Systematic meta-review of supported self-management for asthma: a healthcare perspective

Background Supported self-management has been recommended by asthma guidelines for three decades; improving current suboptimal implementation will require commitment from professionals, patients and healthcare organisations. The Practical Systematic Review of Self-Management Support (PRISMS) meta-review and Reducing Care Utilisation through Self-management Interventions (RECURSIVE) health economic review were commissioned to provide a systematic overview of supported self-management to inform implementation. We sought to investigate if supported asthma self-management reduces use of healthcare resources and improves asthma control; for which target groups it works; and which components and contextual factors contribute to effectiveness. Finally, we investigated the costs to healthcare services of providing supported self-management. Methods We undertook a meta-review (systematic overview) of systematic reviews updated with randomised controlled trials (RCTs) published since the review search dates, and health economic meta-analysis of RCTs. Twelve electronic databases were searched in 2012 (updated in 2015; pre-publication update January 2017) for systematic reviews reporting RCTs (and update RCTs) evaluating supported asthma self-management. We assessed the quality of included studies and undertook a meta-analysis and narrative synthesis. Results A total of 27 systematic reviews (n = 244 RCTs) and 13 update RCTs revealed that supported self-management can reduce hospitalisations, accident and emergency attendances and unscheduled consultations, and improve markers of control and quality of life for people with asthma across a range of cultural, demographic and healthcare settings. Core components are patient education, provision of an action plan and regular professional review. Self-management is most effective when delivered in the context of proactive long-term condition management. The total cost (n = 24 RCTs) of providing self-management support is offset by a reduction in hospitalisations and accident and emergency visits (standard mean difference 0.13, 95% confidence interval −0.09 to 0.34). Conclusions Evidence from a total of 270 RCTs confirms that supported self-management for asthma can reduce unscheduled care and improve asthma control, can be delivered effectively for diverse demographic and cultural groups, is applicable in a broad range of clinical settings, and does not significantly increase total healthcare costs. Informed by this comprehensive synthesis of the literature, clinicians, patient-interest groups, policy-makers and providers of healthcare services should prioritise provision of supported self-management for people with asthma as a core component of routine care. Systematic review registration RECURSIVE: PROSPERO CRD42012002694; PRISMS: PROSPERO does not register meta-reviews Electronic supplementary material The online version of this article (doi:10.1186/s12916-017-0823-7) contains supplementary material, which is available to authorized users.


Background
Asthma is common, affecting 334 million people worldwide, and is responsible for substantial morbidity and an increasing burden on healthcare services globally [1]. In the UK, there are over 6 million primary care consultations, and 100,000 hospital admissions each year, at an estimated cost of £1 billion per year [2].
For a quarter of a century [3], national and international guidelines have recommendedunequivocallythat people with asthma should be provided with selfmanagement education reinforced by a personalised asthma action plan and supported by regular review [4,5], though mode of delivery, personnel delivering the support, the targeted group and the intensity of the intervention vary [6]. The 2014 UK National Review of Asthma Deaths provided a stark reminder of the importance of ensuring that people with asthma respond in a timely and appropriate manner to deteriorating symptoms: only 23% had documented evidence of having been provided with selfmanagement education and 45% of people who died had not sought or received medical attention in their final attack [7].
However, despite self-management being highlighted as a core component of all models of care for people with long-term conditions (LTCs) [8][9][10] and the concept being well established in the context of asthma [4,5], in practice only a minority of people with asthma have an action plan [11]. Effective implementation requires a whole systems approach, combining active engagement of patients with the training and motivation of professionals embedded within an organisation in which self-management is valued [12]. Patient organisations, healthcare professionals, policy-makers, commissioners and providers of healthcare services thus need an up-to-date systematic overview of the evidence to inform decisions about prioritisation of supported selfmanagement and to underpin implementation strategies within diverse healthcare systems.
The data presented in this paper are derived from two parallel programmes of work on supported selfmanagement in LTCs commissioned by the National Institute of Health Research: Practical Systematic Review of Self-Management Support (PRISMS) [13] and Reducing Care Utilisation through Self-management Interventions (RECURSIVE) [14]. In the context of asthma, we aimed to answer questions of importance to clinicians, patient-interest groups, managers responsible for developing healthcare services and policy-makers: can supported self-management reduce the use of healthcare resources and improve asthma control? More specifically, in which target groups has it been shown to work, which components are important, in what healthcare contexts, and at what cost?

