1Action research, interventional research, evaluation, applied research, experimentation: the science of population health intervention takes on different forms, names, and concepts, whether in Europe, Africa, or elsewhere. Behind each of them, paradigms, approaches, methods, and disciplines intersect, or not, and debates persist on how to differentiate among them.

2This chapter presents the different forms of interventional research, their characteristics, differences, and complementarity. The aim is to identify, if not a definition, at least a set of attributes that distinguish this research practice from other forms of research. We also aim to counterbalance a historical trend in the public health and global health literature, which up to now has focused mainly on experimental methods to the detriment of a more open, holistic, and diverse vision of paradigmatic and methodological approaches.

The Concept of Population Health Intervention

3The concept of population health arises from a recognition of the limitations of considering health phenomena solely from an individual biomedical perspective, that is, explaining health conditions on the basis of personal factors, and in particular behavioural factors (e.g. smoking, physical activity), and proposing levers, again individually oriented, that generally involve the healthcare system (e.g. vaccination) (Szreter, 2003). In fact, while this view facilitates the orientation of public health interventions (e.g. health objectives, thematic plans), it limits their impact, since each person is part of a system where very many parameters, such as social status (Burtram, 1996), education level (Berkman & Kawachi, 2014), and living conditions (Eckersley, 2001; Kawachi et al., 1999; Leon & Walt, 2000; Marmot & Wilkinson, 2005), as well as interactions with other people making up the population, influence the person’s health and development (Hosseini Shokouh et al., 2017). Thus, and without setting individual health and population health in opposition, population health results from dynamic and interactive relationships between individuals, between individuals and their environments, and between individuals and the services to which they have access and which they do, or do not, use (Diez Roux, 2016).

4This brings us to two fundamental principles of population health: 1) the need to take into account factors operating at multiple levels of organisation in order to understand health and take action to improve it (involving both social and biological processes), and 2) an explicit concern for health equity, since it is not possible to substantially improve the health of the population as a whole without addressing health inequalities (Diez Roux, 2016). Thus, while this population health approach does not exclude healthcare services (such as preventive clinical practices, particularly in the context of the primary healthcare reorganisation called for in the 1978 Alma-Ata Declaration or the 1989 Ottawa Charter), it does encompass many sectors. Population health will, in fact, strive for a health equilibrium, which calls for looking at all the determinants of health and their inter-influence. For example, while lockdown measures imposed to limit hospitalisations of severe cases of Covid-19 among the elderly and vulnerable are understandable from a public health standpoint, they raise questions from a population health standpoint with regard to potential consequences for the population at large (including the youngest or most vulnerable) in terms of deteriorating mental health, recourse to care, standard of living and quality of life, and health equity (Cambon et al., 2021; Campeau et al., 2018; Turcotte-Tremblay et al., 2017). In this respect, the concept of population health is closely aligned with that of health promotion, which has the distinction of adding a clear objective of strengthening empowerment (i.e. the ability of individuals or groups to be able to act on the social, economic, political, or ecological conditions they face). For example, we saw in Africa that it was the younger populations, who are in the majority, that often spoke out, sometimes violently, against these restrictive measures because they did not feel concerned by the pandemic. Moreover, there are currently heated debates around the objectives of conducting mass vaccinations against Covid-19 rather than focusing on the most vulnerable in the context of a shortage of inputs for West Africa when the majority of the population has already been naturally immunised. Finally, population health is directly interested in translating science into action (on factors and determinants) and considers science and action to be intimately linked and mutually reinforcing (Diez Roux, 2016).

5Table 1 explains the different approaches and their intersections, showing their complexity in a context where there is still no consensus on this matter.

