Meta-analysis and meta-regression of telehealth programmes for patients with chronic heart failure

Abstract

Summary

To assess the effectiveness of telehealth used for chronic heart failure (CHF) patients, we searched for peer-reviewed, randomized controlled trials published between 2001 and 2012. A total of 33 studies met the inclusion criteria. There were 26 studies (79%) which concerned tele-monitoring and 7 (21%) which concerned case management or nurse administered telephone-based management. There were 7530 patients in all, with an average age of 69 years. A meta-analysis showed that telehealth programmes had significant overall effectiveness in reducing all-cause mortality (Fixed effect model risk ratio 0.76, 95% CI 0.66 to 0.88), CHF-related hospitalization (Random effect model risk ratio 0.72, 95% CI 0.61 to 0.85) and CHF-related length of stay (Random effect model mean difference -1.41 days, 95% CI -2.43 to -0.39). In addition, telehealth programmes showed significantly greater effectiveness in reducing mortality and hospitalizations among patients with higher New York Heart Association (NYHA) categories. With age and NYHA held constant, recording questionnaire (symptoms) data could reduce the mortality risk by 34% and the risk of CHF-related hospitalization by 15%; adding a pulse (heart rate) detector could reduce the mortality risk by 40% and the risk of CHF-related hospitalization by 43%. Finally, telehealth programmes showed a tapering effect on mortality reduction: the longer the follow-up period, the less effective they were on decreasing mortality. In conclusion, telehealth programmes demonstrated clinical effectiveness in patients with CHF compared with usual care.

Introduction

Chronic heart failure (CHF) occurs when the heart is unable to provide sufficient pump action to distribute blood flow to meet the needs of the body. CHF can cause a number of symptoms including fatigue, weight gain, shortness of breath, oedema, dizziness and syncope.¹ CHF results in significant medical attention and high hospital costs. In the US, there are more than a million hospitalizations per year for CHF as a primary diagnosis and 3.6 million per year for CHF as a primary or secondary diagnosis.² Approximately 50% of CHF patients are re-hospitalized within 6 months of discharge, a trend that is expected to rise in parallel with the ageing population.^3,4 Ten million people are estimated to suffer from CHF in Europe and 4–7 million in the US.⁵ In 2010, the estimated total cost of CHF was $39 billion in the US.⁶

Exacerbations of CHF signs and symptoms may indicate a worsening clinical status and necessitate medical attention, otherwise decompensation and hospitalization may happen which increase the cost of treatment. If patients had the ability to know more about their illness, recognize changes in signs and symptoms, and could access a healthcare provider promptly the incidence of decompensation and subsequent hospitalization could be reduced.⁷

Telemedicine, based on telecommunication to monitor patients and transmit data related to patient health status,⁸ has been suggested as beneficial, because patients’ signs and symptoms can be assessed remotely and promptly by healthcare providers. Therefore, deterioration can be quickly detected to prevent mortality and re-hospitalization.⁹ However, the extent of the benefit from telehealth is still unclear. Although most research has shown positive outcomes, some studies have found neutral or even negative results. To our knowledge, no prior meta-analyses^9–11 have investigated the impact of patient characteristics and/or which physiological variables are monitored on the clinical effectiveness using a meta-regression approach. We have therefore studied the relation between the clinical outcome and covariates such as duration of tele-monitoring, patient characteristics and the particular physiological variables being monitored.

Methods

We searched in PubMed, Cochrane Review, European Journal of Heart Failure, International Journal of Cardiology and Journal of the American Medical Association for trials that examined the efficacy of telehealth in CHF patients from 2001 to 2012. Telehealth (remote patient monitoring) for CHF patients began developing in the mid 1990s¹² and there were few reports of randomized trials before 2001.¹³ We chose a ten-year period starting from 1 January 2001.

The keywords used in the search were: tele-monitoring, home monitoring, home based solution, telehealth, remote patient monitoring, disease management, medication monitoring, congestive heart failure, chronic heart failure and heart failure. Only randomized controlled trials (RCT) were included that reported at least one of the following outcomes: all-cause mortality, CHF hospitalization or CHF length of stay (LOS). The selection process is shown in Figure 1. Two authors reviewed the articles to ensure that they met the inclusion criteria. After the selection process was complete, each person separately extracted the information from the chosen articles. The results were consolidated and the discrepancies were resolved by a third author.

Figure 1

Study search strategy

For each article, we collected the following information: year of publication, place of study, total number of patients included in the trial, numbers of patients in intervention and control groups, mean duration of follow-up, age, sex, New York Heart Association (NYHA) functional class and types of physiological data recorded by telehealth (e.g. symptoms, bodyweight, blood pressure, electrocardiogram, heart rate, medication adherence, urine output and oxygen saturation). Only a few articles reported any CHF-caused deaths so this was not included.

