Systems approaches to scaling up: a systematic review and narrative synthesis of evidence for physical activity and other behavioural non-communicable disease risk factors

Abstract

Background

Non-communicable diseases (NCDs) are the leading causes of death worldwide. Systems approaches have potential for creating sustainable outcomes at scale but have rarely been used to support scale up in physical activity/nutrition promotion or NCD prevention more generally. This review aimed to: (i) synthesise evidence on the use of systems approaches in scaling up interventions targeting four behavioural risk factors for NCDs; and (ii) to explore how systems approaches have been conceptualised and used in intervention implementation and scale up.

Method

Seven electronic databases were searched for studies published 2016–2021. Eligible studies targeted at least one of four NCD behavioural risk factors (physical inactivity, tobacco use, alcohol consumption, diet), or described evaluation of an intervention planned for or scaled up. Studies were categorised as having a (i) high, (ii) moderate, or (iii) no use of a systems approach. A narrative synthesis of how systems approaches had been operationalised in scale up, following PRISMA guidelines.

Results

Twenty-one intervention studies were included. Only 19% (n = 4) of interventions explicitly used systems thinking to inform intervention design, implementation and scale up (targeting all four risk factors n = 2, diet n = 1, tobacco use n = 1). Five studies (‘high use’) planned and implemented scale up with an explicit focus on relations between system elements and used system changes to drive impact at scale. Seven studies (‘moderate use’) considered systems elements impacting scale-up processes or outcomes but did not require achieving system-level changes from the outset. Nine studies (‘no use’) were designed to work at multiple levels among multiple agencies in an intervention setting, but the complexity of the system and relations between system elements was not articulated. We synthesised reported barriers and facilitators to scaling up, and how studies within each group conceptualised and used systems approaches, and methods, frameworks and principles for scaling up.

Conclusion

In physical activity research, and NCD prevention more broadly, the use of systems approaches in scale up remains in its infancy. For researchers, practitioners and policymakers wishing to adopt systems approaches to intervention implementation at scale, guidance is needed on how to communicate and operationalise systems approaches in research and in practice.

Trial registration

PROSPERO (CRD42021287265).

Introduction

Non-communicable diseases (NCDs), such as cardiovascular disease, cancer and type 2 diabetes, are the leading causes of death worldwide, contributing to more than 41 million deaths globally [1]. Major behavioural risk factors for NCDs include physical inactivity, tobacco use, alcohol consumption and an unhealthy diet [2]. The high burden of NCDs causes substantial economic losses worldwide, and deaths from NCDs disproportionately affect low- and middle-income countries [1]. Whilst there are multiple targeted action plans for the prevention of NCDs, such as the Global Action Plan for the Prevention and Control of Non-communicable Diseases 2013–2030 [2], and for specific risk factors such as the Global Action Plan for Physical Activity (GAPPA) [3] and Global Alcohol Action Plan 2022–2030 [4]; many of the NCD risk factors have reached pandemic proportions. For example, annually, tobacco use causes over 8 million deaths https://www.who.int/news-room/fact-sheets/detail/tobacco, around 3.2 million deaths are attributed to physical inactivity [5], over 3 million deaths result from the harmful use of alcohol [6], and 2.8 million deaths are as a result of being overweight or obese [7]. By 2030, global health-care costs of physical inactivity alone are estimated to exceed INT$520 billion [8]. Increasing implementation of evidence-based solutions to reduce the risk of NCDs at scale is a thus global priority of the World Health Organization (WHO) [9]. Despite burgeoning evidence for the effectiveness of different interventions to prevent NCDs; researchers, practitioners and policymakers have limited access to effective ways of scaling up NCD risk factor interventions, which are essential for global shifts in health [10, 11].

Scaling up presents a complex set of challenges. They are complex, not only due to the factors underpinning NCD risk factors, but also in the nature of the processes required to achieve impact at large scale. Interventions intended for scale up should thus be planned with consideration of complexity [12,13,14]. Addressing one element within a complex public health problem (e.g., through a discrete intervention targeting a specific factor) is unlikely to achieve desirable population-wide effects at scale [10]. For physical inactivity, intervention effects can be attenuated at scale [15], and yet the mechanisms underpinning outcomes of scaling up, which may contribute to these attenuated effects, are often complex and poorly understood [16]. Large-scale interventions span many different community contexts and adapted in response to these contexts, posing problems for the attribution of impact during evaluation [12]. The scaling process itself often involves multiple delivery strategies or systems, and the vast number of settings, contexts and systems affected during scale-up can extend beyond the capacity for data collection [13]. This lack of knowledge of the complex interactions between factors when scaling up poses difficulties when generalising ‘effective’ approaches to successful scaling.

Systems approaches provide a framework for exploring the multitude of interdependent elements that influence a problem, and the scaling up of solutions for that problem, as they can help establish the relations between factors, how they change over time, and acknowledge that effective actions are required across political, social, cultural, economic and scientific domains within the system [17]. All the NCD risk factors themselves have a multitude of interdependent elements that are complex, interconnected and have interacting influences [18,19,20]. Traditional, linear ‘blueprint approaches’ to scaling up that are observed in global health initiatives may not adequately fit the dynamic and unpredictable ways in which health services, organisations, and communities expand and are sustained [21].