6Consequently, in the population health context, the concept of intervention takes a different form from conventional definitions. For example, the World Health Organization classification of health interventions defines health intervention as “an act performed for, with or on behalf of a person or a population, whose purpose is to assess, improve, maintain, promote or modify health, functioning or health conditions” (WHO, 2023). It expresses, in fact, a goal directly focused on health status, whereas population health calls for goals centred on more distal determinants, described as structural. Some describe intervention as a “set of means (physical, human, financial, symbolic) organised in a particular context, at a given time, to produce goods or services to modify a problematic situation” (Brousselle et al., 2018). Plante (1994) describes a programme as a “consistent, organised, and structured set of objectives, means, and people to drive it. It is justified based on needs defined as a deficiency or lack that affects individuals, a community, or a society, and it is under the control of one or more persons responsible for the quality of its formulation and functioning. It is set up to transform things or the state of something” (Plante, 1994). As with conventional definitions of public policy, both definitions posit a fundamental principle of having a problem to solve around which resources are marshalled, with Plante emphasising the notion of transformation. This is often a challenge in countries dependent on international public aid, where standard solutions are usually applied before discerning the details of the problems by involving the people concerned in the reflection. Nevertheless, they effectively dismiss a more salutogenic (i.e. not waiting until a health problem occurs) (Bruchon-Schweitzer & Boujut, 2014) and contextual conception of the intervention. Finally, Hawe and Potvin (2009) describe population health interventions as “policies or programmes that shift the distribution of health risk by addressing the underlying social, economic and environmental conditions”. For them, “these interventions might be programmes or policies designed and developed in the health sector, but they are more likely to be in sectors elsewhere, such as education, housing or employment”. They thus reflect a shift in objectives from the individual to the population, focusing on the reduction of health disparities within the population (rather than on the health of individuals) and stressing the need to study responses outside the healthcare system. As mentioned by Ridde and Guichard (2008), this displacement is not without its own debates around the best strategies for curbing these inequalities (intervening on the social determinants of health at the population level, targeting actions on the most vulnerable populations, combining strategies, etc.) and poses challenges for evaluation. In particular, the latter can no longer be approached in the traditional way of using methods derived from clinical evaluation.

7This notion of intervention then becomes very rich and increasingly complex to define. One solution is to define its attributes, which should be considered as marks of its differentiation from public health interventions. This marking is all the more important in that it can then be used to legitimise the methodological approaches required in research on these interventions.

8Thus, we propose the following attributes of a population health intervention:

Population-based vision

9The population health intervention integrates a population-based conception that looks at the collective and systemic nature of the health process in a population and does not consider the population’s health status to be the sum of the health statuses of the individuals who constitute it. This implies considering intervention goals that are sometimes very indirect (e.g. increasing autonomy, literacy, social ties) and taking into account the disparities that individuals face, whether those are related to personal resources, access to services, or the existence of environments that are (even very indirectly) supportive or detrimental to health.

Intersectoral and operational conception

10If its aim is to maintain, increase, and support equity in health and well-being within the population, the population health intervention can mobilise or leverage multiple strategies characterised by sectoral (i.e. education, employment, income, gender equity, land-use planning) and operational (i.e. access measures, regulations, built or natural environments, communication, education, etc.) diversity, as well as by clearly identifiable (traceable, recognisable) and concrete activities and resources.

Contextual grounding

11The population health intervention is based on, and enters into a relationship with, a context that will shape and transform it. An intervention deployed to accommodate the city of Marseille will not necessarily be accepted by the inhabitants of Bamako. In other words, the elements of this context, which could be related to stakeholder characteristics, including the population, their relationships, their environments, and their individual and collective histories, are part of an interventional system that must be considered when talking about population health interventions (Cambon et al., 2019). This implies, in fact, and as a prerequisite, that the population is one of the stakeholders, and even the main stakeholder, in this intervention, adopting the principles of community health, widely developed in Africa since the 1970s. Organisations, regardless of their nature (community, professional, etc.), are part of this context but do not in themselves constitute population health interventions.

Multiple actors

12Due to its nature and contextual grounding, the population health intervention mobilises a multitude of actors (including the population), who play many roles that influence both its conduct and its impact (such as international aid donors). They can observe, support, curb, or contribute – or even all of these at the same time – depending on the development of the intervention and its organisational context (see Chapter 5).

Objectives of improving health and health equity

13While a population health intervention is obviously aimed at maintaining and improving population health and health equity, these objectives may be focused more or less directly on the latter, depending on the determinants targeted by the strategies and interventions. Moreover, the interdependence of these determinants may, by its very nature, induce unexpected effects related to changes in their equilibrium, which must be taken into account.

