Calls have been made for improving performance of primary care systems in low- and middle-income countries (LMICs) as a necessary condition to achieve universal health coverage in the age of the Sustainable Development Goals. High-performing primary care systems not only are the first point of contact for continuous, coordinated, comprehensive and people-centered health services ¹ , but also provide critical preparedness and response to global, public health threats ² .

There is growing interest in better understanding the ways in which various policies and programs can improve primary care health systems at scale and moving beyond the quick fixes that characterize most efforts at changing complex system change ³ . Large-scale, health system change is described as “coordinated, system-wide change affecting multiple organizations and care providers, with the goal of significant improvements in the efficiency of health care delivery, the quality of patient care, and population-level patient outcomes” ⁴ .

Organizational performance refers to the results generated by an organization, measured against its intended goals and targets. In the private sector, the concept usually refers to profits, efficiency, quality, market-share, and customer satisfaction. In public sector organizations, the definition has shifted with the evolving framings for the role of the State in the production and delivery of public services ⁵ . Governments’ interest has shifted from controlling inputs and compliance with standards, towards reporting quantity and quality of outputs, productivity, efficiency and, more recently, outcomes and policy impacts ^{5, 6} .

Performance measurement and management (PMM) systems are organizational arrangements aimed at measuring organizational processes, outputs and outcomes with the proximal aim of informing the introduction of clinical, managerial, programmatic and policy changes, and the ultimate purpose of contributing to socially valued, population-level health and equity outcomes ⁷ . Forty years of research on PMM systems have shown that such systems can effectively improve performance, although unintended and undesirable effects can also occur ^{8– 16} . While there have been applications of some types of PMM interventions to the health sector of LMICs particularly through the use of financial incentives and pay-for-performance, research that bridges advances made in public administration and organizational science remains largely ignored in health systems research.

In order to address this fragmentation in evidence, we developed a framework that combines a PMM model originally developed by Pollitt to study organizational performance in the public sector ¹⁷ and a taxonomy developed by the Cochrane Collaboration’s Effective Practice and Organization of Care (EPOC) to characterize through the use interventions and outcomes in healthcare delivery ¹⁸ . The former helped us define the main elements of system-wide PMM interventions, while the latter allowed us to identify PMM interventions of relevance to primary care delivery systems in LMICs.

The general PMM system framework contains the following components: (1) An institutional context in which various policies, programs and health interventions are implemented and interact with healthcare stakeholders; (2) a local, socio-economic context where primary care services are delivered; (3) one or more PMM interventions that trigger improvements (or not); (4) a performance measurement process; (5) a sense-making process that allows the transformation of raw data into performance information; (6) a process of dissemination of performance information among system actors and stakeholders with the intent of making it actionable; (7) performance information use, misuse, or non-use; (8) implementation of planned action, leading to measurable organizational improvements (or not); and, (9) the production (or not) of short-term clinical and managerial improvements; intermediate outputs and outcomes; and, distal, societal, and population-level health and equity outcomes (intended and otherwise).

The EPOC taxonomy, in turn, contains various cross-cutting interventions and organizational arrangements of relevance to primary care systems such as implementation strategies, accountability arrangements, and some examples of financial arrangements. Such interventions can induce performance improvements at the level of the workforce, facilities, patients, and populations ^{16, 19– 23} . Furthermore, we hypothesized that PMM interventions may operate at individual (providers, managers, etc.) and/or organizational-levels (facilities, networks of care, local health systems, etc.) and can trigger outcomes across short and long timeframes (desirable as well as undesirable, adverse effects). The main types of PMM interventions and outcomes of relevance to primary care delivery systems are listed in Table 1.

Primary care performance improvement in LMICs has been mostly studied to date by means of research that addresses the effective delivery of health interventions by providers and facilities. Such studies rarely address the effects of PMM interventions on the behaviors of providers and facilities charged with service delivery. We believe that the understanding of health system performance improvement processes requires research that characterizes the context, mechanisms, and processes through which various PMM interventions trigger (or not) process improvements, organizational learning, and system-wide adaptation including but not limited to the emergence of quality supply, patient safety, and population-level equity and health outcomes. Such research is necessary in support of recent calls for a revolution in quality health systems in global health settings ³ .

