Embedded implementation research determinants in Latin American health systems

ABSTRACT OBJECTIVE: To assess the determinants of embedded implementation research (EIR) conduct in seven Latin American and Caribbean countries. METHODS: This qualitative interpretative study conducted and analyzed 14 semi-structured interviews based on a grounded theory approach using Atlas-ti© 7.5.7. We grouped the conditions appointed by interviewees as determinants of EIR conduct into six domains. RESULTS: The participation of high-level engaged decision makers as research co-producers is an important EIR determinant that fosters research use. Nevertheless, EIR faces challenges such as dealing with key personnel changes and fluctuating political contexts. CONCLUSIONS: Despite its limitations, EIR is effective in creating a sense of ownership of research results among implementers, which helps bridge the gap between research and decision-making in health systems.


INTRODUCTION
Most efforts to put scientific evidence in the hands of policymakers and health system managers to ensure evidence-based informed decision making have been based on one-way communication from researchers to decision makers 1 . Generally described as knowledge translation 2,3 , its general objective is to present research results using language and forms accessible to decision makers who, eventually, will use it.
But empirical findings from studies have not always answered policy-relevant questions, mostly because end-users were not involved in the research cycle from the beginning. To meet these challenges, new innovative forms of collaboration between researchers and decision makers have been developed. Initially, they were limited to consulting decision makers in search for a common research agenda and setting priorities 4 .
More recent approaches try to include research results users as active partners throughout the research process, under different names: collaborative research, engaged research, participatory research, research co-production, integrated knowledge translation, or, as in our study, embedded research (ER) 1,[5][6][7] .
Embedded research can also be understood as integrating research itself within organizations to ensure that the entire research process is conducted collaboratively by the health personnel along with professional researchers, creating a synergy based on their mutual expertise to improve implementation. Research and policy social actors thus participate together in identifying and defining research problems, as well as designing and conducting the research, analyzing, disseminating and accepting results 8 .
The main objective of implementation research (IR), which is particularly linked to embedded research, is to improve program implementation and not just to advance knowledge production 9 . According to this perspective, decision makers and implementers have relevant knowledge gained by their daily experience that is directly useful for improving implementation. As shown in several studies, IR is well-suited for embedding research in health programs, policies and services 10 .
On these bases, in 2014 and 2016 the Pan American Health Organization (PAHO) and the Alliance for Health Policy and Systems Research (AHPSR) launched the Improving Program Implementation through Embedded Research (iPIER) initiative to strengthen program implementation and provide support to research projects conducted by decision-makers and researchers in Latin American and Caribbean (LAC) countries 11,12 . It funded implementation research projects, mostly in public health institutions, where a decision maker led a large team including other health staff and local researchers. The initiative focused on capacity strengthening at the individual and institutional levels, based on workshops and ongoing technical support provided by researchers from the National Institute of Public Health of Mexico, co-authors of this paper.
Embedding research in health programs, policies and services is a complex process that requires deeper understanding. Certain conditions may favor or hinder its implementation and have yet to be sufficiently examined, needing further discussion and new perspectives. Existing knowledge gaps have to be bridged to make the most of this kind of efforts 13 . These conditions refer to some of the domains of the Consolidated Framework for Implementation Research -CFIR 14 , mainly the outer and inner settings and the characteristics of the participating individuals.
This article draws on the iPIER 2016 experience with seven decision-maker led teams in Argentina, Bolivia, Brazil, Chile, Colombia, Peru and Dominican Republic. Each team focused their research on existing health programs, policies and services searching for the best ways to improve their implementation (Table 1). Our aim is to analyze the different conditions that influence, facilitate or hinder the development of embedded implementation research in Latin American and Caribbean countries.

Design
Ours is a qualitative-interpretative study based on the methodological postulates of grounded theory 15 , using thematic analysis. The semi-structured interviews explored the perceptions of key informants on the conditions that facilitated or hindered EIR in their contexts, and the possibilities of using the results derived from research projects.
We conducted two interviews per team, one during the first stages of the research projects and a second at the end. As our aim was not to evaluate changes derived from the IPIER initiative itself, we do not differentiate them as baseline and follow-up.
An anthropologist with a master's degree in Public Health and Epidemiology and no ties to the participating institutions conducted both the interviews and qualitative analysis. Before each interview, the researcher introduced himself and explained the purpose of the interviews and the study. The authors had no previous connections with the study participants.

