basis), and was to be delivered as the only aftercare intervention without home visits or more than usual clinic follow-up. We excluded studies in which the remote monitoring was intended primarily to deal with the problems of care givers rather than of patients. We a priori classified programmes as being structured telephone support if they consisted of standardised telephone contact of patients with chronic heart failure and relied on reporting of symptoms alone, or telemonitoring if they consisted of telephone contact for eliciting symptoms and transmission of physiological data.
Our primary outcomes were all cause mortality, all cause rate of admission to hospital (proportion of patients readmitted to hospital at least once during follow-up), and rate of admission to hospital as a result of chronic heart failure (proportion of patients readmitted to hospital at least once during follow-up). Our secondary outcomes were health related quality of life, cost, and acceptability.
Validity assessment and data abstraction
Two investigators (RAC, SCI) independently reviewed the results of the searches for study inclusion and extracted data. We excluded any studies in which additional home or clinic visits (more than usual care) were offered to patients in the intervention or control arms. Study quality (particularly method), randomisation, and intervention, were judged using accepted criteria and compared with the review protocol.9 Disagreements between the two reviewers were resolved by a third reviewer (SS, FAMcA, or JGFC). Data abstraction was carried out independently and blinded by RAC and SCI, with FAMcA checking extracted data. Overall the inter-rater reliability on key inclusion criteria (randomisation and intervention) was strong (κ score 0.73, 95% confidence interval 0.54 to 0.92).27
Study characteristics and data synthesis
Owing to the expected differences in patient populations, programme characteristics, and length of follow-up, we carried out our primary analyses using the DerSimonian and Laird random effects model. Analyses were carried out using RevMan 4.2 (Nordic Cochrane Centre).9 As the outcomes of interest were relatively common we calculated risk ratios and 95% confidence intervals. The risk difference (difference between observed proportion of the event in the treatment and usual care groups9) was calculated by subtracting the risk of the event in the usual care group from that of the treatment group. These data are presented with 95% confidence intervals.
We carried out intention to treat analyses—that is, all patients and their outcomes were analysed in the groups to which they were allocated, regardless of whether they received the treatment. We examined for statistical heterogeneity in each outcome of interest using Cochran's Q test and I2 statistic. Secondary outcomes (expected to be reported less often) were described and tabulated.
Overall 234 of 499 citations were reviewed in detail. Of these, 14w1-w14 randomised controlled trials (4264 patients) were eligible for inclusion (fig 1⇓). One trial was three armedw1; to avoid double counting of the control patients the results for the control arm were shared between the two comparisons for the pooled analysis of all remote monitoring programmes, but all patients in the control arms were counted in each of the sub-analyses (telephone support v usual care, telemonitoring v usual care). Four trials evaluated telemonitoring, w11-w14 nine evaluated structured telephone support,w2-w10 and onew1 evaluated both. Ninety five per cent of the included trials were captured by the Medline search, 2% from CINAHL, and 3% from hand searching and contact with experts.
The length of follow-up of these trials ranged from three to 16 months, the mean ages of participants ranged from 57 to 75 years, and all trials enrolled patients with symptoms (New York Heart Association classification range II-IV, left ventricular ejection fraction <40%). Structured telephone support programmes included monitoring of symptoms, medicine management, and education and counselling on lifestyle. All the telemonitoring programmes included transfer of daily data on weight, pulse, blood pressure, and electrocardiographic findings.
The quality of the studies was evaluated using Cochrane recommendations.9 The 14 included studies were rated as adequate, reporting 61% of the recommended quality variables. Six (42%) described concealment and some level of blinding of patient or outcome assessors, and seven (50%) reported completeness of follow-up (table 1⇓).
Quantitative data synthesis
All cause mortality
All 14 trials reported all cause mortality (15 comparisons, 581 deaths, fig 2⇓) and the pooled estimates showed a statistically significant 20% reduction (95% confidence interval 8% to 31%) with remote monitoring programmes. The benefits were greater with telemonitoring (risk ratio 0.62, 0.45 to 0.85, P=0.003, based on 127 deaths in 807 patients) than with structured telephone support (0.85, 0.72 to 1.01, P=0.06, based on 482 deaths in 3542 patients), although this difference did not achieve significance (P=0.18) using the adjusted indirect comparison method described by Song et al.28 Mortality data from these 14 trials showed little heterogeneity (P=0.56, I2=0%).