This evaluation protocol aims to characterize how, why and under what contextual conditions has the Salud Mesoamerica Initiative (SMI) triggered performance improvements in El Salvador and Honduras through the introduction of various types of PMM arrangements. In the next section we introduce SMI, and in subsequent sections we describe the rationale for the evaluation, the methods to be employed, and further discuss the strengths and limitations of the proposed research.

SMI is a multi-country, large-scale PMM initiative resulting from the partnership between the governments of the eight Mesoamerican nation-states, the Bill and Melinda Gates Foundation, the Carlos Slim Foundation, the Government of Canada, and the Inter-American Development Bank (IADB). SMI is a performance-based financing program that supports participating governments’ production of population-level health and equity outcomes through arrangements that ultimately aim at improving primary care delivery for the poor at scale.

The program was sequenced as three consecutive phases of eighteen to twenty-four months each, for the achievement of progressively complex performance targets in reproductive, maternal, neonatal and child health. Phase 1 programs started in a staggered fashion in 2011 and the final stage started in 2018 and will end in 2020. Performance targets during phase 1 had an initial focus on adherence to standards of care, availability of supplies and, in general, process and output targets. During phases 2 and 3, targets prioritize outcomes such as modern contraceptive prevalence, effective coverage of antenatal care and institutional deliveries, post-partum and post-natal care coverage, and in some countries, reductions in the prevalence of anemia and gains in immunological coverage of measles vaccination ^{24– 26} .

At baseline, IHME collected data from 20,225 households and 479 primary care facilities in the poorest, rural municipalities of all participating countries. Results varied significantly between and within countries, underlying differences in health system performance, availability of inputs, quality of services, and highlighting poverty-related and other disparities in health outcomes ^{26, 27} .

Upon joining SMI, participating governments contributed domestic funds and formally agreed with the IADB to a set of performance targets for each of the three phases. The IADB then matched domestic contributions with grant financing on a 1:1 ratio. Performance contracts between the IADB and each government provided that the former would reimburse half of the initial domestic investment, contingent on the achievement of 80% or more of the agreed-upon targets. Measurement of programmatic performance by an external, mutually trusted agency (i.e., IHME), was required to ensure accountability and credibility in results.

SMI’s original theory of change ( Figure 1) hypothesized that the supply-side financial incentives would target ministries of health (MOH) attention on achieving the agreed performance targets and that the latter would be reinforced by the external measurement of performance. These cycles would be further reinforced by ongoing technical support, policy dialogue, and purposeful dissemination of performance information. Such processes would, in turn, lead to progressive improvements in the availability of quality supply and enhanced, aggregate performance in the primary care delivery system. Additional causal assumptions rested on an increase in domestic pro-poor health spending, and an expansion in the demand for high-impact health interventions among beneficiary populations.

In 2011, the partners agreed on a set of common, high-level principles such as a focus on results, independent performance measurement, and mutual accountability and transparency. These principles established the institutional boundaries that, in turn, allowed the IADB to negotiate country-specific performance contracts, results frameworks, and evaluation plans with each participating government. The implementation approaches through which the program’s PMM interventions would be transmitted downstream into the delivery systems were not prescribed a-priori by SMI and were, instead, left to country-specific, flexible implementation arrangements.

In the two countries under study, El Salvador and Honduras, the focus on country ownership led to each government deciding how to deploy SMI’s non-reimbursable resources and their own domestic financing for the achievement of program targets. El Salvador had gone through a health system reform in the late 2000s, which coincided with the beginning of SMI implementation. There, the government decided to focus its targets on the provision of universal primary care services through Community Health Teams ²⁸ , one of the reform’s central features. Honduras, in turn, had started a large-scale contracting-out and pay-for-performance programs in the late 2000s ²⁹ . The government decided to leverage its experience with those PMM financial arrangements and implemented SMI in primary care systems that had already acquired experience with PMM arrangements. Table 2 lists some of the targets agreed by El Salvador and Honduras.

