Systematic Reviews & Guidelines

PROTOCOL Open Access

Effect of implementation interventions on nurses’ behaviour in clinical practice: a systematic review, meta-analysis and meta- regression protocol Guillaume Fontaine1,2* , Sylvie Cossette1,2, Marc-André Maheu-Cadotte1,2,3, Marie-France Deschênes1,4, Geneviève Rouleau3,5, Andréane Lavallée1,6, Catherine Pépin1,7, Ariane Ballard1,6, Gabrielle Chicoine1,3, Alexandra Lapierre1,2,8, Patrick Lavoie1,2,4, Jérémie Blondin9 and Tanya Mailhot10

Abstract

Background: Practitioner-level implementation interventions such as audit and feedback, communities of practice, and local opinion leaders have shown potential to change nurses’ behaviour in clinical practice and improve patients’ health. However, their effectiveness remains unclear. Moreover, we have a paucity of data regarding the use of theory in implementation studies with nurses, the causal processes—i.e. mechanisms of action—targeted by interventions to change nurses’ behaviour in clinical practice, and the constituent components—i.e. behaviour change techniques—included in interventions. Thus, our objectives are threefold: (1) to examine the effectiveness of practitioner-level implementation interventions in changing nurses’ behaviour in clinical practice; (2) to identify, in included studies, the type and degree of theory use, the mechanisms of action targeted by interventions and the behaviour change techniques constituting interventions and (3) to examine whether intervention effectiveness is associated with the use of theory or with specific mechanisms of action and behaviour change techniques.

Methods: We will conduct a systematic review based on the Cochrane Effective Practice and Organization of Care (EPOC) Group guidelines. We will search six databases (CINAHL, EMBASE, ERIC, PsycINFO, PubMed and Web of Science) with no time limitation for experimental and quasi-experimental studies that evaluated practitioner-level implementation interventions aiming to change nurses’ behaviour in clinical practice. We will also hand-search reference lists of included studies. We will perform screening, full-text review, risk of bias assessment, and data extraction independently with the Covidence systematic review software. We will assess the quality of evidence using the GRADEpro software. We will code included studies independently for theory use (Theory Coding Scheme), mechanisms of action (coding guidelines from Michie) and behaviour change techniques (Behaviour Change Technique Taxonomy v1) with QSR International’s NVivo qualitative data analysis software. Meta-analyses will be performed using the Review Manager (RevMan) software. Meta- regression analyses will be performed with IBM SPSS Statistics software.

Discussion: This review will inform knowledge users and researchers interested in designing, developing and evaluating implementation interventions to support nurses’ behaviour change in clinical practice. Results will provide key insights regarding which causal processes—i.e. mechanisms of action—should be targeted by these interventions, and which constituent components—i.e. behaviour change techniques—should be included in these interventions to increase their effectiveness. (Continued on next page)

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

* Correspondence: guillaume.fontaine@umontreal.ca 1Faculty of Nursing, Université de Montréal, Montréal, Canada 2Research Center, Montreal Heart Institute, Montréal, Canada Full list of author information is available at the end of the article

Fontaine et al. Systematic Reviews (2019) 8:305 https://doi.org/10.1186/s13643-019-1227-x

(Continued from previous page)

Systematic review registration: The protocol has been registered at the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD42019130446).

Keywords: Behaviour change, Implementation strategy, Implementation intervention, Implementation science, Knowledge translation, Theory-based interventions, Nurses

Background Nurses represent the largest group of healthcare profes- sionals that intervenes with patients in all sectors of health systems around the world [1]. Thus, nurses are often ac- tively involved in initiatives aiming to improve service de- livery to enhance patient outcomes [2]. However, changing nurses’ behaviour in clinical practice is a challen- ging and complex endeavor due to the influence of practitioner-level factors, including nurses’ motivational predispositions to change, and organizational-level factors [3, 4]. Multiple barriers specific to nursing practice, in- cluding lack of time, lack of organizational support, com- peting priorities and expanding workloads hinder the implementation of evidence-based nursing practices [5]. In the last decade, we have witnessed the emergence

of implementation science, the scientific study of methods and theoretical approaches to improve health services and health through changes in healthcare pro- fessionals’ and organizations’ practices [6]. Implementa- tion interventions have been associated with more effective health service delivery and improved health outcomes in several clinical practice settings [7–10]. A wide range of clinical behaviours have been targeted by these interventions, including medication prescribing, test ordering, disease screening and management, dis- charge planning and counseling [4, 9, 10]. Although nurses have frequently been the target of implementa- tion interventions, we know little about the effectiveness, theoretical underpinnings and components of these interventions.

Description of implementation interventions An implementation intervention is defined as any strat- egy or program ‘aimed at increasing the use of research- based knowledge in healthcare practice (p. 2)’ [11]. Im- plementation interventions targeting specifically health- care professionals—i.e. practitioner-level implementation interventions—are described in the Cochrane Effective Practice and Organization of Care (EPOC) Group Taxonomy of Health System Interventions [12]. Exam- ples of practitioner-level implementation interventions, also named implementation strategies, include audit and feedback, educational materials, educational games, communities of practice, local opinion leaders, printed educational materials and reminders [12].

