OUP user menu

The effectiveness of disease management programmes in reducing hospital re-admission in older patients with heart failure: a systematic review and meta-analysis of published reports

Jonás Gonseth , Pilar Guallar-Castillón , José R. Banegas , Fernando Rodríguez-Artalejo
DOI: http://dx.doi.org/10.1016/j.ehj.2004.04.022 1570-1595 First published online: 2 September 2004

Abstract

Aims To systematically evaluate the published evidence regarding the effectiveness of disease management programmes (DMPs) reducing hospital re-admissions among elderly patients with heart failure (HF).

Methods and Results Computerised search of MEDLINE (1966 to 31 August 2003) and EMBASE (1966 to 31 August 2003). The Cochrane Library was also searched, and reference lists of review articles on the topic, and of all relevant studies identified, were scanned. Search and selection of studies, data-extraction using standardised forms, and assessment of study quality was performed by two reviewers. The end-point was the proportion of persons who underwent hospital re-admission, and pooled relative risks (RR) were used to summarise the effectiveness of DMPs. The meta-analysis included 54 articles, comprising 27 randomised and 27 non-randomised controlled studies. Randomised studies consistently suggested that, in comparison with usual care, DMP reduced the frequency of re-admission for HF or cardiovascular disease by 30% (pooled RR 0.70; confidence interval (CI) 95% 0.62–0.79), all-cause re-admission by 12% (pooled RR 0.88, 95% CI: 0.79–0.97), and the combined event of re-admission or death by 18% (pooled RR 0.82, 95% CI: 0.72–0.94). The results displayed no substantial variation when only DMPs with home visits, out-patient visits to a clinic, or patient follow-up longer than 6 months were included. For DMPs with out-patient clinical visits, however, the reduction in re-admission for HF or cardiovascular disease, and for all causes, did not attain statistical significance. The magnitude of DMP benefits reported by non-randomised studies was more than double that reported by randomised studies. Practically all the non-randomised studies failed to control for confounding factors, such as severity, co-morbidity and drug therapy.

Conclusion DMPs are effective at reducing re-admissions among elderly patients with HF. Their effectiveness is close to that observed in clinical trials evaluating drugs for HF, such as angiotensin-converting enzyme inhibitors, beta-blockers or digoxin. However, since none of the DMP studies compared different interventions directly, we do not know the relative effectiveness of types of healthcare delivery within the DMP.

  • Heart failure
  • Meta-analysis
  • Re-admission
  • Elderly
  • Disease management programmes

See page 1565 for the editorial comment on this article1

Introduction

Heart failure (HF) is the leading cause of hospitalisation among the elderly in developed countries.1,2 Despite effective treatments to reduce hospitalisations, such as angiotensin-converting enzyme (ACE) inhibitors or beta-blockers, clinical management of these patients remains sub-optimal;3,4 as a result of this and of the natural history of HF, hospital re-admissions are very frequent. Depending on their age and HF stage, 10–50% of patients are readmitted in the 3–6 months following index hospitalisation.5–8 Hospitalisations represent the main cost component of HF care.9

Elrodt et al.,10 defined disease management as a multi-disciplinary approach to care for chronic diseases, that co-ordinates comprehensive care along the disease continuum across healthcare delivery systems. Several systematic reviews have shown that disease management programmes (DMPs) are a potentially useful instrument for reducing hospital re-admissions of HF-patients.11–13 Although DMPs for patients with HF have several components, such programmes always include patient education and support in order to detect signs of decompensation and improve adherence to treatment. Most DMPs rely on nurse participation and include telephone follow-up of patients, yet there is limited knowledge of the influence of the type of healthcare delivery (home medical visits or out-patient visits to a clinic) or the duration of follow-up on the results of DMPs.14 Many re-hospitalisations of patients with HF are due to causes other than HF.14 The possible benefits of DMPs in any given area, such as HF, may be accompanied by unexpected effects in other areas, since most HF patients are at an advanced age and suffer from serious co-morbidities. It is important, therefore, to ascertain the effectiveness of DMPs on hospitalisations due to cardiovascular causes other than HF and, in particular, on all-cause hospitalisations and death. Lastly, it should be noted that a substantial proportion of DMPs in HF have been assessed using non-randomised before-and-after comparisons.

