Background
The implementation of research into healthcare practice is complex [
1], with multiple levels to consider such as the patient, healthcare provider, multidisciplinary team, healthcare institution, and local and national healthcare systems. The implementation of evidence-based treatments to achieve healthcare system improvement that is robust, efficient, and sustainable is crucially important. However, it is well established that improving the availability of research is not enough for successful implementation [
2]; rather, active knowledge translation (KT) interventions are essential to facilitate the implementation of research to practice. Determining the success of KT interventions and the implementation process itself relies on evaluation studies.
In the KT field, experimental designs such as randomized trials, cluster randomized trials, and stepped wedge designs are widely used for evaluating the effectiveness of KT interventions. Rigorous experimental designs can provide strong estimates of KT intervention effectiveness, but offer limited insight into how the intervention worked or not [
1] as well as how KT interventions are mediated by different facilitators and barriers and how they lead to implementation or not [
3‐
5]. KT interventions contain several interacting components, such as the degree of flexibility or tailoring of the intervention, the number of interacting components within the interventions, and the number and difficulty of behaviors required by those delivering or receiving the intervention [
3]. This complexity makes it particularly challenging to evaluate KT intervention effectiveness [
3‐
5]. The effectiveness of KT interventions is a result of the interactions between many factors such as context and mechanisms of change. A lack of intervention effect may be due to implementation failure rather than the ineffectiveness of the intervention itself. KT interventions pose methodological challenges and require augmentations to the standard experimental designs [
6] to understand how they do or do not work.
As a result of these limitations, researchers have started to conduct process evaluations alongside experimental designs for evaluating KT interventions. The broad purpose of a process evaluation is to explore aspects of the implementation process [
7]. Process evaluations can be used to assess the fidelity, dose, adaptation, reach, and quality of implementation [
8,
9] and to identify the causal mechanisms [
10,
11], mechanisms of impact [
12], and contextual factors associated with variation in outcomes across sites [
6,
13]. Furthermore, process evaluations can assist in interpreting the outcome results [
7], the barriers and facilitators to implementation [
14,
15] and sustainability [
16], as well as examining the participants’ views [
17] and understandings of components of the intervention [
18,
19]. Process evaluations are vital in identifying the success or failure of implementation, which is critical in understanding intervention effectiveness.
Notwithstanding the work of Moore and colleagues [
12], there have been scant methodological recommendations to guide KT process evaluations. This deficit has made designing process evaluations in KT research challenging and has hindered the potential for meaningful comparisons across process evaluation studies. In 2000, the Medical Research Council released an evaluation framework for designing and evaluating complex interventions; this report was later revised in 2008 [
4,
20]. Of note, earlier guidance for evaluating complex interventions focused exclusively on randomized designs with no mention of process evaluations. The revisions mentioned process evaluations and the role that they can have with complex interventions, yet did not provide specific recommendations for evaluation designs, data collection types, time points, and standardized evaluation approaches for complex interventions. This level of specificity is imperative for research comparisons across KT intervention process evaluations and to understand how change is mediated by specific factors.
Recently, the Medical Research Council has commissioned an update of this guidance to be published in 2019 [
21,
22]. The update re-emphasizes some of the previous messages related to complex intervention development and evaluation; however, it provides a more flexible and less linear model of the process with added emphasis to development, implementation, and evaluation phases as well as providing a variety of successful case examples that employ a range of methods (from natural experiments to clinical trials). Early reports of the update to the MRC framework highlight the importance of process and economic evaluations as good investments and a move away from experimental methods as the only or best option for evaluation.
In 2013, a framework for process evaluations for cluster-randomized trials of complex interventions was proposed by Grant and colleagues [
20]; however, these recommendations were not based upon a comprehensive, systematic review of all approaches used by others. One study found that only 30% of the randomized controlled trails had associated qualitative investigations [
23]. Moreover, a large proportion of those qualitative evaluations were completed before the trial, with smaller numbers of qualitative evaluations completed during the trial or following it. Given the limitations of the process evaluation work to date, it is critical to systematically review all existing process evaluations of KT outcome assessment. Doing so will aid in the development of rigorous methodological guidance for process evaluation research of KT interventions moving forward.
