Developing the ethics of implementation research in health

Problems addressed by IR must be of high local priority in order to justify the research [16]. One could argue that the requirement for responsiveness is greater from IR compared with clinical research as IR should address priority health needs whereas clinical research is concerned more with proof of principle, and wider application should be tested with subsequent IR. Conversely, when there is an available intervention for an important unmet health need in a community, there is an ethical imperative to conduct IR to try to address the implementation barriers. As such, IR studies are often commissioned by local health authorities. The knowledge of which problems are indeed local priorities relies on epidemiology and health data reporting, which may not always be optimal in resource-challenged environments. Engagement with local health experts and communities is therefore essential in the planning stages of IR to determine whether a health problem is indeed perceived to be a local priority. If a particular problem is not perceived to be a local health priority, the ethics of conducting IR becomes questionable [17].

Equipoise is an important ethical imperative in the conduct of research, it is required to justify any potential risk to research subjects. Clinical equipoise refers to the fact that investigators conducting a randomized controlled trial do not know in advance if an intervention is better than what it is being compared with. In IR, however, such clinical equipoise is generally not present (e.g., a medication is known to cure malaria), but instead, situational or contextual equipoise justifies the conduct of IR, i.e., there remains genuine doubt whether a new and untested package of interventions will work in a specific context [18]. To ethically justify IR, therefore, equipoise regarding the effectiveness of the implementation processes must be preserved.

A balanced discussion about study design is important before embarking on an IR study to weigh the ethical obligation to ensure scientific standards are maintained against the ethical demands of equity and justice [2, 19]. Though the randomized controlled trial is considered the gold standard in clinical research, health system strengthening interventions often do not lend themselves to such a design. Many different study designs, often with multidisciplinary involvement, have been used in IR, each raising particular ethical concerns [17]. Both qualitative and quantitative methods are used in IR and often within the same study. Examples of specific and different ethical considerations may arise with each method as outlined in Table 3. CRTs are often used in health systems research, but when a public health intervention is known to be effective, withholding the intervention from those randomized to the control arm is ethically problematic [4, 20, 21]. As a compromise, a stepped-wedge approach is sometimes justified to address this dilemma in CRTs, as this may mirror the real-world scale-up process [22]. In a stepped-wedge design, an intervention is delivered sequentially to groups of participants with the goal of ultimately including all participants, however in the early stages some groups do not receive the intervention and therefore are analogous to controls, which may pose an ethical challenge even if short-term because for that short period of time the participants are deprived of the proven intervention [23]. Alternatively, quasi-experimental designs where a control group may not be included may be ethically more acceptable in IR, but the scientific rigor and validity may be questioned [24]. The use of random allocation, without consideration of the specific needs and vulnerabilities of the participants, raises concerns of justice and equity. Other appropriate study designs for IR include pragmatic designs, hybrid and mixed methods designs, and open-label demonstration projects which may each have specific ethical issues that will require careful consideration at the planning stage [25‐27]. Engagement with all stakeholders is crucial to develop the most effective and fair study design.

The term stakeholder has numerous definitions, many of which are contextual [28]. Two general definitions are relevant in the context of IR: the first defines a stakeholder as a “person or group with an interest, involvement or investment in something” [29]; the second describes stakeholders as “people who will be affected by a project, or who can influence it, but who are not directly involved in doing the work” [30]. Key stakeholders in IR may include the government, policy-makers, public health functionaries, health care providers, health care managers, financing mechanisms, health care industry, and the community. Communities may include individuals on who interventions are planned, the broader community or social structures to which these individuals belong and the broader society to whom an intervention may eventually be rolled out. Communities and individuals with specific roles are important stakeholders in the research process. Meaningful engagement with stakeholders at all levels is crucial in IR, as a means to identify health priorities, to identify key participants, to communicate transparently and effectively about the goals, design, risks, benefits, and process of a proposed intervention, to gain trust and develop partnerships to enhance success of the study, and to gain feedback and identify unforeseen barriers that could be mitigated at the planning stage [31, 32]. Community engagement is a related but different concept where the members of the community who will benefit from or face the risk of the IR are actively consulted and engaged with, with the goal that they play an active partnership role throughout the IR process.