Participants and Study Setting
Participants' selection criteria included being the research project leader or co-investigator and voluntarily accepting to participate in the study based on an informed consent procedure. We assured privacy for all semi-structured interviews held, interviewing one or more people from each team according to their interest. In the end, 15 informants participated in the interviews and no one declined (Table 2). • Train staff on proper management. • Promote civic and institutional activities to raise awareness on the users' role in their own healthcare.

C6
Malnutrition care in infants Regional • Socio-cultural problems: lack of mothers' time, dislike of the micronutrients' taste, lack of understanding on the micronutrients importance.
• Institutional problems: shortage, lack of time to provide guidance.
• Identify socio-culturally appropriate alternatives for the intake and consumption of micronutrients in infants.
• Collaboration with other programs.

C7
Tuberculosis healthcare in prisons Regional • Lack of access to health services with a right to health perspective.
• Lack of coordination between the provincial health ministry and the provincial justice ministry.
• Provide training on tuberculosis and human rights and design a protocol of actions to detect damage to the inmates' health.
• Propose an inter-ministerial collaboration agreement.
* We replaced the names of countries to assure the anonymity of informants.

Data Collection
Interviews took place between October 2016 and November 2017, during and after research activities. We conducted a total of 14 semi-structured interviews in Spanish a , two for each team, following two interview guides (Appendix 1). The first interview took place while the participants were finishing their protocol or starting fieldwork; the second, when they were writing their final report. According to what each team considered appropriate, the same or different people participated in the first and second interviews. During on average one hour, all interviews were audio recorded and transcribed with the informants' agreement.
At the end of each interview, the researcher wrote analysis notes.

Data Analysis
We performed a thematic analysis 16 based on grounded theory principles to process the narratives with axial coding 15 , using Atlas-ti © 7.5.7 software. After interviewing the participants and codifying the transcriptions, both done by the same researcher, we performed an interpretative triangulation with other researchers reviewing the testimonies.
We decided to use an a priori "self-selected" sample to draw lessons to improve the development of initiatives such as iPIER and to provide further evidence on EIR usefulness. We thus observe that the identified theoretical saturation and explanatory density 17 refer mainly to the specific contexts analyzed. We selected testimonies that best illustrate the actors' different perceptions on the determining conditions to perform EIR. Each testimony received an identifier to help read and interpret them. The countries and informants' names were anonymized, keeping only those characteristics relevant for data interpretation. All testimonies presented here were translated into English from the Spanish transcripts.
We developed 36 codes for analysis and selected six main themes related to the categories of outer and inner settings of the CFIR (Appendix 2) 14 , also considered in the interview guide. We included other subthemes that emerged during data analysis and provided explanatory density.

Ethical Considerations
The research protocol was approved by the Research Ethics Committee of the National Institute of Public Health of Mexico (CI-1454/02-2017). Each project principal investigator a The Brazilian team, however, answered the questions, formulated in Spanish, in Portuguese. The answers were later transcribed and translated into Spanish. received an informed consent form by email, signing and returning it. An oral informed consent was required before each interview. Table 1 summarizes the main findings and recommendations of the seven projects, providing an overview of the health issues they addressed. The implementation problems identified and the recommendations suggested to solve them are consistent with the administrative level of the programs and policies analyzed; a relevant issue related to the feasibility of applying research results and recommendations.

RESULTS
The 15 iPIER initiative participants interviewed, 11 women and four men, were members of health public administration, health services or academic institutions. Interviewees included principal investigators, co-researchers and researchers participating in the research projects from various professional backgrounds: physicians, political sciences, psychology, industrial engineering, anthropology, bacteriology and sociology specialists ( Table 2). Table 3 contains the description of the main themes and subthemes that emerged from the analysis, as well as all the related golden quotes.

Methods development and application
In general, all participating teams lacked clarity regarding health systems research and, to a greater extent, implementation research at the beginning of research activities.
Although some teams included experienced researchers or had them as partners for this project, such professionals usually had an epidemiological research orientation, having difficulty in understanding and adopting a health systems and implementation research perspective (quote 1.1.1).
Participants recognized that both during the research design phase and its results use, one must identify an implementation problem that can be solved with the available resources. More comprehensive problems related to structural issues are usually beyond the institutional influence capabilities of those conducting this type of research (quote 1.2.1).
Different informants mentioned that answering the research questions usually requires using qualitative research methods to reach the depth that will allow providing the correct answers and achieving the objectives (quote 1.3.1). And this is because EIR quite often requires direct consultation with program implementers to consider their experience (quote 1.4.1).
Similarly, several participants mentioned that it is equally important to gather the point of view of health service users to discover social and cultural factors that can affect the implementation of programs and policies (quote 1.5.1).