How implementation interventions might work Implementation interventions aim to ‘produce change in people’s behaviour or the environments in which they operate, or both (p. 2)’ [11]. Importantly, these interven- tions may aim for change at one or many levels (e.g. in- dividual healthcare professionals, teams, organizations, system). Hereafter, we focus specifically on practitioner- level implementation interventions, which target behav- iour change at the level of individual healthcare profes- sionals and teams (i.e. nurses and teams of nurses in this review) (see Fig. 1). Practitioner-level implementation interventions may

be based on a wide range of theoretical approaches (i.e. theories, models, frameworks) [16]. Behavioural approaches to implementation science draw upon de- cades of research in social and health psychology [15]. Theories of behaviour and behaviour change (e.g. theory of planned behaviour, theory of interper- sonal behaviour) appear particularly useful for pre- dicting and explaining nurses’ behaviour in clinical practice. For instance, a researcher could investigate the extent to which nurses’ beliefs, attitudes and sub- jective norms concerning a clinical guideline predict/ explain their adherence to this guideline in practice [16]. Thus, these theories may also be useful for selecting the potential mechanisms of action of behav- iour change in nurses that will be targeted by an intervention to lead to successful implementation [17]. Mechanisms of action represent the causal pro- cesses through which an intervention, or a constituent component, affects nurses’ behaviour in clinical prac- tice. These mechanisms of action ‘can be intraper- sonal psychological processes of the individual (e.g. motivation, skills, attitudes) and/or characteristics of the social and physical environment (e.g. social sup- port)’ [18]. Michie and colleagues have identified 26 mechanisms of action in theories of behaviour and behaviour change that may be targeted by interven- tions [18–20]. Describing the mechanisms of action targeted by implementation interventions could pro- vide insight into the causal pathways leading to be- haviour change in nurses. ‘Implementation intervention’ is an overarching

term used to distinguish the intervention from its constituent components [15]. These components—the

Fontaine et al. Systematic Reviews (2019) 8:305 Page 2 of 10

active ingredients of the intervention—can be de- scribed as behaviour change techniques. For instance, an implementation intervention based on audit and feedback may include multiple behaviour change tech- niques. Behaviour change techniques are “observable, replicable and irreducible components of an interven- tion designed to alter or redirect mechanisms of ac- tion that regulate behaviour; that is, a technique is proposed to be an ‘active ingredient’ (e.g. feedback, self-monitoring and reinforcement)” [21]. A taxonomy of 93 distinct behaviour change techniques, grouped into 16 clusters, has been developed by a Delphi con- sensus research method including a panel of inter- national experts [21]. Some examples of the clusters of behaviour change techniques include ‘feedback and monitoring’, ‘comparison of outcomes’ and ‘repetition and substitution’. Describing behaviour change tech- niques included in implementation interventions would be useful for reporting, replicating and synthe- sizing evidence. Thus, it is hypothesized that implementation inter-

ventions include multiple behaviour change tech- niques altering different mechanisms of action to effect behaviour change in nurses. For example, the implementation intervention ‘printed educational ma- terials’ may include behaviour change techniques such as ‘instruction on how to perform the clinical prac- tice’ to alter mechanisms of action such as ‘know- ledge’, ‘attitudes’, ‘beliefs’ and ‘perceived control’, to effect behaviour change in nurses [22]. The imple- mentation intervention ‘local opinion leaders’, i.e. in- dividuals using their influence to promote and effect behaviour change in clinical practice through leader- ship, will include other behaviour change techniques, such as ‘credible source in favour of the implementa- tion of the clinical practice’ and target mechanisms of action such as ‘social norms’ [23].

Why it is important to do this review So far, implementation interventions have had incon- sistent results with regard to changing nurses’ behav- iour in clinical practice [3, 4, 24]. This may be

explained by several factors. First, studies and reviews examining the effect of implementation interventions have often not addressed key mechanisms of action hypothesized be specific to nursing practice and the clin- ical context [3, 4, 7–10]. Second, it appears that mul- tiple interventions have been theory-inspired rather than theory-based. Indeed, researchers often rely on theoretical approaches only for some part of their intervention rather than adopt a systematic, theory-based intervention de- velopment process [17]. Thus, it appears important to examine the type and degree of theory use (e.g. refer- ence to underpinning theory, measurement of con- structs) in implementation interventions targeting nurses in addition to the effectiveness of such inter- ventions [17, 25]. Third, there has been little research regarding the optimal constituent components—i.e. the behaviour change techniques—of implementation interventions targeting nurses. This limits our ability to make recommendations regarding intervention char- acteristics likely to lead to successful implementation in nurses. To our knowledge, no review has looked into the ef-

fectiveness, theoretical underpinnings (i.e. theory use, mechanisms of action targeted) and behaviour change techniques of practitioner-level implementation inter- ventions aiming to change nurses’ behaviour in clinical practice and, ultimately, improve patient outcomes. Thus, our objectives are threefold:

