Following an initial scoping, we searched for systematic reviews of randomized controlled trials (RCTs) of telehealth interventions incorporating components of supported self-management. The basic search strategy combined “telehealth terms” AND “self-management terms” AND “long term conditions terms” AND “systematic review terms.” The search used a combination of keyword searches and Medical Subject Headings (MeSH). We searched 7 databases for reviews published from January 2000 to May 2016. The search was limited to studies published later than 2000 because telehealth is a relatively recent innovation and rapid advances in technology mean that any earlier work is unlikely to be relevant to contemporary health care. Table 1 summarizes the search strategy and sources, and Table 2 details the outcomes. Multimedia Appendix 1 shows full search terms for the MEDLINE database; we adjusted these for the other databases.

Inclusion and exclusion criteria are summarized using the population, intervention, comparison, outcomes, setting, and study design headings in Table 1 and Table 2. Initial scoping revealed that few reviews of telehealth interventions identified self-management support explicitly as an aim of the telehealth intervention under consideration (eg, self-management was a specified inclusion criterion). We therefore also included reviews in which self-management support was an implied component or mechanism of the telehealth intervention under consideration and in which outcomes relevant to self-management were assessed as part of the review. For the purpose of inclusion, we defined self-management as “any intervention which aimed to empower patients to be active decision makers who deal with emotional, social or medical management of their illness with the aim of improving their independence and quality of life” [6]. We excluded reviews when the telehealth aspect simply involved remote physiological monitoring (eg, of oxygen saturation or blood sugar levels) without an explicit decision-making role on the part of the patient. Reviews in which patients were educated and supported to interpret and act on the clinical information they were recording we considered self-management and therefore included. We included reviews in which self-management was hypothesized as the mechanism by which the telehealth intervention had an impact, but we considered these to be implied self-management unless self-management support was also a specified aim of the intervention.

Inclusion criteria were piloted by 2 authors (PH and HP) and disagreements were resolved by discussion with all authors. PH then assessed all titles and abstracts against the inclusion criteria. Where no abstract was available, articles were retained for full-text assessment. A random sample of 250 abstracts was screened by 2 reviewers (PH and LD). A kappa statistic of agreement was calculated using IBM SPSS version 22 (IBM Corporation) and was high (0.96). Full texts of all potentially eligible articles were assessed independently by 2 reviewers (PH and LD).

We assessed the quality of the included reviews using the revised A Measurement Tool to Assess Systematic Reviews (R-AMSTAR) quality assessment tool. Each included review was assessed independently by 2 reviewers (PH and LD) and disagreements were resolved by discussion. We combined the R-AMSTAR score with the size of the review and an assessment of self-management focus to assign a star-based weighting to the evidence from each review. We awarded 1 star for each of the following:

We then used the weighting of each review to inform the synthesis. Any disagreements in the full-text screening, quality assessment, or data extraction were resolved by discussion, involving a third author when agreement could not be reached.

We grouped outcomes of interest into disease control outcomes (clinical and physiological markers of disease control, unscheduled health care utilization, and validated measures of symptoms and quality of life) and self-management process outcomes (eg, self-efficacy, medication adherence). These are defined in Table 2.

Data were extracted using a piloted data extraction template for included studies (PH), and each study was checked for accuracy (LD). We extracted the description of intervention(s) and component(s) of self-management support; inclusion and exclusion criteria; population of interest; duration and intensity of intervention; outcomes measured; and results as presented in the review. The synthesis and conclusions of each review were collated: we did not analyze results from individual RCTs.