Complexity and systems theory can make a valuable contribution to understanding and addressing population health scale up [13]. Systems approaches are also theorised to be influential in creating sustainable outcomes at scale [13, 14]. However, in public health generally, there is limited evidence for the impacts of adopting a systems approach, with a paucity of detailed descriptions of their operationalisation. In health systems research, a meta-analysis of 35 studies identified a significant improvement to patient and service outcomes when a systems approach informed healthcare design and delivery [22]. However, in NCD behavioural risk factor research, systems approaches have had mixed adoption. For example, in tobacco cessation research, for almost 30 years systems approaches have been recognised and systems level strategies applied in health services [23], whereas for physical activity there are few well described examples of interventions that have been planned, delivered and evaluated using systems approaches and systems analysis methods [24].

For researchers, practitioners and policymakers wishing to achieve population level impact of interventions, and adopt a systems approach to intervention implementation at scale, greater guidance is needed on ways to achieve this and how to operationalise systems approaches in the context of public health scale up. Given that in public health generally, evidence to demonstrate the value of applying a systems approach is still emerging [25], and there is little research that has examined systems-based practice [26] or how systems approaches are applied in scaling up in public health [13, 14]; we sought to contribute to addressing these gaps in the current review.

The objective of this paper is to synthesise the evidence of how systems approaches have been used to inform scaling up in physical activity. However, to explore scaling up of interventions targeting behavioural risk factors for NCDs more comprehensively, the scope was broadened to include the three other key NCD behavioural risk factors: tobacco use, alcohol consumption, and diet. As NCD prevention often operationalises ‘diet’ in terms of obesity, we consider diet and obesity jointly; herein ‘diet/obesity’. Given the current lack of published evidence describing implementation of systems approaches in public health [27], and the fact that interventions targeting physical activity are often designed as multicomponent interventions targeting multiple NCD risk factors combined (i.e., physical activity and diet), this strategy also provides an opportunity to learn from other areas of public health.

The specific aims of this review are as follows. Firstly, to identify how systems approaches have been used to inform and understand: (i) approaches and strategies to scaling up interventions targeting four behavioural risk factors for NCD; and (ii) barriers and facilitators to scaling up of these interventions. Secondly, how the term ‘systems’ has been conceptualised and used in the broader context of intervention implementation and scale up in these studies will be identified.

Methods

This review was prospectively registered with PROSPERO (registration number CRD42021287265) and follows the Preferred Reporting Items for Systemic Reviews and Meta-Analyses (PRISMA) guidelines [28]. The PRISMA checklist is presented in Additional file 1.

Definition of a systems approach to scaling up

Systems-based public health is an evolving field, with no widely agreed definition of what it entails [22, 29]. A ‘systems approach’, however, is generally understood as an approach that takes into account a multiplicity of interacting factors across a system, and the ways in which that system responds and adapts to interventions within it [17]. ‘Scale up’ refers to “deliberate efforts to increase the impact of successfully tested interventions, to benefit a greater number of people and to foster policy and programme development” [30]. Whilst scale up processes commence once an intervention has been developed, in practice settings, scale up can occur before an intervention has been tested within a research trial [31]. ‘Dissemination’ is another term often used when referring to at-scale implementation, as it refers to an active approach to spreading an evidence-based intervention using planned strategies [32]. However, unlike ‘scale up’, dissemination need not include efforts to maximise the scale of interventions. More recently, a ‘systems approach to scale up’ has been defined as an approach that “prioritises the behaviour and function of the system, with a focus on relations between a number of system elements, using system-level levers and dynamic system changes to drive impact at scale” [14]. A systems approach to scale up emphasises that scale up activities (i.e., activities such as obtaining and maintaining the resources and implementation capacity needed for large-scale intervention delivery), should focus on generating changes within a system to achieve the desired population level outcome [14]. For example, the characteristics of the target system(s) that scaling occurs within are considered from the outset of scale up planning, in order to identify how best to reorientate that system to achieve the desired impacts on health [14]. Given that there is huge variance in definitions for terminology used in systems science more broadly [29], for the purposes of this review, we use the definition of a systems approach to scaling up from Koorts and Rutter (2021) [14], which informs the analytical framework for data synthesis and interpretation of our findings.

Eligibility criteria

Studies were eligible for inclusion based on the following criteria: 1) they involved interventions with a primary outcome targeting at least one of the four main behavioural risk factors for NCDs (physical inactivity, tobacco use, alcohol consumption and diet); and 2) interventions were conducted/planned for implementation in a real-world setting with a focus on scale up/scale up outcomes (effectiveness, scale-up, dissemination, translation or implementation studies [including randomised controlled trials with a focus on scale up/scale up outcomes], or protocols) or an evaluation of a previously scaled intervention (i.e., scaled up in a real-world setting). Eligible studies must have also included the term ‘system(s)’ and described either the approach/strategy taken during scale up or barriers and/or facilitators to the scale up process or outcomes. Exclusion criteria included: 1) studies testing intervention efficacy only or without a focus on scale up/scale up outcomes (e.g., randomised controlled trials, feasibility, and pilot studies); 2) reviews; and 3) studies applying or testing a policy (i.e., no intervention was implemented). For the purposes of this review, we defined an ‘intervention’ as “a set of actions with a coherent objective to bring about change or produce identifiable outcomes” [33], and excluded those described as a policy, strategy or government regulation.