Methods
We used established methodology for undertaking a meta-review of systematic reviews (PRISMS) and a systematic review of randomised controlled trials (RCTs) (RECURSIVE) [15]. The PRISMS and RECURSIVE reviews were undertaken during 2012-2013 with initial searches completed in November 2012 and May 2012, respectively. We updated the PRISMS searches in March 2015 with a pre-publication update in January 2017, and the RECURSIVE searches in September 2015. RECUR-SIVE is registered on PROSPERO: CRD42012002694. (PRISMS could not be registered because PROSPERO does not register meta-reviews.) Table 1 summarises the PICOS criteria, search strategies, sources and search dates; further details are in Additional file 1. The PRISMS search strategy involved searching nine electronic databases using the terms: 'self-management support' AND 'asthma' AND 'systematic review'. We defined self-management as 'the tasks that individuals must undertake to live with one or more chronic conditions. These tasks include having the confidence to deal with medical management, role management and emotional management of their conditions' [16]. For the update, we searched not only for systematic reviews published after our initial search date but also for RCTs published after the search dates used by the included systematic reviews (see Additional file 2 for the details of these dates). Included systematic reviews were grouped according to the populations studied (children, adults or ethnic minority groups) and the search dates of the reviews extracted. Dates for the update RCT search were set from the date of the latest review search within each population group.

Search strategy
The RECURSIVE search strategy in nine databases comprised the terms: 'self-management support' AND 'long-term condition' AND 'healthcare use' AND 'randomised controlled trial'. (RECURSIVE included asthma and other LTCs in a single search.) We also specifically sought health economic publications linked to included RCTs. Table 2 summarises the PRISMS and RECURSIVE processes. Following training (repeated cycles of duplicate screening of 100 titles, team discussion and clarification of exclusion rules), one reviewer (HLP or GP for PRISMS; LD for the update; MP for RECURSIVE) reviewed titles and abstracts and selected possibly relevant studies. A random sample of titles and abstracts (10% in PRISMS; 40% in RECURSIVE) was examined by a second reviewer (HP for PRISMS; PB or NS for RECURSIVE) working independently as a quality check. The agreement was 97% for the initial search and 99% for the update in PRISMS and 87% for the initial search and 88% for the update in RECURSIVE.

Identification of relevant papers
After a similar training process, the full texts of all potentially eligible studies were assessed against the eligibility criteria (see Additional file 3) by one reviewer (HLP for PRISMS; LD for update; MP for RECURSIVE). Second reviewers undertook a 10% check for PRISMS (HP) and a 30% check for RECURSIVE (PB or NS), achieving 83% and 85% agreement, respectively. Disagreements were because unclear papers were included by the reviewer pending discussion with a lead investigator. Uncertainties and disagreements were resolved by full team discussion.

Assessment of methodological quality
We used the R-AMSTAR (Revised Assessment of Multiple Systematic Reviews [17]) quality appraisal tool to assess the methodological quality of the systematic reviews included in the PRISMS study. This reflects both the quality of the review process and the rigour with which the review assessed the quality of the studies it included. We used the Cochrane Risk of Bias tool to assess the quality of RCTs included in the updated search [15]. Quality assessment was undertaken by HLP or LD and independently by a second Adults (≥18 years) with asthma (within a wider search of long-term conditions), excluding studies in the developing world.
Intervention Self-management support interventions. Self-management support interventions.

Comparator
Typically 'usual care' or less intense self-management interventions.
Typically 'usual care' or less intense self-management interventions.
Outcomes Unscheduled use of healthcare services (admissions, A&E attendances, unscheduled consultations), health outcomes (asthma control), quality of life, process outcomes (ownership of action plans, self-efficacy).
Healthcare utilisation with comprehensive measures of costs or major cost drivers (i.e. hospitalisation, A&E attendances), quality of life.
Settings Any healthcare setting. Any healthcare setting.
Study design Systematic reviews of RCTs. RCTs published after the date of the last search in the included systematic reviews (see Additional file 2).