Pluralistic nature

14In terms not only of its objective and means of action, but also its grounding and the influence of the actors being mobilised, the population health intervention takes on very variable levels of complexity (simple, complicated, complex) that must be considered and addressed (Brownson et al., 2017). This includes accepting a significant level of uncertainty about the relationships between what is being touched, observed, received, and/or undergone. This characteristic, once again, has implications in terms of the methods needed to understand these interventions and the effects they produce. The use of not only scientific data, but also experiential and local knowledge, to construct the content of the population health intervention is also a key dimension of its pluralistic nature. This plurality of approaches, methods, procedures, and contexts must be called upon during all the processes involved in developing the intervention, from its design, implementation, and evaluation (see Chapter 3) to the methods for using the knowledge produced by this evaluation (see Chapter 6).

15Thus, these attributes remind us that it is illusory to try to isolate an intervention from its context, and that it would be better, during a research process, to focus on the notion of an interventional system (Cambon et al., 2019). This notion of system is important because it considers not only the fact that the same cause does not always produce the same effects, but also that a system learns, adapts, transforms, and changes over time. Thus, this notion of an interventional system takes into account in particular 1) the relationships among the different interventional, historical, processual, and contextual elements, and 2) the notion of a cascade of effects, in particular by making explicit not only the mechanisms of the effects, but also the multiple effects that may be observed over the more or less long term. For example, an intervention may slow the spread of a disease (e.g. Covid-19 lockdowns) but have impacts on other health factors related to life contexts and further degrade the health that was intended to be protected (e.g. the effects of lockdowns on mental health, learning, domestic violence, delayed care, etc.). Some interventions may have an individual but socially differentiated benefit, thereby increasing inequalities within the same population and consequently generating other health problems in a fraction of the population. Thus, considering an intervention in isolation may result in shifting the burden off to the medium or longer term, or in a more or less visible way, depending on the outcomes intended.

Population Health Intervention Research

16With these attributes in mind, it is easier to understand the characteristics of the research that focuses on these interventions: PHIR. Some have proposed that it be defined as the “science of solutions” (Potvin et al., 2014), as opposed to the “science of problems” in relation to studies that characterise population health status and analyse its determinants. Rather, we will talk about the “science of the study of solutions”, to emphasise the fact that it is not a question of producing solutions considered unequivocal, but rather of analysing them from all angles, including their ability to disrupt the balance of the determinants mentioned above.

17This intervention research has certain characteristics that we wish to highlight here.

Positive or negative effects

18The first is that PHIR aims to understand how the different attributes of the intervention combine to produce positive or negative, expected or unexpected, direct or indirect effects on health and the distribution of health within populations and by what mechanisms these effects occur and persist over time. Indeed, as presented in Figure 1 by Turcotte-Tremblay et al. (2017), based on her thesis project conducted in Burkina Faso, four categories of elements can interact and influence the consequences of an intervention: 1) its intrinsic characteristics (degree of complexity, compatibility with needs, benefit, etc.); 2) the characteristics of members of the social system in which it is implemented (e.g. socio-economic status, health status, access to available resources, perceptions and attitudes, etc.; 3) the nature of that social system, including local norms and environment; and 4) the implementation of the intervention.

Figure 1. Framework for the Study of Unintended Consequences.

19This difficulty in understanding how these elements work together to produce an effect requires that we step back from the myths of experimental research (Hawe & Potvin, 2009) and, in particular, from the idea that the objective of the research is to analyse effects exclusively, that it de facto excludes the participation of communities and populations, or that it requires reliance on randomised controlled trials, regarded as the sole guarantors of the demonstration of causality (see Chapter 3). Also, and above all, the specific attributes of clinical research are incompatible with the attributes of population health intervention. For example, the experimental situation may impede the consideration of contextual grounding. Moreover, these specific attributes do not allow us to understand how the intervention works or why it achieves its objectives, or not. This understanding raises a plethora of questions about effects and mechanisms, as well as the conditions under which they arise. Our aim in this book is not to adopt a biased view, or to deny any usefulness to this latter type of evaluative approach, which has historically been put forward in public health and global health, particularly for clinical research, but rather to show that it is not the most appropriate in the context of complex interventions or interventional systems, as is the case with population health. This is particularly true in global health research, where contexts are so unstable and interventions so numerous due to the presence of a myriad of funders, that isolating effects, or even conducting such trials, is almost impossible.