Statistical analysis

The primary end point of the present study was a comparison of (1) the average difference in CHF-related LOS in days and (2) the cumulative incidence of events (number of patients with events/total number of patients per arm) between usual care and telehealth for each of the outcomes considered: all-cause mortality and CHF-related hospitalization.

The effect size used to compare mortality and CHF hospitalization was relative risk or risk ratio (RR). The effect size used to compare CHF-related LOS was the mean difference (MD), because all studies reported LOS in the same scale (days). The RR and the 95% confidence interval (CI) for mortality and CHF hospitalization in each study were calculated. Individual RRs were then pooled according to the fixed effect (FE) model, Mantel-Haenszel fixed-effect (MH) model, and random effect (RE) model.¹⁴ Restricted Maximum Likelihood (REML) was used by default because it was approximately unbiased and quite efficient.¹⁵

Meta-regression is a tool used in meta-analysis to examine the effect of moderator variables on study effect size using regression-based techniques. The goal is to explore which variables (e.g. patient characteristics and monitoring devices) contribute more to the overall effectiveness of telehealth. Most meta-analyses are conducted using either the fixed effect model or the random effect model. Under the fixed effect model we assume that there is an overall true effect size among the studies in the analysis, and differences in observed effects are due to sampling error. The random effect model allows the true effect to vary from study to study¹⁶ The fixed effect model has lower estimated standard error and higher power to detect significance. However the fixed effect model should not be chosen when heterogeneity exists among different studies. Statistical heterogeneity was evaluated by the Cochran Q test and measured by the I-squared statistic. Ignoring heterogeneity could increase Type I error and lead to inaccurate estimations. When the I-squared statistic was >20% we considered the overall effect sizes from random effect models to be preferable.¹⁰ The presence of any publication bias for each outcome was examined by funnel plots. The metafor package in R¹⁷ was used for all computations.

In summary we used empirical studies to suggest the possible covariants for the meta-analysis. For instance, due to the complexity of the literature sources, trials from different studies included different patient population, intervention approaches and study durations. Empirical studies have shown that higher mortality rate occurred among patients with worse symptoms and older ages.^18,19 In addition, some evidence also suggested that the mortality rate was different after 30 days and 1 year duration.²⁰ Therefore it would be interesting to explore whether patients with different backgrounds shared the same telehealth effectiveness. In the meta-regression, we extracted the average age of patients in a given study. For NYHA, the distribution of the patients with different NYHA was used to estimate the study NYHA value, which was defined by a weighted mean. If the NYHA distributions were not available in a study then we assumed that the patients were uniformly distributed.

Results

A total of 33 papers met the inclusion criteria, see Table 1. There were 26 studies (79%) which concerned tele-monitoring and 7 (21%) which concerned case management or nurse administered telephone-based management. The sample sizes varied from 20 to 502 patients (7530 in total) and the average age was 69 years. In most studies (85%) the follow-up period was 6–12 months, while in 2 studies participants were followed up for 24 months and in another 3 studies for less than 6 months.