RCTs
Dates Initial database search: January 1993 (3 years before the publication of the earliest systematic review identified in scoping work) to July 2012. Manual and forward citations were completed in November 2012.
reviewer (HP) with disagreements resolved by discussion within the team (EE, GP, HLP, ST and HP).
To reflect both quality and size of the review, we developed a star weighting system based on (a) the R-AMSTAR score (≥31 was defined as 'high-quality') and (b) the number of participants (≥1000 participants was defined as 'large'): *** Large high-quality review ** Either small high-quality review or large low-quality review * Small low-quality review In the RECURSIVE study, quality assessment of formal economic evaluations was undertaken using the Drummond checklist [18,19]; RCTs reporting healthcare utilisation were assessed by judging allocation concealment (the quality component most associated with treatment effect [20]) as adequate or inadequate according to the Cochrane Risk of Bias tool [15].

Outcomes
The primary outcome in the PRISMS meta-review was unscheduled use of healthcare resources (specifically unscheduled consultations, accident and emergency (A&E) department attendances and hospital admissions). Other outcomes of interest were asthma control, asthmarelated quality of life and process outcomes (specifically, ownership of action plans). Healthcare utilisation rates and costs were the primary focus of the RECURSIVE review, especially major cost drivers (i.e. hospitalisation rates and costs) and comprehensive summaries including multiple sources of cost. The results of formal costeffectiveness, cost-utility and cost-benefit analyses were also of interest. Quality assessment Duplicate quality assessment using: R-AMSTAR [17] for systematic reviews ('high-quality' defined as ≥31), combined with size of the review ('large' defined as ≥1000 participants) to give star rating (1* to 3*). Cochrane Risk of Bias tool for RCTs [15]. Disagreements resolved by discussion.
Duplicate quality assessment using: Drummond for economic evaluations [18,19]. Meta-analysis: Standardised mean differences (random effects model) to examine the effects of selfmanagement support interventions on hospitalisation rates, A&E attendances, quality of life and total costs. Permutation plots of the data from trials reporting both utilisation (hospitalisation rates, A&E attendances or total costs) and health outcomes (quality of life).

Interpretation
Monthly teleconferences to enable synergies between PRISMS and RECURSIVE. End-of-project stakeholder conference to discuss findings and implications for commissioning and providing services for people with LTCs.
A&E accident and emergency, LTC long-term condition, R-AMSTAR Revised Assessment of Multiple Systematic Reviews, RCT randomised controlled trial

Extraction of data
Data for the PRISMS review were extracted by HLP and LD (update) using a piloted data extraction form, and checked independently by HP for integrity and accuracy. Disagreements were resolved by team discussion. We extracted data on review rationale, the self-management intervention under review, review methodology, summary details of included RCTs (participant demographics, comparison groups, settings, service arrangements, components, duration/intensity of the intervention, follow-up arrangements) and the results of meta-analyses and narrative syntheses. We extracted the findings and conclusions as synthesised by the authors of the systematic reviews, specifically avoiding going back to the individual primary studies. The RCTs in the update review were extracted using similar headings. A piloted data extraction sheet was devised for RE-CURSIVE that included descriptive data (characteristics of studies, populations and interventions) and quantitative data (for use in meta-analyses). All the descriptive data and approximately 40% of the quantitative data were double-extracted by two members of the research team working independently.

Data analysis
Meta-analysis is inappropriate at the meta-review level owing to the overlap of included RCTs between reviews. However, for the primary outcome, where two or more systematic reviews (including the RECURSIVE metaanalyses) present pooled statistics, we displayed the results graphically by creating 'meta-Forest plots'. We undertook narrative syntheses to answer our key questions: Does supported self-management reduce use of healthcare resources and improve asthma control? For which target groups does it work? Which components contribute to effectiveness? and In what contexts does supported self-management work? We categorised the reviews and RCTs included in the PRISMS meta-review according to the question(s) that they answered (see Tables 3 and 4: column 3) and synthesised the findings within these categories.th=tlb= The final question (What is the effect of selfmanagement on healthcare utilisation and costs?) was answered by a meta-analysis of the RECURSIVE RCT data. The primary analysis explored whether selfmanagement support could reduce utilisation without compromising outcomes. Standardised mean differences (SMD) were computed using a random effects model meta-analysis due to anticipated heterogeneity. Four meta-analyses examined the effects of self-management support interventions on hospitalisation rates, A&E attendances, quality of life and total costs, respectively. We then constructed permutation plots of the data from the subset of trials reporting both utilisation (hospitalisation rates, A&E attendances or total costs) and health outcomes (quality of life). Further details about the analytic approach are described in the RECURSIVE report [14]. Forest plots and permutation plots [21] for the subset of studies reporting both health outcomes and utilisation outcomes were constructed in STATA version 14.
Interpretation and end-of-project workshop The PRISMS and RECURSIVE teams worked independently, but held regular teleconferences to enable synergies between the findings of the parallel reviews to be developed. Frequent meetings of the multidisciplinary teams aided interpretation of the emerging findings. Finally, we held an end-of-project stakeholder conference at which the findings and over-arching conclusions from PRISMS and RECURSIVE were presented to 34 multidisciplinary stakeholders, including people with LTCs, clinicians, commissioners, providers of healthcare services and policymakers. Small discussion groups discussed and advised on practical implications for commissioning and providing services for people with LTCs.