The overlap in included RCTs (ie, reviews with similar inclusion criteria may include the same RCTs) precluded meta-analysis of the review findings; we therefore undertook a narrative synthesis. We used harvest plots to illustrate the disease control outcomes related to telehealth-supported self-management components [17,18]. Harvest plots use bars representing individual reviews placed on a plot matrix to indicate whether the review intervention showed an overall positive, negative, or no consistent effect for the outcome in question. To construct the harvest plots, we needed to judge each review as to whether it showed an overall positive effect for each outcome or group of outcomes. Given the heterogeneity in outcomes and methods of data synthesis among the included reviews, and to ensure objective and consistent assessment of each review, we devised a set of rules to underpin decisions about whether a review showed a positive, negative, or no effect. These were devised and refined by discussion between the authors and are described in Table 3. Interpretation of the findings was aided by regular discussion within the research team.

The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flowchart shown in Figure 1 illustrates the search results and review selection.

A total of 53 systematic reviews met the inclusion criteria [16,19-70]. These presented data from 231 unique RCTs (119 diabetes, 58 heart failure, 28 asthma, 23 COPD, and 3 cancer). The year of publication ranged from 2000 to 2016. Information on the geographical spread of the RCTs included within the SRs was incomplete, but included studies from North and South America, Europe, Asia, and Oceania. Multimedia Appendix 2 lists the RCTs included in each systematic review.

Multimedia Appendix 3 shows details of participant demographics, interventions and comparison, setting, and content and intensity of interventions for each of the included systematic reviews.

R-AMSTAR scores for the included systematic reviews ranged from 19 to 43 out of a possible 44. Multimedia Appendix 4 gives scores for the individual components of the R-AMSTAR score. Taking into account quality assessment, explicit self-management focus, and total population size, 8 reviews received an evidence weighting of 3 stars [16,21,25,26,38,45,53,56], 23 received 2 stars [22,24,27-31, 33,37,40,42,44,46,49,50,52,54,55,58,62,63,67,68], 21 received 1 star [19,20,23,32,34-36,39,43,47,48,51,57, 59-61,64-66,69,70], and 1 received no stars [41]. The first column of the table in Multimedia Appendix 5 displays these criteria.

Multimedia Appendix 5 provides an overview of the focus, quality, findings, and conclusions of each of the included reviews. It also displays how the interventions described map to Pearce et al’s taxonomy of self-management support [14]. Additional detail is shown in Multimedia Appendix 6. The text that follows synthesizes the findings against the 2 aims of the metareview.

The overall findings of the reviews for impact on disease control outcomes of telehealth-supported components of self-management are illustrated in the harvest plot shown in Figure 2.

All of the interventions in the included systematic reviews were complex interventions with multiple components. Reporting of the details of the components of interventions was highly variable, and we therefore limited analyses of which specific components were associated with improvements in disease outcomes to a subset of the included systematic reviews. No single self-management component was found to be consistently effective, or consistently ineffective.

The individual long-term conditions considered in this metareview differed both in the quantity of evidence for telehealth interventions supporting self-management and in the findings and conclusions of the included systematic reviews. Diabetes and heart failure constituted the greatest evidence base, with available data on cancer being very limited.

The impact of telehealth-supported self-management on disease control and health care utilization was inconsistent, with positive outcomes more frequently identified in type 2 diabetes and heart failure and often no effect demonstrated in other conditions. The highest-quality evidence for heart failure showed an overall improvement in mortality in meta-analyses of telemonitoring and telephone support [58]. In contrast, none of the reviews assessing mortality in COPD showed any significant improvement with telehealth [65,66,68,69]. None of the reviews, however, reported a negative impact of interventions employing telehealth for any condition. This should be treated with some caution, however, as few reviews specifically considered or assessed for publication bias and, of those that did, some found evidence to suggest bias [38,46]. Findings varied by disease group and by components of telehealth delivery. The highest-weighted evidence showed improvement in HbA_1c in type 2 diabetes with interventions remotely monitoring blood glucose and in some more intensive telephone interventions [31,38,53]. Physiological telemonitoring and telephone support for heart failure were associated with reduced mortality and heart failure-specific hospital admissions [58]. For both of these conditions, however, findings were inconsistent between reviews, and analyses of similar interventions reached different conclusions [25,55,57]. Interventions for type 1 diabetes did not improve glycemic control [49]. There was some evidence for reduced hospitalization with telehealth interventions in more severe asthma and COPD [63,68], but analyses of more specific, self-management-focused interventions showed insufficient or inconsistent evidence of benefit [27,62]. The interventions described incorporated a range of self-management components. No single component was consistently effective in any disease area, although none were associated with harm. Interventions with multiple components, or more intensive interventions, may be associated with greater benefits [30,53,63,68]. In most reviews, however, the description of self-management components and analysis of their relation to clinical outcomes was not sufficiently detailed to draw firm conclusions about which components or combinations were most beneficial, or in what conditions.

