A scoping review and systems thinking synthesis of scale-up frameworks Authors: Cole Hooley, Katherine E Marçal, Kayden Barlow, Mia T Vogel Published: 2026 ABSTRACT AbstractBackgroundThere is a pressing need to scale-up health and human services. While there are many scale-up frameworks in the broader literature, most of the general scale-up frameworks treat scale-up as a linear process. There are calls to bolster research viewing scale-up as a dynamic process. The purpose of this review is to consolidate and apply systems thinking to existing scale-up frameworks to illustrate feedback relationships among the drivers of scale-up.MethodsWe used a scoping review informed by hermeneutic principles to search the literature for scale-up determinants. We searched six scholarly databases (Medline, Web of Science, Embase, CINAHL, PsycInfo, and Google Scholar) for review articles of scale-up frameworks that yielded key determinants of scale-up. We also included other scale-up papers that yielded determinants. Articles that did not meet these criteria were excluded during the title/abstract and full-text screening process. Two coders assessed each article before it was excluded. Two coders extracted the scale-up determinants and definitions. We used thematic analysis to code and consolidate the determinants. Then we used an ex-post development approach to create a causal loop diagram (CLD) visualizing the scale-up determinants and feedback structures.ResultsOur search yielded 2693 articles which resulted in a usable sample of 40 articles after the screening process. The 40 articles contained 1053 scale-up determinants. Our coding procedures and CLD reduced these determinants to 25 in our model. There are six reinforcing loops and three balancing loops that are at play in scale-up efforts depicted in the final CLD. Among the determinants, motivation, capacity, enabling environment, and alignment were prominent.ConclusionsThe resulting interpretation and CLD provide a dynamic conceptualization of scale-up complexity, highlighting key leverage points in the scale-up process.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12913-026-15107-8.Keywords: Scale-up, Implementation science, System dynamics, Complexity, Scoping review, Causal loop diagram, Systems thinking Background Scale-up is “the deliberate efforts to increase the impact of successfully tested health innovations so as to benefit more people and to foster policy and programme development on a lasting basis” [1]. There are various elements that constitute successful scale-up such as sustained inputs, sustained implementation, sustained positive impact in the lives of service recipients, increased equitable reach, enhanced organization capacity, routinization, attending to dynamic considerations, and a shift in stakeholder views [2]. Getting to successful scale-up is a challenging and elusive process particularly given the complex systems within which health and human services are embedded [3, 4]. Numerous frameworks have been developed to describe the scale-up process and factors associated with it [5, 6]. These frameworks point out that scale-up is a complex process with various interacting pieces. Articulating the feedback relationships among these interacting pieces has been challenging. This review article responds to this need by exploring how system dynamics, an approach to articulating complex feedback dynamics, can inform scale-up frameworks. Conceptualizations of scale-up range along a continuum from linear to complex [7]. In the linear, or intervention-oriented, approach scale-up is viewed as a stepwise process in which an intervention progresses from efficacy trials to effectiveness studies, to widespread scale-up. In contrast, in the complex, or systems-oriented, approach scale-up is viewed as a process of transforming the broader system in which the intervention is embedded to realize the hoped-for outcomes. This perspective emphasizes the importance of feedback, adaptation, and system emergence [7]. Both approaches—and variations between them—have value [3]. The choice of which approach to use should be informed by the complexity of the problem, intervention, system, and context [7]. Greater complexity typically favors a systems-oriented approach. To date, intervention-oriented scale-up has been the dominant paradigm within implementation science [8]. Most frameworks depict a linear progression while acknowledging that scale-up is complex and shaped by multiple determinants [7, 9–11]. Furthermore, a review of public health interventions found that most empirical scale-up studies used a linear approach [12]. However, there is a growing body of work advancing systems-oriented approaches [3, 4, 7, 8, 10, 11]. This study contributes to that literature. From a complexity science perspective, scale-up can be viewed as a complex adaptive system—a network of interdependent components that interact in non-linear, dynamic, and often unpredictable ways [4, 11, 13–15]. These systems are characterized by internal rules, feedback loops, and emergent behaviors [4]. Components influence each other’s behavior and evolve in response to both internal dynamics and external pressures [13, 14, 16]. Boundaries are fluid, interactions are recursive, and outcomes are often disproportionate to inputs [16, 17]. For example, if a behavioral health system rapidly expanded a program to meet an urgent need, the reach of the service would increase. However, those actions could lead to workforce strain and eroding fidelity thereby diminishing the effectiveness of the services. This dynamic could feed back into motivation and funding or other external support to reduce