The aim of this paper was to systematically review the experimental evaluations of DMPs among elderly patients with HF and to ascertain the effectiveness of such programmes on three variables: hospital re-admission for HF or other cardiovascular causes, all-cause re-admission and re-admission or death. In addition, this paper examined the influence of healthcare delivery and of the type of DMP assessment (randomised or non-randomised) on the results of DMPs.

Methods

Search of the literature

A search of MEDLINE (1966 to 31 August 2003) and EMBASE (1966 to 31 August 2003) was undertaken. Because there are numerous terms to name DMP and their use is not standardised in the literature, a very sensitive strategy was pursued to identify the greatest number of papers. The following textual terms and MeSH headings were used: cardiac failure or heart failure was combined with re-admission or re-hospitalisation or hospitalisation or discharge. Language restrictions were not applied. We also searched the Cochrane Library and scanned reference lists of review articles on the topic and of the relevant studies identified. For very recent studies, we completed the information from congress proceedings.

Selection of studies and extraction of data

Studies were included if they met the following criteria: randomised and non-randomised controlled studies assessing DMPs targeted, among others, at patients aged ⩾65 years with principal or secondary diagnosis of HF. Studies were subsequently excluded if: they failed to furnish information to quantify the association between DMPs and re-admission, no evidence of, or Math75% of, hospitalisations were re-admissions, Math75% of subjects had a diagnosis of HF, Math50% of patients were aged over 60 years, or the mean age of patients studied was Math60 years; the data supplied were partially reported in another paper already included.

As used by Weingarten et al.,15 our working definition of DMP was `an intervention designed to manage HF and reduce hospital re-admissions using a systematic approach to care and potentially employing multiple treatment modalities'. To identify a systematic approach to care, we searched for keywords such as guidelines, clinical pathways, protocols, algorithms, care plans, quality improvement activities, and patient support and education.15 Indeed, we expected that all DMPs would include teaching and support activities addressed to patients.

For randomised studies, study quality was assessed with the scale developed by Jadad et al.,16 while for non-randomised studies quality was approximated by the degree of control for confounding factors. The Jadad scale assesses three aspects, namely, randomisation procedure, blinding of the intervention, and patient attrition in the follow-up. The authors advocate that a score of 3 or more on the Jadad scale should be taken to indicate “high quality”.16

The search and selection of studies, data-extraction (using standardised forms), and assessment of study quality was performed by two reviewers (JG and PGC): disagreements were discussed with a third reviewer (FRA) and settled by consensus.

Statistical analysis

Randomised and non-randomised studies were examined separately. We chose as end-point the proportion of persons readmitted over the follow-up in the DMP and control groups. Studies were eligible for meta-analysis if such an end-point could be ascertained from the published reports. For each study, relative risk (RR) was calculated as the measure of the effect of the intervention. Heterogeneity of the RR across the studies was tested using the Math test. Where the results were homogeneous, RRs were combined using the Mantel–Haenszel fixed-effects model.17 Where the heterogeneity test was statistically significant but the RRs were nevertheless consistent in their direction (above or below 1), these were combined using Der Simonian and Laird's random effects method.17

We produced funnel plots (scatter plots of DMP effects against their standard error across the studies), and looked for the presence of asymmetry as a means of examining the `small studies effect', including the potential for publication bias. We tested the asymmetry of funnel plots with the method of Egger et al.18

We conducted sensitivity analyses, defined a priori, to examine whether the results of the meta-analysis varied with types of organising care within the DMP (home visits, out-patient visits to a clinic, telephone follow-up), duration of follow-up (longer than 6 months), Jadad score, and study size.

All statistical tests were two-sided. The meta-analysis was performed with the RevMan programmeme (Review Manager Version 4.2. Oxford, England: The Cochrane Collaboration, 2000) and the method of Egger et al., was implemented with Stata (Statistics Data Analysis v 7.0. Texas: Stata Corporation, 2001).