The social utility

20The second characteristic of PHIR is its social utility. Indeed, if PHIR is a science of the study of solutions, then the ultimate goal is surely to transform the system by ensuring that it can adopt the solutions developed. PHIR should support and guide policy decisions, professional practices, and, where appropriate, changes in people’s behaviours. This mandate has direct implications for how research is conducted, since it should lead to concrete conclusions that can be directly used by the actors in this system (see Chapter 5), that is, conclusions that are acceptable, adaptable, viable, and sustainable (see Chapter 3). In this context, the concepts of transferability (i.e. the ability of an intervention to achieve the same results in another context) (Wang et al., 2006) and viability (i.e. the ability of an intervention to meet stakeholders’ needs) (Chen, 2010) must be considered when choosing methodologies for designing and analysing an intervention and when drawing conclusions. These methodological choices relate not only to options for data collection and analysis, but also to the manner in which the research is conducted and the stakeholders are involved in it (see Chapter 4).

The plurality of methods

21The third characteristic relates to the plurality of methods used in a PHIR (see Chapter 4), whereas randomised controlled trials often use single, quantitative methods. Indeed, in a PHIR, stakeholders and researchers ask a multitude of questions (see Chapter 2), each of which calls for specific methods (Chapters 2 and 3). The question should guide the method, not the other way around. Thus, if the objective is to study solutions (e.g. to fight malaria or reduce road accidents) and how they work, then multiple methods are needed, because some methods quantify while others seek to understand, and some establish correlations while others differentiate each element’s relative share in the production of a mechanism or an effect. Using these methods in combination is thus necessary to understand the interventional system under study. This methodological plurality requires a multidisciplinary approach to PHIR, without which the interventional phenomenon can only be viewed in a fragmentary manner, thereby producing erroneous conclusions.

The paradigmatic cross-fertilisation

22The fourth characteristic refers to the paradigmatic cross-fertilisation called for by PHIR. The methods chosen depend on the paradigms in which the researchers are embedded, i.e. their epistemological field (i.e. their worldview, the distance they maintain from their analyses, and the legitimacy they assign to those analyses to describe what they observe), their ontological field (i.e. their comprehension of the lived world through a single reality as opposed to several), their methodological field (i.e. the techniques they use to apprehend the lived world), and their teleological field (i.e. the purposes and benefits of PHIR) (Gendron, 2001). The multiplicity of questions inherent in the ambition to understand how it works rather than to observe what works calls for this paradigmatic cross-fertilisation, thus reinforcing again the necessary multidisciplinarity of this research. Here, in concrete terms, the point is to observe the intervention and its effects (expected or not) through a network of differentiated analyses and postures, rather than favouring one.

23Moreover, since PHIR consists of (often) identifiable and concrete activities aimed at populations, research that focuses, for example, on healthcare organisations (health services research), therapeutics (clin-ical research), or technologies (research on health technologies) is not actually included. This is because such research (even if it may also be interventional) is more concerned with patients than with populations and their subgroups and therefore constitutes another type of research, equally essential to science, but not considered in this book as part of PHIR. However, these types of research and their experimental approaches receive the major portion of health research funding, as is the case in France, for example. This shows all the challenges and needs for developing PHIR in the French-speaking world.

24Finally, a distinction must be made between PHIR and the evaluation of health interventions. For some, the two are identical under a common term – applied health research – on the grounds that they use the same methods and lead to the same conclusions (reinforced by the social utility dimension of PHIR) (Barker et al., 2016; Donaldson et al., 2015). Hawe and Potvin (2009) offer two arguments for distinguishing between evaluation and intervention research.

25The first is that PHIR encompasses multiple research questions that go beyond the outcomes and process questions to which evaluation is often confined. These questions aim to capture the full range of attrib-utes of population health interventions in the context in which they are embedded.

26The second is that the conclusions of PHIR are much broader than those of an intervention evaluation. They produce a body of knowledge whose scope is more extensive because it is less specific than that produced in an evaluation.