The aim of our systematic review is to synthesize the existing evidence on process evaluation studies assessing KT interventions. The purpose of our review is to make explicit the current state of methodological guidance for process evaluation research with the aim of providing recommendations for multiple end-user groups. This knowledge is critically important for healthcare providers, health quality consultants, decision and policy makers, non-governmental organizations, governmental departments, and health services researchers to evaluate the effectiveness of their KT efforts in order to ensure scarce healthcare resources are effectively utilized and enhanced knowledge is properly generalized to benefit others.
Discussion
Our findings provided many insights into the current practices of KT researchers conducting integrated or separate process evaluations, the focus of these process evaluations, the data collection considerations, and the poor methodological quality and a lack of theoretical guidance informing these process evaluations.
The majority of included studies (60.2%) conducted a separate (stand-alone) rather than integrated process evaluation. As Moore and colleagues suggest, there are advantages and disadvantages of either (separated or integrated) approach [
12]. Arguments for separate process evaluations focus on analyzing process data without knowledge of outcome analysis to prevent biasing interpretations of results. Arguments for integration include ensuring implementation data is integrated into outcome analysis and using the process evaluation to identify intermediate outcome data and causal processes while informing the integration of new measures into outcome data collection. Our findings highlight that there is no clear preference for separate or integrated process evaluations. The decision for separation or integration of the process evaluation should be carefully considered by study teams to ensure it is the best option for their study objectives.
Our findings draw attention to a wide variety of terms and foci used within process evaluations. We identified a lack of clear and consistent concepts for process evaluations and their multifaceted components, as well as an absence of standard recommendations on how process evaluations should be developed and conducted. This finding is supported by a literature overview on process evaluations in public health published by Linnan and Steckler in 2002 [
29]. We would encourage researchers to employ terms that are utilized by other researchers to facilitate making meaningful comparisons across studies in the future and to be mindful of comprehensively including the key components of a process evaluation, context, implementation, and mechanisms of impact [
12].
Our findings highlight two important aspects about process evaluation data collection in relation to timing and type of data collected. In terms of data collection timing, almost half of the investigators collected their process evaluation data post-intervention (46%) without any pre-intervention or during intervention data collection. Surprisingly, only 17.7% of the included studies collected data pre- and post-intervention, and only 18 studies collected data pre-, during, and post-intervention. Process evaluations can provide useful information about intervention delivery and if the interventions were delivered as planned (fidelity), the intervention dose, as well as useful information about intervention reach and how the context shaped the implementation process. Our findings suggest a current propensity to collect data after intervention delivery (as compared to before and/or during). It is unclear if our findings are the result of a lack of forethought to employ data collection pre- and during implementation, a lack of resources, or a reliance on data collection approaches post-intervention. This aside, based upon our findings, we recommend that KT researchers planning process evaluations consider data collection earlier in the implementation process to prevent challenges with retrospective data collection and to maximize the potential power of process evaluations. Consideration of key components of process evaluations (context, implementation, and mechanisms of impact) is critically important to prevent inference-observation confusion from an exclusive reliance on outcome evaluations [
12]. An intervention can have positive outcomes even when an intervention was not delivered as intended, as other events or influences can be shaping a context [
30]. Conversely, an intervention may have limited or no effects for a number of reasons that extend beyond the ineffectiveness of the intervention including a weak research design or improper implementation of the intervention [
31]. Implicitly, the process evaluation framework by Moore and colleagues suggests that process evaluation data collection ideally needs to be collected before and throughout the implementation process in order to capture all aspects of implementation [
12].
In terms of data collection type, just over half (54.4%) of the studies utilized qualitative interviews as one form of data collection. Reflecting on the key components of process evaluations (context, implementation, and mechanisms of impact), the frequency of qualitative data collection approaches is lower than anticipated. Qualitative approaches such as interviewing are ideal for uncovering rich and detailed aspects of the implementation context, nuanced participant perspectives on the implementation processes, and the potential mediators to implementation impact. When considering the key components of a process evaluation (context, implementation, and mechanisms of impact), by default, it is suggestive of multi-method work. Consequently, we urge researchers to consider integrating qualitative and quantitative data into their process evaluation study designs to richly capture various perspectives. In addition to individual interviews, surveys, participant observation, focus groups, and document analysis could be used.