A strength of metareview methodology is that it allows a relatively rapid synthesis of a large body of primary literature and enables a broad overview of a subject area [15]. There are, however, limitations inherent in metareview methodology. Any systematic review is limited by the time delay from completion of a primary study to publication, and subsequently conducting the review itself. By reviewing systematic reviews, this lag time is further extended and, as such, the results risk being out of date, although we updated the review before completing the paper. The observation in 1 review that some more recent studies showed greater benefits in clinical outcomes [28] highlights the importance of up-to-date evidence in a fast-moving field such as telehealth. That we were unable to differentiate the RCT findings by year of publication is therefore a limitation of our methodology.

By relying on systematic review findings, the evidence is 1 step removed from the empirical evidence, and thus reliant on the interpretation of the review authors. This was mitigated to some extent by assessing the methodological quality of the reviews and using this to weight the evidence. This limitation is most evident, however, when addressing questions about strategies for developing and implementing telehealth interventions. For example, no reviews specifically addressed how potential participants were identified, recruited, and retained, a key issue if telehealth interventions are to be successfully implemented [73]. Few explicitly considered the importance of patient decision making in the interventions. Our consideration of the impact of the intensity of the interventions is also limited to a few systematic reviews, as the majority did not analyze the impact of this on disease control outcomes. By relying on the analysis of the review authors, we were also unable to adjust for important factors such as geographic location, age of participants, and socioeconomic variables such as educational status. A metareview such as this is thus well suited to forming an overview of the topic, but loses granularity and detail of the evidence, a limitation particularly evident when attempting to analyze components of self-management support.

There is also overlap in the included RCTs, which risks overrepresenting the results of a few RCTs included in several systematic reviews and giving a false impression of consensus. Overlap precludes meta-analysis, but the use of harvest plots provides a visual synthesis of the findings of each individual review. We designed the rules used to determine how each review was displayed in the harvest plots to ensure that the assessment of reviews was consistent and transparent. Specifying that more than 50% of RCTs in a narrative synthesis must show positive results for a review to be considered as having consistently positive findings avoids overly optimistic interpretation, but risks overlooking individual RCTs that may be particularly relevant to the review question.

There are also limitations specific to this metareview. This metareview considered only 6 long-term conditions and may have reached different conclusions had we selected a different set of conditions. However, we selected common conditions, 5 of which had a good evidence base for telehealth or self-management, or both, and included 1 area (cancer) in which these concepts are still at an early stage. The relative lack of data for cancer may reflect a lack of research in this field or simply that the available evidence has yet to be synthesized in a systematic review.

Not all the included systematic reviews explicitly focused on self-management. It could be argued that the inclusion of systematic reviews with an implied self-management focus was based on our subjective assessment. However, we developed clear rules for the inclusion criteria and undertook duplicate full-text screening, but this might have resulted in some inappropriate inclusions or exclusions due to limited description of interventions. In addition, we identified reviews that focused specifically on self-management and incorporated this into the weighting system, so as to minimize the impact of this limitation on the interpretation of the findings.