Information sources and search strategy

The following online databases were searched online for peer reviewed English language articles published on or after January 1st, 2016, until 31st October 2021: EBSCOHost, Medline, CINHAIL, Sportdiscus, Global Health, PsychINFO and EMBASE. This search time frame (2016–21) was chosen in response to recent calls in public health for the use of systems approaches to address complex population level problems [20], and it would enable us to capture more recent interventions that were scaled post publication of key global action plans (e.g., [2, 3] and [4]). Grey literature was searched via Google Advance and the first ten pages were screened for inclusion. The search strategy (Additional file 2) was developed and tested in consultation with a Deakin University research librarian, drawing on previous systematic reviews of related topics (e.g., [22, 34,35,36]), and informed by the PICO(T) (participants/population; intervention; comparator; outcomes, time) methodological approach, as recommended by Cochrane reviews [37, 38]. In this paper, Participants/populations were any age; the Intervention needed to target at least one of the four main behavioural risk factors for chronic disease (physical inactivity, tobacco use, alcohol consumption, diet); the Comparator was not required as this review focused on studies conducted in real-world settings; the Outcomes included the approaches and strategies to scaling up, barriers and facilitators experienced, and how ‘systems’ has been conceptualised and used; and the Time was 2016–2021.

Study selection

Search results were imported into data management software Covidence (https://www.Covidence.org), and duplicates from the search were automatically removed. One researcher (NR) screened article titles against inclusion and exclusion criteria. All abstracts and full texts were screened by two authors independently (CS, JM). Where discrepancies in study inclusions occurred, a consensus agreement was made by four authors (CS, JM, HK and KB). Where there was incomplete information to determine scale up approaches or scale up frameworks used, reference lists and forward searching was undertaken by JM and HK.

Quality appraisal

The Mixed Methods Appraisal Tool (MMAT) version 2018 [39] was used to appraise the quality of included studies as it encompasses multiple study designs (e.g., qualitative research, randomized controlled trials, non-randomized studies, quantitative descriptive studies, and mixed methods studies) which was reflective of the included study designs. JM and KB conducted the quality appraisal independently using the MMAT screening questions and relevant checklist questions (by each of the five study design categories) in the Covidence software. Options for each question were yes, no or can’t tell. Any disagreements in quality appraisal were resolved by discussion between JM and KB and coming to a consensus decision. As per MMAT instructions, an overall score from each question was not calculated [39].

Data extraction

Data were extracted independently by two authors (CS, JM), with other authors (HK, KB) consulted for clarification where necessary. Data extraction included: author, year of publication, title, country or region of intervention scale-up, study design, aim, adaptions made (e.g. intervention, approach, setting), level of scale up (e.g., state/national level), intervention duration, target population, target behaviour (physical inactivity, tobacco use, alcohol consumption, diet), implementation setting, target intervention, name/number of organisations/stakeholders involved in scale up, role of organisations/stakeholders involved in scale up (e.g., funder, evaluator), how the term ‘system(s)’ has been conceptualised and used in the broader context of intervention implementation and scale up, framework/definition of scale up, method/approach to scale up, evaluation design for intervention effectiveness and evaluation of scale-up, data collected, and reported barriers and facilitators to scale up. Extracted data were tabulated (by JM and HK) to present study characteristics and results.

Data synthesis

Following guidance on narrative synthesis methodology [40], JM created a qualitative textual description for each included study, containing information on the scale up approach described, and how the term ‘systems’ and systems approaches were conceptualised, used or informed each study. Studies were required to have included the term ‘system(s) in order to be eligible for inclusion in the review and were not required to have described or implied any use of a systems approach to scaling up. Based on the textual descriptions and in accordance with the main aims of this review, two authors (HK and KB) independently created an initial set of categories (n = 8 and n = 10, respectively) to capture how systems approaches had been used to inform and understand: (i) approaches and strategies to scaling up public health interventions; (ii) barriers and facilitators to scaling up; (iii) the evaluation of scale up processes and outcomes; and (iii) how the term ‘systems’ was conceptualised and used in each study. Initial categories were discussed and refined by HK, KB, CS and JM until consensus was reached, to produce a final set of six categories that comprised the final analytical framework for data synthesis (see Table 1 below).

Table 1 Characteristics of high, moderate and no use of systems approaches to scaling up, based on [14]

Based on the definition of a systems approach to scaling up [14], the six categories within the analytical framework were assigned to one of three groups, according to their alignment with the definition: (Group 1; Analytical framework category 1 and 2) High use of systems approaches in scaling up (e.g., studies in which systems thinking informed the intervention design, implementation and scale up approach, the intervention had a focus on system changes); (Group 2; Analytical framework category 3 and 4) Moderate use of systems approaches in scaling up (e.g., studies in which the intervention had a focus on system changes or the role/influence of system factors, but did not explicitly adhere to the definition of a systems approach to scale up) and; (Group 3; Analytical framework category 5 and 6) No use of systems approaches in scaling up (e.g., studies in which the intervention may involve multiple strategies in multiple settings/sectors targeting different levels, but did not explicitly adhere to the definition of a systems approach to scale up). Table 1 presents the six categories within the analytical framework, against the three levels of a systems approach to scaling up.

Using the analytical framework (Table 1), a narrative synthesis was undertaken by JM with HK. Key findings and descriptions in each study were coded in line with the analytical framework. Themes were used to describe the approach to scale up and conceptualisation/use of the term ‘systems’ and a systems approach. Comparisons between themes were identified to present a synthesis of findings across studies, individual findings are only reported where themes were unique to an individual intervention or study. Qualitative textual descriptions relating to the barriers and facilitators to scaling up were summarised using an inductive thematic analysis approach. Descriptions were aggregated according to major themes, which informed the key barriers and facilitators reported in the Results.