Table 1

Study characteristics

	Country	Study duration, months	Sample size	Age, years (SD)	Males, %	NYHA	Physiological variables measured	Primary outcomes
Antonicelli 2008²⁴	Italy	12	57	78 (7)	61	II–IV	scale, ECG, symptoms, pulse, urine output	Mortality
Balk 2008²⁵	Netherlands	9.6	214	66	70	I–IV	scale, medication dispenser	Mortality
Blum 2006; Blum 2007^26,27	USA	24	204	72(9)	72	II–IV	scale, symptoms, pulse	Mortality
Boyne 2011²⁸	Netherlands	14	382	71.5 (11.2)	59	II–IV	scale, pulse	All-cause hospitalization
Capomolla 2004²⁹	Italy	11	133	57(10)	88	II–IV	scale, medication dispenser, pulse	Mortality, CHF hospitalization
Cleland 2005³⁰	Germany, Netherlands, UK	8	426	67	77	I–IV	scale, pulse	Mortality, CHF hospitalization, CHF LOS
Dar 2009³¹	UK	6	182	72	66	II–IV	scale, symptoms, pulse, SpO₂	CHF hospitalization, CHF LOS
de Lusignan 2001³²	UK	12	20	75.2	NA	II–IV	scale, symptoms, pulse	Mortality
Dendale 2012³³	Belgium	6	160	76 (10)	65	NA	scale, pulse	Mortality, CHF LOS
Giordano 2009³⁴	Italy	12	460	57(10)	85	II–IV	scale, ECG, medication dispenser, symptoms	Mortality, CHF hospitalization
Goldberg 2003³⁵	USA	6	280	59	68	III–IV	scale, symptoms	Mortality
Jerant 2001³⁶	USA	6	25	69.5	48	II–IV	scale, symptoms	CHF hospitalization, CHF LOS
Jolly 2009³⁷	UK	12	169	67.96	75	I–IV	scale	mortality, CHF hospitalization, CHF LOS
Kashem 2008³⁸	USA	12	48	53.5	74	II–III	scale, symptoms, pulse	Mortality, CHF hospitalization
Kielblock 2007³⁹	Germany	12	502	74	51	NA	scale	Mortality, CHF hospitalization
Koehler 2012⁴⁰	Germany	24	710	66.9	81	III–IV	scale, ECG	Mortality, CHF hospitalization
Lyngå 2012⁴¹	Sweden	12	344	73 (10.2)	75	III–IV	scale	Mortality, CHF hospitalization
Mortara 2009⁴²	UK, Poland, Italy	12	461	60 (11)	85	II–IV	scale, ECG, symptoms, pulse	Mortality, CHF hospitalization
Scherr 2009⁴³	Austria	6	120	78.1	69	II–IV	scale, medication dispenser, pulse	Mortality, CHF LOS
Schwarz 2008⁴⁴	USA	3	102	78.1	48	II–IV	scale, symptoms	Mortality
Soran 2008⁴⁵	USA	6	315	76	35	II–III	scale, symptoms	Mortality
Villani 2007⁴⁶	Italy	12	120	69	75	II–III	scale, symptoms, pulse, urine output	Mortality, CHF hospitalization
Wade 2011⁴⁷	USA	6	316	76.7	52	NA	scale, symptoms	Mortality, CHF hospitalization
Weintraub 2010⁴⁸	USA	3	188	69	66	I–IV	scale, symptoms, pulse	Mortality, CHF hospitalization
Woodend 2008⁴⁹	Canada	12	121	68	74	II–IV	scale, ECG, symptoms	Mortality
Zugck 2008⁵⁰	Germany	12	88	58.1 (10.2)	82	II–IV	scale, ECG, SpO₂	Mortality, CHF hospitalization
Blue 2001⁵¹	UK	12	165	75.0	48	II–IV	case management	Mortality, CHF hospitalization, CHF LOS
Dunagan 2005⁵²	USA	12	151	70.0	44	II–IV	nurse administered telephone based management programme	Mortality, CHF hospitalization
Kasper 2002⁵³	USA	6	200	61.9	61	II–III	telephone support	Mortality, CHF hospitalization
Krumholz 2002⁵⁴	USA	12	88	73.8	57	NA	case management	mortality, CHF hospitalization, CHF LOS
Laramee 2003⁵⁵	USA	3	287	70.7 (11.8)	54	I–IV	case management programme	CHF hospitalization
Riegel 2002⁵⁶	USA	6	358	75.0 (12)	49	II–IV	education management	CHF hospitalization, CHF LOS
Riegel 2006⁵⁷	USA	6	134	72.1 (11)	46	II–IV	telephone case management	CHF hospitalization, CHF LOS

The baseline trial characteristics were similar among the three outcome groups, see Table 2. Study publication dates were almost equally distributed during the ten year period, and most research took place in Europe or the US (Figure 2 and 3). In most of these studies, the CHF diagnosis had already been established. The patients had had one or more CHF-related hospital admissions in the previous year before the study was conducted. This criterion excluded low-risk patients. The standard of CHF care in the home refers to tests on the left ventricle of the heart, an ACE inhibitor, instructions on how to care for patients at home or advice on how to stop smoking.²¹ Tele-monitoring devices were used to collect several types of physiological data (Figure 4). While all studies had recorded scales (bodyweight, BP, or both), none included a fall detector. More than a half of the studies recorded symptoms and pulse (heart rate) but less than 10% detected patients’ SpO₂ or urine output. Finally, 27 papers (82%) reported all-cause mortality, 23 (70%) papers reported CHF-related hospitalization and 10 (31%) reported the CHF-related LOS (Table 2).

Figure 2

Publications

Figure 3

Countries of publications

Figure 4

Variables monitored in the tele-monitoring programmes

Table 2

Baseline characteristics

Clinical outcome	Number of studies	Average trial duration (months)	Total number of patients	Average age (years)	Average percentage males	Average NYHA grade
Mortality	27	10.2	6162	68.0	68	2.7
CHF hospitalization	23	9.6	5675	68.2	65	2.7
CHF LOS	10	8	1960	72.8	58	3.1

