Application of the Expert Recommendations for Implementing Change (ERIC) compilation of strategies to health intervention implementation in low- and middle-income countries: a systematic review

The past two decades have been marked by rapid growth in the field of implementation science to address the large research-to-practice gap across contexts and health areas [1]. In recent years, the field’s focus has shifted from defining barriers and facilitators of implementing evidence-based practices to identifying strategies that effectively address and overcome these barriers. Implementation strategies are generally defined as the approaches or techniques used to enhance the adoption, implementation, sustainment, and scale-up of an evidence-based practice [2, 3]. These strategies vary in complexity and can target determinants at the intervention-, patient-, provider-, organization-, community-, policy-, and funding levels [4, 5].

Though the evidence base on implementation strategies is growing, current data on strategy effectiveness is mixed, with high variation in strategy effects observed across studies and outcomes [6‐11]. Several reasons may contribute to this variation. It may be that certain strategies are not sufficient to improve implementation outcomes across contexts; it may also be that strategies were not appropriately matched to the contextual determinants or tailored to the setting [12]. However, reporting on implementation strategies often lacks the necessary information to determine why a strategy was or was not effective; for example, information on how a strategy was selected, adapted, and operationalized and whether or not the strategy was carried out as intended [7, 13‐17]. As such, calls for consistent, detailed reporting of implementation strategies have emerged in tandem with calls for increased research on strategy effectiveness [2, 3, 18‐20].

To aid reporting efforts, the field has developed taxonomies of implementation strategies [21, 22] and a methodology for specifying implementation strategies [3]. One strategy taxonomy, the Expert Recommendations for Implementing Change (ERIC) project, built upon a narrative review of the literature [23] and used a modified Delphi process to develop a compilation of implementation strategies, comprising 73 discrete strategies [21] which can be further grouped into nine thematic clusters [24]. The ERIC compilation has become the most commonly used taxonomy of strategies in the field of implementation science, with over 3000 citations. It has enabled a standardized language for naming implementation strategies that have been used to characterize implementation efforts both in prospective and retrospective analyses [25‐30]. To support standardized strategy specification, Proctor et al. [3] developed guidelines to help stakeholders operationalize strategies based on specific domains, including the strategy actor, action, action target, temporality, dose, implementation outcome affected, and justification. These specification guidelines are consistent with the Patient Centered Outcomes Research Institute's Standards for Studies of Complex Interventions [31], and have the potential to not only improve our understanding of implementation strategy mechanisms, but also the required parameters for replication in other research and practice settings.

In low- and middle-income countries (LMICs), implementation science has become a key tool for helping to bridge the research-to-practice gap that is larger than that of high-income countries [32, 33]. As such, the need to identify effective implementation strategies for efficient, effective delivery of evidence-based practices is imperative in these settings. In recent years, formal implementation research in LMICs has expanded as funding sources increasingly recognize the utility of this work. There is now a growing evidence base for using implementation frameworks [34, 35] and measures [36, 37], and specific determinants to implementation [38] in LMICs have been identified. However, little is known about effective implementation strategies in these settings [39].

The purpose of the present study is to report on the application of the ERIC compilation of implementation strategies in LMICs and provide recommendations to the field for improving its application moving forward. Our aims are twofold: (i) to systematically review the literature on the use of ERIC strategies in LMICs, including which specific strategies have been included in the research, how they were selected, how the strategies were used (i.e., specification and targeted intervention/health condition/population), and how they were adapted; and (ii) to assess evidence for the effectiveness of specific ERIC implementation strategies in LMICs.

We registered our systematic review protocol in the International Prospective Register of Systematic Reviews (PROSPERO # CRD42021268374) and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [40]. See Additional file 1 for the completed PRISMA checklist.

The database search yielded 659 articles, of which 441 were duplicates. We screened the remaining 218 article titles and abstracts and excluded 88, leaving 130 for full-text review. Of these, 41 were excluded that did not use ERIC (i.e., only cited ERIC in the manuscript introduction or discussion, without application to the present study), 12 that did not take place in an LMIC, 11 that did not meet multiple inclusion criteria, 5 that were not peer-reviewed, and 1 that was not primary research (Fig. 1).

The final sample included 60 studies (Table 1, see Additional file 3 for individual study characteristics), all published in English. The first study using ERIC in an LMIC was published in 2016, and the number of studies using ERIC in LMICs increased over the years, from just 1 in 2016 to 20 in 2022. Studies included data collected in all six WHO Regions, with the majority being conducted in the African region (n = 36, 60.0%). Most studies focused on healthcare settings (n = 52, 86.7%), while just 11 (18.3%) focused on community and four (6.7%) on policy-level settings. The most common health conditions targeted in studies were infectious diseases (n = 19, 31.7%), maternal and child health (n = 10, 16.7%), and mental health and substance use (n = 10, 16.7%). Two studies did not focus on a specific health condition, but rather on strategies for implementing clinical trial recruitment and building a national framework for research.