Results

The search generated 22 eligible papers, corresponding to 21 intervention studies (two papers [protocol and outcomes] addressed the same intervention). Figure 1 presents the PRISMA flowchart, which displays the number of studies screened, assessed, and included/excluded for the final review. Results reported correspond to data contained in the 22 eligible papers.

Fig. 1

PRISMA flow diagram

Study sample characteristics

A description of the included papers (n = 22) is presented in Table 2. Of the 21 intervention studies we included in this review, 19 were discrete interventions [12, 41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60], one study involved evaluation of 16 discrete interventions [61], and one study involved evaluation of six recommended evidence-based strategies [62]. For the purposes of this review, studies that included multiple discrete programmes or strategies (e.g., [61, 62]), are reported as one intervention. Twenty of the 21 interventions were already being implemented at scale, only one intervention was planned for scale up in a real-world setting, with their evaluation focussing on potential scalability for future roll out [59].

Table 2 Study characteristics

According to World Bank categorisations [64], the 21 interventions were implemented in five high income countries (United States of America [41, 42, 49, 56, 57, 61, 62], Australia [43, 44, 48, 50, 51, 59, 60], Canada [45, 47, 52, 53], New Zealand [12] and the UK [58]), and one low income country (Ghana [46, 63]).

Seven interventions targeted physical activity and diet [27, 42, 44, 47, 57, 58, 60], four targeted diet [41, 45, 51, 59], four targeted physical activity [50, 52, 53, 62], two targeted tobacco use [49, 56], two targeted all four behaviour risk factors (physical inactivity, tobacco use, alcohol consumption and diet) [12, 48], and one targeted diet and malnutrition [46]. One study targeted physical activity, tobacco use, diet, UV exposure and preventive care (e.g., cancer screening and human papillomavirus vaccination) [61], but the tobacco control and HPV vaccine elements of the intervention were reported as under way and were not included by the authors as part of their evaluation [61].

Interventions targeted health improvements or behaviour change among children/adolescents and families [41,42,43,44, 46, 47, 50, 51, 53, 55, 60, 61, 63], older adults [52], smokers [49, 56], men [58], and general community members, including community based partner organisations/practitioners (e.g., food banks and the YMCA) [12, 45, 47,48,49, 57, 59, 61, 62]. Interventions targeted multiple community-based settings, such as schools [41, 43, 44, 46, 50, 51, 53, 61, 63], childcare settings [44, 47, 61], faith-based organisations [41, 57], local health districts and local governments [44, 48], workplaces [43], professional football clubs [58], supermarkets [59], coordinated care organisations [49], cancer centres [56], and in the community at a broader environmental/policy level [12, 42, 45, 48, 49, 52, 60,61,62].

Of the 22 papers reviewed, there was one qualitative study [45], one type 2 hybrid effectiveness implementation trial [52], one group randomised trial [57], two quasi-experimental designs [43, 51], three case studies [12, 44, 48], four cluster randomised controlled trials [46, 50, 53, 63], five mixed method evaluations [41, 47, 59, 61, 62], and five summary/descriptive articles describing intervention and scale up processes/outcomes generally [42, 49, 56, 58, 60].

Approaches and strategies to scaling up

There were varying approaches and strategies used to reach at-scale implementation (Table 2). Fifteen (71%) of the 21 interventions were described as ‘designed for scale’. These interventions were developed based on prior evidence from other programmes, and scaling required strong research-practice partnerships and existing resources customised for stakeholders. Of the 15 interventions designed for scale, 10 (67%) were implemented without the reported need for small scale pilot trials, and were scaled across a single community [62], multiple communities within one state [48, 60, 61], across a whole state/province [41, 44, 47], across multiple states [42], or nationally [12, 45]. Five (33%) of the 15 interventions followed a traditional translation pathway of a phased approach from small-scale controlled efficacy testing to plan for scale up [59], and from small-scale controlled efficacy/feasibility testing to then effectiveness and implementation testing at a state [50, 57] and province level [52, 53].

Four interventions were expanded due to earlier effectiveness outcomes, and ongoing government funding and support. Two of these three interventions started as smaller community projects that were incrementally expanded to reach multiple communities or settings state-wide [43, 51], and two began as pilot trials that were incrementally expanded to reach nationally [46, 58]. Two interventions targeting tobacco reduction were scaled as part of existing health system infrastructure. One was the extension of more than two decades of established state level policies and evidence-based initiatives [49], and the other was integrated as part of an existing large, comprehensive health system [56].

Barriers and facilitators to scaling up

Forty-one textual descriptions relating to barriers and facilitators to scaling up were summarised from nineteen included studies (Table 2). Six themes emerged, which included four facilitators (committed stakeholder engagement, capacity building in the local workforce, flexibility in delivery and implementation, and ongoing programme feedback and improvement) and two barriers (competing interests and priorities, and insufficient time and resources).

Facilitator 1: Committed stakeholder engagement

Full engagement of stakeholders was key for scale-up success, as reported in several studies [12, 41,42,43, 45, 47,48,49, 53, 56,57,58,59,60, 62]. Early involvement and participation of stakeholders in the planning and implementation of initiatives were also identified as important [12, 57, 58]. End users and leadership were the most mentioned stakeholders/aspects that played a key role in the scale up process. A weight management programme targeting men with overweight or obesity leveraged the popularity of professional football among target users, and existing organisational structure of local football clubs, as delivery systems to provide a sustainable model for wider implementation of the programme [58]. Several studies reported that the initiatives, and implementation process, were driven by the local contexts and needs and resulted in scale-up successes [12, 41, 43, 48, 62]. Strategically engaging leadership and facilitating alignment in goals and missions across different organisations and communities was associated with a greater degree of collective actions [12, 45, 56, 59, 60].