Results

Studies included and excluded

We identified 5324 papers in EMBASE, 2735 in MEDLINE, and 20 from other sources, chiefly secondary references. After reading the title or abstract of all papers identified, the complete text of 286 was retrieved. Of these, 82 fulfilled the inclusion criteria, but 28 were excluded19–46 (Table 1 ). Hence, this review included a total of 54 studies, comprising 27 randomised47–74 and 27 non-randomised trials.75–101

View this table:
Table 1

Excluded articles

Randomised trials

The studies were published from 1993 through 2003 (Table 2 ). Of the 27 studies, 13 came from the United States of America. Study size ranged from 34 to 1966, with half of studies not exceeding 200 subjects. Moreover, though the study by Hughes et al.,61 enrolled 1966 patients, only 30 had HF. Similarly, of the 363 patients included in the study of Naylor et al.,64 only 60 had HF. Although the success of randomisation among HF subjects was not tested in these two studies, they were nevertheless included since they provided a breakdown of the results for this type of patient. Thus, we decided to consider them as `independent substudies' specific for heart failure patients. Only 11 studies47,50,52,53,58,59,63,64,66,69,73 attained a score of 3 on the Jadad scale.

View this table:
Table 2

Randomised trials evaluating the effect of disease management programmes on hospital re-admission of older patients with heart failure

With regards to patient characteristics, an average of only 26% (range 5–67%) of initially screened patients was included in the trials. Among the included patients the mean age was over 70 years in most studies. Patients presented with ejection fractions Math40% in 13 studies, and serious co-morbidity was frequent, principally acute myocardial infarction, arterial hypertension, heart valve disease and diabetes. The majority of patients received ACE-Inhibitor (range 22–97%) and diuretics (range 70–100%); digoxin (range 3–97%) and beta-adrenoceptor antagonists (range 10–63%) were also given to a lesser extent.

Interventions had an important patient education and support component; in many cases, subjects were informed about the nature of HF, alerted to signs of early decompensation, and taught disease management (Table 2). Moreover, in all but 4 studies49,61,62,70 mention was made of the participation of nurses in the administration or co-ordination of the intervention, and in all but 3 studies,51,62,70 mention was made of the use of telephone calls as part of the intervention. However, interventions were more variable with regard to place of commencement (in the hospital or after discharge), type of organising care (home visit or out-patient visit to clinic), and duration (ranging from a single home visit or duration of a single week to interventions lasting 12 months).

As the end-point was hospital re-admission, a simple variable obtainable even without contact between investigators and patients (e.g., from administrative sources), loss to follow-up was very small. In all studies the loss was below 5%, with the exception of the study by Laramee where it was 8%.49 Follow-up of subjects tended to be from 3 to 12 months, though the study by Stewart and Horowitz,57 reported a median 4.2-year follow-up. Results usually revealed a trend towards improvement in most of the end-points studied: percentage of re-admissions, number of re-admissions per patient and month, re-admission-free time, days of re-admission, etc. (Table 2).

Fig. 1 shows a forest plot of the studies eligible for meta-analysis. With respect to re-admission for HF or cardiovascular cause, six studies reported an homogeneous and significant reduction in re-admission47,48,54–56,58,60 (Fig. 1(a)). On the basis of the 3160 patients covered, the studies suggest that DMP reduces the frequency of re-admission for HF or cardiovascular cause by 30% (pooled RR 0.70; 95% confidence interval (CI) 0.62–0.79).

Fig. 1

Meta-analysis of the effectiveness of disease management programs on three outcomes among older patients with heart failure. Randomised studies. (a) Heart failure or cardiovascular readmission, (b) all-cause readmission, (c) readmission or death.

With regards to all-cause re-admission, only three studies48,56,73 reported a significant reduction (Fig. 1(b)). The results of the trials showed certain heterogeneity (Math), due mainly to the study by Weinberger et al.,72 which reported a statistically significant increase in re-hospitalisations (Fig. 1(b)). A random-effects model based on 4440 patients showed that DMP reduced all-cause re-admissions by 12% (pooled RR 0.88, 95% CI: 0.79–0.97). The study by Weinberger et al.,72 included patients with an extremely low quality of life; moreover, the intervention simply afforded patients enhanced access to primary-care services, without improving the healthcare delivery structure.102 Once we excluded this paper, because it dealt with a disease management programme with very limited treatment modalities, the homogeneity of the results across trials rose (Math) and the pooled RR, calculated for 3936 patients, was 0.85, 95% CI: 0.79–0.92. It should be highlighted that the study by Stewart et al.,57 is not shown in Fig. 2(b) because, despite being the only study with a follow-up of more than 4 years, it embraced subjects from two previous studies66,71 which had a six-month follow-up and had already been included in Fig. 1. The study by Stewart et al.,57 was nevertheless incorporated into the sensitivity analysis, with those studies having a follow-up of more than 6 months (see Table 3 below). An earlier study by Stewart et al.,67 was likewise not included, as it consisted of an 18-month follow-up of patients from another previous study.71