This study protocol addresses two research questions: (1) What are the effects of using supply-side incentives on the performance of primary care systems in El Salvador and Honduras? How are those effects produced and under what contextual conditions? And, (2) What are the effects of external measurement of performance on the primary care systems of El Salvador and Honduras? How are those effects produced and under what contextual conditions?

We recognize that the evaluation of a program as complex as SMI needs to be informed by methodological approaches that go beyond the measurement of progress against agreed-upon performance targets and should also attempt to further explore the lessons that can be learned from the flexible, adaptive nature imagined by SMI and its sponsors. By providing governments with a high degree of flexibility in implementation, SMI introduced important distinctions with other global program partnerships. It explicitly attempted to increase government buy-in, and encouraged local adaptation and learning which, in turn, make concerns with fidelity in implementation less important than, for instance, characterizing the adaptations that worked or not, and why. Therefore, in this study protocol we follow the approach used in the evaluation of other whole-system transformational reforms which suggest that “program fortunes can be shaped and constrained by interactions between the program and the context” ³⁰ in each participating country and, also, by the necessary adaptations and responses to dynamic and changing environmental conditions.

To address the research questions and the complex dynamics introduced by SMI’s adaptive approach, we decided to use realist evaluation. Realist evaluation is based on the premise that an evaluation needs to answer “what worked, how, in what circumstances and for whom” ^{31, 32} . It is a form of theory-driven program evaluation that has been used in evaluation studies and in health systems research for the evaluation of complex policies, programs and interventions in various socio-economic settings, including LMICs ^{33– 44} . The appeal of this approach, compared to other theory-driven methods, lies in its explicit foundations in critical realism – an epistemology located between positivism and relativism. Such perspective contends that program interventions bring about change through underlying, usually hidden, causal mechanisms, and considers the role of context as indispensable in explaining causality.

The starting point in a realist evaluation is the development of a Program Theory (PT). In this study protocol, the preliminary PT was developed based on previous research identified through literature review, document review, and consultations with experts involved in SMI’s design, implementation, and evaluation. The PT will be used to inform the process of data collection and to the completion of a refined PT. The latter will provide explanations of why, how and under what conditions do SMI interventions trigger causal mechanisms that, in turn, lead to specific outcomes, intended and otherwise.

This evaluation is an 18-month study running from May 2017 to December 2018 and executed contemporaneously with the finalization of SMI’s phase 2 in El Salvador and Honduras. The evaluation seeks to maximize diversity in institutional and policy context to increase the likelihood of identifying variations in policy and program conditions and thus characterizing the process of change generated to date by the program.

At the country-level, a case-study design with contrasting cases was selected as the primary study design. We defined each country’s primary care system as the unit of analysis. Furthering the purpose of this evaluation to understand high-performance at large-scales, we will also purposefully identify and study outlier, high-performing primary care delivery systems which “can reveal a great deal about intense manifestations of the phenomenon of interest” ⁴⁵ . Contrasting case approaches align well with the realist evaluation proposition that contexts can trigger to-be-identified mechanisms that, in turn, interact with program interventions and contribute to generating outcomes (or not).

This step has already been completed. For the development of the preliminary PT, we first reviewed the literature to identify social science theories and empirical evidence that explicitly addressed PMM interventions and outcomes in public and private organizations; characterized the mechanisms of change of relevance for large-scale health system change; and, explored the scarce evidence that exists about the role of context in triggering or obstructing health system transformation. Given that one of the authors (WM) was involved in the production of an evidence gap map on PMM systems in primary care delivery in LMICs, we used their systematic search of various academic databases ⁷ to inform this protocol’s preliminary PT ( Supplementary File 1 contains the MEDLINE search strategy). The combination of a scoping review of the PMM literature and the systematic search required by the evidence gap map mentioned above, helped us identify several social science theories that provide causal explanations about the mechanisms through which PMM interventions produce organizational change at multiple levels within a primary care system. A paper summarizing the findings from the evidence gap map will be published separately.

