This study identified characteristics of E. coli AMR interventions that affect their implementation in social-ecological AMR systems. We used the AMR-Intervene framework to describe the interventions in their diversity regarding the social-ecological factors of complex systems, as well as reporting quality in scientific publications. The study of 42 unique E. coli AMR interventions highlighted a lack of diversity of interventions and their action profile, a lack of reporting regarding implementation strategy, and a need in new, adaptive, flexible and complex interventions to tackle AMR.
Lack of diversity of interventions
The interventions coded in the database had limited diversity, mainly in their social system and governance. Table 1 highlights the homogeneity within the description of the intervention such as spatial homogeneity, small diversity of actors, targeted groups and settings, and similar funding sources. All interventions also targeted the individuals (human or animals) and, except for environmental studies, did not target the cell, gene, or ecosystem level. Also, the sector targeted by the intervention was primarily either human or animal health, and One Health interventions remain rare or not reported. Goals and challenges of collective actions are similar and remain in the same categories; typically, one goal and challenge aiming at reducing the state of AMR through surveillance. The homogeneity of the results of the current research limits our capacity to find successful interventions. As diversity of action is a key driver of resilience in social-ecological system, this might be an obstacle to build societal resilience to AMR.
Moreover, the use of system thinking has highlighted the need for new forms of co-evolutionary governance to tackle AMR (25,26). Numerous studies acknowledge the social-ecological context of AMR, its diversity, its co-dependence and co-evolution with our societies, practices, policies, and technologies (11,19,27). In the same way that we increasingly rely on a specific AM to treat an infectious disease, we are using the same type of interventions to respond to growing AMR. Diversification in actions towards the fights against AMR has been made one of the priorities for successful actions (11,27). Thus, the lack of diversity is also proof of a lack of understanding and/or acknowledgment of the complexity of the system, which in turn can decreases the power of actions and interventions (18,28). More geographical diversity is also desired, as most of the interventions identified were implemented in high-income countries, while AMR interventions in low- and middle- income countries have been rapidly increasing over the last decade (29).
Black box of intervention implementation
From the 52 publications, it seems that factors related to the implementation, sustainability and success of an intervention are part of a black box (30), i.e., these factors are poorly or not reported in scientific publications. The aim of publishing AMR interventions is obviously to report on their effectiveness. Maximizing the usefulness of intervention requires more attention to implementation, including the description of the intervention, its implementation strategy, or the sharing of experience with other people involved in implementation, (11). Also, assessment of the interventions was mainly focused on quantitative indicators, which rationalize and simplify the description process. While this facilitates comparison of intervention results, it obscures the drivers of successes. Furthermore, many different indicators of success or failure were identified in this review. Units and definitions for indicators are not harmonized between evaluations. And still, the main indicators are quantitative, and authors did not explore qualitative methods to assess the interventions. Qualitative assessment of interventions would allow the publication of unusual results or feedback about the intervention and things that may not typically be measured in quantitative assessments of interventions (e.g., social capital, acceptability of the intervention) but may be important to intervention success (28). It could also help assessors to identify unexpected consequences of the intervention. We recognize that authors might not have yet the capacity to report the implementation and assessment of interventions for various reasons (e.g., time constraints, publication restrictions, reluctance to use non validated tools) but believe that the involvement of interdisciplinary team in the design, implementation, evaluation and reporting of interventions can fills this gap.
In most publications we identified a lack of contextualization, as the intervention was not implemented and reported in a global system perspective and usually seen as an individual action. All readers would benefit in knowing the context, the means, and other details of the intervention (31–35). Many studies developed frameworks for reporting and/or assessing interventions (36–41), but few included parameters linked to the social-ecological context of interventions, specifically in an AMR system. The scientific literature is in need of a new definition and understanding of AMR interventions, including a global contextualization of it as we defined an intervention primarily in our study. A new format of publication for health interventions with a systematic description following criteria to ensure minimum details would be an asset for the scientific community and decision makers (17,42).
Limitations – methodology
This study characterized 42 E. coli AMR interventions using the AMR-Intervene framework. Several limitations could have influenced the results of the study. First, coding bias has been largely prevented, but some variables of the framework still rely on subjective assessment by the coder. Therefore, we implemented a versioning system of AMR-Intervene that can be updated depending on research findings and needs of users. Second, the identification of interventions can also be affected by selection bias and affect our capacity to reach them via the online review process. Some interventions might not have clearly mentioned E. coli as a targeted pathogen. For example, interventions targeting many pathogens including E. coli among others, interventions based on clinical syndromes (e.g., urinary tract infections, frequently caused by E. coli ), as well as all interventions targeting ESBL (extended spectrum beta-lactamase) resistance. Finally, our study does not avoid the classical publication bias such as the year of publication, the publication of successful interventions only, publication habits and the dominance of high-income countries in the scientific literature. Details of unsuccessful interventions or even failure factors or difficulties that the implementers faced are rarely reported. Thus, this study should be complemented by a parallel survey to directly reach implementers and better capture success and failure factors, non-published or ongoing interventions that are not covered by this study.
Towards a better understanding and implementation of AMR interventions
One publication mentioned that activities to fight AMR might have diminishing returns and the reduction of AMU could not be improved by the current means (i.e., AMR interventions as imagined nowadays) and we might need a shift to other more ambitious interventions and/or improvement of intervention implementation.
Improving implementation of interventions is necessary for increasing our capacity to tackle the global issue of AMR. Indeed, there is a need of interventions that are better adapted to their social-ecological context, interventions that are diverse and flexible, with various angles of attack, at all levels of the social and bio-ecological system. However, if interventions are not detailed and reported in a more systematic approach, we might not have sufficient understanding of the current situation of AMR interventions.
A successful intervention should also be able to evolve by integrating feedback and adapting to change. Kruk et al. (43) mentioned that interventions should be integrative, adaptive, self-regulating, diverse, and aware. This would lead to a new generation of complex interventions that would improve how we tackle AMR (11). Diversity in actions and actors may be relevant at the intervention-level, and also at the country or sector-level (e.g., studying interventions from the same sector). On the other hand, diversity should not compromise the use and application of interventions that are effective and work in different context.
Future interventions are also in need of qualitative data about implementation, such as feedback about the experience, success and failure factors, and contextual factors that can impact the effectiveness or continuity of interventions. In this study we gathered a certain number of different success and failure factors, described in the publications. All factors were identified in previous literature but no publications reported factors about the characteristics of individuals involved, the implementation political and economic climate or how the process of executing the intervention among other factors classified by Darmschroder et al. (24). Therefore, there is a need to improve how interventions are assessed, possibly mixing methods of evaluation (28). While contextual factors cannot always be included in the assessment of intervention, mentioning concomitant interventions, even in a different sector, may help understand of the context. A better understanding and overview of a common fight against the AMR issue can also be the aim of a resilient governance system, e.g., listing, gathering, and enhancing AMR interventions in a country.