1. To examine the effectiveness of practitioner-level implementation interventions in changing nurses’ behaviour in clinical practice and in improving pa- tient outcomes;

2. To identify: a. The types—i.e. individual theory items,

categories of theory use—and degree—i.e. total theory use score—of theory use in the development and evaluation of these interventions according to the Theory Coding Scheme [25];

b. The causal processes—i.e. mechanisms of action—targeted by these interventions to bring

Fig. 1 Causal modelling approach to the development of theory-based practitioner-level implementation interventions inspired by Hardeman [13], Michie [14] and Presseau [15]

Fontaine et al. Systematic Reviews (2019) 8:305 Page 3 of 10

about behaviour change in nurses according to guidelines of Michie and colleagues [18–20];

c. The constituent components—i.e. behaviour change techniques—included in these interventions according to the Behaviour Change Technique Taxonomy v1 [21];

3. To examine whether using theory, targeting specific mechanisms of action and including specific behaviour change techniques increase implementation intervention effectiveness in changing nurses’ behaviour in clinical practice.

Methods This systematic review protocol is based on the Effective Practice and Organization of Care (EPOC) Cochrane Group guidelines [26, 27] and reported according to the Preferred Reporting Items for Systematic review and Meta-Analysis Protocols (PRISMA-P) Checklist [28] (see Additional file 1). This protocol was prospectively regis- tered on the International Prospective Register of Sys- tematic Reviews (PROSPERO; CRD42019130446; available from: https://www.crd.york.ac.uk/prospero/dis- play_record.php?ID=CRD42019130446).

Criteria for considering studies for this review Types of studies We will include all experimental studies (i.e. randomized controlled trials (RCTs), cluster RCTs, crossover RCTs) and quasi-experimental studies (i.e. non-randomized controlled trials, cluster non-randomized controlled tri- als). We will exclude all qualitative, cross-sectional, ob- servational studies, case reports, discussion papers, editorials, knowledge syntheses, dissertations and theses. We will only include studies published in English or in French, regardless of the geographic location, in a peer- reviewed journal and in peer-reviewed conference proceedings.

Types of participants We will include studies conducted with registered nurses (RNs), clinical nurse specialists (CNSs), nurse practi- tioners (NPs), licensed practical nurses (LPNs) or regis- tered practical nurses (RPNs). We will include studies conducted in any type of clinical setting (e.g. hospitals, ambulatory clinics, community health centres). We will exclude studies including other groups of healthcare professionals and/or undergraduate nursing students.

Types of interventions We will include studies reporting practitioner-level imple- mentation interventions targeting nurses. We define a ‘prac- titioner-level implementation intervention’ as any strategy aimed at increasing the use of research-based knowledge in healthcare through changes in nurses’ clinical practice [6,

29]. More specifically, we will consider for inclusion studies which report an intervention including at least one imple- mentation strategy targeting specifically nurses as described in a subsection of the Cochrane Effective Practice and Organization of Care (EPOC) Group Taxonomy of Health System Interventions [12] (see Additional file 2). We will in- clude studies combining multiple implementation strategies listed in the EPOC Group Taxonomy of Health System Interventions. However, we will exclude studies including financial interventions, patient-oriented organizational interventions, struc- tural organizational interventions and regulatory inter- ventions, which are beyond the scope of this review. We will include studies with all types of

comparator(s).

Types of outcome measures Primary outcome We will include studies reporting on at least one out- come related to a change in nurses’ behaviour in clinical practice. More specifically, we will include studies reporting an objective measure of nurses’ behaviour (e.g. clinical interventions reported in patients’ medical files, number of tests ordered) or a subjective measure of nurses’ behaviour (e.g. self-reported performance of clin- ical interventions).

Secondary outcomes We will also collect data related to the following outcomes:

� Other outcomes in nurses � Objective or subjective measures of nurses’

intention to change behaviour in clinical practice and other hypothesized mechanisms of action, including knowledge, attitudes, beliefs, subjective norms and skills.

� Patient health behaviour, health status and well- being � Objective measures of patient health behaviour,

health status and well-being, including physical health and treatment outcomes, psychological health and psychosocial outcomes, as long as they can be associated with nurses’ interventions per- formed in clinical practice.

Search methods for identification of studies Electronic searches We developed the search strategy with a graduate stu- dent in librarianship and information science (JB). The search strategy was then validated by an experienced li- brarian. It includes a combination of three major con- cepts: (1) implementation interventions; (2) nurses; (3) study design (see Additional file 3). We first developed

Fontaine et al. Systematic Reviews (2019) 8:305 Page 4 of 10

the search strategy for PubMed (see Additional file 4), then tailored it to each database. We refined the search strategy over a period of 2 months to ensure specificity, sensibility and replicability in all databases. The search strategy targets six databases:

� Cumulative Index to Nursing and Allied Health Literature (CINAHL), via EBSCOhost (1980 to present);

� Excerpta Medical Database (EMBASE), via Ovid SP (1947 to present);

� Education Resources Information Center (ERIC), via Ovid SP (1966 to present);

� PsycINFO, via APA PsycNet (1967 to present); � PubMed (including MEDLINE), via NCBI (1946 to

present); � Web of Science—Science Citation Index (SCI)

Expanded and Social Sciences Citation Index (SSCI), via Clarivate Analytics (1900 to present).