Context - We hypothesize that individual and interpersonal actions and reactions to SMI interventions will be influenced by the context in which providers, facilities and MOH organizational units are embedded, a feature that is particularly relevant in complex programs such as SMI ^{46– 48} . In this evaluation, context includes the institutional and policy setting that formalizes the laws and rules that govern the public sector in general, and the primary care delivery system, in particular. It also encompasses the internal organizational environment and the related practices, routines, and collective norms that drive organizational culture; and, the socio-economic local environments where primary care supply and demand interact for the production (or not) of outcomes (intended and otherwise). Finally, global agendas can also influence the choices and actions of high-level policy-makers ^{49– 51} possibly through their interactions within various policy networks ^{52– 56} .

Program actors’ actions and reactions to context factors and to SMI interventions will likely vary according to their levels of interest, engagement, and resistance; relevant antecedents and experiences; the degree of system readiness among providers, managers and policy-makers; and on inherent features of program interventions and reforms, all of which have been empirically studied from the perspective of the theory of diffusion of innovations ^{4, 30, 57– 61} .

Mechanisms – We identified performance-driving mechanisms at three levels within a primary care system: individual, interpersonal and collective. From such multi-level perspective, primary care system states cannot solely be attributed to the behaviors of individuals but to the triggering of up to three types of interrelated causal mechanism: situational, action-oriented, and transformational mechanisms ^{62– 64} . Situational mechanisms refer to the macro, organizational-level environment in which system actors and their social interactions occur including, among others, the social institutions and collective norms that can exert influence on individual actors (macro-to-micro change). Action-oriented mechanisms explain how individual actors’ ideas, actions and reactions influence other individuals’ behaviors across the system, usually through diffusion from one to many actors (micro-micro change). Finally, transformational mechanisms explain how the sum of new behaviors by multiple actors bring about larger-scale changes in macro institutions and social norms (micro-to-macro transformation). In this evaluation, we propose to study individual and interpersonal mechanisms only.

At the individual level, we hypothesize that the motivation of healthcare providers has the potential to play a catalytic role in the generation of performance gains in facilities and local primary care delivery systems. At the interpersonal level, we theorize that social connections, imitation and the diffusion and dissemination of new beliefs and behaviors will further trigger their internalization and assimilation within primary care organizations. Downstream, the institutionalization of new organizational routines and practices through top-down policies and regulations will further normalize pro-performance behaviors across the primary care system. Program effects would accrue at any and all of these three levels of system change. Also, at each of these levels of potential system transformation, passive or active resistance by system actors may hinder, delay or entirely block the process of change, leading to any combinations of underperformance or performance failure.

Mainstream research in economics, psychology, organizational behavior, and public administration, among other fields, tends to assume that incentives and rewards serve as powerful motivators for the achievement of desirable behaviors among utility-maximizing, rational individuals ^{65, 66} . This approach has fueled the design of various types of accountability-driven, PMM interventions that borrow performance approaches from the private sector and apply them in LMIC public settings, particularly in health and education. Such interventions attempt to reduce the misalignment of incentives between principals (voters, legislative bodies, executive-level leadership, funders, etc.) and their agents (program implementers, care providers, etc.), ^{67– 70} . Many public-sector reforms in LMICs and various global health partnerships have been influenced by this body of knowledge and by the adoption of PMM approaches by various global health program partnerships including SMI itself, the Global Finance Facility, GAVI, and the Global Fund to Fight AIDS, Tuberculosis and Malaria, among others.

Theoretical and empirical developments in public administration research also suggest that workforce motivation can be explained by intrinsic motives such as public service motivation, a socially learned set of preferences prevalent among individuals working in the public sector ^{71, 72} . In our search for substantive theories and evidence that could explain workers behaviors in primary care settings in LMICs, we decided to focus on self-determination theory, a macro theory of human motivation that has been used in recent years to study workforce motivation in various contexts, including LMIC ^{44, 73} . The theory has good cross-cultural validation and has demonstrated that individual workers who satisfy internal needs for competence, autonomy and relatedness feel intrinsically motivated and committed ^{74– 77} .