Meta-analysis

The meta-analysis showed that telehealth was associated with a significantly lower risk of deaths (RR 0.76, 95% CI 0.66 to 0.88, P = 0.001) compared with usual care (Figure 5). No significant heterogeneity between studies was detected according to the Cochran Q test (P = 0.49) so the fixed model was used (Figure 5). Studies were distributed symmetrically around the mean effect size based on visual inspection on the funnel plots in Figure 8, which indicated that the sampling error was random, i.e. no publication bias was observed. Meanwhile, positive effects on CHF hospitalization were found (RR 0.72, 95% CI 0.61 to 0.85, P <0.001) according to the random effect model because significant heterogeneity (66%, P < 0.001) was present (Figure 6 and Table 3). In addition, the telehealth group had shorter CHF-related LOS (MD -1.4 days, 95% CI -2.4 to -0.4, P = 0.007) with extremely high heterogeneity among studies (71%, P < 0.001), but still without losing the significance in the random effect model (Figure 7 and Table 3). No publication bias was found on CHF hospitalization or CHF LOS according to the symmetric funnel plots in Figure 8.

Figure 5

Forest plot for all-cause mortality

Figure 6

Forest plot for CHF hospitalization

Figure 7

Forest plot for CHF LOS

Figure 8

Publication bias for the three outcomes

Table 3

Meta-analysis summaries

	Measure	Model	Overall	95% CI	P-value	Heterogeneity (I-squared)	Q (p)	Publication bias	Effectiveness
Mortality	RR	FE	0.76	0.66, 0.88	<0.001	18.27%[a]	25.4 (0.49)[b]	No	24% reduction
CHF hospitalization	RR	RE (REML)	0.72	0.61, 0.85	<0.001	66.26%[c]	61.8(<0.001)[d]	No	28% reduction
CHF LOS	MD	RE (REML)	-1.41	-2.43, -0.39	0.007	71.32%	38.6 (<0.001)	No	1.41 day reduction

[a]

I-squared <20% indicated small heterogeneity

[b]

Q test was not significant so that no significant heterogeneity was presented; fixed effect model was able to be used

[c]

I-squared >20% indicated high heterogeneity

[d]

Q test was significant so that there was significant heterogeneity among studies; random effect model had to be used

In a sensitivity analysis, there was similar effectiveness of tele-monitoring and telephone-assisted case management, see Table 4. However, case-management did not have significantly better effects on reducing mortality or CHF-related LOS compared with usual care, but did have a significantly reduced rate on CHF hospitalization (RR: 0.66, P<0.001).

Table 4

Sensitivity analysis

Outcome	Measure	Model	Tele-monitoring vs. usual care			Case-management vs. usual care
Outcome	Measure	Model	Overall	95% CI	P-value	Overall	95% Cl	P-value
Mortality	RR	FE	0.74	0.61, 0.89	0.002	0.86	0.63, 1.19	0.37
CHF hospitalization	RR	RE (REML)	0.74	0.58, 0.95	0.016	0.66	0.55, 0.78	<0.001
CHF LOS	MD	RE (REML)	-1.41	-2.79, -0.03	0.045	-0.7	-2.52, 1.11	0.45

In summary, telehealth had a significant overall effect on reducing all-cause mortality, CHF hospitalization and CHF LOS. There was a mean decrease in patient mortality of 24%, CHF hospitalization of 28% and CHF LOS of 1.4 days. To our knowledge, the present study is the first meta-analysis which attempts to find the relationship between effectiveness of telehealth programmes and the types of physiological information recorded.

All-cause mortality

Residual heterogeneity was tested to see if the effect sizes of studies were identical. The Q-value (Q = 18.0) indicated that there was no significant residual heterogeneity after the influence of duration was taken into consideration for mortality. (The Q-value is a measure of weighted squared deviations to test residual heterogeneity in meta-analysis.¹⁶) The fixed effect model was therefore used. The regression model of the mortality risk ratio was: exp(-0.69 + 0.03x), where x was the study duration in months.

According to the model the 95% CI of the regression coefficient was 0.0087 to 0.0539 (P = 0.007). Therefore it was significant. An interpretation could be that the log relative risk of mortality increased significantly when the study duration increased, in other words, the longer the duration, the less effective telehealth was. The RR for duration x was exp(-0.69 + 0.03x), and so the first month's RR could reach 0.52. This suggests that telehealth programmes decreased patient mortality by 48% for the first month compared with usual care.

The increasing trend of mortality RR and the decreasing trend of effect in a one-year telehealth period based on the results of fixed effect regression model are shown in Figure 9.

Figure 9

Variation of all-cause mortality effectiveness in one year

CHF hospitalization

The test for residual heterogeneity had Q = 60.6 (P<0.001), so that residual heterogeneity was significant. Duration did not explain the heterogeneity in the dataset. The random effect model for RR was exp(-0.41 + 0.0002x), where x indicated study duration in months. The random effect model provided non-significant results for the regression coefficient. Thus we concluded that there was no significant difference in CHF hospitalization due to the different follow-up durations. Therefore it was appropriate for us to use overall RR as the RR given any duration for CHF hospitalization.