Lay involvement
The PRISMS project (which reviewed evidence from 14 LTCs) benefited from a lay collaborator who was involved from the inception of the project. She and other lay representatives from a range of LTC interest groups (including Asthma UK) contributed to an initial stakeholder workshop at which the choice of LTCs studied in the project and self-management interventions of interest were discussed. Lay members also participated in the end-ofproject workshop (described above), which aided interpretation and guided dissemination. The PRIMER patient and public involvement group at the University of Manchester, UK, collaborated with the RECURSIVE project.

Updating of searches prior to publication
We updated our PRISMS searches in January 2017 by undertaking forward citation of the original included reviews using Web of Science. Forward citation has been shown to be an efficient and effective method of identifying relevant papers in systematic reviews of complex and heterogeneous evidence [22]. We considered it was very unlikely that a subsequent systematic review or RCT would be published without citing at least one of the previously published reviews. One reviewer (HP) undertook focused data extraction of key findings, which were checked by MP. The additional data were added into the syntheses as appropriate. Had we identified studies that substantially changed our conclusions we planned to undertake full duplicate data extraction, quality assessment and revise our synthesis.

Results
Description of the studies in the meta-review Figure 1 illustrates the PRISMA flow chart for both reviews. After removal of duplicates, 9633 references were identified from the initial PRISMS search and an additional 6321 from the update search. From these, 25 systematic reviews  were included in the PRISMS meta-review, representing data from 240 unique RCTs.
In the 18 systematic reviews that reported the duration of follow-up in their included RCTs [23-25, 27-29, 33, 35, 38-40, 42-47, 61], the modal duration (in 10 of the reviews) was 12 months, with only 3% of reported RCTs falling outside the range of 3-24 months. The update RCTs had a similar profile, with 6 of 13 update RCTs having a duration of 12 months (range 3-30 months).

Overview of presentation of results
Tables 3, 4, 5 and 6 provide summaries of the studies included in the PRISMS meta-review, update RCTs, the RECURSIVE review and pre-publication update with more detailed tables in Additional file 5.
Can supported self-management reduce the use of healthcare resources and improve asthma control? Use of healthcare resources Figure 2 is a meta-Forest plot illustrating the metaanalyses (including three PRISMS 3* reviews and RECUR-SIVE) that report relative risks of admissions, A&E attendances and/or unscheduled consultations [27,31,38]. Treatment event rates from the meta-analyses are in Table 7. These results suggest similar effects in adults [38], children [27] and mixed populations [31].
Six reviews reported a reduction in days missed from school or work [24, 29-31, 38, 41]. Two 3* meta-analyses [31,41], two small reviews each with only one RCT [24,29] and five of the 13 RCTs in a 2* narrative synthesis of school-based interventions [30] concluded that self-management interventions reduced absenteeism. A single RCT reported in a 3* narrative review in adults concluded that asthma education following A&E attendance had no effect on absenteeism [38].
In which target groups has supported self-management been shown to work?

Cultural groups
Four reviews explored the impact of self-management in cultural groups [25,29,35,46]. A 2* meta-analysis reported that culture-specific programmes reduced hospitalisations in children and improved quality of life in adults compared to generic interventions [25]. A 3* narrative synthesis found only two RCTs testing culturally tailored interventions, one of which improved quality of life [46]. The involvement of community health workers reduced use of healthcare resources in two thirds, and improved symptoms in all seven RCTs included in a 2* narrative review [35]. An inpatient visit from a lay educator to Black or Latino children improved self-efficacy and n/a n/a     Parents of children 2-10 y with asthma and/or eczema.

RCT. FU: 6 mo
Between-group improvement in parents' self-efficacy and childs' 'eczema behaviour', but not equivalent asthma outcomes.