Nearly three-quarters of studies described empirical research (n = 43, 71.7%), the others being protocols for studies not yet completed (n = 17, 23.3%). Study populations included patients (n = 31, 51.7%), providers (n = 43, 71.7%), policymakers (n = 14, 23.3%), community members (n = 8, 13.3%), and researchers (n = 6, 10.0%); 34 (56.7%) studies included more than one of these study populations. Nineteen (31.7%) articles described formative implementation strategy design and nine (15.0%) described retrospective strategy specification. The majority (n = 36, 60%) of studies included an impact evaluation, the most common design being quasi-experimental with no control (n = 13, 21.7%) and a cluster randomized control trial (n = 9, 15.0%). Of the 17 studies that included a control, n = 9 tested the intervention and n = 7 tested the implementation strategy. Implementation frameworks were used by 47 (78.3%) studies, with 36 (60.0%) citing implementation determinant frameworks, 24 (40.0%) citing evaluation frameworks, and seven (11.7%) citing a process framework. A total of 44 (73.4%) studies evaluated implementation outcomes, most commonly adoption (n = 29), acceptability (n = 27), and fidelity (n = 26); 34 (56.7%) studies evaluated multiple implementation outcomes. Under half of studies (n = 25, 41.7%) evaluated health outcomes.

Across the 60 studies, 548 strategies were proposed, 282 (51.5%) of which were planned for future delivery and 266 (48.5%) had been delivered in the study. The total number of strategies described per study ranged from 1 to 46 (median = 6). Despite all 60 studies referencing the use of ERIC, just 32 (53.3%) studies explicitly matched specific implementation strategies used in the study to a specific ERIC strategy (Table 2). One study described strategy development being guided by other frameworks in addition to ERIC [47‐49]; all other studies cited only ERIC as the guiding framework. The most commonly utilized ERIC strategies were (1) conduct educational meetings (n = 16, 2.9% of all 548 strategies proposed across studies); (2) audit and provide feedback (n = 15, 2.7%); (3) assess for readiness and identify barriers and facilitators (n = 13, 2.4%); and (4) build a coalition (n = 12, 2.2%). Of the 73 ERIC strategies, 9 (12.3%) were not cited at all. These include altering patient/consumer fees, changing liability laws, developing disincentives, making billing easier, preparing patients/consumers to be active participants, starting a dissemination organization, using capitated payments, using other payment schemes, and visiting other sites. Strategies from all ERIC strategy categories were described, though 5 of 9 (55.5%) strategies in the Utilize Financial Strategies category were not used.

Eight (13.3%) studies did not include any component of strategy specification (i.e., they named strategies but did not describe them at all), representing 88 (16.1%) of the 548 strategies described across studies. Among the strategies that were specified (Table 3), the most common components described were action (n = 412, 75.2%), action target (n = 333, 60.8%), and actor (n = 314, 57.3%). The study team itself comprised the majority of actors specified for implementation strategies, accounting for 185 (58.9%) of the 314 actors specified; action targets were most commonly healthcare providers, accounting for 196 (58.9%) of 333 action targets specified. The least commonly specified components were fidelity (n = 17, 3.1%), adaptation (n = 100, 18.2%), action mechanism (n = 129, 23.5%), and targeted implementation outcomes (n = 148, 27.1%). Only 12 (20.0%) studies described adaptation of implementation strategies, of which two (3.3%) described the use of Implementation mapping and one the use of human-centered design (1.7%) to guide adaptation, while nine (15.0%) described more general stakeholder engagement without the use of a specific framework of the process. Only one (0.5%) strategy was tested for independent effects: Gachau et al. found that audit and feedback significantly improved 24 of 34 indicators of pediatric guideline adherence in Kenya [50].

In the present systematic review, we found 60 studies that cited the use of ERIC strategies in LMIC settings. These studies included data from all WHO regions and focused on diverse health issues, with over 35% published in the final year of our review period, indicating a growing application of ERIC in LMICs. However, just over half explicitly matched implementation strategies employed to an ERIC strategy and 16% of strategies did not include any components of strategy specification. Moreover, a minority of studies employed randomized trials or high-quality quasi-experimental designs with controls, and just one study evaluated implementation strategy effectiveness.

While nearly half of the included studies did not explicitly match implementation strategies used in the study to a specific ERIC strategy, several notable points arose from the ERIC strategies that were reported. First, a wide variety of ERIC strategies were identified, with 88% of the 73 ERIC strategies represented. This suggests that almost all ERIC strategies can be applied to LMIC contexts. Yet, several strategies that seem critical to implementation in LMIC contexts—including capturing and sharing local knowledge, conducting local needs assessment, providing local technical assistance, and tailoring strategies—were not commonly cited in any of the included studies. This may be because these strategies were considered irrelevant or incompatible [35], redundant with other ERIC strategies [34], or part of routine processes for implementing health interventions [51‐53] rather than discrete implementation strategies. Additionally, other ERIC strategies rarely, if ever, cited included those that required systems-level changes—such as change fees/incentives/billing, change policy, and enhancing local data systems and analysis—despite data suggesting organizational- and policy-level strategies are effective [54] and a critical component to closing the research-to-practice gap [55]. Finally, a number of strategies not used were those related to privatized health systems (e.g., alter consumer fees, change liability laws, use other payment schemes), which are less common in LMICs than centralized, public health systems, and thus may not be applicable or require adaptation in LMIC contexts. Further research is needed to explore why certain ERIC strategies have not been used in LMICs and if there are additional implementation strategies relevant to LMIC settings not currently included in the ERIC compilation.