CHF LOS

The test for residual heterogeneity had Q = 34.4 (P<0.001) so that residual heterogeneity remained significant after considering the duration when modelling CHF LOS. Similarly the significant residual sums of squares indicated that duration did not explain all of the heterogeneity in the dataset. As a result, the random effect model was appropriate. The random effect model of MD was -1.37–0.01x, where x was the study follow-up duration in months. However, it did not provide a significant result for the regression coefficient. Thus there was no significant difference in CHF LOS among different follow-up durations. We were able to use a constant effect size (the overall MD) as a reduction for CHF LOS given any follow up duration. The meta-regression results are summarised in Table 5.

Table 5

Meta-regression on follow-up duration

Clinical outcome	Dependent variable	Independent variable	Model	Beta	95% CI	P-value	Residual heterogeneity	Effect size given follow up duration x (in months)
Mortality	RR	Duration	FE	0.03	0.009, 0.054	0.007	No	exp(-0.69 + 0.03x)
CHF hospitalization	RR	Duration	RE	0.0002	-0.037, 0.038	0.99	Yes	0.72
CHF LOS	MD	Duration	RE	-0.01	-0.454, 0.441	0.98	Yes	-1.41

Age, NYHA and variables monitored

We applied the fixed effect model for mortality, but the random effect models to analyse CHF hospitalization and LOS based on the heterogeneity among the studies. The test for residual heterogeneity of meta-regression models on mortality was not significant. However residual heterogeneity did exist for meta-regression models on CHF hospitalization and LOS. Only the meta-regression models with at least one significant predictor are listed in Table 6. Age was the only significant covariate for CHF hospitalization but not for mortality or CHF LOS. Telehealth was less effective in older people. NYHA was a significant covariate in predicting mortality and CHF hospitalization. The results indicated that the higher the NYHA, the lower the risk ratio. Telehealth had a larger effectiveness in reducing mortality and CHF hospitalization when patients had higher NYHA categories. In addition, recording questionnaires (symptoms) or pulse (heart ratio) information significantly decreased the risk ratio for mortality and CHF hospitalization. It was surprising that ECG had significant negative effectiveness on CHF hospitalization, which might be the result of data bias or some clinical reason. There was no significant covariate for CHF LOS.

Table 6

Meta-regression models with significant predictors

Clinical outcome	Dependent variable	Independent variables	Model	Residual heterogeneity	Effect size for significant predictors[a]
All-cause mortality	RR	Age NYHA Variables monitored	FE	No	exp(0.93–0.003A - 0.33N - 0.42S) exp(2.44–0.001A - 0.98N - 0.51P*)
CHF hospitalization	RR	Age NYHA Variables monitored	RE	Yes	exp(-3.59 + 0.06A* - 0.48N + 0.54E) exp(-2.1 7 + 0.04A - 0.26N - 0.24M) exp(-2.15 + 0.04A - 0.44N - 0.1 6S) exp(-1.7 + 0.05A - 0.72N* - 0.57P*)
CHF LOS	MD	Age NYHA Variables monitored	RE	Yes	No model with significant predictors

[a]

Predictors: A, age N, NYHA E, detected ECG in telehealth or not (1 = yes, 0 = no) M, detected Medication in telehealth or not S, detected Symptoms in telehealth or not P, detected Pulse in telehealth or not

significant at P<0.05

In summary, after controlling for age and NYHA, adding a questionnaire (symptoms) could significantly reduce all-cause mortality by 34% and CHF hospitalization by 15%; adding a pulse (heart rate) detector could significantly reduce all-cause mortality by 40% and CHF hospitalization by 43%, see Table 7.

Table 7

Summary of meta-regression on age, NYHA and variables monitored

Clinical outcome	Model	Outcome	Covariates	Number of models with significant covariates	Significant monitored covariates	Coefficients	Impact on effect size
All-cause Mortality	FE	RR	Age NYHA Variables monitored	2	Symptoms Pulse	0.66 [a] 0.6 [b]	34% reduction 40% reduction
CHF Hospitalization	RE (REML)	RR	Age NYHA Variables monitored	4	ECG Symptoms Pulse	1.72 [c] 0.85 [d] 0.57 [e]	72% higher 15% reduction 43% reduction
CHF LOS	RE (REML)	MD	Age NYHA Variables monitored	0	N/A	N/A	N/A

[a]

Calculated by exp(-0.42)

[b]

Calculated by exp(-0.51)

[c]

Calculated by exp(0.54)

[d]

Calculated by exp(-0.16)

[e]

Calculated by exp(-0.57)

Discussion

The present study demonstrates that telehealth significantly reduces all-cause mortality by an average of 24%, CHF-related hospitalization by 28% and CHF-related LOS by 1.4 days. The meta-regression shows that the reduction of all-cause mortality was even more pronounced during the early months of follow-up. A larger effect on all-cause mortality and CHF hospitalization was observed for patients in the higher NYHA categories. In addition, the benefit was greater when symptoms or heart rate were recorded by telehealth.