Facilitator 2: Capacity building in local workforce

Efforts in capacity building [12, 41, 53, 57, 58] were common in the included scale-up initiatives. Capacity building for the implementation workforce was necessary, especially in systems thinking guided initiatives. This could include providing adequate resources and funding to support existing infrastructures. Understanding and responding to the adaptive nature of systems were recognised as a key skillset for local workforce [12, 48].

Facilitator 3: Flexibility in delivery and implementation

Initiatives that had a multi-component or flexible delivery model were more acceptable to local implementation as they allow for autonomy [12, 53]. The flexibility also enabled initiatives to be appropriately adapted to fit local context and place characteristics [58, 62]. By contrast, rigid and controlled approaches in the initiatives appeared to hinder implementation [45, 48].

Facilitator 4: Ongoing programme feedback and improvement

Establishing and utilising an ongoing evaluation and feedback system was perceived instrumental in scale up [57, 59, 60]. In a whole-of-community child obesity prevention programme, ongoing monitoring and feedback of programme achievements to community organisations informed quality improvement and innovation, which resulted in a significant increase in the promotion of community settings implementing the evidence-based practice [60].

Barrier 1: Competing interests and priorities

Implementation of initiatives was challenged by conservative/risk averse attitudes of local organisations [45], competing priorities of local organisations [43], and competing interests of industries [12, 59]. For example, organisations were reluctant to adopt the initiative due to a low perception of risk versus benefits, i.e., potential low relative advantage [45]. Another healthy supermarket intervention described the challenges in promoting sales of healthy foods due to perceived customer demand and supplier contract agreements [59].

Barrier 2: Insufficient time and resources

Lack of appropriate resources such as funding was identified as a barrier to successful scale-up [12, 45, 53, 55, 59]. For one intervention that targeted different points of the system [12], siloed and competitive funding approaches in the government impacted the financial security of the initiative.

Use of systems approaches to scaling up

Tables 3, 4 and 5 summarise how studies categorised in the high, moderate and no use of a systems approach groups, respectively, articulated how ‘systems’ was conceptualised or used, in what ways a systems approach was adopted, and the methods/theoretical frameworks or principles applied to study scale up processes or outcomes.

Table 3 Studies categorised as ‘high use of a systems approach’ (n = 5)

Table 4 Studies categorised as ‘moderate use of a systems approach’ (n = 7)

Table 5 Studies categorised as ‘no use of a systems approach’ (n = 9)

High use of a systems approach to scaling up

Five studies were categorised as demonstrating high use of a systems approach to scaling up [12, 44, 48, 51, 56] (Table 3). Interventions in this group targeted physical activity and diet (n = 2), diet (n = 1), tobacco use (n = 1) and all four behavioural risk factors (n = 1). Table 6 provides a case example of a high use systems approach. In these studies, systems thinking informed the intervention design or implementation and scale up approach, or the intervention had a focus on system changes. In these studies, there was an explicit focus on relations between system elements and using system changes to drive impact at scale. Four of the five studies explicitly used systems thinking to inform the intervention design, implementation, and scale up processes, with the goal of systems change (Table 3).

Table 6 Case example of a ‘high use’ systems approach to scaling up

Conceptualisation and use of ‘systems’ and a systems approach

All five studies in the high use group involved engagement and collaboration with communities and stakeholders in the process and evaluation of scaling up. Four studies explicitly used systems thinking to inform the intervention design, implementation, and scale up processes, with the goal of systems change [12, 48, 51, 56]. Systems thinking was reflected in this process by exploring connections, priorities, and common interests between community groups and organisations [48, 56], identifying systems elements and existing infrastructure that supports scale-up [51], conceptualising systems change in the prevention paradigm and infrastructure [12], and quantifying the connection between systems players and how it contributes to the success of scale-up [44].

Methods, theoretical frameworks and/or principles adopted to study scale up processes or outcomes

Three studies in the high use category employed theoretical frameworks to guide the systems thinking practice. One obesity prevention study applied the Analysis Grid for Environments Linked to Obesity (ANGELO) framework [65] to guide the development and implementation of scale-up, defining the key priorities while considering existing capacity within a system [51]. Another study adapted the Cancer Care Continuum as a systems framework to identify the necessary resources and players needed for tobacco treatment integration in each of the cancer care stages (e.g., prevention, diagnosis) [56]. This same study used the Consolidated Framework for Implementation Research (CFIR) [66] to describe and evaluate implementation. The WHO’s Prevention System Building Blocks [67] was applied to one study that targeted all four behavioural risk factors, to identify facilitators and areas for improvement in scale-up of obesity prevention in the domain of workforce, leadership, relationships and networks, resources, and knowledge and data [12]. Three studies in the high use category defined the measures or system level changes resulting from the scale up process [12, 44, 48]. For example, one study used a quantitative systems analysis approach (i.e., social network analysis) to evaluate the diffusion of knowledge of the intervention among stakeholder groups [44], whereas another developed ad-hoc qualitative indicators to capture the presence and/or absence of system-level changes from stakeholders’ perspectives [12].