27However, the boundaries may appear thin and porous if we look mainly at the methods and analyses, provided the evaluation is conducted with rigour and is not vague on certain aspects, given the social utility dimension of PHIR. In fact, PHIR can: 1) result in scientific publications for the international community (e.g. contribute to debates on the Sustainable Development Goals) and an internal report addressing key questions raised by decision-makers (e.g. the evaluation of an intervention); 2) pursue an objective of producing knowledge that also addresses societal questions and issues (e.g. evaluation); 3) lead to research perspectives and recommendations for action (e.g. evaluation); and 4) be funded by calls for research projects, international global health organisations (Unitaid, Global Fund, Echo, WHO, etc.), and grants from health operators or non-governmental organisations (e.g. evaluation) as part of an evidence-based policy. Consequently, the distinction may lie more in the fundamentals of the science guiding the scientific posture and approach of a PHIR, and particularly the fact that it:

• produces only original knowledge that contributes to advancing science and the state of knowledge in the field;

• builds on a preliminary analysis of this state of the art, and in particular on a (systematic) review of the scientific literature;

• grounds its hypotheses and analyses in theoretical bases and proven conceptual frameworks;

• remains objective and transparent in the choice of methods (each one justified with regard to its scientific strengths and weaknesses);

• is rigorous in their application and their analysis;

• and remains neutral in the interpretation and presentation of results (results are not truncated, transformed, or hidden).

28Of course, these fundamentals also apply to scientific approaches in public health other than PHIR.

The different forms of PHIR

29Given the complexity of the interventions studied and the multiplicity of questions that can be asked, a PHIR implementation faces many challenges. Thus, researchers interested in this field have contributed to developing the types of evaluative approaches by comparing, combining, hybridising, and sometimes even “tinkering” with paradigms and methods.

30Thus, several types of research exist. The aim here is not to be exhaustive but to present the best known and most widely used, as well as their specific features, in order to understand this field of research not only in its contradictions, but also its complementarities, and even its overlaps.

Controlled trial

31The controlled trial is still considered by many health research teams and funders as the best research approach (design, specifications) to identify a causal relationship between an intervention and an effect, all else being equal. For example, the French Development Agency (AFD) has just launched the Fund for Intervention in Development for the poorest countries, where it strongly recommends the use of these cluster randomised controlled trials.

32The focus is on evaluating the effectiveness of the intervention by attempting to standardise both the intervention modalities and the effect that the context (including population characteristics) might have on the outcomes. The context is considered as both a variable and a bias that must be eliminated to produce generalisable conclusions of causality. Similarly, the elements of the intervention and the way the population is exposed to it are under the control of the research team to compare the outcomes with a population that is not exposed. The design with the highest internal validity according to proponents of this approach is the individually randomised controlled trial (Campbell & Stanley, 1966).

33Because this study design is modelled on the clinical field (e.g. therapeutic drug trials), it has many ethical or methodological limitations in population health (Tarquinio et al., 2015). Thus, adaptations of this research design have been created to consider its limitations (see Chapter 3). Our objective in this book is not to discredit its use, but rather to explain its poor suitability for population health interventions, whether in Europe, Africa, or elsewhere.

34Despite these adjustments, the very premises upon which this type of design is based, and in particular that of developing universal laws while deliberately ignoring contextual elements and the heterogeneity of effects, raise questions when, as in the case of PHIR, we need to consider issues of viability and transferability. In other words, what is the point of producing conclusions on an intervention conducted under conditions that will never reoccur or that would occur very differently? The causal inference is indeed validated, but because the conditions under which this intervention operates are neither studied nor taken into account in the conclusions, it is unlikely that, when it is generalised, the same results will be observed. Thus, while the intellectual exercise may make sense, its value in PHIR is low, as it cannot guide decisions and practices. One response to this significant limitation is to back up the experimental studies with an evaluation of the process (including mechanisms) to understand how this intervention works (see Chapter 3).

Evaluation research

35Evaluation research aims to produce knowledge about a specific intervention in order to inform a decision. As such, there is a very close link between the object of the research and the decision to be taken (Clarke, 1999; Collins et al., 2004; Patton, 1990). This research is often the result of a request from outside the research team by health managers or actors implementing field interventions. It aims to mobi-lise scientific methods and tools to study a decision, a transformation, or a public health practice.