enthusiasm in continued service expansion. By contrast, the same service could experience a different scale-up outcome based on the interactions among the constellation of factors. Such complexity helps explain why the same intervention can produce different outcomes across seemingly similar systems. Contextual variation, feedback mechanisms, and emergent dynamics can make standardized, one-size-fits-all approaches to some scale-up efforts ineffective [4]. These features underscore the importance of a systems-oriented approach [8]. System dynamics offers tools to make sense of such complexity. It uses visual and computational methods such as causal loop diagrams (CLDs) and simulation modeling to represent feedback relationships and system behavior [18–20]. System dynamics attends to both individual components and the whole system, capturing interactions and their subsequent whole-system emergent patterns [20]. CLDs, in particular, provide a qualitative way to map interconnections and explain how and why systems behave as they do [18, 19]. These behaviors are often articulated as feedback loops. These loops are categorized as reinforcing where the sequence of behaviors perpetuates an outcome that is either a virtuous or vicious cycle, or balancing loop where behaviors counteract each other to return to the original goal outcome. This tool could be used to understand the feedback relationships in scale-up efforts. Recent studies have called for scale-up frameworks that more explicitly integrate feedback mechanisms [8, 10]. Our study answers that call by applying system dynamics methods to conceptualize scale-up. Specifically, we aim to (1) identify key determinants of scale-up among framework reviews, and (2) postulate feedback relationships among them. By doing so, we seek to illuminate how these interdependencies influence scale-up trajectories and outcomes. There are valid critiques about the proliferation of frameworks in implementation science [21, 22]. The focus of this paper is showcasing a way to infuse systems thinking within scale-up frameworks. Methods We used three methods to meet the aims of this paper. Specifically, we used a scoping review informed by hermeneutic review principles to identify key determinants of scale-up. Then, we consolidated the determinants using qualitative coding. Finally, we used a causal loop diagram (CLD) method to articulate possible feedback relationships among those determinants. Search for scale-up determinants We used a scoping review method informed by hermeneutic principles to search the literature for scale-up determinants [23, 24]. A scoping method is appropriate for the aims of our paper because we sought to identify the determinants of scale-up across a complex and heterogenous literature [24]. Scoping reviews involve five required phases with an optional sixth phase [23]. This study was not pre-registered and no protocol has been published. To minimize the potential for post-hoc bias, we developed the search strategy a priori, identified exemplar articles before conducting the review, applied the same search strategy for each search update, used dual screening and extraction procedures, verified coding decisions across multiple team members, and maintained traceability between extracted determinants, qualitative codes, and CLD variables throughout model development. Stage 1. The first stage of a scoping review is to identify the research question. The question guiding this portion of the study is what are the determinants of scale-up? Stage 2. The second stage of a scoping review is to identify the relevant studies. Following guidance refining the scoping review method, we used the Population, Concept and Context (PCC) approach to guide our study identification efforts [24]. We sought review articles that involved any population, any substantive area, across any setting that were reviews of scale-up frameworks that resulted in a distillation of scale-up determinants. This wide cross-section of reviews gave us a fuller picture of the broad field of scale-up and helped to identify core determinants that spanned service type, sector, and population. We developed a search strategy combining terms from our previous work and other reviews e.g. [25]. and executed the search in various databases in June 2020, November 2021, and March 2025 [26]. No date restrictions were applied to the first search, and the two search updates captured newly indexed articles since the previous search. The search strategy included three hedges matching our population, concept, and context (scale-up, framework, and review) all combined with “and”. The same search strategy was used across all three search waves. Information about the articles we pulled search terms from is in the supplement (S4). We searched Embase, MEDLINE, Web of Science Core Collection CINAHL, PsycINFO, and Google Scholar to maximize the capture of unique references [26]. See the supplementary file for the full search strategy and additional notes (S2, S4). We prioritized precision by limiting our search to titles, consistent with other scale-up framework reviews [27]. We identified five exemplar articles a priori [9, 27–30], with the goal of our search capturing 80% of them [31]. Using a building block approach [32], we iteratively refined the strategy to ensure relevance and inclusion of the exemplars. We screened results from Google Scholar page by page until citations became irrelevant, yielding the first 250 citations [33]. The final strategy captured all exemplar articles. Stage 3. The third stage is to select the studies from those pulled by the search. The first two stages followed standard scoping review