Stakeholder and community engagement are cross-cutting processes which must be carried out during the planning, implementation, and post-research phases of IR. Particular ethical underpinnings of engagement with policy-makers and health financers at the planning stage include determination that the intervention will address a local priority health need and to gain buy-in and commitment for the scale-up and sustainability of an effective intervention. Scale-up is considered by some to be an ethical corollary of IR. Important ethical goals for engagement with communities as partners in planning and design of IR include to determine acceptability of a proposed intervention, maximize uptake, ensure inclusion of vulnerable populations, establish accountability processes, and particularly when individual informed consent is not feasible due to the research design, to ensure individuals are aware of the rationale and opt-out possibilities. The process of decision-making in IR should consider the power differential between the researchers and the community and allow adequate representation of the research participants and the community at large [33‐35].

Often community representatives are selected to facilitate communication between researchers and the community. The selection of community representatives must be an inclusive and fair process of democratic election or nomination, guided by the community itself, to ensure appropriate, acceptable, and comprehensive representation of all sectors of the community irrespective of class, race, gender, sexual orientation, or ethnicity and to avoid any potential conflicts of interest-specific individuals may have [33‐36]. Engagement with disadvantaged and marginalized groups is imperative up-front and throughout the IR process to ensure acceptability and equitable participation in IR, and importantly, to identify any specific unanticipated barriers they may face and to develop strategies to mitigate further marginalization or stigmatization through the research [8, 37].

The clinical efficacy and safety of an intervention is generally known before IR is conducted. In clinical research, an individual participant can personally weigh the relative risks and benefits before giving their informed consent, and the risks and benefits are usually borne by the same individual. For example, in clinical research testing, the efficacy of a new vaccine, the benefit of personal protection, and the risk of side-effects are borne by the individuals who participate in the study. In IR, for example in mass drug administration interventions, the community may benefit from large-scale treatment of individuals, but an individual may experience side-effects from a medication they may not personally have required. In addition, the potential risks of an IR intervention may also result from the modality of implementation [2]. For example, a community wide public health screening campaign for sexually transmitted infections which had been successful in one low-income country may carry different risks of stigmatization, religious ostracism, and social discrimination if implemented in an underdeveloped and religiously orthodox country, leading to a different risk-benefit balance. The risks associated with IR may not always be obvious up-front as health systems are complex adaptive systems, and interactions between the components in the health system are not often clearly understood [38]. Diligent situational analysis must therefore be conducted during the planning phase of IR to identify potential risks before harm is done [19]. In addition, a particular feature of IR is that at times an intervention is implemented in one group, but the benefit may accrue to another group [2]. For example, IR studying the implementation and uptake prior to scale-up of a malaria transmission-blocking vaccine exposes vaccinated individuals to the risks of vaccination, but unless a large proportion of the community is vaccinated, the individuals vaccinated will protect others from malaria transmission, but will not be protected themselves. How to balance the risks experienced by one group against the benefits gained by another requires ethical deliberation and effective communication with the research participants. The ethical deliberation should be based on the solidarity principle and should be transparent, involving communities and all stakeholders [39]. To what extent individuals within a group should be exposed to risks for the benefit of others cannot be clearly defined, but it should be decided based on community and stakeholder consultations [2]. A line which cannot be crossed is knowingly exposing one group to harm or significant risk for the benefit of another.