36The research can therefore have multiple objects: an intervention’s effectiveness, its efficiency or cost effectiveness compared to others, the feasibility of its implementation, its acceptability, the conditions for its sustainability, etc. (Clarke, 1999; Collins et al., 2004; Patton, 1990). Consequently, this research can be carried out at different points in the intervention process.

37It can thus be useful for improving a programme or a system, in which case it is focused on the design of the intervention. This form of research is usually carried out during the action to improve alignment with the intended goals. For example, this could involve supplementing the development of a programme or service with an analysis of how it operates in real life (Collins et al., 2004). It engages people who are not involved in the design and implementation of the intervention, but it can also mobilise actors who take an active part in the action. It can also be used to validate an intervention or compare it to another, in which case it is oriented towards the decision to continue or stop the intervention. It is conducted at the end of the intervention and is usually entrusted to an external person or organisation that is not directly involved in developing the action or system. It is research aimed at informing a decision that is contingent on policy issues, such as stopping, continuing to fund, or reorienting an intervention.

38The limitations of this type of evaluation lie in the difficult balance between the posture of researchers, whose natural curiosity and need for knowledge can take them beyond the questions posed by the decision-maker, and the problematic convergence of research and decision-making agendas, which has been widely studied in research on stakeholder – researcher partnerships (Baker et al., 2004; Bryant, 2002; Dagenais et al., 2009).

Action research

39Action research is defined as an iterative process of collaboration between researchers, practitioners, and the population working together in a series of activities that include problem identification, planning, implementation, and evaluation of solutions, as well as reflection on the process (Avison et al., 1999). Its philosophical foundations owe much to Latin American and African thinkers such as Freire in Brazil, Fals-Borda in Colombia, or Ki-Zerbo in Burkina Faso.

40This type of research is rooted in a pragmatic paradigm in which research aims not only at explanation, but also at social change through a cyclical theory/practice interaction (whereas traditional research is often more linear); the theory supports or emerges from the action and is used to understand and act on the real problems encountered on the ground (Hart & Bond, 1995; Sylvestre et al., 2019).

41This process relies on the strong participation of the people concerned because it requires consensus among all stakeholders (including the population and researchers) on the objectives and the means. It is focused primarily on the actors’ concerns, helping them to transform their practices through the interweaving of research and action (Hagger et al., 2020). Baum speaks of participatory action research to emphasise this characteristic of mobilisation (Baum et al., 2006), as can be seen in much research in India or Africa, for example.

42It is thus a particularly valuable type of research in health promotion, where community participation is an objective in itself, beyond intervention (Whitehead et al., 2003). In this context, the researcher is an agent of change, in the same way as the other stakeholders (Reason & Bradbury, 2008).

Developmental evaluation

43Developmental evaluation focuses on the use of results by stakeholders (Patton, 2021). Its objective is to make the evaluation relevant and directly useful to the main users by promoting effective use of results. In this way, evaluators can support programme improvement, which is one of their main activities at the heart of the evaluation process.

44This type of evaluation therefore involves active collaboration between the evaluator and the main expected users of the evaluation and, above all, the parties involved in organising the activities. It helps prepare the change linked to using the results, and supports decision-making and the development of interventions (Gamble, 2008; Patton & LaBossière, 2012). As with action research, the underlying theory (which makes it a type of research, even if it is called evaluation) is modified to allow for the emergence of new evaluative knowledge and to support adaptation to changes introduced by the intervention. It therefore takes into account the elements of the context and reports on them over the course of the development. Again, the process is not linear but dynamic. Users react continuously to the data, adjust the intervention with the support of the researchers, and engage in reflexive work on their lived experience (Fagen et al., 2011).

45Thus, this type of evaluation is suitable for interventions implemented in complex environments, such as population health interventions. These contextual elements and the needs of users are constantly feeding into the reflections of researchers (and their team). In this type of evaluation, therefore, the cross-fertilisation of methods and viewpoints is a sine qua non condition, since the aim is to understand the dynamics linked to the context and the complexity of the intervention within that context, and above all, to devise, in an iterative process, innovative strategies to support the development of the intervention (Dozois et al., 2010; Patton, 2010).