Our finding of inconsistent evidence of benefit for disease control and health care utilization with telehealth-supported self-management is similar to the findings reported by other overviews of telehealth interventions that were not specifically focused on self-management. A recent metareview of heart failure telemonitoring interventions showed a reduction in all-cause mortality (relative risk 0.60-0.85) and heart failure hospitalizations (relative risk 0.64-0.86) in an analysis of 15 meta-analyses [74]. An overview of telehealth across a number of conditions noted modest improvements in HbA_1c in some reviews with others noting no overall effect [2]. A consistent conclusion, reflected in our findings, is that telehealth is not associated with worse outcomes. Taken together, it appears that, while not consistently superior to usual care, telehealth is a safe alternative mode of delivery for self-management support, particularly in conditions such as heart failure and type 2 diabetes, where the evidence base is more developed. The research agenda may therefore shift to demonstrating equivalence, or understanding the impact of offering choice, or evaluating other potential benefits of telehealth (such as improved access).

While components of the reported interventions mapped to Pearce et al’s taxonomy of self-management support [14], few reviews analyzed how these related to the self-management process or clinical outcomes. In common with other overarching analyses of self-management support [12], we found no evidence to support 1 component as being consistently effective or essential for improved outcomes. The greatest volume of evidence was for educational interventions and monitoring of clinical data incorporating feedback, and several reviews, notably those with higher weighting, demonstrated improved outcomes in some (notably type 2 diabetes) but not all conditions (Figure 2). Clinical review, adherence support, and lifestyle interventions also showed some positive results, but with substantial inconsistency. There was some evidence for telehealth-delivered adherence support in asthma and diabetes, but this was variable and inconsistent. In heart failure, evidence linking knowledge, self-efficacy, or self-care behaviors to improvements in disease outcomes was either lacking or inconsistent [56]. Greenwood et al concluded that a range of components (including educational, self-monitoring, goal-driven, and interactive elements) were important in improving glycemic control [30] and emphasized the multifaceted nature of self-management support.

Analysis of the reasons for effectiveness (or lack thereof) of the interventions is limited in the included reviews. Such questions have been addressed using qualitative methodologies [75], which have highlighted that users valued convenience [76], a sense of being “watched over” [11,77], and finding the telemonitoring data reassuring [78]. Some felt empowered by these interventions to use the health care system more effectively [76,77]. Such findings correlate with the quantitative findings of this review showing that interventions with active monitoring of blood glucose data with feedback in type 2 diabetes were often associated with improved outcomes. However, qualitative explorations of some asthma and COPD interventions suggested that daily professional monitoring of telemonitored parameters (such as pulse oximetry or symptom scores) could engender reliance on professionals rather than supporting self-management [11,76,78,79]. In contrast, self-monitoring (including oximetry) without daily professional oversight gave patients greater confidence in following their self-management plans [80].

While telehealth interventions were not consistently found to be superior to usual care, none of the reviews reported any negative effects, suggesting that telehealth is a safe alternative mode of delivery for self-management support, particularly in conditions such as heart failure and type 2 diabetes, where the evidence base is more developed. Improvements may be more readily seen in those with more severe disease. The decision to adopt telehealth strategies will be determined not only by clinical or demographic circumstances but also by patient or clinician preference. Findings vary within and between different conditions, and further investigation is required to establish the role in conditions (such as cancer) where current evidence is limited.

We found little explicit evidence of a mediating role for self-management in telehealth interventions and the specific components that may encourage self-care. While some evidence suggests that, in the context of type 2 diabetes, more intensive interventions may be associated with greater improvements in glycemic control, such observations are limited to specific analyses from a small number of included reviews [22,30,53], and it is not clear whether this applies to other disease areas [58]. Larger-scale trials of self-management interventions delivered by telehealth, based on explicit self-management theory [81,82], linked with process evaluations that explore intermediary outcomes such as self-efficacy, and providing detailed description of the interventions, are needed before the extent to which telehealth technologies may be harnessed to support self-management at scale can be established.