Timeline and human resources availability
Despite one of the main conditions of the EIR being that it must be led by implementers themselves, due to the heavy workload that implementation imposes on decision makers and frontline staff, participants recognized the need to outsource professional researchers to carry out some research activities (quote 2.1.1).
Likewise, considering these activities would be added to their regular tasks, interviewees recognized the need to generate a key motivation for the research and its results among the personnel participating in EIR projects (quote 2.2.1).
Several participants mentioned that the research team profile is a relevant issue, since its members need to know not only how the program or policy is implemented, but also how to do research (quote 2.3.1). In this sense, the inclusion of high-level decision makers in the research team appeared as strategic to facilitate fieldwork (quote 2.3.2). In the municipalities they do very little on health issues, they do very little while they could be taking a more important role in prevention and search for a solution (…) This is going to be a good tool at the national level that will help the local and regional authorities to act. (PK, C2, 2°, F, CI, Medicine) 6.1.2 I believe that [with our research] the process [of the elaboration of a drug bioequivalence policy] will become visible since it worked, as little as it worked, but it did work… and I think that we will also generate inputs for more critical references for these processes so they can follow-up and monitor future policies in the domain that we have evaluated. (PG, C4, 2°, M, CI, Anthropology) 6.1.3 It is very possible that the results be applied, the health system is strengthening its steering system, so this kind of research is timely and pertinent… in fact, the Health Ministry is currently in the process of clinical practice guides and protocols development. (PO, C1, 2°, F, CI, Medicine) 6.2 Different strategies to apply EIR results.
6.2.1 We are still working in the second phase of the project, which consists of the establishment of a deliberative dialogue among social actors that really are responsible of the implementation of the TB prevention, treatment and control policy in the prisons of (our)the province… (PB, C7, 2°, F, CI, Political Sciences) 6.2.2 I think that the results will help, but we have to present them in different decisionmaking levels, for instance, at the different health centers, this is super important (…) and to the regional authorities of the programs.  Although iPIER teams received financial support to conduct their research projects, participants noted that the availability of financial resources is essential to conduct EIR (quote 3. 1.1 & 3.1.2). The lack of resources allocated or earmarked for research, especially EIR, limits decision-makers' and implementers' capabilities to perform it, mainly because of the implicit workload in their roles and the lack of resources in their institutions (quote 3.1.3).

Institutional dynamics and organization
Participants recognize that, in certain circumstances, there are regulatory limitations to conducting research within government offices linked to program implementation. In such cases, the project outsourced research activities to academic institutions (quote 4.1.1).
According to the interviewees, the instability of some high-level decision makers' positions is a contributing factor to the lack of institutionalization of certain health policies and programs and also affects EIR (quote 4.

Political environment
Despite the relevance that EIR has at the national level when backed by international technical and financial support, according to some interviewees, the inclusion of certain health issues on the political agenda is a key aspect for improving policies and programs that can determine the possibilities of conducting EIR. In a way, health issues are prioritized according to their visibility to decision makers (quote 5.1.1).
But how one addresses a health issue (even the decision to address it) depends on the political perspective of the group in power (quote 5.1.2).
Sometimes research has to be reallocated due to the arrival of high-level decision makers with particular views on what health issues should be addressed and how (quote 5.1.3). These situations directly impact the possibility of conducting EIR and also determine being able to introduce changes in the programs based on research results.

Perception on the use of research results
Participants noted that EIR results can be used in different ways: a. As evidence to generate changes in the actions of program implementers (quote 6.1.1).
b. As a contribution to different sectors of society lobbying for better public policies and generation of policy options (quote 6.1.2).
c. As a useful tool to develop or improve healthcare guides and protocols (quote 6.1.3).
In this sense, interviewees mentioned different strategies that allow using EIR results to promote changes in the policies and programs studied, for example: a. Conduct deliberative dialogues with implementers (quote 6.2.1).
b. Disseminate results at different decision-making levels (quote 6.2.2).
c. Raise awareness among high-level decision makers on the relevance of the results (quote 6.2.3).
Finally, to create changes in health policies and programs based on EIR results, participants remarked the need to involve different social actors (quote 6.3.1).