All cause admission to hospital
Of the eight trials (nine comparisons) that reported rates of all cause admission to hospital (fig 3⇓), none reported a statistically significant result. Even the pooled estimates did not show a significant benefit on this end point with remote monitoring programmes (0.95, 0.89 to 1.02).
Hospital admissions as a result of chronic heart failure
Nine trials (10 comparisons) reported rates of admission to hospital as a result of chronic heart failure, and although only one reported a statistically significant benefit, all trials showed similar relative reductions (P for heterogeneity 0.76, I2=0%) and the pooled results showed a reduction of 21% (11% to 31%) with remote monitoring programmes (fig 4⇓). Although no appreciable difference was found between the relative reductions seen with telemonitoring and telephone support programmes, evidence from randomised trials was insufficient to conclusively state that telemonitoring programmes reduce admissions to hospital since only one of these trials reported this outcome.
Quality of life, cost, adherence, and patient acceptability
Only six trials examined the effect of the intervention on health related quality of life (table 2⇓). Of these trials, three reported a significant and substantial improvement in quality of life between the intervention and control groups at the end of follow-up. The effect of the intervention on healthcare costs was reported in only four of these trials (all of structured telephone support); however, three of the four trials reported lower healthcare costs for patients randomised to the intervention (table 2⇓). The cost of the intervention was infrequently reported and variation existed between programmes (table 2⇓). None of the trials on telemonitoring reported the cost of the intervention or its effect on healthcare costs. Acceptability of the intervention to the patient was under-reported, with only four trials reporting this outcome (table 2⇓). Three of these trials reported that the intervention was acceptable to patients, with patients from one trial of telemonitoring considering the video link not useful.w11
An analysis of publication bias using funnel plots showed an unlikely possibility of bias within studies showing a reduction in mortality after remote monitoring.
This systematic review found that remote monitoring programmes for patients with chronic heart failure living in the community reduced admissions to hospital and all cause mortality by nearly one fifth while improving health related quality of life, but had no significant effect on all cause admission to hospital. Although few studies have examined economic outcomes, the three studies on structured telephone support suggested that the interventions were economically cost effective Thus, this systematic review builds on earlier ones5 6 of multidisciplinary interventions for chronic heart failure by tackling key issues and uncertainties relating to the specific effect of telephone based programmes. This review is particularly important as remote monitoring programmes provide a potentially feasible option for dealing with the expanding population of patients with chronic heart failure that cannot be accommodated within existing multidisciplinary chronic heart failure clinics owing to constraints caused by geography or resources.
The significant effect of structured telephone support on the risk of admissions to hospital for chronic heart failure (risk differences ranged between 2% and 35%) can be attributed in part to the triage of patients by telemonitoring nurse at the first sign of clinical deterioration, and the consequent immediate intervention of a primary care doctor.w5 w14 Similarly, all trials on telemonitoring in this review involved daily transmission of vital signs, weight, and symptoms at various time points to healthcare providers, thus potentially leading to earlier detection and management of clinical deterioration by the patient or managing health professional. A recent study of rapid up-titration of β blockers in 49 patients with chronic heart failure reported that deterioration in symptoms, including weight gain, oedema of the legs, and increasing dyspnoea, were usually present eight to 12 days before admission to hospital.29
The lack of effect of remote monitoring programmes on all cause admissions to hospital may require further exploration. This observation is consistent with an earlier meta-analysis.6 Importantly this result does not simply reflect a paucity of data as there were more events for this end point (1561 admissions in 3586 patients) than for deaths and admissions to hospital. Reduced mortality will increase the duration of exposure to the risk of admission and will reduce the effect of intervention on this outcome. However, telemonitoring is likely to produce false alarms and pre-emptive admissions in patients who are deteriorating but not yet in crisis and also to lead to early discharge because the patient still has a high level of monitoring at home. Consequently, telemonitoring may be more effective at shortening hospital stay than at reducing admissions. Increased survival and admissions for common comorbid conditions (for example, chronic respiratory disease, fractures from falls, and cancer among participants who were typically elderly)30 may also prevent a reduction in the frequency of admissions associated with telemonitoring. Finally, remote monitoring in patients with chronic heart failure focuses on indices specific to that disease and treatment; it may have little effect on other reasons for admission. Whether extending the range of monitoring to provide more comprehensive support will result in a further improvement in health outcomes is yet to be determined.