Regarding interpersonal mechanisms, we theorize that diffusion mechanisms within socially connected individuals can trigger interpersonal, action-oriented mechanisms that spread ideas, perceptions and behaviors from a few individuals to many more. We based this hypothesis on the diffusion of innovations theory ^{59– 61, 78– 81} , neo-institutional theory ^{82– 87} and, in particular, on recent characterizations of the processes of change triggered by PMM interventions in public sector organizations ^{82, 83} , and in healthcare settings ^{88– 90} .

Collective, whole-system change is the least theorized and empirically studied type of system transformation in healthcare. Given that we will study SMI at its mid-term, our assumption is that such types of transformational mechanisms may not be observable. However, based on the scarce number of studies that have addressed large-scale system change in healthcare settings ^{4, 30, 91– 93} , we propose that, were individual and interpersonal efforts at primary care system change sustain through time, such process may lead to the accumulation of new practices, routines and organizational behaviors beyond individual groups and teams through the norming of pro-performance, pro-social organizational cultures and social learning and modelling ^{94– 96} . Such system-level changes could lead to the emergence of quality and safety effects across the health primary care. The repetition of these cycles of improvement and learning would also lead to the generation of population-level health and equity outcomes, intended and otherwise.

Based on the evidence and theoretical framework discussed above, the preliminary PT was developed as a series of linked propositions ( Box 1) and was also represented in graphical form ( Figure 2) to highlight the interrelated linkages between system elements and to avoid perceptions of linearity in causal reasoning.

Realist evaluation is method neutral. The nature of the phenomenon under study, the research questions, and the preliminary PT are the main factors that define study design and data collection methods ^{31, 32} . In this study protocol, the primary data collection methods will be key informant interviews, non-participatory observation, and document review. Data collection will proceed between May 2017 and December 2018.

Study participants and sample. Study participants (or “key informants”) will be recruited based on their deep knowledge of and involvement in SMI, the central phenomenon of interest in the study ^{45, 97} . Therefore, while the sample size cannot be determined a-priori, for planning purposes we estimate the need to conduct approximately eighty (80) key informant interviews in the two countries. The adequacy of the final sample size will be continuously assessed during the research process.

Key informant interviews will be collected from four sets of actors: 1) Country policy- and program implementation actors; 2) Health care providers at primary care facilities; 3) Performance verification and evaluation stakeholders; and, 4) Program designers.

Country policy- and program implementation experts will have been involved in the governing of each country’s health system and/or in the design and implementation of phases 1 and 2 of the SMI program. These interviews will help understand the institutional and policy context; any relevant antecedents to the SMI program; and, also, concurrent investments by the government or external financing agencies in the same areas where SMI was implemented. Health care providers included will belong to high-performing primary care facilities directly involved in the delivery of health services in SMI areas of influence. Interviews will characterize the delivery of services; the relations between providers and the communities they serve; the features of the implementation of SMI; and, the perceptions and behaviors triggered by the latter. Performance verification, evaluation and initiative-wide management stakeholders (IADB and IHME) will be interviewed to acquire information about SMI’s interventions from their perspectives. Respondents will be invited to participate voluntarily in the study; and, no compensation will be provided.

Data collection. Specific questions will explore reasons for policy-makers to join SMI; respondent perceptions about the interventions under study; reactions triggered by the use of supply-side incentives and performance measurement and management arrangements and interventions; knowledge of and perceptions about context-specific factors that hinder or contribute to the actions and reactions among program actors, (local socio-economic context, institutional factors, and internal organizational environment); description of additional interventions that could explain program effects; and, effects or outcomes generated in an unintentional fashion. Interviews with country actors and stakeholders will be conducted in Spanish by bilingual members of the research team. IADB and IHME respondents will be interviewed in English. All interviews will be recorded and transcribed verbatim and, when applicable, professionally translated into English. Semi-structured interview guides will be used for data collection ( Supplementary File 2).