Fig. 2

Funnel plots for the three outcomes among older patients with heart failure. The vertical line in the graphs corresponds to the pooled relative risk across studies. (a) Heart failure or cardiovascular readmission, (b) all cause readmission, (c) readmission or death.

View this table:
Table 3

Effectiveness of a disease management program (DMP) in reducing hospital re-admissions or death in older patients with heart failure, by DMP characteristics and duration of follow-up

For the combined end-point of re-admission or death, four studies48,54,55,65 reported a statistically significant reduction although three of them had a very small size.54,55,65 Study results for this end-point were fairly heterogeneous (Math), although more so in terms of the magnitude of the effect measured than in terms of its direction (all studies but one52 registered an RR of less than 1) (Fig. 1(c)). Based on a total of 2985 patients, DMPs reduced the frequency of this end-point by 18% (pooled RR 0.82, 95% CI: 0.72–0.94). The study by McDonald et al.,60 was not included in this analysis because it consisted of a one-month follow-up of most of the patients who had been followed up over three months in another, already included, study by the same authors.55 The study by McDonald et al.,60 reported the elimination of 1-month hospital re-admission in both regular and multi-disciplinary care. The authors stated that this unexpected result represented a dramatic improvement, both for this patient cohort (20% 30–day pre-enrolment re-admission rate reduced to 0% following index admission in both care groups), and in comparison with available data. Lastly, the only study that examined long-term mortality (4.2 years) observed a marginally significant reduction due to the DMP (56% in the DMP versus 65% in the usual care group; Math).57

There was no substantial variation in results when only DMPs with patient follow-up longer than 6 months,47,50,51,54,57,58,61,63,65,68 home visits,52,54,57,58,63,64,66,71,73,74 or out-patient visits to a clinic50,51,53–55,68,69,72–74 were included (Table 3). In the case of DMPs with out-patient visits to a clinic, however, the reduction in the frequency of re-admission for HF or cardiovascular disease, and for all causes, failed to attain statistical significance (Table 3). When only studies with a Jadad score of 3 were selected,48,50,52,53,58,63,64,66,69,73 the results again did not vary substantially, but they were now homogeneous for the end-points of all-cause re-admission, and re-admission or death (Table 3).

There was some indication of asymmetry in the funnel plots for the three outcomes considered, in that there was a gap in the right bottom side of the graphs. The asymmetry was statistically significant for heart failure or cardiovascular re-admission, and for re-admission or death (Fig. 2). Therefore the results were compatible with certain publication bias. However, if such bias really exists, its magnitude is likely to be small, because sensitivity analyses showed that pooled RR for the three outcomes did not vary significantly when small studies (e.g., less than 100 subjects) were excluded (Table 3).

A number of studies reported that DMPs has a greater effect reducing multiple versus first re-hospitalisations.51,54,56,73 In addition, Stewart et al.,67 and Rich et al.,74 found that in the subgroup of patients with a worse clinical situation, DMPs increased re-hospitalisation.

Of the 27 studies included in the review, 13 assessed the cost of DMP versus that of usual HF management. Two of these studies61,64 were not considered because they did not report data specifically for subjects with HF and their patients were mostly elderly subjects with other diseases. Ten studies48,49,54,56,59,66–68,70,73 estimated that the implementation of a DMP reduced costs and only one study53 reported similar costs in the DMP and usual-care groups. Lastly, an economic analysis of the data from the trial by McDonald et al.,55 published subsequently, suggested that intervention was cost-effective.103 Similarly, Stewart et al.,104 carried out an economic analysis of DMP in HF based on data yielded by previous trials, and concluded that DMP, particularly those including nurse-coordinated home visits, were very effective.