Searching other resources Using a snowball method, we will manually screen the reference list of included studies to identify additional studies by looking at titles. In addition, we will search the Cochrane Database of Systematic Reviews (CDSR) and Google Scholar for related systematic reviews to find additional studies.

Data collection and analysis The different stages of data collection will be conducted by review authors in teams of two. Five teams of two were formed: team A (GF and CC), team B (AB and ALavallée), team C (MAMC and CP), team D (GR and GC) and team E (ALapierre and MFD) (see Table 1). The teams were formed based on the experience of each review author in a particular field (e.g. screening titles and abstracts, assessing risk of bias, coding studies using qualitative research software).

Selection of studies We will manage the records obtained with the search strategy with the Covidence systematic review software v1430 (Veritas Health Innovation, Melbourne, Australia; www.covidence.org) [30]. Covidence is the primary

screening and data extraction tool for Cochrane authors, streamlining the production of intervention reviews. Ten review authors, in teams of two, will independently screen all titles and abstracts retrieved by the search strategy and apply the eligibility criteria. We will con- duct a full-text review for the citations who will be rated as relevant, potentially relevant or with unclear relevance by at least one of the two reviews authors. Ten review authors, in teams of two, will independently screen full- text articles and identify studies for inclusion and iden- tify and record reasons for the exclusion of the ineligible studies. At any time during the review process, we will resolve disagreements through discussion and consen- sus. An author not involved in the study selection process will make a decision in case of a persistent dis- agreement. We will record the process of study selection in a PRISMA flow chart [31].

Data extraction and management A modified version of the Cochrane EPOC Review Group data collection form [32] was developed specific- ally for this review. This form will be iteratively validated by the whole team to ensure its completeness and clar- ity. Before data collection, we will calibrate our data col- lection form on a random sample of five full-text articles. The data collection form will be revised for clar- ity, as needed. Subsequently, ten review authors, in teams of two, will conduct all data collection for each study independently. We will collect data at the follow- ing levels:

� Study level: study design, year of study conduct, sample size, power analysis (yes/no), type of randomization, setting, country of study conduct, study funding source(s) and contact author;

� Participant level: type and number of participants, inclusion criteria, withdrawals and exclusions (loss to follow-up), age, sex, level of instruction, practice setting;

� Intervention level: implementation strategies included in each intervention according to the EPOC Taxonomy (see Additional file 2), framework(s), model(s) or theory(ies) underlining the intervention, clinical topic(s), target clinical practice(s) in nurses, timing (frequency, duration of the intervention), mode of delivery, providers, economic variables (e.g. intervention cost), description of control group(s) intervention(s); � The types—i.e. individual theory items, categories

of theory use—and degree—i.e. total theory use score—of theory use, the mechanisms of action targeted, and the behaviour change techniques included in implementation interventions will be

Table 1 Review stages and review teams involved

Review stages Review teams

Selection of studies Teams A, B, C, D and E

Data extraction Teams A, B, C, D and E

Risk of bias assessment Teams A, B, C, D and E

Theory coding Teams A, D and E

Mechanisms of action coding Teams A, B, C, D and E

Behaviour change technique coding Teams A, B, C, D and E

Fontaine et al. Systematic Reviews (2019) 8:305 Page 5 of 10

identified during a coding phase after data extraction;

� Outcome level: name, time points measured, definition, unit of measurement, scales, validation of measurement tool, missing data, results according to our primary and secondary outcomes, intention to treat (yes/no).

Theory coding We will conduct a theoretical analysis of included stud- ies using an amended version of the Theory Coding Scheme [25]. As Garnett et al. [33] suggested, we re- moved the items ‘quality of measures’ and ‘randomization of participants to condition’ because they relate to methodological issues rather than theory use. The amended Theory Coding Scheme has a total of 17 items (three of which have sub-items) (see Additional file 5). Six review authors in teams of two will code each study independently using QSR International’s NVivo version 12 qualitative data analysis software [34] for spe- cifying if each Theory Coding Scheme item is present (1) or absent (0). We will resolve differences through discussion, and we will involve another review author if a consensus is not reached. Rounds of testing will be performed initially until the inter-rater reliability (IRR) reaches a substantial level of agreement (prevalence-ad- justed bias-adjusted kappa (PABAK) statistic greater or equal to .70 [35, 36]). A total theory use score will be calculated (i.e. the sum of all 17 items and sub-items, which will result in a maximum possible score of 22). A higher score will be indicative of a highest degree of the- ory use.