We will use non-participant observation to collect information about the process of dissemination of results from the external measurement of performance at the end of phase 2, in 2018. We will document the process followed in the policy dialogue session, the agenda, components, objectives, and the reactions by domestic stakeholders. Summary memos of the observations will be generated to be maintained in the project files.

To further understand policy and program context, the study will also review key program documents pertinent to the design, implementation and evaluation of SMI in El Salvador and Honduras. Specific attention will be given to documents that describe the policy and program context in each country, the implementation strategies, and the performance and evaluation frameworks. Also, we will identify sources of secondary data that may be used for future triangulation. A complete list of reviewed documents will be maintained and included as a supplemental file with the final report of findings.

The analysis from the interviews will be conducted using an integrative methodology that merges both inductive and deductive approaches ⁹⁸ . We will construct a set of a-priori codes drawing from the theory-driven perspective used in realist evaluation to develop PTs, as described above. This will be combined with emergent inductive codes identified from a rigorous open coding process.

Realist evaluation collects data from various sources and, based on that, aims to build plausible accounts of key program events, adjustments in implementation, and on their intended and unintended effects ⁹⁹ . In an initial stage of data analysis, two coders will review a sub-set of transcripts in an iterative and systematic manner using the constant comparison method, and afterwards finalize the codebook through negotiation. Subsequent transcripts will be coded by three experienced coders using the final codebook.

The coded data will be appraised using complementary analytic approaches. Coders will use iterative conceptual and pattern coding to identify major themes within and across cases. Within-case analysis will proceed as follows: deductive codes from each transcript will be aggregated into tables to identify preliminary linkages whereby certain outcomes are related to specific context-intervention-mechanism configurations. Coders will also scan each deductive coding category across the entire sample of responses to identify commonalities and differences, e.g. multiple combinations of contexts that could facilitate/inhibit the interventions; or a confluence of interventions that are catalytic and reinforce one another. We expect these analytic approaches to be complementary, and to allow building context-mechanism- outcome (CMO) configurations that will then be gauged to determine which patterns plausibly explain how program interventions generated the observed effects, expected and otherwise.

At the conclusion of this stage, the resulting data will be integrated into preliminary analysis documents and diagrams to reflect the team’s visualization of emergent, alternative theories of system change. The final thematic structure will be used to refine the preliminary program theory for developing within-case theories of system change at the delivery and policy levels. Also, for refining across-case theories at the same levels of analysis (e.g., policy- and primary care delivery-levels) to assess the extent to which the same mechanism may explain different outcomes in different contexts ³⁰ . Data analysis will be done with QRS nVivo. Furthermore, analysis will be complemented by contrasting evaluation results and emergent causal patterns with other relevant SMI studies and data sets ^{25– 27, 100, 101} . The presentation of findings will be made following the standards developed for the reporting of realist evaluations ¹⁰² .

A set of measures will be taken to increase the validity of the study in terms of reflexivity, credibility and confirmability, and enhance the trustworthiness, transparency, and accountability of the research. All researchers will engage in the introspective practice of maintaining ‘personal biases memos’ to make explicit all self-identified biases and pre-conceptions that may affect the research process ^{103– 105} . All analytic decision notes and memos, biases memos, document analysis syntheses, interview guides, research team meeting agendas and minutes, and analysis outputs including coded transcripts, conceptual frameworks, tables, etc. will be preserved to provide a verifiable audit trail.

The study’s protocol was reviewed and declared exempt by the George Washington University’s Institutional Review Board (study number 041733). The Ministries of Health of El Salvador and Honduras were informed of the proposed research by the IADB and provided written approval for the research activities.

Ethical approval documentation will be made available on request. The study will employ scrupulous adherence to the highest ethical standards, and current international and local legislation pertaining to research governance. The data collection will operate under explicit informed consent, which will be preserved in study records. Respondents will be given the choice to provide consent verbally on tape before the interviews, or in writing. To maintain anonymity, respondents will reserve the right to review the study outputs and withdraw consent if necessary. All identifying information will be removed from transcripts and stored separately with access restricted to the research team. All transcripts will be stored electronically in password protected cloud services, and physical documents will be securely stored at George Washington University, Milken Institute School of Public Health.