Non-randomised trials

Table 4 summarises the 27 non-randomised trials included in the review. Patient characteristics and type of intervention were similar to those of randomised studies. In the great majority of studies DMP were also associated with favourable results for several of the end-points considered.

View this table:
Table 4

Non-randomized trials evaluating the effect of disease management programs on hospital readmission of older patients with heart failure

No study adjusted for confounding factors as important as co-morbidity or type of drug treatment, and only one was adjusted for disease severity.79 In 14 studies75,78,79,82,86,88,91–93,95,97–99,101 before-and-after comparisons were made using the same subjects, so that both sex and (as the follow-up period was relatively short) age were deemed to be adjusted. Some studies matched control subjects for sex and age,81 functional status, co-morbidity and age,87 and sex and medical claims.90

Of the 12 studies that furnished information on re-admissions for HF, five were eligible for meta-analysis (Fig. 3(a) ). Three studies78,92,97 reported a significant reduction (Math) in re-hospitalisations in the intervention groups, and the other 2 showed a trend to reduce hospitalisations. However, there was heterogeneity in the magnitude of the benefit across studies (P<0.001). Based on a total of 1875 patients, the pooled RR was 0.38; 95% CI: 0.16–0.93. Favourable results for DMPs even improved (pooled RR 0.2; 95% CI: 0.09–0.64) after exclusion of two trials in which the interventions (HF observation unit in hospital emergency ward78 and use of a clinical pathway to reduce hospital stay)80 were somewhat different to those in the remaining studies.

Fig. 3

Meta-analysis of the effectiveness of disease management programs on three outcomes among older patients with heart failure. Non-randomised studies. (a) Heart failure or cardiovascular readmission, (b) all cause readmission, (c) readmission or death.

Of the 16 studies that reported data on all-cause re-admission (Fig. 3(b)), eight were meta-analysed. Only three studies82,84,85 showed a significant reduction (Math) in re-hospitalisations, and the pooled RR (Math patients) across all studies was 0.50, 95% CI: 0.34–0.74. These results were not substantially modified (pooled RR 0.41; 95% CI: 0.30–0.79) by exclusion of the study by Luzier et al.,85 which differed from the remaining DMP, in that it consisted of optimization of ACE-inhibitor dosage by a clinical pharmacist. Lastly, only two studies76,77 supplied information on the combined end-point of re-hospitalisation or death, and both suggested that DMPs had a favourable effect (pooled RR 0.37; 95% CI: 0.24–0.58) (Fig. 3(c)).

Of the 27 studies reviewed, only 12 assessed the cost of DMPs versus that of usual HF management.79,81–83,85–87,89–91,94,101 All concluded that there was a reduction in cost with DMP, though in two studies85,87 costs increased in two subgroups of patients.

Discussion

This review provides evidence that DMPs reduce re-admissions for HF or cardiovascular cause, all-cause re-admissions, and the frequency of the combined end-point of re-admission or death among older patients with heart failure. These results were observed regardless of the type of healthcare delivery within DMPs, such as being home-based or clinic-based, and the duration of follow-up. In addition, these results were obtained in heterogeneous groups of patients, with different degrees of morbidity and severity, and in settings with diverse healthcare systems ranging from publicly funded systems of the type found in the UK to other, fundamentally privately funded systems like that of the USA.

The benefits of DMPs are of a magnitude close to that observed in clinical trials evaluating drugs for HF, such as ACE-inhibitors,105–107 beta-blockers108–110 or digoxin.111 Furthermore, there is evidence that the programmes can be cost-effective and even lead to financial savings.13 However, since none of the DMP studies compared different interventions directly, we do not know the relative effectiveness of types of organising care within the DMP. Accordingly, clinical trials should be conducted to compare these directly (e.g., either by comparing tele-management with home visit care,112 or by comparing interventions in which nurses change drug therapy according to protocols with interventions in which patients are referred to a physician for this purpose, etc.).