Mechanism of action coding We will code the mechanisms of action of behaviour change in clinical practice targeted by implementation interventions using coding guidelines from Michie and colleagues [18–20]. We will use the labels and defini- tions of the 26 mechanisms of action listed on the The- ory and Technique Tool (www.theoryandtechniquetool. humanbehaviourchange.org/tool) associated with the three publications mentioned above [18–20] (see Add- itional file 6). Each mechanism of action will be coded as either present (1) or absent (0) in the experimental and comparator interventions. To be coded as ‘present’, the mechanism of action will have to be explicitly men- tioned/used to select or develop intervention techniques (as specified in the item 5 of the Theory Coding Scheme [25]). Mechanism of action coding will be conducted using QSR International’s NVivo version 12 qualitative data analysis software [34]. Ten review authors in teams of two will code each study for mechanisms of action in- dependently, differences will be resolved through discus- sion and we will involve another review author if a

consensus is not reached. Rounds of testing will be per- formed initially until the IRR reaches a substantial level of agreement (PABAK greater or equal to .70 [35]).

Behaviour change technique coding We will use the labels, definitions and examples of the 93 behaviour change techniques included in the Behav- iour Change Technique Taxonomy v1 [21] to code stud- ies for behaviour change techniques. In addition, we will use the coding tool developed by Pearson, Byrne-Davis [37] illustrating behaviour change techniques applied to health professional training. A coding manual and in- structions will be given to review authors. Review au- thors involved in the behaviour change technique coding will complete the Behaviour Change Technique Tax- onomy Online Training (www.bct-taxonomy.com) prior to coding. The training, lasting approximately 6 h, is a resource where researchers can familiarize themselves with behaviour change technique labels, definitions and examples, and learn how to accurately, reliably and con- fidently apply the taxonomy. When review authors iden- tify a behaviour change technique in the experimental intervention or in the comparator intervention, they will code the behaviour change technique as either present in all probability (+) or present beyond all reasonable doubt (++). Behaviour change technique coding will be conducted using NVivo version 12 [34]. Ten review au- thors in teams of two will code each study for behaviour change techniques independently, differences will be re- solved through discussion and we will involve another review author if a consensus is not reached. Rounds of testing will be performed initially until the IRR reaches a substantial level of agreement (PABAK greater or equal to .70 [35]).

Assessment of risk of bias in included studies Ten review authors in teams of two will assess risk of bias independently for each study using the criteria out- lined in the revised Cochrane Collaboration Risk of Bias Tool (RoB 2.0) [38]. Any disagreement will be resolved by discussion or by involving another review author. For individually randomized trials (including crossover trials) and non-randomized controlled trials, we will assess the risk of bias according to the following domains: (1) bias arising from the randomization process; (2) bias due to deviations from intended interventions; (3) bias due to missing outcome data; (4) bias in measurement of the outcome; (5) bias in selection of the reported result. For cluster-randomized trials, we will include an additional domain: (1b) bias arising from identification or recruit- ment of individual participants within clusters. Non- randomized studies will be considered at high risk of bias. We will summarize the ‘risk of bias’ judgments across different studies for each of the domains listed

Fontaine et al. Systematic Reviews (2019) 8:305 Page 6 of 10

using the risk of bias graph and the risk of bias sum- mary. We will not exclude studies on the grounds of their risk of bias but we will report them when present- ing the results of the studies.

Unit-of-analysis issues We anticipate the inclusion of cluster RCTs. Thus, we will evaluate the analysis methods of these studies by de- termining the level of analysis and if statistical correc- tions were used (e.g. generalized estimating equations). We will conduct analyses adjusting for clustering if we observe unit-of-analysis issues by dividing the original sample size by the design effect, as suggested by the Cochrane Handbook for Systematic Reviews of Interven- tions [27]. For studies with multiple intervention groups, we will include each pairwise comparison relevant to this review separately, but with shared intervention groups divided out approximately evenly among the compari- sons [27].

Dealing with missing data We will contact investigators to obtain missing data when necessary. In the case where investigators do not answer our request, data imputation will be performed using the statistical formulas recommended by the Cochrane Handbook for Systematic Reviews of Interven- tion [27] when applicable. In the case where missing outcome data cannot be obtained and data imputation cannot be performed, we will exclude the study for the outcome in question.

Assessment of heterogeneity We will assess heterogeneity by examining the charac- teristics of included studies, the similarities and dispar- ities between the types of participants, the types of interventions and the types of outcomes. We will then use the chi-square statistic and the I2 to assess statistical heterogeneity for analyses including two studies or more within the Review Manager (RevMan) software (version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). For the chi-square statis- tic, we will use a statistical significance level (p value) of 0.10 instead of the conventional level of 0.05, as this test is known to have low statistical power [27]. A statisti- cally significant result will indicate a problem of hetero- geneity [27]. For the I2 statistic, as suggested by Higgins et al. [27], we will interpret the values as follows: 0–40%, might not be important; 30–60%, may represent moder- ate heterogeneity; 50–90%, may represent substantial heterogeneity and 75–100%, considerable heterogeneity.

Assessment of reporting biases We will assess reporting biases using funnel plots if more than 10 studies are included in the meta-analysis

for a specific outcome. We will follow the guidelines re- garding funnel plot asymmetry as described in the Cochrane Handbook for Systematic Reviews of Interven- tions [27]. We will also perform Egger’s regression to further assess a publication bias [27, 39]. Egger’s regres- sion is a linear type of regression between each study standard normal deviate (i.e. mean difference between the groups in a single pairwise comparison divided by its standard error) and its precision (i.e. inverse of the standard error). Egger’s regression will be performed using IBM SPSS Statistics (Version 25, IBM Corpora- tions). An asymmetrical funnel plot at visual inspection and a p value ≤ to 0.05 for the constant of the regression will be considered as indicative of publication bias.