Embedded research

46Embedded research, like action research and developmental evaluation, is aimed at facilitating the integration of evidence into practice (McGinity & Salokangas, 2014). It has recently been strongly promoted by the World Health Organization for health systems research in Latin America or the Middle East. In this type of research, researchers work within health management organisations to identify, design, and conduct studies and share results that address the needs of the organisation’s agents and are in line with their professional objectives and contexts. This research is therefore directly linked to the organisation’s mandate (Marshall et al., 2014; McGinity & Salokangas, 2014). Through this insertion into the organisation, the researcher is able to share in the worldview of the organisation and its partners and interact closely with the users of the research results, since they are involved in the reflexive process and have rapid access to data emerging from the research.

47It is a particularly effective knowledge transfer strategy, as it promotes co-production of evidence (Armstrong et al., 2013; Buffett et al., 2007; Gervais et al., 2013; Souffez & Laurendeau, 2011). The researcher is nevertheless obliged to back up this collaboration, and in particular this vision of the world, with an academic theoretical reflexivity. In this respect, the researcher differs from knowledge brokers (see Chapter 6), who promote networking and knowledge transfer within the organisation and help strengthen the capacity to use it (Burnett et al., 2002; CHSRF, 2003). Even though, like knowledge brokers, they are on the border between two worlds (Lewis & Russell, 2011), “embedded” researchers focus on the (co)production of knowledge and are not responsible for its actual use. To this end, researchers and host organisation staff work together to co-create, refine, implement, and assess the impact of new and existing knowledge in relation to the context (Langlois et al., 2017). This approach differs from action research in that the researchers have a somewhat larger role in the process and the end goal is not necessarily social change and empowerment.

48Here, the challenge for the researcher is to maintain a reflexivity that is independent of the context and of the political agenda of the organisation implementing the population health intervention, even though these must be taken into account, in the sense that the research programme is shared with the organisation in a mutually advantageous relationship.

Natural experiments

49Natural experiments are studies in which the intervention is neither decided nor organised by the researchers (Craig, Cooper et al., 2012). Examples of this are a study of the effect on population health in France of a decision to ban smoking in public places and another of a ban on travel between regions in Senegal during the Covid-19 pandemic. Natural experiments are of particular interest because they broaden the range of interventions that can be usefully evaluated when it is neither ethical nor possible to implement the intervention specifically for research (Benmarhnia & Fuller, 2019).

50Such studies should follow good practice in the conduct of observational studies, such as the prior specification of hypotheses, clear definitions of target populations, explicit sampling criteria, and valid and reliable exposure and outcome measures (West, 2009). They require comparison of exposed and non-exposed groups (or groups with varying degrees of exposure) to identify the effect of the intervention (Meyer, 1995). When studying the intervention’s effectiveness, the challenges of causality can be addressed by statistical methods of accounting for confounding factors.

51The absence of any manipulation of the intervention inherently introduces biases that limit the study’s internal validity and, consequently, the strength of the evidence for causal inference, such as would be observed in experimental studies. For this reason, these studies are conducted when the intervention can reasonably be expected to have a significant impact on population health but there is still scientific uncertainty as to the extent or nature of the effects and whether the intervention or the underlying principles can be replicated, extended, or generalised (Craig, Cooper et al., 2012).

52Thus, Petticrew et al. (2005) consider this type of research particularly appropriate to study interventions on the structural determinants of health (e.g. employment, housing, pricing of noxious products, etc.) at the heart of population health.

Conclusion

53The most notable difference between these types of studies, which we clearly have not inventoried exhaustively and which are the subject of many other writings, is that experimental studies are aimed particularly at producing universal and immutable laws, whereas the others, contextually anchored, take a more pragmatic view of what they observe in order to support operational decision-making. In this respect, their ambitions overlap with those of two other streams of research: 1) implementation science, which aims to understand how interventions produce their effects by uncovering the factors associated with effective implementation and the intervention’s capacity to adapt to actors and contexts; and 2) implementation research, which aims to produce knowledge on how interventions take into account and integrate evidence into the formulation of their content to be more effective (Ridde & Turcotte-Tremblay, 2019).