Quality of life, acceptability, and cost benefits were infrequently reported in these trials. Although those reporting these outcomes showed significant improvements with remote monitoring, future studies of telemonitoring or structured telephone support programmes should be encouraged to incorporate such measures and outcomes in their reporting. The same caveats and recommendations apply to cost data arising from these studies.
Limitations of the study
A limitation of this review is the relatively small number of studies (n=14) and participants (n=4264). In addition, few trials had follow-up beyond six months. Thus our observations on the positive, short term benefits of remote monitoring programmes may not extend to longer term outcomes. However, the hazard ratio for admission to hospital in patients with chronic heart failure is not linear as the greatest risk of readmission in such patients occurs in three to six months. Nevertheless, it is expected that the body of evidence on remote monitoring for chronic heart failure will expand considerably in the next decade as strategies on communicating information become normalised into medical practice and a better understanding is gained of the content of care provided by remote monitoring.
Furthermore, it is anticipated that the following studies will add to the evidence base in this specialty: the as yet unpublished reports on the home or hospital in heart failure31 trial, which showed an overall neutral effect on mortality and admissions to hospital; the recent work by the Scalvini et al team in Italy,32 which showed a significant reduction in risk of readmission (risk ratio 0.50, 95% confidence interval 0.34 to 0.73; P=0.01); and Riegel et al'sw10 most recent paper, which questions the effect of this type of intervention in non-Caucasian ethnic groups.
Although we have shown substantial and statistically significant benefits with remote monitoring for patients with chronic heart failure, monitoring is not a treatment but rather a different way of systematically organising effective care. Thus programmes that include remote monitoring should not be seen as a replacement for specialist care8 or multidisciplinary chronic heart failure clinics (two interventions that improve outcomes6 33). However, remote monitoring may be of particular benefit to patients who have difficulty accessing specialised care because of geography, transport, or infirmity.8 34
What is already known on this topic
Systematic reviews on telephone support and telemonitoring in patients with chronic heart failure have provided inconclusive evidence of benefit
What this study adds
Remote monitoring has the potential to improve clinical outcomes in community dwelling patients with chronic heart failure
We thank librarian Margaret Goedhart (University of South Australia) for her knowledge and skill in navigating bibliographies and electronic sources, and the following investigators for further data and information from their studies: WA Gattis, LR Goldberg, A Laramee, B Riegel, RT Tsuyuki, A Woodend, and S Scalvini.
Contributors: RAC conceived and designed the study. RAC and SCI reviewed the literature, developed the study protocol, and searched for and abstracted the data. RAC, SCI, SS, and JGFC analysed and interpreted the data. FAMcA assessed the quality of extracted data and was responsible for synthesis and analysis of the data. He will act as guarantor. JGFC hand searched the literature and referred experts to RAC and SCI. All authors contributed to the drafting of the article and revising it for important intellectual content.
Funding: RAC is supported by the National Institute of Clinical Studies and the National Heart Foundation of Australia. SCI and SS are supported by the National Health and Medical Research Council and the National Heart Foundation of Australia. FAMcA receives salary support from the Alberta Heritage Foundation for medical research population health scholar programme, the Canadian Institutes of Health Research new investigator programme, and the University of Alberta/Merck Frost/Aventis chair in patient health management. JGFC (within the past five years) was principal investigator transEuropean network homecare management system study and received research funds from the European Union and Phillips Healthcare to carry out the study and has received honorariums from Phillips for speaking on telemonitoring. JGFC is a consultant to the EU sponsored MyHeart project which is part funded by Phillips and other manufacturers of telehealth devices.
Ethical approval: Not required.