A key principle driving the ethics of clinical research in humans is individual autonomy. In public health research autonomy has two dimensions, one concerns individual autonomy and the other concerns relational autonomy in the context of the community to which the individual and the health system belong [40]. Informed consent is the process through which a research participant can exercise their autonomy. In clinical research, a fully informed individual can determine whether or not they wish to freely participate in a study and can usually opt-out of the research at any stage. In IR, there may be difficulties in operationalizing informed consent [2, 20, 41, 42]. For example, an individual in a cluster in a CRT may not have the chance to decide and give consent to randomization as randomization happens at the cluster level. In the case of non-excludable cluster level, interventions such as environmental modifications, an individual may not be able to exercise a meaningful refusal to participate. In such situations in IR therefore there is a need to articulate informed consent differently from traditional individual consent in a clinical trial. At one end of the spectrum is a complete waiver of individual informed consent where the ethical risks are minimal, and the interventions are largely at a cluster level (i.e., no individual can opt-out), rendering refusal meaningless. For example, in an IR study of an ultraviolet wave system to provide safe drinking water to a population, the harms are considered minimal, and it is not possible for any individual participant in a cluster randomized to the intervention arm to easily opt-out of the study [43]. Such a waiver of consent does not however preclude the need for meaningful community engagement and provision of information. At the other end of the spectrum is the example where individuals have the opportunity to refuse participation in the research project, even though the intervention will occur at the cluster level. This could occur for example when public health professionals wish to test whether community health workers can be trained to provide injections in the community, and individuals have the option to refuse participation and visit the health facility to receive the injection from a nurse. Individual informed consent would be the norm in this case, even though consent for randomization cannot be provided by the individual and it is operationalized at the cluster level prior to individual contact [20, 43]. In the middle of the spectrum is dual consent from gatekeepers and individuals. Community agreement relies on the identification of an appropriate gatekeeper, who should have a keen interest in the welfare of the community and represents the community in a fair manner [36]. There are several challenges in selection of the gatekeeper. In traditional communities where collective decision-making is practiced, selection of a gatekeeper may not be problematic. But in more complex societies, or more complex studies, selection of one voice to represent the community is often challenging. The community leader may not be the most appropriate person to make decisions on whether a community or its members should participate in a study or not. For example, an elderly male village leader may not be an appropriate gatekeeper to consent to an IR intervention on pregnant women in his community. Selecting community representatives fairly requires inclusion of a variety of representative stakeholders, especially those from the target groups, and ensuring transparency of the process [20, 36]. The agreement of the gatekeeper, however, cannot replace individual consent or assent where relevant as discussed above [4]. Ultimately, it is important that proper ethical oversight is in place through Institutional Ethics Committees to ensure that the appropriate informed consent process is followed, maximally respecting autonomy of individuals in the study [44].

Challenges in operationalizing informed consent in the context of IR also include whether the beneficiaries are individuals or populations, and appropriate identification of who the actual research participants are [18]. For example, when implementing a taxi voucher system to increase the rates of institutional deliveries and reduce maternal mortality, should consent be obtained from the pregnant women, the taxi drivers, or the health care workers whose performance will also be evaluated? The Ottawa statement on ethical conduct of cluster randomized controlled trials define a research participant as: the intended recipients of the experimental or control intervention; the direct targets of experimental or control environmental alterations; persons with whom researchers interact to collect their data; persons whose identifiable private information is accessible to the researcher for collection of their data [21]. As such, in the example above, patients and health care workers should provide consent, but whether this should extend to the taxi drivers is questionable and may be difficult to operationalize. A further important ethical issue in the informed consent process is the extent of information to be revealed to the participants in the intervention and control arms, where applicable. In IR, especially when behaviour change interventions are being studied, knowledge of the intervention itself may change the outcomes and implementation process. There is, therefore, often the need to conceal some information about the intervention. The ethical justification for this is debatable, and it must be balanced against the risks/benefits and the potential impact on study validity as discussed in the CIOMS guidelines [45]. The informed consent process in IR therefore may be quite different from that in clinical research and requires thorough consideration to ensure optimal ethical conduct of IR.