procedures. We modified Stage 3 (study selection) by incorporating principles of a hermeneutic review [32]. The hermeneutic method has been used in other scale-up framework research [34] and has been combined with other review methods in framework research [35]. In addition to the explicit inclusion criteria, we included articles that meaningfully contributed to understanding scale-up determinants [36]. This interpretive criterion was applied consistently across all search waves to ensure that the final sample captured literature relevant to the study aims. We included articles written in English, that were reviews of scale-up frameworks, which yielded determinants of scale-up, erring on the side of inclusion when methodological details were unclear. And, as per the hermeneutic principles, we included other scale-up articles that contributed to identifying scale-up determinants using our interpretive criterion. We exported the search results from the three waves to EndNote v9 and deduplicated them [37]. We imported the deduplicated citations to Rayyan for screening [38]. The first author screened the title/abstracts for inclusion, and a second screener reviewed all of the articles marked for exclusion [39]. Any conflicts were resolved through discussion between the two screeners. We used the snowball method to identify additional citations from the articles we included for analysis [40]. Qualitative coding of scale-up determinants Stage 4. The fourth stage is to extract data from the final sample of articles. First, we took the articles and disaggregated them into analyzable units [41, 42]. The analyzable units were the main determinants from the articles [29]. This included those words or phrases used by the authors as substantive headings in their narrative descriptions, and/or the labels used in their summative figures. We also extracted any definitions or explanations of the determinants. One coder extracted the data into a spreadsheet, and a second coder checked and verified the extraction [31]. Any differences between the first and second coder were resolved through discussion. Stage 5. The fifth stage is to collate and summarize the data and to present the results. We used an iterative thematic analysis to summarize the determinants from the scale-up framework review articles [43–45]. First, one author inductively open-coded the determinants generating a body of summative codes [46, 47]. A second author reviewed these summative codes. Then, three members of the team applied these codes to the balance of determinants, refining and adding codes as needed [48]. Codes were verified by at least two members of the team. The authors engaged in a form of axial coding to consolidate constructs in the CLD [46, 47]. The supplemental file has a table that shows an example of raw constructs, their intermediate code, and the finalized CLD variables (supplemental file table S5). Causal loop diagram construction of scale-up feedback loops To articulate the feedback loops among the determinants, we used an ex-post development approach [49] to create a CLD [18, 19, 50]. We began the construction of the CLD by identifying key outcome variables that could be used as reference modes (behavior over time graphs) for scale-up [18, 51]. We selected scale-up outcomes of contact coverage, effective coverage, and equity from the Health Services Coverage framework to begin [52, 53]. Contact coverage is the proportion of the target population receiving an indicated service. Effective coverage is the proportion of the target population who receive the intended health benefit of that service. Equity is a measure to ensure that contact and effective coverage do not favor or disfavor particular segments of the target population. Outcomes beyond coverage are essential for scale-up [2, 11]. We used these three variables to begin our CLD, and we incorporated other important scale-up outcomes in the full CLD (e.g., quality, impact, capacity, embeddedness, sustainability) [2]. We created the CLD using Stella Architect v3.4.1. In a CLD, arrows indicate causal influence between variables. A positive (+) link indicates that variables move in the same direction (an increase in one variable leads to an increase in the other, or a decrease leads to decrease in the other), whereas a negative (–) link indicates that variables move in opposite directions. Reinforcing loops represent self-amplifying dynamics that can generate virtuous or vicious cycles. For example, increased dissemination of an effective service may increase motivation to adopt that service, which increases implementation. Increased implementation leads to more clients receiving the service (contact coverage), and when coupled with quality delivery, more clients improving (effective coverage). The more clients improve may then further increase dissemination, creating a self-reinforcing cycle. Balancing loops, in contrast, represent goal-seeking or stabilizing dynamics. Instead of amplifying, the effect stabilizes. These loops are identified using an “R” or “B” notation with a clockwise or counterclockwise arrow to help the reader know which direction to read the loop. Then, we used an iterative process to build the CLD by pulling determinants from our coding efforts that influence the scale-up outcomes [18]. The first phase of CLD construction ran in parallel with the qualitative coding of determinants. Two authors independently evaluated an initial set of scale-up framework review articles. We extracted and consolidated key constructs from the article’s frameworks to develop a seed structure with relevant outcome and initial variables. One of the authors who specializes