A major finding from this systematic review is the lack of methodological rigor in many of the process evaluations. Almost 40% of the studies included in this review had a MMAT score of 50 or less, but the scores varied significantly in terms of study designs used by the investigators. Moreover, the frequency of low MMAT scores for multi-method and mixed method studies suggests a tendency for lower methodological quality which could point to the challenging nature of these research designs [
32] or a lack of reporting guidelines.
Our findings identified a lack of theoretical guidance employed and reported in the included process evaluation studies. It is important to note the role of theory within evaluation is considered contentious by some [
33,
34], yet conversely, there are increasing calls for the use of theory in the literature. While there is this tension between using or not using theory in evaluations, there are many reported advantages to theory-driven evaluations [
29,
33,
34], yet more than 60% of the included studies were not informed by theory. Current research evidence suggests that using theory can help to design studies that increase KT and enable better interpretation and replication of findings of implementation studies [
35]. In alignment with Moore and colleagues, we encourage researchers to consider utilizing theory when designing process evaluations. There is no shortage of KT theories available. Recently, Strifler and colleagues identified 159 KT theories, models, and frameworks in the literature [
36]. In the words of Moore and colleagues who were citing the revised MRC guidance (2008), “an understanding of the causal assumptions underpinning the intervention and use of evaluation to understand how interventions work in practice are vital in building an evidence base that informs policy and practice” [
9].
Limitations
As with all reviews, there is the possibility of incomplete retrieval of identified research; however, this review entailed a comprehensive search of published literature and rigorous review methods. Limitations include the eligibility restrictions (only published studies in the English language were included, for example), and data collection did not extend beyond data reported in included studies.
Conclusions
The current state of the quality of evidence base of process evaluations in KT is weak. Policy makers and funding organizations should call for theory-based multi or mixed method designs with a complimentary process evaluation component. Mixed method designs, with an integrated process evaluation component, would help to inform decision makers about effective process evaluation approaches, and research funding organizations could further promote theory-based designs to guide the development and conduct of implementation studies with a rigorous process evaluation component. Achieving this goal may require well-assembled implementation teams including clinical experts, as well as strong researchers with methodological expertise.
We recommend that future investigators employ rigorous theory-guided multi or mixed method approaches to evaluate the processes of implementation of KT interventions. Our findings highlighted that to date, qualitative study designs in the form of separate (stand-alone) process evaluations are the most frequently reported approaches. The predominant data collection method of using qualitative interviews helps to better understand process evaluations and to answer questions about why the implementation processes work or not, but does not provide an answer about the effectiveness of the implementation processes used. In light of the work of Moore and colleagues [
12], we advocate that future process evaluation investigators should use both qualitative and quantitative methods (mixed methods) with an integrated process evaluation component to evaluate implementation processes in KT research.
We identified the timing of data collection as another methodological weakness in this systematic review. It remains unclear why almost half of the included process evaluation studies collected data only post-implementation. To provide high-certainty evidence for process evaluations, we advocate for the collection of pre-, during, and post-implementation measures and the use of statistical uncertainty measures (e.g., standard deviation, standard error, p values, and confidence intervals). This would allow a rigorous assessment of the implementation processes and sound recommendations supported by statistical measures. The timing of pre-evaluations also helps to address issues before implementation occurs. There is widespread acceptance that the generalizability of quantitative trials of KT interventions would be significantly enhanced through complementary process evaluations. Most data collection occurred post-intervention undermining the ability to evaluate the process of implementation.
Strong science and methodological guidance is needed to underpin and guide the design and execution of process evaluations in KT science. A theory-based approach to inform process evaluations of KT interventions would allow investigators to reach conclusions, not only about the processes by which interventions were implemented and the outcomes they have generated, but also about the reliability of the causal assumptions that link intervention processes and outcomes. Future research is needed that could provide state-of-the-art recommendations on how to design, conduct, and report rigorous process evaluations as part of a theory-based mixed methods evaluation of KT projects. Intervention theory should be used to inform the design of implementation studies to investigate the success or failure of the strategies used. This could lead to more generalizable findings to inform researchers and knowledge users about effective implementation strategies.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.