DISCUSSION
Our study points to different factors that can either facilitate or hinder EIR performance and the eventual use of its results to improve health policies, programs and services, mainly in the context of Latin America and the Caribbean. Focusing on the results of the 2016 IPIER initiative, this study first contrasts its findings with the published materials on the 2014 initiative, searching for new lessons. In general, one of the main barriers for conducting EIR is the lack of certain research capabilities among implementers. According to the EIR approach, these actors should lead the research efforts, emphasizing their prominence and collaboration with the professional researchers supporting them. This finding is consistent with the published literature 12 and directly affects the development of research projects.
While the EIR approach contributes to improving decision-makers' research capabilities, the goal is not to train them as researchers, but to improve their understanding of the usefulness of research evidence in improving health system performance 12 . Realizing the value of the EIR helps generate a sense of ownership of its results and implementers can identify them as a product of their own efforts 6,12,18 . This may even reduce the need for knowledge translation and promote immediate improvement of health policies and programs 12,19,20 .
Different participants agreed on the great advantage that having a high-level decision makers in the research team brings, echoing previous evidence in this regard 12,19,21,22 . Their administration skills facilitates executing EIR as well as the immediate or later use of results 12 . Although they usually represent a strategic force, under certain circumstances their participation can also become an obstacle. Among the most relevant are the great mobility of their posts (also causing the lack of program continuity), their heavy workload, and, eventually, their limited research capacity and experience 12 . Sometimes, these limitations were resolved by including professional researchers as close collaborators of the teams and, mainly, of the decision-makers.
When regulations prevented institutions from conducting research, direct relations with academic institutions and the outsourcing of professional researchers allowed to conduct the EIR projects. But with such research experts always working under the implementers' leadership.
Although the projects in this study received financial support, interviewees repeatedly remarked the lack of resources to do research as a major limitation for the development and continuity of EIR in LAC 5 .
Participants highlighted political events as important phenomena that affect the health system and, consequently, the conditions in which EIR is developed. The 2014 iPIER participants also faced similar situations 23,24 . Such events are also related to the place health issues occupy in the political agendas and how they are prioritized in economic terms and in the media 25 . Similarly, the political views of some groups can influence not only how health issues are addressed, but even if they are addressed at all. The particular interests of social actors with economic power can also limit the possibilities to conduct EIR and use its results.
As such, the willingness to change and different political views and positions at the highest government levels or "political will," as some informants describe it, can hamper implementers' efforts to improve health programs. But as we identified in several testimonies, possibilities for using EIR results are related to what Reich 26 describes as political feasibility. Social actors with enough power and capacity (civil society organizations, health professional associations, organized health services users, academic institutions or private entrepreneurs) can outweigh the opposition to address certain issues. Thus, most participating teams considered the advantages of including such actors to support the use of EIR results, as was successfully done by the 2014 IPIER teams 19,27,28 .
On other occasions, decision makers and health professionals are unwilling to collect information that may highlight deficiencies in the programs they operate 12,20,29 . This is certainly one of the many reasons why professional research is still very important and needs to be fostered within the academic domains where it is normally developed. External and independent research is the basis for objective assessments that cannot be addressed from the perspective of decision makers who have to provide quick and efficient answers to pressing issues.
Our study shows that the resources and support provided by the iPIER initiative were important facilitators of the EIR process. The initiative funded a neglected area of health research (implementation research) for which the participation of health staff is a crucial element 9 . And the technical advice and guidance it offered was based on an iterative and ongoing approach to capacity building, which helped decision makers and their teams develop a broader health system perspective, often absent in their backgrounds.
Another relevant issue is that while EIR aims to identify contextualized solutions to local implementation problems, it also allows to identify systemic health system dysfunctions that can be addressed by interventions at this higher and broader level. This points to a clear tension over smaller and easier to apply improvements in implementation (quick wins) versus longer-term changes in health system performance that are usually beyond the scope and responsibility of program managers.
Even if EIR aims to find solutions for implementation problems from the implementers' point of view, the users' perspective on what works and what can be improved is a recurring theme raised by the interviewees, as in the 2014 IPIER initiative 24,25,27,28,30 .
Strengthening the decision makers' and implementers' research capabilities increases the value and sense of ownership of EIR results among them, which can ensure greater and more direct use of scientific evidence to improve implementation and can shift the need for knowledge translation. EIR can be a good option to increase the relevance and impact of research. As it gains recognition, support and funding, more decision-makers will be willing to participate in this type of collaboration and apply research results to improve health system's performance.
Since no sufficient time has passed since the IPIER experience, the actual changes that could have been promoted by using the results of the implementation research conducted by the participating teams need further research, as is also true for the deepening of capacity building and policy collaboration to advance EIR.
In short, the timely application of certain strategies allows to overcome important barriers, make the most of facilitating factors and promote the use of results. Other challenges are beyond the influence of researchers, health professionals and even decision makers. But promoting and enabling the involvement of several social actors in research can be a powerful strategy in the hands of the health system itself.