Why should different interventions in relatively heterogeneous patients and settings all yield favourable results? Although the data in this review do not allow for a definitive answer, it is possible to speculate on two possibilities. First, rather than residing in what most differentiates some DMPs from others, such as the mode of delivering care, the key elements of success lie in what is common to all of them: programme content, in particular patient education to increase self-efficacy in managing their disease. Second, trial patients tend to be strictly selected: of those who are initially screened, the percentage of subjects finally included does not reach 50% in the majority of randomised trials (Table 2). This suggests that in each study, researchers have chosen those patients who can benefit most from the intervention (e.g., patients who have a high risk of re-admission or display a considerable degree of ignorance of the nature of the disease and its management), or that patients refuse to participate in interventions which they perceive as uncomfortable or for which they are not prepared (e.g., fatigued patients refuse to participate in programmes that require visits to a clinic). Future studies must clarify these points and, in particular, ascertain which specific type of intervention is best suited to each type of patient.

What does our study add to previous reviews on this topic? A meta-analysis of randomised clinical trials of DMPs in HF was published in 2001.11 The review, which included 11 studies published up to 1999, concluded that DMP for the care of HF-patients which involved multi-disciplinary teams, reduced hospitalisations and appeared to be cost saving. From 1999 through August 2003, a further 16 randomised trials have been published. Moreover, the review did not include non-randomised trials. Another systematic review on this topic was published in 2003.12 The review included 32 randomised and non-randomised studies published up to March 2002. By August 2003, however, a total of 54 studies (randomised and non-randomised) on DMPs in HF had been published. This recent review12 made excessive emphasis on “vote counting”, a technique whereby the number of studies with positive results for a number of outcomes are counted as effective. This technique yields no estimate of effect size, and does not directly assess homogeneity of effect across studies. Most importantly, it weights studies of all sizes and effects of all magnitudes equally. “Vote counting” should be restricted to situations where effect measures are not presented and cannot be obtained,17 which is not the case for most studies evaluating DMP in HF. Lastly, another review has been published in 2004,13 after our manuscript was submitted for publication. This study evaluated interventions described as comprehensive discharge planning plus post-discharge support for inpatients with HF and mean age ⩾55 years. This review did not consider non-randomised trials but covered 3304 patients from 18 randomised trials published from 1966 to October 2003. The review yielded results for DMPs that are even more favourable than those obtained by us. In fact, Phillips et al.,13 estimated a 25% reduction in all-cause re-admission (pooled RR 0.75, 95% CI: 0.64–0.88) while we only observed a 12% reduction.

For the correct interpretation of our results, some features of the studies must be commented upon. Firstly, the Methods section often provides insufficient information to judge the quality of the interventions, their intensity or duration. In most instances where patient education is provided, insufficient information is given to understand how the educational process was carried out and how to replicate it. Furthermore, the descriptions supplied do not allow for clear comprehension of the key elements for success of any given programme (e.g., improvement in patient education and support, or improvement and optimization of treatments). This is probably due to the space constraints imposed on manuscripts. Thus, it might be useful to create a repository of detailed descriptions of DMP, so that these could be reproduced and compared.

Secondly, such dearth of information is particularly marked for the groups who underwent so-called “usual care”. This is a key aspect, because the effectiveness of any intervention diminishes with the quality of care dispensed in the group assigned to usual care. Hence, a possible explanation for the modest results observed in the trial by Laramee et al.,49 is that patients in the usual-care group also received medication and CHF education by staff nurse. Furthermore, these patients received nurse home-visits with the same frequency as did the DMP group.

Thirdly, with the single exception of the recent DIAL study,47,48 the size of randomised trials is moderate at best, limiting the statistical power to show relevant reductions (e.g., 15–20%) in hospital re-admissions. Furthermore, only one study reports on the effectiveness of DMP beyond one year follow-up,57 and it requires confirmation. Hence, future studies should have a larger size and duration; interventions should perhaps also have a wide spectrum, being targeted, not merely at HF, but rather at better management of all co-morbidities,57 since many re-admissions in patients with HF are due to causes other than HF.

Fourthly, randomised studies are inevitably non-blinded. This allows for a certain degree of co-intervention by healthcare staff, other care-givers or the patients themselves to compensate for not being in the DMP group, or to optimise the situation of the intervened subjects and better illustrate the effectiveness of DMP. Thus, it is not easy to anticipate the effect of non-blind design on study results. Furthermore, many of the trials did not state whether the randomisation sequence was concealed until allocation of patients to trial groups. Although failure to do so could lead to overestimation of the benefit of the intervention,113 no such overestimate was observed in our review when the results of studies with a Jadad score of 3 were compared against the results of all the randomised trials (Table 3).