in system dynamics plotted those determinants in the CLD. We then discussed the relationships between the determinants, resolving any discrepancies. We then integrated the coding from the thematic analysis [47]. We plotted new determinants and their relationships in the CLD. We followed this iterative process, adding variables until we reached saturation of determinants and until all the articles had been coded [54]. The objective of the CLD was to synthesize existing theoretical knowledge into a coherent systems representation of scale-up. The causal relationships depicted in the model therefore represent conceptually supported hypotheses synthesized from the literature rather than empirically validated causal effects. Although individual relationships vary in the amount of direct empirical evidence available, each linkage was included only when judged to be theoretically plausible and supported by at least one relevant source. Several measures promoted confidence in the model. We used a modified version of rigorously interpreted quotation analysis where we tracked the qualitative codes, the connection between those codes and CLD variables, and verified the linkages between them [55]. Key feedback mechanisms were reviewed against relevant theoretical and empirical literature both within and outside the original sample of articles to ensure all causal links were adequately justified [49]. Linear processes articulated by existing frameworks were overlaid onto the CLD to ensure consistency and accurate representation. Results Search and coding results We identified a total of 2693 records across the six databases and across the three search time points. The screening process yielded 35 articles for extraction. During the full-text screening process we identified an additional 5 articles, bringing the final sample of coded articles to 40 (see PRISMA flow diagram in supplemental file, S1). We extracted and coded 1053 scale-up determinants from the 40 articles. The list of articles and associated determinants are in the supplemental file (S3). Through our coding procedures, we refined these determinants into the final pool of variables for the CLD (see Table 1). In total, there are 25 variables in the CLD. The full list of those variables and definitions is in Table 1. When support for linkages between variables was not provided in the extracted articles, we confirmed their relationship using other studies; the studies used to support our linkages are listed in Table 2. For example, the CLD variables “discrimination” and “commitment to inclusivity” emerged from comments within the extracted articles about the importance of honoring human rights and working against discrimination. Those articles, though, did not offer sufficient detail to guide us on the linkages between those variables and others in the CLD. So, we offer one possible set of connections that we were able to find literature to support. Those connections, while supported, should be viewed as preliminary and are open to further development and study. The tentative positioning of these equity-related pathways reflects a limitation of the available scale-up literature rather than a diminished role of equity in scale-up efforts. Although equity-related concepts were frequently identified in the reviewed literature, the mechanisms through which they interact with other determinants were often only implicitly discussed. This is why those connections are demarked using dashed lines in our CLD figure. Table 1Causal loop diagram variables and definitionsCLD variableVariable definitionClient demandClients’ (end user’s) perceived need, desire to receive a service, and seeking that serviceCompeting demandsMultiple needs on a community or system diluting attention/resourcesLeadership supportDegree of scale-up engagement, attitude towards, commitment, involvement, and accountability demonstrated by leaders or other institutional championsFit to current systemPerceived compatibility and applicability of the intervention with the service systemComplexity of interventionPerceived difficulty to understand and use the intervention; components/resources required to implement or use interventionOrganizational capacity to implementAbility to implement the intervention given organizational resources/knowledge/abilityKnowledge of interventionAwareness of and information about the interventionExternal supportFinancial, policy, political, or other instrumental support from external institutionsDissemination about benefitsSpreading information about intervention benefitsMotivationOrganization’s willingness/desire to provide interventionAdoptionCommitment to implement the interventionImplementationThe process (e.g., energy, resources, time) put toward using/integrating an intervention within a settingContact coverageNumber of target population who receive the indicated interventionQuality controlEnsuring that the intervention is delivered thoroughly, effectively, efficiently, and reliablyIntervention effectivenessThe degree to which the intervention produces the hoped-for effectEffective coverageNumber of the target population who receive the intended benefit of the interventionUrgency of needIntensity of need/desire for a servicePerceived benefitsPerceived effectiveness of a serviceMaintenanceEfforts and resources allocated to sustain the practiceEquityAttaining the highest level of health for all peopleDiscriminationDifferential access to the goods, services, and opportunities of societyBarriers to servicesFactors that hinder the use of servicesHealth disparitiesUnfair differences in health outcomes between social groupsDissatisfaction with status quoGap between