54Clearly, these different types of research have certain characteristics that either overlap or diverge, such that they cannot be unambiguously differentiated, as these differences can be quite subtle. To our knowledge, there is no international consensus in this respect. For example, action research, like developmental evaluation and embedded research, is based on very strong collaboration among the actors involved (researchers, population, users), but it considers the transformation of practices and situations to be an objective in itself (it is the participation that produces change, more than the results produced), whereas, in the other two, transformation comes from the use of the results, whose genesis is determined by the actors’ agendas. Some study designs are more suited to evaluating “simple” or even complicated interventions, while others allow for the analysis of complexity, related to the types of intervention and their more or less direct effects, to the malleability of the intervention components within the context, to the impact of this context on those components, as well as to the roles of the various actors and the population in the research process, which can certainly influence the researcher’s reflexivity.

55Table 2 presents the main characteristics of each type of research design according to six parameters of a PHIR: 1) the roles of the actors in the research (including the population); 2) the objective; 3) the purpose; 4) the role of the context; 5) the type of validity emphasised; and 6) the data collection methods (see Table 1). However, this list is arbitrary because in real life the issues, like the methods, may be hybridised. For example, developmental and embedded evaluations ultimately differ only in the researcher’s position, external in one case and agent of the organisation in the other, which in the first case will produce conclusions specific to an intervention, and in the second, conclusions about an intervention within a specific organisational vision. Civil society organisations and non-governmental organisations (NGO) working in the field of development aid often favour these latter two approaches. Along the same lines, evaluative research does not preclude active stakeholder participation, which could, over the course of the process, foster reflexivity and a change in practices, as is observed in action research. International global health organisations often set up evaluation monitoring committees but prefer expert approaches in which the participation of those concerned is reduced. Similarly, in each of the designs, except for the experimental study, stakeholders' involvement and the production of concrete recommendations for decision-makers can be introduced.

Table 2. Key characteristics of types of population health research.

56In the PHIR context, each type of study is relevant, insofar as its development and/or hybridisation can be conceptualised in terms of the following characteristics:

• Convergence between the originality of the knowledge produced and its social utility: This involves combining, in the objectives, the aim of producing original knowledge and the requisite pragmatism to be applied in producing, and even supporting, the recommendations emerging from this research (see Chapter 6).

• The consideration of context as a determinant of results: This means rejecting methods that would exclude from the conclusions the influence of contextual components and considering transferability and sustainability as matters of necessity (see Chapter 3).

• The participatory dimension of research: This involves recognising the multiplicity of actors involved in the research and the value of their role in studying the complexity of the intervention, rejecting the illusion of personal neutrality and objectivity (i.e. the researcher being external and outside of the action) in favour of collective neutrality and objectivity (i.e. the convergence of different views and subjective perceptions) in the interpretation of the phenomenon under study (see Chapter 5).

• The need to address a wide range of interdependent issues and consequently multidisciplinarity: This means shifting the object of the research away from the effectiveness of interventions alone, as would be done in clinical research or controlled trial methods, and recognising that this notion of effectiveness only makes sense when coupled with other equally important questions (how, under what conditions, with/by whom, to what extent, etc.). This implies the need to develop this research in a multidisciplinary manner to bring together different views and interpretations of the same phenomenon (see Chapter 2).

• Acceptance of all methods: To study the entire interventional system, methodological hybridisation is necessary, provided that the methods used are justified and rigorously implemented (see Chapters 3 and 4).

57Thus, PHIR can only be conducted and be meaningful if it can be done without dogmatism (which often arises in debates about methods), and in a multidisciplinary and pragmatic manner, while reconciling the particular characteristics of population health interventions with the specific requirements of science. As such, Figure 2 presents the particular characteristics of PHIR in terms of the equilibrium created in combining these two elements (see Figure 2).

Figure 2. Population Health Intervention Research.

58Following this introductory text, the following chapters describe possible research questions (Chapter 2), the approaches (Chapter 3), the methods (Chapter 4), the types of actors involved and their roles (Chapter 5), and ways of using PHIR results (Chapter 6).