This paper describes the protocol for a realist evaluation of PMM interventions introduced by SMI in the primary care systems in El Salvador and Honduras. The protocol proposes a contrasting case study design with outlier sampling, to understand how, why, and under what contextual conditions do SMI’s PMM interventions trigger high-levels of performance at scale. To our understanding, this is one of the few realist evaluations addressing PMM systems in primary care systems in LMICs. However, faced with budgetary and operational constraints, the researchers made design choices which generate challenges that are discussed below.

The first challenge relates the scarcity of theory-driven evidence on PMM systems initially identified when we scoped the health systems research literature in general, and in LMICs in particular. Many studies were undertheorized, disregarded context and mostly focused on dimensions of performance that addressed the effective delivery of specific health interventions. Various impact evaluations tried to isolate the factors that contributed to the achievement of specific health outcomes, but under-theorized and infrequently measured the causal contributions of the PMM interventions of interest to this evaluation. Due to such design choices by conventional impact evaluations, many studies that have attempted to study primary care performance to date provide a poor understanding of the potential effects that PMM interventions may have on the behaviors of providers, facilities, or higher-levels of LMIC primary care systems. Similarly, if and when mechanisms or context factors are theorized or reported, such factors tended to be framed in terms of their contributions to health outcomes of interest, not to causal effects on the attitudes and behaviors of system actors at individual, interpersonal, or collective levels. Also, much of the health systems research tradition appears to have developed in isolation from other disciplines that have a rich tradition of PMM theorizing, research and practice including public administration, organizational studies, and social psychology, among others. To address the issues above, we decided to, first, develop a conceptual framework that is informed by multi-disciplinary research and theory mostly arising from experiences in industrialized settings.

Another challenge arose from the choice of a cross-sectional study design. While it would have been desirable to accompany SMI and its domestic partners in a continuous process of sense-making and performance information use, such approach was not feasible with the resources available. Therefore, this evaluation may serve as the starting point in theorizing multi-level, complex and dynamic processes of large-scale system change in high-performing primary care systems in SMI; we also expect that its findings will further support and complement SMI’s learning and evaluation plans, and contribute to generating theory and evidence of relevance to other global contexts.

Issues of country case selection also need to be explicitly addressed. As this is the first realist evaluation done in the SMI context, the research team chose to focus on characterizing high-performance systems and thus only included positive outliers at the national and primary care delivery levels. Such decision was made to maximize information power and richness from studying similar, extreme cases. While acknowledging that a contrasting case of high- and low-performer countries and/or primary care delivery systems would have been desirable, such design was not feasible at this stage, yet can be undertaken in the future. Furthermore, understanding how and why system improvements sustain (or not) through time remains a valuable yet complex research endeavor that requires further theorizing and additional empirical studies that are outside the scope of this evaluation. Given the unusual duration of SMI’s implementation period, the initiative offers a unique learning space from which to acquire new knowledge about the processes underlying system inertia, resistance to change, and the dynamics of systems that ‘learn” and adapt through time ^{106– 114} .

The refined PT and other results from this evaluation have several anticipated uses and applications. For instance, we expect that program implementers will use the findings to assess program adjustments in its third and final phase (2018–2020); also, to identify options for re-designing domestic health policies and new evaluation priorities; and, to inform the design of longitudinal, experimental or quasi-experimental evaluation designs that may deepen one or more of the various casual patterns identified.

There is growing concern that high-performing primary care systems are needed to prevent global pandemics and to deliver on the promises of Universal Health Coverage and the Sustainable Development Goals ² . This realist evaluation aims to contribute to such ambitious goals by conducting this study in El Salvador and Honduras, two of the top-performing low-income countries in SMI. The findings of this evaluation in regard to how, why and under what conditions have these two low-income countries transformed their primary care PMM systems, will provide learning opportunities for spreading insights, evidence and new theory to other countries trying to address similar challenges.

No data is associated with this article.