With the inclusion of the latest published telehealth RCTs for CHF patients, we investigated the effect of telehealth on three treatment outcomes: all-cause mortality, CHF-related hospitalization and CHF-related LOS. Three types of meta-analysis were employed: FE, MH (a special type of FE) and RE (REML). The appropriate method was selected based on the heterogeneity among the trials. For example, RE (REML) was used to model CHF-related hospitalization and CHF-related LOS; while all-cause mortality was modelled by MH and FE. The different meta-analysis methods allowed a fair comparison between our results and those published in previous meta-analyses: we verified that the results produced by the same method using the same set of trials were the same. In addition, all three meta-analysis methods yielded consistent evidence on the clinical effectiveness of telehealth programmes, which reduced the risk of bias induced by the method itself. There did not appear to be a significant effect on effectiveness depending on the geography of the publication. The results confirmed, extended and updated previous systematic meta-analyses.^1,9,10,22

Moreover, our meta-regression analysis examined the relationships between clinical effectiveness and trial durations using meta-regression models. These showed that the trial duration negatively influenced the benefit on all-cause mortality but had no significant effect on the CHF-related hospitalization or LOS. The lack of significance in the latter case might be explained by papers with documented outcome on hospitalization and LOS having very similar trial durations, so they did not generate meaningful regression coefficients.

Our meta-regression analysis also attempted to answer questions such as which physiological variables monitored would be more effective. For instance, we found that adding questionnaires (symptom monitoring) or a pulse detector significantly enhanced the benefits of telehealth with respect to all-cause mortality and CHF hospitalization. Although the results showed that adding ECG had a negative effectiveness, this might have been because patients equipped with ECG suffered from more serious CHF. These results may be helpful in designing a telehealth programme to produce the greatest benefits. However, due to lack of statistical power for certain meta-regressions we are not able to draw a general conclusion. More evidence is needed to ascertain the correlations in certain covariates. Future work could include more well-designed cohort studies to increase sample size and possibly higher statistical power to detect significant differences.²³

References

Maric

, Kaan

, Ignaszewski

, Lear

SA.

A systematic review of telemonitoring technologies in heart failure. Eur J Heart Fail 2009; 11: 506–17.

Giamouzis

, Triposkiadis

, Butler

, Westermann

, Giannakoulas

Heart failure. Cardiol Res Pract 2011; 2011: 159608

Giamouzis

, Kalogeropoulos

, Georgiopoulou

Hospitalization epidemic in patients with heart failure: risk factors, risk prediction, knowledge gaps, and future directions. J Card Fail 2011; 17: 54–75.

Mujib

, Giamouzis

, Agha

SA.

Epidemiology of stroke in chronic heart failure patients with normal sinus rhythm: findings from the DIG stroke sub-study. Int J Cardiol 2010; 144: 389–93.

Giamouzis

, Mastrogiannis

, Koutrakis

Telemonitoring in chronic heart failure: a systematic review. Cardiol Res Pract 2012; 2012: 410820

Klersy

, De Silvestri

, Gabutti

Economic impact of remote patient monitoring: an integrated economic model derived from a meta-analysis of randomized controlled trials in heart failure. Eur J Heart Fail 2011; 13: 450–9.

Mueller

, Vuckovic

, Knox

, Williams

RE.

Telemanagement of heart failure: a diuretic treatment algorithm for advanced practice nurses. Heart Lung 2002; 31: 340–7.

Kleinpell

, Avitall

Telemanagement in chronic heart failure: a review. Dis Manage Health Outcomes 2005; 13: 43–52.

Clarke

, Shah

, Sharma

Systematic review of studies on telemonitoring of patients with congestive heart failure: a meta-analysis. J Telemed Telecare 2011; 17: 7–14.

10.

Klersy

, De Silvestri

, Gabutti

, Regoli

, Auricchio

A meta-analysis of remote monitoring of heart failure patients. J Am Coll Cardiol 2009; 54: 1683–94.

11.

Polisena

, Tran

, Cimon

Home telemonitoring for congestive heart failure: a systematic review and meta-analysis. J Telemed Telecare 2010; 16: 68–76.

12.

Jacobus

, Lichtenstein

The benefits of remote patient monitoring. In: Business Briefing: Global Healthcare 2003. London: Business Briefings Ltd, 2004

13.

Martinez

, Everss

, Rojo-Alvarez

, Figal

, García-Alberola

A systematic review of the literature on home monitoring for patients with heart failure. J Telemed Telecare 2006; 12: 234–41.