Moderate use of a systems approach to scaling up

Seven studies were categorised as moderate use of a systems approach to scaling up [41,42,43, 49, 52, 53, 57] (Table 4). Interventions in this group targeted physical activity and diet (n = 3), physical activity (n = 2), diet (n = 1), and tobacco use (n = 1). Table 7 provides a case example of a moderate use systems approach. All studies in the group, as described in the papers, recognised the importance of systems thinking in resolving scale-up challenges, however, systems thinking was not embedded from the outset to identify what would be required to achieve system-level changes. System analysis methods were not necessarily used to study or address barriers and facilitators to the scale up process or outcomes. In these studies, there was an implicit focus on systems change, as reflected in the intervention design and evaluation that targeted multiple sectors and components.

Table 7 Case example of a ‘moderate use’ systems approach to scaling up

Conceptualisation and use of ‘systems’ and a systems approach

For the moderate use group, where systems were involved during scale up, this included via the intervention: (i) targeting different system levels (i.e., targeting multilevel organisations) (e.g., [41, 43]), or points of the system (i.e., improving system barriers to access and to affordability) (e.g., [49]); (ii) involving organisations that had an influence at different system levels (e.g., schools) (e.g., [53]); or (iii) including intervention strategies that had a specific focus on system changes at an organisational (i.e., changing faith based organisations’ engagement in health promotion) (e.g., [57]) or policy/environmental level (i.e., empowering communities) (e.g., [42]). For example, two studies recognised the wide range of determinants on diet and/or physical activity and designed strategies to promote an environment that supports health behaviours (e.g., improve built environment for more opportunities in physical activity) [42, 43]. Further, one study recognised the interconnectedness of multiple sectors in smoking cessation and planned for accountability and incentives to support the continued partnership to improve access to cessation services [49].

Methods, theoretical frameworks and/or principles adopted to study scale up processes or outcomes

Of the seven included studies, two studies utilised conceptual models of scaling up (i.e., [30, 68]) to guide implementation of interventions targeting physical activity in older adults [52], and in schools [53]. Four of the seven studies described or defined the system level changes resulting from the scale up process [41,42,43, 52], with a focus on community capacity as an indicator of system level changes. All studies utilised a series of strategies (e.g., educational activities) across multiple sectors (e.g., community, education) to strengthen the community capacity that supports healthy eating and/or physical activity. None of the studies, as reported in the papers, mentioned the measure on the change in community capacity.

No use of a systems approach to scaling up

Nine studies were categorised as not using a systems approach to scaling up [45,46,47, 50, 58,59,60,61,62] (Table 5). Interventions in this group targeted physical activity and diet (n = 3), diet (n = 2), physical activity (n = 2), diet and malnutrition (n = 1), and physical activity, tobacco use, diet, UV exposure, and inadequate preventive care (n = 1). Table 8 provides a case example. These studies adopted an approach to scaling up that included multiple sectors, settings and intervention components, although, as reported in the papers, the complexity of the system and the relations between the system elements were not explicitly targeted or articulated.

Table 8 Case example of a ‘no use’ systems approach to scaling up

Conceptualisation and use of ‘systems’ and a systems approach

Where systems were involved during scale up in the ‘no use’ group, this included via the intervention: (i) targeting different system levels (i.e., targeting multilevel organisations, e.g., [62]) or settings within the system (e.g., [60]); (ii) involving organisations that had an influence at different system levels (e.g., representing food retailers, Non-farmers, advertisers) (e.g., [45, 46, 59]); or (iii) including intervention strategies that had a specific focus on system changes at an organisational level (i.e., changing childcare provider capacity to engage in health promotion or increasing football club coaches capacity to deliver health promotion within their existing organisations) (e.g., [47, 58]). All studies in this group involved multiple strategies in multiple sectors and may have involved targeting different systems; however, the strategies and sectors were not considered in a relational way or tended to focus on one of few points of within the system. For example, in one study, only school meals service and distribution were focussed on, even though the intervention was designed as a multi-sectoral strategy to increase food production, household income, and food security [46]. Several multicomponent interventions did not describe or explicitly target systems change (e.g., [50]), but acknowledged the role and influence of system factors (e.g., [47]). Whilst some studies acknowledged that there must be sufficient organisational or system support for effective scale up, where the term ‘system(s)’ was mentioned, this was referred in context of: ‘evaluation systems’ (e.g., [62]), ‘monitoring systems’ (e.g., [46, 60]), implementation ‘delivery system’ (e.g., [50, 58]), or in the broader general context of the social/childcare/health system that the intervention took place (e.g., [45, 59]). Where ‘systems’ were not explicitly mentioned, factors relevant to systems change could still be utilised (i.e., establishing community infrastructure and developing community action plans to support capacity at scale) (e.g., [61]) (Table 5).

Methods, theoretical frameworks and/or principles adopted to study scale up processes or outcomes

None of the nine studies included in the ‘no use’ group described using a scale up framework. One study retrospectively applied the PRACTIS guide [69], which is a framework that incorporates a systems thinking perspective on effective implementation and scale up, to describe the scale up of an intervention targeting professional football clubs [58]. However, the authors acknowledge that the PRACTIS guide was not available to guide intervention scale up from the initial stages. Concepts within the social system, relevant to Rogers’ Diffusion of Innovations [70], informed the evaluation of two studies, to establish scalability of an intervention [59] and actual scale up outcomes [45]. Five of the nine studies described system level changes resulting from scale up, mainly on outcomes as a result of multiple strategies implemented in interventions [47, 50, 61, 62] or as a result of strong policy alignment [60].