Strengthening the evidence base for implementation strategies requires that their operationalization be reported in detail and that modifications to implementation strategies be systematically documented. While the included studies’ lack of ERIC strategy matching complicates the interpretation of specific ERIC strategy applicability to LMICs in this review, strategy matching itself may not be a requirement for understanding if and how certain implementation strategies are effective. However, implementation strategy components need to be specified in a way that allows their replication in research and practice and comparison across contexts [56, 57]. Among the studies reviewed here, all but eight included some type of strategy specification. However, most strategies only included specification of their action (75%), action target (61%), and actor (57%). Strategy details, such as temporality and dose, are necessary to replicate them in further testing. Moreover, mechanisms of action and fidelity, described for just 24% and 3% of strategies, respectively, are required for generating theory and selecting strategies that appropriately target contextual determinants [19]. Also concerning are how few strategies (18%) included a description of their adaptation process, which is likely necessary to meet the specific needs of the study context [2, 58‐60]. Of the strategies that did include specification, most relied heavily on providers or research teams as actors and providers as action targets. This echoes our finding that few strategies addressing the organizational and policy levels were used and further highlights the dearth of research on implementation strategies that support population-level health improvements previously observed in high-income [61] and LMIC settings [39]. Moreover, the reliance on research team members as strategy actors threatens the generalizability of implementation strategies used in the studies to real-world settings.

Finally, though we had hoped to be able to assess ERIC strategy effectiveness in LMIC, just one strategy was evaluated in the studies included in our review. This is likely related to few studies defining strategy justification, mechanisms, and targeted implementation outcomes in their research. Instead, most studies looked at implementation outcomes without directly linking them to an identified determinant and a theory of change, precluding them from evaluating the individual strategy’s impact. Research teams may have lacked the resources to conduct a randomized control trial to rigorously test a strategy, as less than 20% of studies employed this design. However, while randomized control trials are the gold standard for research on intervention effectiveness, research on implementation strategies can successfully employ alternative designs, including interrupted time series, factorial, adaptive, and rollout designs [62, 63]. These designs can provide more flexibility and feasibility in resource-limited settings, while simultaneously maximizing external validity [64, 65].

The present study has several limitations. For one, we only focused on articles that used the ERIC compilation to inform study strategies and not more broadly defined implementation strategies. Therefore, our results may not be generalizable to all implementation strategy research in LMIC. We chose to focus on ERIC strategies as it is the taxonomy most commonly used in the field of implementation science, with over 3300 citations for the original [23] and refined [21] ERIC strategy papers and was developed using an expert review of existing strategy compilations and reviews. While we recognize that this biases toward research connected to non-LMIC academics, we note that around two-thirds of included studies were published in the final 2 years of our review period, suggesting that the use of ERIC is becoming more widespread in LMICs and highlighting the need to understand how to promote improved application and testing of these strategies. Second, we only included articles published in peer-reviewed journals. Owing to factors such as publication cost and language of these journals, this likely biased our findings to represent international and well-funded research. Further research is needed to explore non-ERIC strategy application in LMICs across languages and in the grey literature. Third, as we included study protocols in our sample, we cannot say with certainty that all strategies proposed will be applied in the research phase. However, the objective of this review was to provide a comprehensive description of the current state of this rapidly growing field (e.g., 35% of included papers published in the last year, 28% of studies still in the protocol phase), and inclusion of study protocols allows us to capture the most recent data. Finally, as many strategies employed were not matched to ERIC by the study authors, we were unable to draw conclusions about which may be the most relevant ERIC strategy(ies) in LMICs. However, those that did match indicated the individual ERIC strategies are applicable in LMICs.

This systematic review demonstrated the broad and growing use of the ERIC strategy taxonomy in LMICs, with inconsistency in application and very limited testing of ERIC strategy effectiveness. Moving forward, we provide the following recommendations to promote the development of implementation strategies to more rapidly close the research-to-practice gap in LMICs. First, research in LMICs must move beyond merely describing strategies to evaluating their effects on implementation outcomes. Moreover, strategies that are tested need to be better specified so that their effectiveness may be compared across studies and contexts and their mechanisms of change can be understood. Researchers should also consider reporting how the strategy would be deployed under routine, non-research-related conditions as well to promote application beyond the study period. Finally, strategies targeting policy, organizational, and community-level determinants should be explored to encourage change that supports scale-up and sustainability of individual, research-based implementation efforts in LMICs. To catalyze these lines of research, there is a need for greater capacity-building among researchers in LMICs to gain training in implementation research. Moreover, this training should directly involve and/or emphasize methods for the engagement of diverse local stakeholders, such as policymakers and community members, who may be better situated to develop and implement strategies at the systems level. Research funders, governments, and other implementers should consider encouraging work that includes each of these components.