14.

Petitti

DB.

Meta-Analysis, Decision Analysis, and Cost-Effectiveness Analysis. Oxford: Oxford University Press, 2000

15.

Viechtbauer

Bias and efficiency of meta-analytic variance estimators in the random-effects model. J Educ Behav Stat 2005; 30: 261–293.

16.

Borenstein

, Hedges

, Higgins

JPT

, Rothstein

HR.

Introduction to Meta-Analysis. Wiley-Blackwell, 2009

17.

Viechtbauer

Conducting meta-analyses in R with the metafor package. J Stat Softw 2010; 36: 1–48.

18.

Cowie

, Wood

, Coats

AJ.

Survival of patients with a new diagnosis of heart failure: a population based study. Heart 2000; 83: 505–10.

19.

Horwich

, Fonarow

, Hamilton

, MacLellan

, Borenstein

Anemia is associated with worse symptoms, greater impairment in functional capacity and a significant increase in mortality in patients with advanced heart failure. J Am Coll Cardiol 2002; 39: 1780–6.

20.

Lee

, Austin

, Rouleau

, Liu

, Naimark

, Tu

JV.

Predicting mortality among patients hospitalized for heart failure: derivation and validation of a clinical model. JAMA 2003; 290: 2581–7.

21.

Maryland Health Care Commission. Quality Care for Congestive Heart Failure. See http://mhcc.maryland.gov/consumerinfo/hospitalguide/patients/technical_information/chf_care.html (last checked 28 May 2013)

22.

Clark

, Inglis

, McAlister

, Cleland

, Stewart

Telemonitoring or structured telephone support programmes for patients with chronic heart failure: systematic review and meta-analysis. BMJ 2007; 334: 942

23.

Cohn

, Becker

BJ.

How meta-analysis increases statistical power. Psychol Methods 2003; 8: 243–53.

24.

Antonicelli

, Testarmata

, Spazzafumo

Impact of telemonitoring at home on the management of elderly patients with congestive heart failure. J Telemed Telecare 2008; 14: 300–5.

25.

Balk

, Davidse

, Dommelen

. Tele-guidance of chronic heart failure patients enhances knowledge about the disease. A multi-centre, randomised controlled study. Eur J Heart Fail 2008; 10: 1136–42.

26.

Blum

, Janowick

, Gottlieb

SS.

One year changes in quality of life for heart failure patients in a home telemonitoring program. J Card Fail 2006; 12: S122

27.

Blum

, Gottlieb

SS.

Morbidity and mortality benefits of reliable instrumental support. J Card Fail 2007; 13(Suppl. 2): S164

28.

Boyne

, Vrijhoef

, Nieman

FH.

Telemonitoring in patients with heart failure: results from a multicenter randomized controlled trial (the TEHAF study). J Am College Cardiol 2011; 57: E389

29.

Capomolla

, Pinna

, La Rovere

MT.

Heart failure case disease management program: a pilot study of home telemonitoring versus usual care. Eur Heart J Suppl 2004; 6(Suppl. F): F91–F98

30.

Cleland

, Louis

, Rigby

AS.

Noninvasive home telemonitoring for patients with heart failure at high risk of recurrent admission and death: the Trans-European Network-Home-Care Management System (TEN-HMS) study. J Am Coll Cardiol 2005; 45: 1654–64.

31.

Dar

, Riley

, Chapman

A randomized trial of home telemonitoring in a typical elderly heart failure population in North West London: results of the Home-HF study. Eur J Heart Fail 2009; 11: 319–25.

32.

de Lusignan

, Wells

, Johnson

, Meredith

, Leatham

Compliance and effectiveness of 1 year's home telemonitoring. The report of a pilot study of patients with chronic heart failure. Eur J Heart Fail 2001; 3: 723–30.

33.

Dendale

, De Keulenaer

, Troisfontaines

Effect of a telemonitoring-facilitated collaboration between general practitioner and heart failure clinic on mortality and rehospitalization rates in severe heart failure: the TEMA-HF 1 (Telemonitoring in the Management of Heart Failure) study. Eur J Heart Fail 2012; 14: 333–40.

34.

Giordano

, Scalvini

, Zanelli

Multicenter randomised trial on home-based telemanagement to prevent hospital readmission of patients with chronic heart failure. Int J Cardiol 2009; 131: 192–9.

35.

Goldberg

, Piette

, Walsh

MN.

Randomized trial of a daily electronic home monitoring system in patients with advanced heart failure: the Weight Monitoring in Heart Failure (WHARF) trial. Am Heart] 2003; 146: 705–12.

36.

Jerant

, Azari

, Nesbitt

TS.