Particular issues relating to privacy and confidentiality in IR relate to the fact that IR often requires that facility level data on patient outcomes be available or that individual level data from facility health records be obtained. For example, if a public health intervention is implemented to regulate institutional deliveries and improve the quality of skilled institutional deliveries, there may be interventions at the health facility level, but confidentiality restrictions on access to data from women who deliver in the facilities may hamper effectiveness analyses of the health system impact [19]. In such cases, the data that is obtained from facilities should either be anonymized, or the individuals about whom data is being sought should provide consent for their data to be reviewed by the researchers. Where such consent is not possible, it is the responsibility of the researcher to obtain a waiver of consent from the respective ethics board, and put in place mechanisms to ensure that the confidentiality of the patient information is respected. A proactive strategy of informing patients about potential data collection for research and quality improvement purposes up-front, but reassuring them about privacy and confidentiality could also serve strengthen the patient-researcher partnership and build trust [46].

There are two approaches to decide on standard of care or prevention to be given to a control group [47‐49]. One approach is to allocate the local de facto existing standard, which in some situations may be grossly insufficient, making it ethically unacceptable based on justice and fairness principles. For example, in an IR trying to study prevention of mother to child transmission of HIV in a country where routine anti retroviral therapy is not available, the local de facto standard of care is no treatment. Having a placebo control arm in the study is not acceptable in spite of the local de facto care being no treatment, because an effective treatment which reduces transmission is available and should be accessible to the mothers. The second approach is to provide the local de jure standard of care or prevention, which is agreed upon by public health experts of that region and is acceptable to the community. This approach may still be unfair in that this standard may be unsustainable for the local health system after the IR is completed. For example, in a public health behaviour change implementation study focused on hand washing among schoolchildren, the intervention group receives a school-based lunchtime hand washing program, and the control group receives soap and water in all schools, but without any emphasis on hand washing before eating. In this case, the standard of care is provision of hygiene tools, and the intervention is emphasizing the use of these tools. In this context, allowing the control group to have no intervention can be considered ethical. The consideration of standards of care or prevention may therefore identify new gaps as targets for future IR.

Ancillary care refers to the identification of problems that may contribute to ill-health that are beyond the scope of the study in question, for example, researchers studying home management of malaria may come across household members with other diseases needing attention [11, 50, 51]. Sometimes ancillary care responsibilities can be foreseen at the design stage and at others they are encountered only during the conduct of the IR. Ancillary care obligations are present when the need is serious in terms of severity or urgency or both and when there is a possibility of provision of care within the scope of the research [11]. For example, in the school-based hand washing behaviour change IR study, uninterrupted tap water supply may be lacking and this is an ancillary care requirement. However, this example illustrates that it may not be realistic to expect implementation researchers to assume all ancillary care responsibilities. Researchers may not have the expertise to provide the ancillary care; the provision of the care may be costly or may require system-level interventions. The researchers must, however, establish process of accountability for ancillary care need identified through the research, determine which needs may realistically fall within the scope of responsibility of the researchers, and proactively engage with the local government or non-governmental organizations during the planning and conduct phases of IR to identify who will be able to meet other needs [52].

Well conducted IR should lead to strengthening of research capacity of the local institutions as well as individuals’ capacity to conduct research in settings where such capacity is weak [53]. Research capacity strengthening can range from creating a trained workforce of researchers and research volunteers up to training and capacity building of research experts and infrastructure to permit independent conduct of locally responsive IR in the future. Based on the need in the area where the IR is being conducted, appropriate research capacity strengthening should be facilitated. This can be facilitated by appropriate stakeholder engagement ensuring commitment by donors and governments to build sustainable research capacity. Not only is it important for the IR to strengthen local research capacity, it should also strengthen the health system within which it is conducted. For example, true partnership in an implementation research study of rapid diagnosis of tuberculosis resistance should build sustainable infrastructure in technology required, expertise to run and maintain the technology, strengthen the local health information system to track data acquired, train local researchers in design, conduct, analysis and reporting of study findings as well as participation in post-intervention scale-up, thereby strengthening the local research and health system capacity. IR projects focusing on specific health gaps may, however, create vertical program structures within the health system which may be disempowering to the system through inefficient resource utilization [13]. It is ethically important that the conduct of IR should focus on horizontal integration of public health interventions into the health system such that a project empowering any component of the health system may have positive repercussions for the entire system. Strengthening the capacity to translate research findings into health policy is a specific imperative in IR and must be a component of all phases of the IR process [53].

