This comprehensive systematic review draws together LHS approaches to improvement with explicit synthesis of guidance to support the implementation, scale-up, and sustainment of evidence-informed health policies and interventions. We developed a consolidated guidance that comprises six pillars that provide the infrastructure, capacity, and expertise to support ongoing improvement activities. It also identified ten sequential yet flexible steps, organised across three LHS phases which comprised the continuous improvement process. It is hoped the work will contribute to the strengthening of health systems by supporting their generation and application of implementation (inclusive of scale-up and sustainment) research to more rapidly improve the safety and quality of healthcare, patient outcomes and community health.

A range of excellent LHS frameworks have been published^10–14. These tend to focus on descriptions of the core infrastructure, or pillars needed to support improvement and ongoing learning. For example, Reid et al¹¹ conceptualised an LHS as an engine with learning gears (such as implementation and evaluation) and factors such as leadership and funding providing the “fuel” needed for the gears to produce equity-centred outcomes. It also identifies several “moderators” and “brakes” that can restrict or even impede such outcomes. Similarly, Menear et al¹² proposed a conceptual framework for value-creating LHSs that integrates four main dimensions – core values, pillars, processes, and outcomes. Their framework highlights seven pillars that support LHS functionality: scientific, social, technological, policy, legal, ethical, and socio-political infrastructure. These pillars are underpinned by core values such as transparency, equity, and collaboration, and are critical to enabling the learning processes that connect data to knowledge and action. Such ‘pillars’ provide an environment conducive to emergent learning, feedback and improvement within health systems. We saw value in consolidating such guidance and found remarkable consistencies in the attributes and functions of pillars across a diverse range of LHSs, and guidance documents. Such findings are similar to other recent reviews of LHS frameworks¹⁰. LHS frameworks provide less guidance, however, on the actual process of improvement - that is, the specific steps needed to enhance healthcare. Our review sought to address this gap by synthesising the steps recommended by process models of three important and distinct outcomes that help maximise the impact of evidence-based healthcare: implementation, scale-up, and sustainment. It is often suggested that each of these outcomes are considered across each phase of an improvement process. For example, sustainment should be considered when selecting an intervention to implement, and in the process of implementation. We are not aware, however, of guidance that has attempted to bring these together. While the application of the process may differ, such as the specific barriers, strategies employed to address them, or measures used to assess outcomes, we found little differences in the broad steps or process reported in documents reviewed between those with an emphasis on implementation, scale-up or sustainment. We believe this is useful contribution, offering a structured, but consistent approach to improvement. Importantly, the process is also not intended to be static. In alignment with the continuous nature of LHS, it can be re-applied over time to support different stages of improvement – from initial implementation through to scale up or sustainment. In this way, it offers both common process and practical flexibility for strengthening evidence-informed practice in dynamic, real-world settings.

Our findings closely align with the theoretical domains of evidence-informed policy-making (EIP) institutionalisation described by Kuchenmuller et al³⁶. Their framework identifies six domains essential for embedding evidence use within policy systems: governance, standards and routinised processes, partnership, collective action and support, leadership and commitment; resources, and culture. These domains map closely onto our identified Pillars. However, while the EIP institutionalisation framework primarily conceptualises these domains as structural and cultural mechanisms that confer legitimacy and sustain evidence use, our synthesis focuses on their practical operationalisation within LHSs, specifically in relation to implementation, scale-up, and sustainment of health interventions. In particular, our findings place emphasis on the systems and processes needed to generate new practice-based evidence, with dedicated Pillars for data systems and evidence synthesis. Such elements are not foregrounded in Kuchenmüller et al.’s framework³⁶, which assumes the availability of existing evidence for use. This focus on evidence generation, as well as use, underscores the dual role of LHSs as both users and producers of evidence. Additionally, while Kuchenmüller et al. frame institutionalisation as a process unfolding in stages, our findings describe a cyclical, practice-oriented improvement process, comprising actionable steps. These two perspectives - the institutionalisation of evidence use and the operationalisation of implementation improvement cycles - offer complementary pathways toward strengthening LHSs. Future research could explore how these perspectives can be integrated to support durable, evidence-informed health systems globally.

Synthesising the steps or stages across guidance documents was a complex process. Unlike the pillars, which do not follow a prescribed order, the steps were typically presented as sequential activities with variation in how different guidance documents presented this sequence. Some did not present any explicit ordering at all. Despite this challenge, we were able to generate a coherent and comprehensive set of steps through iterative coding processes and remaining grounded in the overall objectives and intended users of the review. In cases where ambiguity existed, we prioritised placement that aligned with the practical needs of policy-makers and targeted end users. To enhance usability and relevance across settings, we deliberately kept the steps broad. In some cases, this meant including several key actions (i.e., sub-steps) within a single step, to better reflect the fluid and often non-linear nature of real-world improvement processes.