Data synthesis Descriptive synthesis We will synthesize the characteristics of included studies at four levels—i.e. study level, participant level, interven- tion level, outcome level—in table format. We will quan- tify the types—i.e. individual theory items, categories of theory use—and degree—i.e. total theory use score—of theory use, the types, categories and number of identi- fied mechanisms of action, and the type and number of identified behaviour change techniques across studies.

Quantitative synthesis All summary intervention effects estimates will be pre- sented using a random-effects model using a 95% confi- dence interval (CI) as we anticipate clinical and methodological heterogeneity across included studies. For continuous outcomes, we will analyze data using the stan- dardized mean difference (SMD) since it is not expected studies will have the same outcome measures/scales to evaluate implementation. We will ensure that an increase in scores for continuous outcomes can be interpreted in the same way for each outcome, and report where the di- rections will be reversed if this is necessary. For dichotom- ous outcomes, we will pool events between groups across studies using risk ratios and 95% CIs. We will undertake meta-analyses that will compare

changes between intervention and control participants in primary and secondary outcomes only if: (1) the im- plementation interventions, targeted clinical practices and the underlying clinical question are similar enough for pooling to make sense; (2) there is at least two stud- ies available for each outcome of interest. Meta-analyses will be conducted in RevMan version 5.3 software (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) [40]. The significance of the effect sizes will be determined using Cohen’s classifi- cation (< 0.2 = negligible; 0.2—0.49 = small; 0.5—0.8 = moderate; > 0.8 = large) [41]. We will define a statisti- cally significant result by a two-sided alpha of 0.05. If it

Fontaine et al. Systematic Reviews (2019) 8:305 Page 7 of 10

is not possible to conduct a meta-analysis, we will present a narrative summary of the results.

Meta-regression We will undertake random-effects meta-regression ana- lyses if at least 10 studies report enough data to compute a SMD regarding the primary outcome (clinical practice change). We will conduct meta-regression analyses to: (1) examine the association between the Theory Coding Scheme covariates (i.e. individual theory items, categor- ies of theory use and total theory use) with intervention effectiveness; (2) examine the association between type, categories and number of mechanisms of action with intervention effectiveness; (3) examine the association between type and number of behaviour change tech- niques with intervention effectiveness. Meta-regression analyses will serve to investigate unex-

plained heterogeneity in the SMDs between studies. Each study will be weighted in the regression models using the inverse of its variance; studies with the lowest amount of variance will be given a bigger weight in the regression model than those with the largest amount of variance. The association between each variable of interest and the primary outcome will be illustrated in table format where, for each variable, we will report its regression coefficient (B), standard error, 95% CI and statistical significance. Meta-regression analyses will be conducted in IBM SPSS Statistics version 25.0 [42]. Wilson’s SPSS macros will be used to build all regression models [43, 44].

‘Summary of findings’ table and GRADE We will create a ‘summary of findings’ table for the main intervention comparison(s) and include the most important outcomes (e.g. nurses’ behaviour in clinical practice) to draw conclusions about the certainty of the evidence. Two review authors will assess the qual- ity of the evidence independently for each outcome according to the five domains (risk of bias, inconsist- ency, indirectness, imprecision, publication bias) established by the Grading of Recommendations As- sessment, Development, and Evaluation (GRADE) guidelines [45]. Review authors will use the GRADE profiler Guideline Development Tool software (GRA- DEpro; 2015, McMaster University and Evidence Prime Inc.) [46], based upon the data extracted with the data collection checklist.

Subgroup analysis and investigation of heterogeneity We plan to carry out subgroup analyses to investigate heterogeneity when ten or more studies are available in the underlying outcome. If there are a sufficient number of studies, we will explore the following potential effect modifiers:

� Implementation intervention types according to EPOC taxonomy [12];

� Practice setting; � Clinical practice(s) targeted in nurses; � Study design.

Sensitivity analysis We will conduct a sensitivity analysis by excluding stud- ies deemed at high risk of bias. We will also conduct a sensitivity analysis to exclude studies with imputed data.