Reducing the cost of frequent hospital admissions for congestive heart failure: a randomized trial of a home telecare intervention. Med Care 2001; 39: 1234–45.

37.

Jolly

, Taylor

, Lip

GY.

A randomized trial of the addition of home-based exercise to specialist heart failure nurse care: the Birmingham Rehabilitation Uptake Maximisation study for patients with Congestive Heart Failure (BRUM-CHF) study. Eur J Heart Fail 2009; 11: 205–13.

38.

Kashem

, Droogan

, Santamore

, Wald

, Bove

AA.

Managing heart failure care using an internet-based telemedicine system. J Card Fail 2008; 14: 121–6.

39.

Kielblock

, Frye

, Kottmair

, Hudler

, Siegmund-Schultze

, Middeke

Einfluss einer telemedizinisch unterstützten Betreuung auf Gesamtbehandlungskosten und Mortalität bei chronischer Herzinsuffizienz. Dtsch med Wochenschr 2007; 132: 417–22 [German]

40.

Koehler

, Winkler

, Schieber

Telemedicine in heart failure: pre-specified and exploratory subgroup analyses from the TIM-HF trial. Int J Cardiol 2012; 161: 143–50.

41.

Lyngå

, Persson

, Hägg-Martinell

Weight monitoring in patients with severe heart failure (WISH). A randomized controlled trial. Eur J Heart Fail 2012; 14: 438–44.

42.

Mortara

, Pinna

, Johnson

Home telemonitoring in heart failure patients: the HHH study (Home or Hospital in Heart Failure). Eur J Heart Fail 2009; 11: 312–8.

43.

Scherr

, Kastner

, Kollmann

Effect of home-based telemonitoring using mobile phone technology on the outcome of heart failure patients after an episode of acute decompensation: randomized controlled trial. J Med Internet Res 2009; 11: e34

44.

Schwarz

, Mion

, Hudock

, Litman

Telemonitoring of heart failure patients and their caregivers: a pilot randomized controlled trial. Prog Cardiovasc Nurs 2008; 23: 18–26.

45.

Soran

, Piña

, Lamas

GA.

A randomized clinical trial of the clinical effects of enhanced heart failure monitoring using a computer-based telephonic monitoring system in older minorities and women. J Card Fail 2008; 14: 711–7.

46.

Villani

, Malfatto

, Delia Rosa

Disease management for heart failure patients: role of wireless technologies for telemedicine. The ICAROS project. G Ital Cardiol (Rome) 2007; 8: 107–14 [Italian]

47.

Wade

, Desai

, Spettell

CM.

Telemonitoring with case management for seniors with heart failure. Am J Manag Care 2011; 17: e71–9

48.

Weintraub

, Gregory

, Patel

AR.

A multicenter randomized controlled evaluation of automated home monitoring and telephonic disease management in patients recently hospitalized for congestive heart failure: the SPAN-CHF II trial. J Card Fail 2010; 16: 285–92.

49.

Woodend

, Sherrard

, Fraser

, Stuewe

, Cheung

, Struthers

Telehome monitoring in patients with cardiac disease who are at high risk of readmission. Heart Lung 2008; 37: 36–45.

50.

Zugck

, Frankenstein

, Nelles

Telemedicine reduces hospitalisation rates in patients with chronic heart failure - results of the randomized HiTel trial. Eur J Heart Fail Suppl 2008; 7(Suppl. 1): 9

51.

Blue

, Lang

, McMurray

JJ.

Randomised controlled trial of specialist nurse intervention in heart failure. BMJ 2001; 323: 715–8.

52.

Dunagan

, Littenberg

, Ewald

GA.

Randomized trial of a nurse-administered, telephone-based disease management program for patients with heart failure. J Card Fail 2005; 11: 358–65.

53.

Kasper

, Gerstenblith

, Hefter

A randomized trial of the efficacy of multidisciplinary care in heart failure outpatients at high risk of hospital readmission. J Am Coll Cardiol 2002; 39: 471–80.

54.

Krumholz

, Amatruda

, Smith

GL.

Randomized trial of an education and support intervention to prevent readmission of patients with heart failure. J Am Coll Cardiol 2002; 39: 83–9.

55.

Laramee

, Levinsky

, Sargent

, Ross

, Callas

Case management in a heterogeneous congestive heart failure population: a randomized controlled trial. Arch Intern Med 2003; 163: 809–17.

56.

Riegel

, Carlson

, Kopp

, LePetri

, Glaser

, Unger

Effect of a standardized nurse case-management telephone intervention on resource use in patients with chronic heart failure. Arch Intern Med 2002; 162: 705–12.

57.

Riegel

, Carlson

, Glaser

, Romero

Randomized controlled trial of telephone case management in Hispanics of Mexican origin with heart failure. J Card Fail 2006; 12: 211–9.