A&E attendees
Two 3* meta-analyses demonstrated reduced use of healthcare resources (admissions, A&E attendances and unscheduled consultations) in adults recruited during A&E attendance (13 RCTs) [38] and in children with a history of A&E attendance in the previous 12 months (38 RCTs) [27]. Neither review found improved markers of asthma control [27,38], though an update RCT in paediatric A&E attendees (low risk of bias) found that children discharged with an action plan had fewer symptoms at 28 days compared with usual care [50].

Specific age groups
School-based interventions [30], often using information technology-based programmes [30] or delivered  Fig. 2 Meta-Forest plot of healthcare resource use from meta-analyses. This meta-Forest plot displays the summary data from the PRISMS systematic reviews that reported relative risk (RR). Note that meta-analysis is inappropriate at meta-review level owing to the overlap of included randomised controlled trials between reviews by peers [48,56], improved quality of life and, in some cases, reduced absenteeism [30,48,56,61]. Generic parenting skills initiatives improved selfefficacy in families struggling to manage young children with asthma, with inconsistent effect on asthma outcomes [64,67]. Two update RCTs reported interventions in older people that improved control and quality of life [49,51], and one reduced use of unscheduled care [49]. A key feature of both complex interventions was a structured approach to tailoring in order to meet personal goals or address individual problems.
Which components of supported self-management are important?
A 3* meta-analysis (36 RCTs; 6090 participants of all ages recruited from primary and secondary care settings) defined optimal self-management as education including Table 8 Tailoring of self-management support for targeted populations (Continued) 16 short group educational sessions, including strategies for problem solving, delivered in the school lunch break.

Peer-led programmes
Year 11 pupils were trained to deliver the school-based asthma educational lessons to younger pupils.

Al-Sheyab 2012 [48] Adolescents
Compared to children in control schools, knowledge and QoL improved. Also increased self-efficacy to resist smoking.
Asthma self-management skills and psychosocial skills taught at a day camp by peer leaders followed by monthly peer telephone contact.
Rhee 2011 [56] Adolescents 13-17 y The intervention group had improved QoL and positive 'attitude to illness' compared to those attending adult-led camps.

Technology-based
Internet-based interventions, delivered at home, clinic or school, which delivered a psycho-educational programme involving information and skills training modules targeting improved health outcomes.

**Stinson 2009 [47] Children 4-17 y
The majority of studies reported improvement in symptoms, but impact on other outcomes was inconsistent.
Theoretically based asthma computer programme with core modules (adherence, inhaler use, smoking reduction), with tailored sub-modules to address specific behavioural traits.
Internet-based self-management programme covering education, self-monitoring and an electronic action plan, and encouraging regular medical review. Supported by 2 face-to-face groups. Abbreviations: A&E accident and emergency, PAAP personalised asthma action plan, QoL quality of life, RCT randomised controlled trial advice on self-monitoring and a written action plan that was supported by regular professional review [31]. There is evidence that reducing the intensity of self-management education or level of clinical review may reduce its effectiveness [36].

Components of an action plan
The components of an action plan were further defined in two 3* and three 2* reviews [23,24,32,36,39]. In adults, self-monitoring based on peak flow or symptoms is equally effective [32,36,39]. In a comparison in children, symptom-based plans were more effective at reducing unscheduled healthcare [23], and equally effective at improving most measures of asthma control; the exception was days with symptoms, which were reduced more by peak-flow-based than symptom-based plans [23]. A 3* review concluded that action plans with between two and four action points, including recommendations on increasing inhaled corticosteroids and initiating oral corticosteroids, were consistently effective in reducing admissions and A&E attendances [32].

Behavioural change techniques
One 3* meta-analysis demonstrated that self-management interventions that incorporated specific behaviour change techniques reduced unscheduled care and improved control [43]. Meta-regression of the data from the 38 RCTs (7883 participants) concluded that active involvement of participants in the intervention was a key factor in reducing unscheduled healthcare [43]. More specifically, identifying individual behavioural traits (e.g. rebelliousness, low perceived emotional support) in adolescents enabled targeted use of behavioural change techniques [54]. A goal-setting approach proved challenging to implement in primary care settings [63].

Technology
Two 1* narrative reviews investigated computer-or internet-based interactive self-management programmes [28,47]. The effect on healthcare utilisation was inconsistent, confirmed by a recent review identified in the pre-publication update [62], though both showed improvement in symptoms [28] and/or quality of life [28,47]. Two update RCTs of web-based selfmanagement programmes for adolescents also showed improved asthma control [54,57], and an extended follow-up of RCT participants concluded that these effects could be sustained 18 months after conclusion of the trial [59]. Several school-based programmes used technology-based interventions to improve control and reduce absenteeism [30]. Supported self-management using mobile phone technology currently has a limited and inconclusive evidence base [42,45], though a recent RCT in pregnancy demonstrated improved asthma control and quality of life [68].