Discussion and dissemination Results of this systematic review, meta-analysis and meta-regression will inform knowledge users (e.g. practi- tioners, policy-makers) and researchers regarding the ef- fectiveness of practitioner-level implementation interventions in changing nurses' behaviour in clinical practice. In addition, data regarding the theory use, tar- geted mechanisms of action and included behaviour change techniques in studies will be useful for reporting, replicating and synthesizing evidence. Results will be dis- seminated through publications, conference presenta- tions, website postings and interactive knowledge exchange events with key stakeholders. This review has potential limitations. First, this review

will build exclusively on published studies, whereas un- published studies, grey literature and non-peer- reviewed literature will be excluded. Although includ- ing unpublished, grey and non-peer-reviewed literature has potential benefits in terms of comprehensiveness, it can introduce bias in the results of the systematic re- view and meta-analysis. Unpublished studies are usually of lower methodological quality than published studies [47]. Second, we anticipate that outcome measures for nurses’ behaviour in clinical practice will vary signifi- cantly across studies. Thus, we will conduct a meta- analysis using the SMD. This will allow us to standardize the results of studies to a uniform scale be- fore pooling them. However, this method also has downsides since it assumes that the differences in standard deviations among studies reflect differences in measurement scales and not differences in variability among study populations [27]. Review authors deemed the use of the SMD appropriate for this review since it focuses on nurses, minimizing the risk of bias. Third, this review focuses exclusively on practitioner-level im- plementation interventions and their effect on nurses’ behaviour in clinical practice and patient outcomes. Other types of implementation interventions (e.g. finan- cial interventions, patient-oriented organizational inter- ventions, structural organizational interventions, regulatory interventions) may have important effects on nurses’ behaviour in clinical practice. However, we

Fontaine et al. Systematic Reviews (2019) 8:305 Page 8 of 10

believe these interventions differ in scope and deserve their own review.

Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s13643-019-1227-x.

Additional file 1. PRISMA-P ChecklistR1.

Additional file 2. EPOC TaxonomyR1.

Additional file 3. Concept PlanR1.

Additional file 4. PubMed Search StrategyR1.

Additional file 5. Theory Coding SchemeR1.

Additional file 6. List of Mechanisms of ActionR1.

Abbreviations CI: Confidence interval; CINAHL: Cumulative Index to Nursing & Allied Health Literature; CNSs: Clinical nurse specialists; EMBASE: Excerpta Medical Database; EPOC: Effective Practice and Organization of Care; ERIC: Education Resources Information Center; GRADE: Grading of Recommendations Assessment, Development, and Evaluation; IRR: Inter-rater reliability; LPNs: Licensed practical nurses; NCBI: National Center for Biotechnology Information; NPs: Nurse practitioners; PABAK: Prevalence-adjusted bias- adjusted kappa; PRISMA-P: Preferred reporting items for systematic review and meta-analysis protocols; RCT: Randomized controlled trial; RNs: Registered nurses; RoB: Risk of bias; RPNs: Registered Practical Nurses; SCI: Science Citation Index; SPSS: Statistical Package for the Social Sciences; SSCI : Social Sciences Citation Index

Acknowledgements GF was supported by the Vanier Canada Graduate Scholarship (Canadian Institutes of Health Research), and scholarships from the Fonds de recherche du Québec—Santé, the Canadian Nurses Foundation, the Montreal Heart Institute Foundation and Research Center, Quebec’s Ministry of Higher Education, and the Faculty of Nursing at the Université de Montréal. We wish to thank Stéphane Ratté at the Université de Montréal for validating the search strategy.

Authors’ contributions GF conceptualized the study, designed the study, drafted the article and is the guarantor of the review. SC conceptualized the study, designed the study and drafted the article. All other authors helped conceptualize the study, draft and revise the article. All authors read and approved the final manuscript.

Funding Not applicable.

Availability of data and materials No additional data are available.

Ethics approval and consent to participate Not applicable.

Consent for publication Not applicable.

Competing interests The authors declare that they have no competing interests.

Author details 1Faculty of Nursing, Université de Montréal, Montréal, Canada. 2Research Center, Montreal Heart Institute, Montréal, Canada. 3Research Center, Université de Montréal Hospital Center, Montréal, Canada. 4Center for Innovation in Nursing Education, Faculty of Nursing, Université de Montréal, Montréal, Canada. 5Faculty of Nursing, Université Laval, Québec, Canada. 6Research Center, CHU Sainte-Justine, Montréal, Canada. 7Institute of Public Health Research, Université de Montréal, Montréal, Canada. 8Research Center,

Hôpital du Sacré-Coeur de Montréal, Montréal, Canada. 9School of Librarianship and Information Science, Université de Montréal, Montréal, Canada. 10Department of Pharmacy and Health Systems Sciences, Bouvé College of Health Sciences, Northeastern University, Boston, USA.

Received: 16 April 2019 Accepted: 11 November 2019

References 1. World Health Organization. The 2017 update, Global Health Workforce

Statistics. Geneva: World Health Organization; 2017. 2. World Health Organization. Global strategic directions for strengthening

nursing and midwifery 2016–2020. Geneva: World Health Organization; 2016.

3. Wuchner SS. Integrative review of implementation strategies for translation of research-based evidence by nurses. Clinical Nurse Specialist. 2014;28(4): 214–23.

4. Scott SD, et al. Systematic review of knowledge translation strategies in the allied health professions. Implementation Science. 2012;7(70):1–17.

5. Melnyk BM, et al. The state of evidence-based practice in US nurses: critical implications for nurse leaders and educators. Journal of Nursing Administration. 2012;42(9):410–7.

6. Eccles MP, Mittman BS. Welcome to implementation science. Implementation Science. 2006;1(1):1–3.

7. Tricco AC, et al. Effectiveness of quality improvement strategies for coordination of care to reduce use of health care services: a systematic review and meta-analysis. Canadian Medical Association Journal. 2014; 186(15):E568–78.