perceived and desired state of the system that motivates actionCommitment to inclusivityWillingness and action to ensure everyone in need of a service receives it and its intended benefits Causal loop diagram variables and definitions Table 2Causal loop diagram variable linkages and associated supportVariable 1PolarityVariable 2Representative support from literatureLeadership support+Organizational capacity to implement[34, 59, 80]Fit to current system+Organizational capacity to implement[8, 61, 81]Complexity of intervention-Organizational capacity to implement[61, 69, 82, 83]Knowledge of intervention+Leadership support[69]Knowledge of intervention+Client demand[69, 84–86]Dissemination about benefits+Knowledge of intervention[81, 87, 88]Leadership support+Commitment to inclusivity[89]Dissemination about benefits+External support[69, 87, 90]External support+Organizational capacity to implement[28, 61]Dissemination about benefits+Motivation[61, 69, 80]Organizational capacity to implement+Motivation[64, 80, 91]Organizational capacity to implement+Implementation[61, 70, 91, 92]Organizational capacity to implement+Quality control[59, 70, 93–96]Competing demands-Organizational capacity to implement[90]Competing demands-Fit to current system[61, 97, 98]Competing demands-Commitment to inclusivity[99, 100]Commitment to inclusivity1+Dissatisfaction with status quo[61]Health disparities+Competing demands[90, 101–103]Client demand+Motivation[61, 69, 104]Motivation+Adoption[64, 69, 105]Adoption+Implementation[69, 80]Implementation+Contact coverage[106–108]Contact coverage+Effective coverage[52, 53, 109]Effective coverage-Urgency of need[52, 110–113]Urgency of need+Motivation[59, 114]Urgency of need+Dissatisfaction with status quo[8, 115]Urgency of need+Client demand[116–121]Client demand+Maintenance[59, 69, 122–129]Dissatisfaction with status quo+Motivation[63, 130, 131]Health disparities+Dissatisfaction with status quo[132–134]Effective coverage+Dissemination about benefits[69, 113, 135]Contact coverage-Quality control[136–141]Quality control+Intervention effectiveness[59, 142]Intervention effectiveness+Effective coverage[92, 111, 113]Effective coverage+Perceived benefits[86, 143, 144]Perceived benefits+Maintenance[29, 145]Maintenance+Effective coverage[141, 146]Perceived benefits+Client demand[8, 69, 147]Equity-Health disparities[28, 67, 148]Effective coverage+Equity[149, 150]Barriers to services-Contact coverage[141, 151]Equity-Barriers to services[71, 152, 153]Discrimination+Barriers to services[72, 154, 155]Discrimination-Equity[156]Commitment to inclusivity1-Barriers to services[157–161]Commitment to inclusivity1-Quality control[58, 162]1= extracted articles would mention the importance of inclusion efforts. There was insufficient guidance in those papers on how to model that in the CLD. We reviewed the extant literature to provide one possible idea of how it could fit into the larger model. This tentative positioning suggests that greater commitment to inclusion influences increases dissatisfaction with the status quo. It also suggests that increased commitment to inclusion reduces barriers to services and may exert an initial demand on limited quality control resources which could reduce quality control in the short term Causal loop diagram variable linkages and associated support 1= extracted articles would mention the importance of inclusion efforts. There was insufficient guidance in those papers on how to model that in the CLD. We reviewed the extant literature to provide one possible idea of how it could fit into the larger model. This tentative positioning suggests that greater commitment to inclusion influences increases dissatisfaction with the status quo. It also suggests that increased commitment to inclusion reduces barriers to services and may exert an initial demand on limited quality control resources which could reduce quality control in the short term Results of feedback model Figure 1 depicts the overall feedback structure identified through the review and synthesis process. Reinforcing loops amplify behavior, whereas balancing loops counteract change and promote stability. At the center of the diagram are relationships among motivation, adoption, implementation, contact coverage, effective coverage, and urgency of need. Surrounding these core relationships are reinforcing and balancing feedback loops involving dissemination, capacity, maintenance, client demand, external support, and equity-related factors. In causal loop diagrams, the spatial placement of variables is intended to improve readability and reduce line crossings rather than communicate the relative importance of constructs within the system. Consequently, variables located near the center or periphery of the figure should not be interpreted as more or less influential. Influence is determined by the pattern of causal relationships and feedback loops in which variables participate. For readability, some variables are represented more than once in the figure; duplicated variables represent the same underlying construct. The final CLD contained six reinforcing (“vicious or virtuous cycle”) loops (R1-R6) and three balancing (“goal-seeking” or compensatory) loops (B1-B3; Fig. 1). Scale-up causal loop diagram R1: “Stimulating system appetite” The first reinforcing loop describes the process by which a health or human service scale-up depends on effective dissemination of benefits. This loop first captures the linear process showing that motivation to bring in a new intervention or service drives adoption, and stronger adoption leads to implementation. More provision of the service increases contact coverage which, all else being