Which contextual factors influence effectiveness?
Resonating with the concept of 'optimal' self-management (education, an action plan and regular review) [31], a 3* meta-analysis identified that omitting regular review (1 RCT) or reducing intensity of education (1 RCT) was associated with a smaller reduction in unscheduled consultations [36]. A 2* meta-analysis analysed the findings of 18 RCTs (3006 participants) according to the components of the Chronic Care Model [92].
Interventions that included all four components had a greater effect on adherence to inhaled corticosteroids compared to trials including self-management unsupported by the organisational components [33].

Organisational role in promoting supported self-management
A 3* narrative review of 14 RCTs (4588 participants) concluded that proactive organisational systems can increase action plan ownership by promoting uptake of asthma reviews and implementing (and monitoring) structured management systems for asthma care [37]. A recent RCT of a structured approach to self-management education in both primary care and specialist units improved asthma control and reduced unscheduled care [65], and a large cluster RCT at low risk of bias showed an increased adherence to guidelines and reduced asthma symptoms by systematically providing individualised prompts to general practitioners and parents of children with asthma [52].
Automatically linking an action plan to prescriptions given to patients being discharged from A&E improved clinician management and patient uptake of steroid courses [50].
What is the effect of supported self-management on healthcare utilisation and costs?
The RECURSIVE meta-analysis confirmed that selfmanagement support interventions for people with asthma are associated with significant improvements in quality-of-life outcomes (SMD 0.26, 95% CI 0.12-0.39), significant small decreases in hospitalisation rates and costs (SMD −0.21, 95% CI −0.40 to −0.01), significant small decreases in A&E visits (SMD −0.25, 95% CI −0.49 to −0.01), and non-significant small increases in total healthcare costs (SMD 0.13, 95% CI −0.09 to 0.34). Figure 3 shows a Forest plot of the total costs.
What is the evidence that supported self-management for asthma can reduce costs without compromising outcomes? Figure 4 shows the overall permutation plot of the studies (n = 21) reporting data on both quality of life and healthcare utilisation. The majority of the studies on quality of life versus costs related to hospitalisations and A&E attendances were in the right-down quadrant, indicating cost-effectiveness (reduced healthcare utilisation and improved quality of life). However, in terms of total costs (n = 7), the picture was mixed with more studies around zero and the right-up quadrant, indicating that similar costs or small cost increases are necessary to achieve better quality of life.
What is the evidence that supported self-management for asthma is cost-effective?
Four studies applied formal economic analyses; two showed that self-management support interventions were dominant (i.e. significantly better health outcomes with significantly lower costs) [72,86], and two produced non- Fig. 3 Meta-analysis of total costs. CI confidence interval, ES effect size Fig. 4 Permutation plot. Quality of life (x-axis), hospitalisations (y-axis blue) and total costs (y-axis red). In this permutation plot, the effects of self-management interventions on outcomes (quality of life) and utilisation (hospitalisations and total costs) can be visualised simultaneously by placing them in quadrants of the cost-effectiveness plane depending on the pattern of outcomes. Such plots identify studies in the appropriate quadrant (i.e. those that reduce costs without compromising outcomes) and those in problematic quadrants (i.e. those that reduce costs but also compromise outcomes, or those that compromise both outcomes and costs).
significant ratios between costs and benefits at levels likely to appeal to decision-makers (better outcomes with nonsignificant increases in costs) [75,89] (see Additional file 5 for more details). Thus, the benefits derived by supported self-management interventions are associated with reductions in key areas of healthcare utilisation such as hospitalisations and A&E attendances and can be delivered at similar levels of total costs to usual care.

Summary of findings
Extensive evidence (n = 270 RCTs) derived from a broad range of demographic and healthcare settings reveals that supported self-management can reduce hospitalisations, A&E attendances and unscheduled consultations, and improve markers of control and quality of life for people with asthma. Core components of effective selfmanagement are education, provision of an action plan and the support of regular professional review. Effectiveness has been demonstrated in diverse cultural, clinical and demographic groups, with evidence that tailored programmes have greater impact than generic interventions. A range of modes of delivery (including telehealthcare) may be employed to suit preferences and context. The cost of providing self-management support is offset by a reduction in hospitalisations and unscheduled healthcare.