Quality appraisal findings

Quality assessment findings are presented in Additional file 3. In general, there were a lack of details in the reporting of the sampling strategy and whether the study sample is representative of the target population, and the risk of nonresponse bias for quantitative survey was rarely considered. For mixed methods studies, it was often unclear how divergences and inconsistencies between quantitative and qualitative results were addressed.

Discussion

Systems approaches and complexity science have potential for enhancing both the development and the scaling up of health interventions [20, 21, 71], but there is a paucity of knowledge of whether and how systems approaches have been adopted when scaling up in public health [14]. In this paper, we systematically explore the use of systems approaches to scaling up prevention of four behavioural risk factors for NCDs; physical inactivity, tobacco use, alcohol consumption and unhealthy diet. Of the studies included in this review, interventions targeted a mix of behavioural risk factors across different ages and settings, however, almost all studies were conducted in high income countries. Despite the fact that the majority of interventions in this review were described as having been designed for scale, only four (19%) of the 21 interventions explicitly used systems thinking to inform the intervention design, implementation, and scale up processes, with the goal of systems change. Whilst studies often included multiple sectors and intervention components; recognition of the complexity of the system and the relations between the system elements were not explicitly targeted or articulated.

Systems approaches are not imperative for scale up, and there is a lack of robust evidence that a systems approach leads to better outcomes for sustainability and impact at scale. However, this review illustrates that current conceptualisations of what constitutes designing an intervention for scale does not necessarily include a consideration of the impact of systems or principles of systems thinking. This is despite the fact that successful scale up includes when an evidence-based intervention becomes embedded in a system(s) to achieve long-term, sustainable health impact [10]. We also identified that interventions described as designed for scale were reliant on strong research-practice partnerships. Facilitators to scale up included committed stakeholder engagement, capacity building in the local workforce, and flexibility in delivery and implementation. Conversely, barriers included competing interests and priorities, and insufficient time and resources. These barriers and facilitators are consistent with previous scale up research [72, 73]. Studies in this review also reported challenges relating to measuring system changes, including the time required to detect long-term changes [51], and challenges with appropriate measures for population reach and engagement [48], which is consistent with research that explored some of the tensions of scaling up in physical activity [13].

More than two thirds of interventions included were described as designed for scale, whereas less than one third of interventions followed a traditional translation pathway of small-scale controlled efficacy testing, to effectiveness and implementation at scale. This finding suggests that there may have been a move away from the traditional translation pipeline that begins with controlled research trials, to one that considers effective real-world translation, as has been recommended for more than two decades [74]. It also suggests that there is greater consideration of translation and population impact early in the research process, which has also been recommended for over a decade [75]. Nonetheless, despite recommendations for the use of complex systems models in public health [20], recognition of the impact of system factors varied greatly across studies. Of those studies that did adopt a systems approach to scale up (the ‘high use’ group), most of the interventions targeted all four behavioural risk factors for NCDs we included in this review. In this high use group, there was an explicit focus on relations between system elements and using system changes to drive impact at scale. Of the studies that did not adopt a systems approach (the ‘no use’ group), most of the interventions targeted physical activity and diet.

The varied appreciation, adoption and implementation of systems approaches for the four NCD behavioural risk factors we included in this review may reflect the overall lack of scale up approaches that adopted systems perspectives. For example, in physical activity promotion research, systems approaches have largely been underutilised and systems concepts need to be engaged more robustly in physical activity interventions [24]. In physical activity scale up research in particular, systems approaches are perceived as important, but may not be seen as feasible to achieve in practice [10]. In obesity prevention research, which can include interventions targeting physical activity and diet, there is a lack of evidence for whole of systems approaches [34]. Complex systems perspectives have been applied to study alcohol reduction and associated harms, however their application has remained predominantly at the individual or local level [19]. Consistent with public health promotion more broadly [20], alcohol reduction interventions have often been reductionist with a focus on easily modifiable risk factors and high risk groups [76], despite advocacy for a focus on the real-world systems that alcohol consumption and harms are created and shaped by [19]. For smoking cessation, the benefits of approaches that systemically integrate into or incorporate clinical settings have long been recognised, however, their adoption has been inconsistent and implementation slow [77]. For several decades, tobacco control has become increasingly complex and thus the use of systems thinking has long been encouraged [78]. However, the effectiveness of system change interventions on tobacco cessation rates and system level outcomes at scale, remains less clear [79].

Whilst our review showed that socio-ecological models have been an integral part of designing, implementing, and evaluating physical activity interventions, the interconnectedness among these elements in scale up was rarely considered. Implementation of a suite of activities across multiple settings or levels need not mean a systems approach has been adopted [80]. Our findings support this, as studies in both the high and no use systems approach groups included interventions targeting multiple settings and levels. Yet, only studies in the high use group (e.g., [48]) explicitly applied systems thinking in practice and embedded systems approaches from the outset.

Consistent with the definition of a systems approach to scale up that we used in this paper [14], scale up exists on a continuum and scale up need not adopt a systems approach. As is shown in this review, scale up approaches can include linear, intervention-orientated expansive approaches that prioritise the spread of interventions into existing systems, through to approaches that sit within a complex systems paradigm that begin by considering the characteristics of the target system(s) that scaling occurs within [14]. Scaling health interventions in a traditional, linear way through efficacy to effectiveness and scale up trials is well documented (e.g., [73]). Our review draws attention to the fact that there is a lack of published evidence for ways to operationalise systems approaches when scaling interventions, including how to select and apply relevant theories and frameworks. Of the 21 interventions we included, only three employed a theoretical framework to guide systems thinking practice. Of the studies that did, to some extent, adopt a systems approach to scale up (i.e., the high and moderate use groups), system-level change and the scale up processes and outcomes that drive the change at scale were not well defined, and often the reporting was brief and lacked theoretical explanation. When a systems approach was applied and concisely reported, it provided a more comprehensive view of a problem and can help identify potential barriers and opportunities for scaling up interventions (e.g., [56]), however, our quality appraisal highlighted that most studies lacked information on key parameters.