equal, increases effective coverage. When more clients benefit from services, these benefits are disseminated to other providers and clients. Dissemination about the service drives motivation for others to adopt the intervention, thus increasing adoption in a virtuous cycle. It is important to note that reinforcing loops may work in either the positive or negative direction; thus, low motivation reduces adoption and implementation, resulting in poor coverage. When few benefits of an intervention are observed, little positive information is disseminated, further depressing motivation to adopt. R2: “Building support” Successful health or human service scale-up depends on adequate organizational capacity for implementation, which often requires external support in the form of political will, funding, and other forms of instrumental support. According to this loop, implementation cannot occur without adequate organizational capacity; implementation leads to contact and, subsequently, effective coverage. Dissemination of intervention benefits drives external support in the form of political will and/or funding; these resources contribute additional organizational capacity for expanded implementation. R3: “Perceived benefits motivating maintenance” Effective coverage can either increase or decrease over time; thus, scale-up frameworks must consider maintenance efforts to sustain a health or human service. Effective coverage leads to perceived benefits by both clients and providers, who are therefore more inclined to invest resources in maintaining a program; ongoing maintenance prevents effective coverage from declining. R4: “Rapid implementation eroding quality” A risk of widespread implementation is fidelity drift and lack of quality control for a health or human service. This reinforcing loop details the process by which urgent needs can drive motivation, subsequently driving adoption, implementation, and contact coverage. Without additional organizational capacity, however, increased contact coverage risks erosion of intervention quality and effectiveness, which decreases effective coverage. Thus, urgency of need does not decline, and motivation for adoption continues to increase. To maintain service quality as service access increases, there must be enough organizational capacity to meet the increased demand. R5: “Increased coverage through increased equity” Multiple barriers exist to achieving the equitable scale-up of health or human services. Structural discrimination, residential segregation, and historical mistreatment may make services inaccessible or inappropriate for marginalized communities. These barriers can reduce contact and effective coverage, which further reduces equity and compound barriers to accessing services. R6: “Perceived benefits driving client demand” In this reinforcing feedback loop, more perceived benefits of health or human services by potential clients increase the demand for health or human services. That increased demand increases the service system’s motivation to adopt the service. The increased motivation leads to increased adoption, implementation, contact and effective coverage. This increase in effective coverage—more members of the target population receiving the service and improving–further increases perceived benefits. B1: “Reducing need through effective coverage” B1 details the process by which efficient and effective health or human service scale-up can undermine its own sustainment. According to this loop, the urgency of need increases motivation for a solution, such that the more dire or overwhelming a particular problem becomes, the more motivated the service system is to adopt a new strategy or program. Increased motivation increases the likelihood of adopting an intervention. With more providers and agencies adopting an intervention, more also implement the intervention. More implementation leads to greater contact coverage which, all else being equal, leads to great effective coverage. When more clients are effectively served, over time, the urgency of need and subsequent motivation for providing the service declines. B2: “Reduced need undermining maintenance” A balancing process acts on maintenance. Over time, effective coverage reduces urgency of need, which subsequently reduces client demand for a service, which may eventually reduce motivation for and necessity of maintenance efforts for that health or human service. When interventions are not maintained, effective coverage declines. Thus, effectively meeting needs can undermine efforts at sustainment. B3: “Reducing need through building equity” The final loop describes the balancing loop by which addressing health disparities drives dissatisfaction with the status quo, as marginalized groups are excluded from a health or human service access or benefits. Dissatisfaction drives motivation for adoption; as adoption increases implementation, contact and effective coverage follow. Effective coverage increases equity; as more clients benefit from services, health disparities decline and reduce the dissatisfaction that drives motivation for change. Theoretical extension: commitment to inclusivity Limited material from the modified scoping review articulated organizations’ roles in directly combatting exogenous discrimination, reducing barriers to services, and promoting equity. As such, the pathways described below should be viewed as a theoretical extension of the model rather than a direct synthesis finding from the reviewed literature. Although the review identified equity, discrimination, and inclusion-related concerns, it provided limited guidance regarding the mechanisms through which organizations