Interpretation of findings
The literature on asthma self-management is particularly well developed and may thus be an exemplar for other LTCs [13,14]. The 16 systematic reviews reporting effectiveness were typically large (five included data from >5000 participants [27,30,31,41,43]) and had consistently positive results, suggesting a mature evidence base, unlikely to be influenced by further trials. Outcomes in subgroups were more often the subject of the update RCTs as the field moves on from demonstrating overall effectiveness to investigating the impact in specific target groups [48-58, 60, 61, 72], demographic contexts [52][53][54]66], or mode of delivery [54,59,62,72].
Self-management support for asthma is a complex intervention and successful interventions were multicomponent, including education, trigger avoidance, teaching self-monitoring, optimal treatment strategies, promotion of adherence and behaviour change techniques, many of which are common to self-management in other LTCs [6]. Appropriately in a variable condition [4], the hallmark of asthma self-management is the provision of an action plan with advice on recognising and responding to deterioration in control [4,32]. People with asthma, however, have broader concerns as they accommodate the condition within their lives and the action plan needs to be embedded in support for 'living with asthma' [93].
Self-management support interventions are an integral component of high-quality care for people with LTCs [8][9][10]. Several of the systematic reviews demonstrated the synergy between self-management education and regular clinical review [31,33,36], and supported selfmanagement is most effective when delivered within a proactive asthma management programme [33,37,65], or integrated within organisational routines [50,52]. Only a minority of trials had follow-up periods over 12 months, and studies are needed to confirm long-term sustainability. Costs associated with self-management interventions are similar to usual care.

Strengths and limitations
Meta-reviews have some intrinsic strengths and limitations. The methodology enables the efficient review of a large body of evidence and thus provision of a comprehensive overview to inform policy and practice. However, it relies on the quality of the included systematic reviews (e.g. comprehensive search strategies, accurate data extraction and synthesis). We used the validated R-AMSTAR instrument to assess the quality of included systematic reviews [17]. In contrast to GRADE [98] (now recommended by the Cochrane Handbook [15]), R-AMSTAR assesses the overall quality of the review, rather than assessing the quality of evidence individually for each outcome.
Re-synthesising materials that have already been synthesised risks further loss of detail and has the potential for erroneous assumptions, especially if the primary focus of the review did not directly align with the questions of the meta-review. Overlap between the RCTs included in the systematic reviews may result in undue emphasis on commonly cited papers.
Whilst some reviews and update RCTs directly compared interventions with or without specific components [23-25, 32, 36, 39, 43], or a specific mode of delivery [28,29,41,45], often the different interventions were compared to usual care, allowing only indirect comparison [31,33,35,37,42,46,47]. A further limitation is that 'usual care' is rarely defined in RCTs [99], and the definition is even more unclear at metareview level. Typically usual care is enhanced in the context of a trial, reducing the apparent impact of an intervention [100].
Systematic reviews are only as current as their most-recent search, and meta-reviews add an additional time delay. In the PRISMS meta-review we therefore not only updated our search for systematic reviews, but also searched for RCTs published after the date of the last search used by the included systematic reviews. In addition, prior to publication we undertook forward citation on all the included systematic reviews, which identified two recent systematic reviews and six RCTs [61][62][63][64][65][66][67][68]. None of these changed our conclusions, confirming the maturity of the evidence base.
The two reviewers who undertook the screening and data extraction were not working independently; however, both projects ensured all the reviewers were fully trained and instituted random checks at every stage. Restricting inclusion to reviews with extractable RCT data maintained the quality of evidence, but may have resulted in some lower-grade but useful evidence being rejected.
RECURSIVE was not restricted to formal costeffectiveness studiesit had a broader focus and included studies reporting data on healthcare utilisation only, without a full effectiveness analysis including costs and quality of life. Some of the RCTs in the RECURSIVE meta-analysis used a more comprehensive definition of 'total costs' (e.g. based on societal perspective) compared to others; to account for this inconsistency, we also present the results on key sources of costs such as hospitalisation and A&E attendance rates.
The PRISMS and RECURSIVE teams worked independently, but met regularly throughout the studies to optimise synergies. A further strength was the multidisciplinary team, including backgrounds in public health, general practice, epidemiology and health psychology, enabling a balanced interpretation.