Given the important public health impact of IR, there is an ethical obligation to disseminate the research findings widely, including feeding back to the communities and stakeholders who participated in the research [54, 55]. If an implementation strategy had a negative or positive impact in a certain context, either finding may be important for researchers planning similar interventions elsewhere. Therefore, irrespective of the results, the findings of the IR should be disseminated. Furthermore, resource utilization globally could be enhanced by an imperative for dissemination of IR findings, as once an intervention has been tested in many different local contexts, its findings may be presumed to be generalizable and obviate the need for new IR studies and delays in scaling-up of the intervention in new contexts.

Due to the inherent nature of IR, there is an ethical obligation for IR findings to be used to inform effective and equitable public health action. This necessitates timely consideration and uptake as relevant of the IR findings into public health policy and practice. Potential barriers to translating knowledge into action include lack of prior consultation with policy-makers, lack of funding, weak health systems, poor communication of findings by researchers to policy-makers, and absence of a culture of evidence-based decision-making among others [53]. Therefore, in order to translate the research into public health action, implementation researchers should engage with policy-makers and health system officials, important stakeholders in IR, upfront to ensure commitment to sustainability should the intervention prove successful, and must communicate their research findings rapidly, clearly, and concisely to engage and inform policy-makers in a timely fashion. Researchers should also propose actionable suggestions based on the research findings to facilitate uptake and scale-up of successful interventions. Barriers identified during IR may require further study to develop strategies to overcome them. Effective communication between researchers and policy-makers, as well as education of the public are important social justice obligations in IR, ensuring that expectations raised during the research are met, and those who participated as control subjects gain access to interventions withheld from them during the study.

Scalability and sustainability are important ethical considerations at both planning and post-study phases, as ultimately these are the goals of IR [13, 56]. The duty to ensure sustainability post-study cannot only lie with the researchers. Multiple stakeholders must come together to promote this goal which requires ongoing stakeholder engagement throughout the IR process. The researchers should ensure through effective engagement during the planning and conduct stages of IR that the non-research stakeholders such as policy-makers, local providers, and health system officials remain committed to sustaining implementation of an intervention if found to be effective. If access to a proposed public health intervention cannot be ensured for a community after the IR, it may not be ethical to carry out such a research activity. If specific interventions are provided during IR without a plan for sustainability, this could lead to exacerbation of inequity and harmful effects to the community as well as loss of trust in the health system.

Irrespective of the context in which IR is conducted, LMIC or developed countries, there is an ethical obligation to share benefits of the IR with the community [57]. There are various classifications of the benefits that can be achieved as a result of conduct of the IR. The benefits may be direct as a result of the intervention being studied, or indirect and not related to the intervention per se. The benefits may accrue to individual participants or to the community at large. For example, IR may be conducted in communities where the local health system is weak, therefore success of an intervention may result in introduction of a new intervention that was effective in the local context, providing individual benefit. In addition, the IR likely identified and overcame barriers which would have contributed to some strengthening of the local health system that would have a broader impact. IR researchers can facilitate sharing of direct benefits by advocating for sustainable translation of research findings into action, and sharing of indirect benefits through building research capacity and health system strengthening. The unique nature of IR where the individuals who bear the risks of the IR are not always the ones who enjoy the benefits is a challenge as discussed above. Optimal benefit sharing can be promoted through proper pre-IR planning and ethical conduct. Community and stakeholder engagement plays an important role in achieving benefit sharing as when adequately informed they can advocate for access to proven benefits. Benefit sharing has important social justice implications, and it is the obligation of the researcher to achieve a balance of risks and benefits to both individuals and communities [58, 59].