The Phase 1 steps were especially challenging to position. For example, the identification and establishment of an operational team appeared at various points across guidance documents. In some cases, it was presented more as a pillar – a resource either assumed to be already present or meant to be established outside of the guidance application. For example, the Competitive Intelligence Process described a network of people (including executives and those with designated positions) as an infrastructure to operationalise the process, but not as step within the process itself ³⁷. Another challenge was interest holder identification and engagement – this was a recurring activity across all phases of the guidance. As such, we addressed this aspect by including it within Step 1.2 as well as within one of the six pillars, recognising its importance as an infrastructure across the entire improvement process. We also included the key interest holder-related information and activities within respective steps – for example, engaging interest holders to help identify and understand the problem (Step 1.1), and the formal identification of interest holders needed for improvement activities as part of planning in Step 1.2. Despite challenges in synthesis, the overarching LHS phases²⁹ provided a highly useful structure for organising the steps. These phases were unchanged from the outset of synthesis, offering a strong scaffold for coding and situating the diverse improvement activities described across the guidance documents.

A number of methodological strengths of the review are worth noting. The coding process was primarily inductive, allowing for findings to emerge directly from the source data, supported by strong LHS-based codebooks and a collaborative coding team. Where gaps or inconsistencies arose, deductive searches and peer discussions were used to clarify concepts and validate emerging interpretations. This experience highlights the value in having content expertise within the coding team and involved across the synthesis process, to identify potential gaps and ensure thorough coding. All coded data were reviewed by at least three members of the research team: original coders, secondary reviewers, and a third reviewer during the final data cleaning process. This ensured a high level of consistency and rigor.

Unlike traditional reviews and syntheses that appraise the quality of included documents to determine the strength of the evidence base, our approach treated all included guidance documents equally. Given the purpose of this review, we did not apply differential weighting based on document quality or comprehensiveness. Instead, all documents were reviewed in full, and all relevant content was extracted, regardless of whether a document provided a comprehensive framework or focused on a single component (e.g., strategy development or evaluation). While some documents contributed more coded content than others, this did not lead to greater influence on the results. The final steps and pillars were formed using an iterative, team-based synthesis process, using the breadth of relevant guidance across all included documents.

However, the comprehensive nature of steps and pillars also presents a potential limitation. At first glance, the steps and pillars may appear complex or even infeasible, particularly for settings with limited resources or capacity. Yet this reflects its purpose: to present a thorough version of what LHSs focused on implementation, scale-up, and sustainment can entail. Importantly, the review findings do not assume that all components must be enacted in full, and while the ‘steps’ are presented linearly, in complex adaptive systems like healthcare, we acknowledge this will rarely be the case. The application of findings from this review will be different by setting and context; the steps and pillars should be interpreted and tailored to the context in which they are applied. This tension between ideal and feasible is not unique to our synthesis - most individual guidance documents included in this review explicitly acknowledged the challenges posed by contextual constraints and resourcing limitations. For example, the IDEAS (Integrate, Design, Assess, and Share) framework drew attention to the variability of its application based on resource constraints ³⁸.

Nonetheless, there are likely to be challenges in operationalising comprehensive LHSs consistent with the steps and pillars describe in this review. ³⁹. Health systems operate within complex, resource-constrained environments, and establishing and sustaining multi-disciplinary teams, building the necessary infrastructure, and embedding continuous learning cycles into routine practice can be resource- and time-intensive. It may also require shifts in culture, processes, and roles. Many of the included guidance documents acknowledged the importance of a supportive organisational culture in enabling reorientation for improvement efforts. For example, the LHS Architectural Framework emphasised the need for internal alignment and a culture of inquiry to embed continuous learning into routine operations⁴⁰. For academics and public health researchers, it also represents a new way of working²⁹, one that requires closer, ongoing partnerships and a willingness to adapt timelines and traditional processes. Potential risks include insufficient leadership buy-in, fragmented data systems, or a lack of resourcing to enact the full cycle. The potential for health systems to improve the quality and impact of healthcare is unrealised. There are a range of challenges to developing learning health systems, even in high income countries and well-resourced healthcare organisations. This review sought to facilitate the development of LHSs by consolidating diverse guidance, to enhance the generation and application of research to support better implementation, scale-up and sustainment of effective health interventions. Here we provide key findings of our synthesis. The WHO will soon publish a practical guide to support to help operationalise LHSs. Further efforts should focus on testing and refining the framework in real world contexts to further strengthen its relevance and impact.