To progress the field and better equip researchers, practitioners and policymakers to invest in efforts for scaling up, it has been recommended that NCD prevention adopts new paradigms and perspectives that incorporate systems thinking [18, 20, 81]. This includes using frameworks that incorporate a systems thinking perspective on how to achieve effective outcomes at scale [13] and adopt new ways of accounting for the complex systems in which interventions are implemented [82]. Complexity and systems theory can be used to understand and approach population health scale-up by considering the system as a whole prior to intervention design, development and implementation. For example, Scaling Readiness assessments involve both quantitative and qualitative data collection, and can be used to ascertain the characteristics of the system that influence prospective interventions prior to investment [83]. The Intervention Scalability Assessment Tool (ISAT) [84] has also been used to assess prospective scalability of interventions into existing systems, through a participatory process with stakeholders. For frameworks that guide scale up planning, the PRACTIS guide [69] adopts a systems thinking perspective to scaling up that can inform both scale up planning [85] and evaluation [58]. Specifically, PRACTIS workshops (i.e., [86]) involve a participatory process and co-design process with stakeholders, which allows for the systematic identification and documentation of data that are influential for intervention uptake, political support and community sustainability across multiple levels of the system. For data collection that can account for the complex systems in which interventions are implemented, resources such as the Consolidated Framework for Implementation Research (CFIR) [66] outline key factors influencing implementation of interventions, including ways of collecting data to account for these multilevel factors at scale. A key component of the Designing for Dissemination and Sustainability (D4DS) logic model [85], is designing a research product with the end in mind. This includes, understanding the characteristics of systems by using methods (e.g., system dynamics modelling [87]) to anticipate and plan for adaptation of interventions in response to changes in context over time [85]. Use of these frameworks and tools is recommended for improving operationalisation of systems approaches in the context of public health scale up.

Building the evidence base in systems approaches to scaling up, for physical activity and other NCD behavioural risk factors, has the potential to improve how we communicate and operationalise systems approaches when scaling for widespread and sustainable impact, in both research and in practice.

Strengths and limitations

A key strength of this review is that we utilised a narrative synthesis approach which enabled us to account for different scale up approaches and the varied terminology used to describe systems and systems approaches. Due the complicated and understudied nature of the topic we address in this paper, and thus the extended time required to extract, analyse, and reach consensus on the use of a systems approach to scaling up; we acknowledge the gap between data extraction and publishing. However, by combining a narrative synthesis approach with systematic screening, independent data categorisation, and quality assessment of methodological rigour, the robustness of our review is enhanced. The data were extracted by two independent researchers, and the analytical framework was developed, and data synthesised by three independent researchers. Given the lack of a universally agreed-upon definition of a 'systems approach' and the broad use of systems language, our methodology enabled us to capture implicit and explicit systems methods and approaches. Our analytical framework also allowed us to identify whether scaling up adhered to a systems approach, even if the term 'systems approach' was not explicitly used in the study.

Our systematic search process meant that we were able to identify relevant articles, however, as our data synthesis was reliant on information contained within the published papers, it is unknown to what extent systems or systems approaches were considered in the scale up process by decision-makers but were simply not reported in the published articles. We undertook reference list searches for studies where information was incomplete or unclear, however, this was only conducted for the scale up strategy and framework used, and not for all aspects of scale up. There is thus the potential that additional information related to scale up may have been published elsewhere. In addition, we also only report on the country of scale up directly relevant to the papers included, whereas interventions may have been expanded to other countries elsewhere. For example, for one intervention we acknowledge that adaptations had occurred for implementation in other countries (e.g., [58]), which we do not report on in our results. In this review, we focused on the four major behavioural risk factors for NCDs [2]. We acknowledge that the WHO physical activity guidelines also include recommendations related to sedentary behaviour [88]. Sedentary behaviour is considered a distinct health risk factor to physical inactivity [89, 90], and thus future reviews may wish to broaden the scope of included studies to align with global guidelines. To our knowledge, this review is the first to categorise the use of systems approaches in scale up, however, future research is needed to quantify different levels. Finally, non-English publications were excluded from this review.

Conclusion

Systems approaches allow for consideration of complexity at scale, facilitating different ways of planning for and interpreting challenges that are associated with scale up. Systems approaches also align with some of the key facilitators to successful scale up. By acknowledging the interconnectedness among various components of a system and ensuring their efficient and effective collaboration; systems approaches can potentially lead to more successful and impactful scale up outcomes. This review showed that the use of systems approaches when scaling up interventions targeting key behavioural risk factors for NCDs (physical inactivity, tobacco use, alcohol consumption and diet) is still in its infancy and there is a need for high quality studies. In particular for population level physical activity promotion, this presents a huge gap in knowledge. For decision-makers wishing to adopt or support a systems approach to intervention implementation at scale, greater guidance is needed on what is required to achieve this, and how to communicate and operationalise systems approaches in both research and practice.