Fifth, approximately half the studies reviewed were non-randomised and had a very low degree of adjustment for confounding factors. The magnitude of DMP benefits as reported by non-randomised trials were more than double that reported by randomised trials. Yet the consistence in the direction of the results in the two types of studies lends the non-randomised studies a certain degree of credibility.

Lastly, some decisions we made in our review also call for comment. First, the studies which evaluate DMPs usually assess several end-points; thus, there is a reasonably high likelihood of achieving statistical significance in one or more of the many comparisons made. Consequently, we confined ourselves to studying hospital re-admission, which is an easily understandable, clinically and socially relevant variable. In addition, procedures to measure re-admission are widely standardised, rendering this end-point easily combinable. Second, fairly different interventions have been combined; this is inevitable due to the very definition of DMPs, which accepts multiple treatment modalities and, because in all likelihood, no two DMPs are alike. In an attempt to mitigate this limitation, DMP who share some types of healthcare delivery were grouped in the sensitivity analysis. Third, the use of scores to assess the quality of studies included in the meta-analysis is controversial; although they are appropriate for comparing groups of trials, as we did, their results depend on the choice of the scale and some of them give more weight to quality of reporting than to actual methodology. In fact, the latter is one of the common criticisms of the Jadad score. We have tried to overcome this problem by indicating specific components of quality, supported by both empirical and theoretical evidence, such as blinding of outcome assessment or control for confounders. Lastly, since the quality of the economic assessments was not a principal objective of our review, this aspect was not evaluated. Nevertheless, it should be pointed out that, in most trials, the cost analyses did not include all the relevant components of cost. It is possible, as Ekman and Swedberg suggest,14 that a potential saving stemming from a reduction in re-admissions may conceivably lead to an increase in costs in other categories of care, such as hip surgery or cataract operations, because the target population are elderly and closer follow-up by healthcare staff may well facilitate the provision of such services.

Acknowledgments

This work has been supported by FIS grant 00/0035, by an unrestricted educational contract with Roche Pharmaceuticals, Spain, and by ISCIII (red de Centros RCESP C03/09). The funding bodies had no role in data extraction or analyses, writing of the manuscript, or in the decision to submit the paper for publication.

References

  1. [1]
  2. [2]
  3. [3]
  4. [4]
  5. [5]
  6. [6]
  7. [7]
  8. [8]
  9. [9]
  10. [10]
  11. [11]
  12. [12]
  13. [13]
  14. [14]
  15. [15]
  16. [16]
  17. [17]
  18. [18]
  19. [19]
  20. [20]
  21. [21]
  22. [22]
  23. [23]
  24. [24]
  25. [25]
  26. [26]
  27. [27]
  28. [28]
  29. [29]
  30. [30]
  31. [31]
  32. [32]
  33. [33]
  34. [34]
  35. [35]
  36. [36]
  37. [37]
  38. [38]
  39. [39]
  40. [40]
  41. [41]
  42. [42]
  43. [43]
  44. [44]
  45. [45]
  46. [46]
  47. [47]
  48. [48]
  49. [49]
  50. [50]
  51. [51]
  52. [52]
  53. [53]
  54. [54]
  55. [55]
  56. [56]
  57. [57]
  58. [58]
  59. [59]
  60. [60]
  61. [61]
  62. [62]
  63. [63]
  64. [64]
  65. [65]
  66. [66]
  67. [67]
  68. [68]
  69. [69]
  70. [70]
  71. [71]
  72. [72]
  73. [73]
  74. [74]
  75. [75]
  76. [76]
  77. [77]
  78. [78]
  79. [79]
  80. [80]
  81. [81]
  82. [82]
  83. [83]
  84. [84]
  85. [85]
  86. [86]
  87. [87]
  88. [88]
  89. [89]
  90. [90]
  91. [91]
  92. [92]
  93. [93]
  94. [94]
  95. [95]
  96. [96]
  97. [97]
  98. [98]
  99. [99]
  100. [100]
  101. [101]
  102. [102]
  103. [103]
  104. [104]
  105. [105]
  106. [106]
  107. [107]
  108. [108]
  109. [109]
  110. [110]
  111. [111]
  112. [112]
  113. [113]
View Abstract