promote inclusivity. Drawing on insights from the review as well as extant literature, we suggest a possibility of how an endogenous commitment to inclusivity might interact with the newly articulated causal feedback model. First, we posit that a commitment to inclusivity must be championed by a dedicated leader in the organization [56], who is guided by knowledge of both the intervention and the merits of inclusive practice. At the same time, a commitment to inclusivity might be undermined by competing demands; when multiple groups present with multiple needs, organizations may be less motivated or equipped to take on the challenge of adapting systems and services for marginalized groups. A commitment to inclusivity may likewise drive multiple outcomes. First, a greater commitment to serving a broad population with diverse identities and needs may increase dissatisfaction with the status quo; inclusive practices, both motivated and challenged by inequity in health or human service systems, may condition organizations and champions to view more critically existing inequities. Effective commitment to inclusivity can reduce barriers to services and increase access opportunities for marginalized groups [57], but a tradeoff may emerge by which increased coverage can strain resources and impede quality control [58]. The full CLD thus contextualized extant linear processes within a feedback framework that highlighted counterintuitive, or unanticipated, system behaviors. Furthermore, the CLD pointed to important factors which could be adjusted to improve efficiency and reach of scale-up efforts. Discussion There are pressing needs to scale-up health or human services. The process to scale-up a service is complex and, as such, requires tools that can capture and make sense of that complexity. System science methods offer a suite of tools that can be used to understand this complexity. In this paper, we demonstrated how one of those tools, causal loop diagramming, can be applied to scale-up frameworks to examine feedback structures within the scale-up process. Substantial work has been done to understand scale-up broadly, including various reviews synthesizing frameworks, determinants, and outcomes [2, 5, 6, 9]. Our paper focused on review papers that resulted in scale-up determinants. The search yielded 40 review articles that produced over a thousand determinants that influence scale-up. The list of determinants represented a range of granularity from very specific to general. These various levels of granularity can serve different purposes and can be beneficial depending on the needs of the scale-up project. For our purposes, we sought to understand high-level patterns and feedback structures, so we synthesized the 1,053 determinants to 25. Understanding core influences of scale-up can help focus energy and resources to essential tasks while still providing flexibility in approach [59, 60]. At the highest level, four determinants appear essential for scale-up: motivation, capacity, enabling environment, and alignment [59, 61–63]. This parallels Michie et al.’s COM-B model, where behavior change depends on capability, opportunity, and motivation and Scaccia’s model that notes motivation, organizational capacity, and innovation-specific capacity are required [62, 63]. Just as individuals must want, be able, and have the chance to act, organizations—and the actors within them—require the same conditions to change [61, 64]. The complexity of scale-up arises from its factorial nature. Instead of working within one organization there are many organizations involved, each with their own collective of actors whose willingness and ability are needed [3, 4, 8]. And there are more external influences wrapped around the variety of involved organizations that need to be aligned to produce an enabling environment. A systems approach makes these interdependencies explicit, identifying leverage points and unintended effects of specific scale-up efforts [19, 65, 66]. The construction of a formalized system dynamics simulation model of these and related factors is a consideration for future research [18, 66]. The CLD modeling process highlighted complexity in scaling up interventions. Although several linear frameworks articulate the scale-up process, the present study applies systems science to investigate the complex feedback relationships not captured by linear process models [3, 4, 8]. Findings point to unintended consequences of scale-up efforts, unintuitive insights about the necessary ingredients for scale-up [58, 65, 66]. The model also highlights the importance of equity to successful scale-up, consistent with broader calls to center equity within implementation and scaling efforts [67, 68]. From a systems perspective, successful scale-up efforts can be both self-perpetuating by garnering support for positive outcomes (R1-R3) as well as self-limiting by meeting need and reducing population-level desire for intervention (B1-B2) [18, 66]. Understanding and incorporating these insights into scale-up planning may help planners anticipate feedback dynamics that shape reach and sustainment, while making equity considerations more explicit in scale-up design. Considering scale-up as dynamic rather than linear incorporates insights from existing frameworks while extending understanding of the organizational and human interactions that shape outcomes [3, 8]. For example, motivation to adopt a new intervention is not turned on or off but rather accumulates to reach a tipping point at which the decision to adopt is made; motivation is endogenous to a range of factors, including outcomes of prior adoption and implementation efforts [66, 69]. Furthermore, motivation is influenced by levels of dissatisfaction with the status quo, pulls from client demand, knowledge about the benefits of new practice, and the perceived capacity of the organization to implement the new practice [62–64]. Motivation alone is insufficient to support successful scale-up but is required in concert with other factors – all of which are also endogenous to the system [62, 63]. Thus, efforts to increase motivation for adoption may be secondary to documenting need or gaps in current services or bolstering organizational capacity for implementation [60, 63, 70]. Equity emerges as a key outcome as well as driver of successful scale-up. Equity here refers to striving for the highest possible health for all people, with particular attention to those at greatest risk given social conditions [67]. Whereas other frameworks indicate that equity is essential, our CLD postulates where equity fits and how it potentially interacts with other scale-up variables [68]. The relatively tentative positioning (i.e., dashed line) of our “commitment to inclusivity” variable and its pathways in the CLD reflects limitations in the available scale-up literature rather than a diminished role of inclusion in scale-up efforts. Although inclusion and related concepts were recurrent themes in the reviewed literature, the mechanisms through which inclusion constructs interact with other determinants of scale-up were often discussed implicitly. Consequently, the inclusion-related pathways should be viewed as preliminary and open to further empirical development. Increasing contact and effective coverage can promote equity by serving more people in need; meeting needs through expanded effective coverage can lower barriers to additional service engagement by marginalized groups [52, 68]. Lower barriers to service utilization can promote contact and effective coverage over time in a virtuous cycle. Likewise, failure to reach marginalized populations heightens barriers to services access, eroding contact and effective coverage over time in a vicious cycle [71–73]. Promoting equity also may help reduce health disparities; when organizations have fewer competing demands for their attention and resources, there exists greater capacity to successfully adopt, implement, and scale interventions [71, 74–76]. These equity-related feedback loops point to the importance of dismantling structural barriers and disparities impacting marginalized groups, who benefit from targeted outreach and adapted service models that consider their unique circumstances and needs [77–79]. Strengths and limitations We have provided an example of using system science methods to inform scale-up framework thinking. This perspective operationalizes calls to increase dynamic thinking in scale-up [7, 8, 10]. We have also attempted to incorporate potential pathways of how discrimination and inclusion may operate in a scale-up process [68]. The absence of preregistration represents a limitation. Although several procedures were used to enhance transparency and consistency, including dual screening, multi-coder verification, and explicit linkage of codes to model variables, future work would benefit from prospective protocol registration. While we tried to be as generous as possible in our article inclusion process, there is a chance we may have missed relevant articles. However, during our coding procedures we felt like we reached a point where no new concepts were emerging. We opted to take a high-level approach to point out key drivers of scale-up. Depending on need, this sort of model could be made even more granular. As a conceptual synthesis, the CLD represents one theoretically informed interpretation of how determinants of scale-up may interact. Although individual causal relationships were grounded in the literature where possible, the model as a whole should be viewed as a conceptual hypothesis that warrants empirical testing and refinement. Conclusions Viewing scale-up as a dynamic process reveals multiple pathways of stoking success and anticipating potential complications in scale-up. Our review and subsequent CLD are an example of how systems thinking can be applied to scale-up research, particularly in health or human services, and integrates the consideration of health equity and social determinants of health which are important for reaching intended audiences. This initial offering, we hope, will be refined by subsequent empirical research to strengthen our understanding of what leads to successful scale-up. Beyond scale-up, the resulting CLD may provide a useful framework for examining broader health and human service system reform efforts characterized by complex feedback relationships. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements Not applicable. Abbreviation Causal loop diagram Author contributions CH conceptualized the study. CH, KB, and MV helped with search and extraction. CH & KB and other research assistants performed coding. KM created the CLD in collaboration with CH. CH led manuscript drafting. All authors read and approved the final manuscript. Funding Mia T. Vogel was supported by grant number T32 HL130357 from the National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. This manuscript is the result of funding in whole or in part by the National Institutes of Health (NIH). It is subject to the NIH Public Access Policy. Through acceptance of this federal funding, NIH has been given a right to make this manuscript publicly available in PubMed Central upon the Official Date of Publication, as defined by NIH. Data availability The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.