8. Tricco AC, et al. Effectiveness of quality improvement strategies on the management of diabetes: a systematic review and meta-analysis. The Lancet. 2012;379(9833):2252–61.

9. Ivers, N., et al., Audit and feedback: effects on professional practice and healthcare outcomes. Cochrane Database of Systematic Reviews, 2012(6): p. Art. No.: CD000259.

10. Forsetlund, L., et al., Continuing education meetings and workshops: effects on professional practice and health care outcomes. Cochrane Database of Systematic Reviews, 2009(2): p. Art. No.: CD003030.

11. Cahill, L.S., et al., Implementation interventions to promote the uptake of evidence-based practices in stroke rehabilitation. Cochrane Database of Systematic Reviews, 2017(3): p. Art. No.: CD012575.

12. Effective Practice and Organisation of Care (EPOC). EPOC Taxonomy; 2015. 2015; Available from: https://epoc.cochrane.org/epoc-taxonomy.

13. Hardeman W, et al. A causal modelling approach to the development of theory-based behaviour change programmes for trial evaluation. Health Education Research. 2005;20(6):676–87.

14. Michie S, et al. From theory to intervention: mapping theoretically derived behavioural determinants to behaviour change techniques. Applied Psychology. 2008;57(4):660–80.

15. Presseau J, et al. Using a behaviour change techniques taxonomy to identify active ingredients within trials of implementation interventions for diabetes care. Implementation Science. 2015;10(55):1–10.

16. Nilsen P. Making sense of implementation theories, models and frameworks. Implementation Science. 2015;10(1):53.

17. Michie S, et al. From theory-inspired to theory-based interventions: a protocol for developing and testing a methodology for linking behaviour change techniques to theoretical mechanisms of action. Annals of Behavioral Medicine. 2017;52(6):501–12.

18. Carey RN, et al. Behavior change techniques and their mechanisms of action: a synthesis of links described in published intervention literature. Annals of Behavioral Medicine. 2019;53(8);693-707.

19. Connell LE, et al. Links between behaviour change techniques and mechanisms of action: an expert consensus study. Annals of Behavioral Medicine. 2019;53(8);708-720.

20. Johnston M, et al. Linking behavior change techniques and mechanisms of action: triangulation of findings from literature synthesis and expert consensus. Annals of Behavioral Medicine. 2018. https://psyarxiv.com/ur6kz/

21. Michie S, et al. The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions. Annals of Behavioral Medicine. 2013;46:81–95.

Fontaine et al. Systematic Reviews (2019) 8:305 Page 9 of 10

22. Giguère, A., et al., Printed educational materials: effects on professional practice and healthcare outcomes. Cochrane Database of Systematic Reviews, 2012(10): p. Art. No.: CD004398.

23. Flodgren, G., et al., Local opinion leaders: effects on professional practice and health care outcomes. Cochrane Database of Systematic Reviews, 2011(8): p. Art. No. CD000125.

24. Johnson MJ, May CR. Promoting professional behaviour change in healthcare: what interventions work, and why? A theory-led overview of systematic reviews. BMJ Open. 2015;5(9):e008592.

25. Michie S, Prestwich A. Are interventions theory-based? Development of a theory coding scheme. Health Psychology. 2010;29(1):1–8.

26. Effective Practice and Organisation of Care (EPOC). EPOC resources for review authors. 2017; Available from: http://epoc.cochrane.org/resources/ epoc-resources-review-authors.

27. Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]: The Cochrane Collaboration; 2011.

28. Moher D, et al. Preferred reporting items for systematic review and meta- analysis protocols (PRISMA-P) 2015 statement. Systematic Reviews. 2015;4(1): 1-9.

29. Damschroder LJ, et al. Fostering implementation of health services research findings into practice: a consolidated framework for advancing implementation science. Implementation Science. 2009;4(50):1–15.

30. Veritas Health Innovation Ltd. Covidence [Computer program]. Melbourne: Veritas Health Innovation Ltd; 2019.

31. Liberati A, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Medicine. 2009;6(7):e1000100.

32. Effective Practice and Organisation of Care (EPOC). Data collection form. EPOC resources for review authors. 2013; Available from: epoc.cochrane.org/ epoc-specific-resources-reviewauthors.

33. Garnett C, et al. Reported theory use by digital interventions for hazardous and harmful alcohol consumption, and association with effectiveness: meta- regression. Journal of Medical Internet Research. 2018;20(2):e69.

34. QSR International Pty Ltd. nVivo [Computer program]. Doncaster: QSR International Pty Ltd; 2019.

35. Byrt T, Bishop J, Carlin JB. Bias, prevalence and kappa. Journal of Clinical Epidemiology. 1993;46(5):423–9.

36. Streiner DL, Norman GR, Cairney J, editors. Health measurement scales: A practical guide to their development and use. 5th ed. Oxford: Oxford University Press; 2015.

37. Pearson E, et al. Behavior change techniques in health professional training: developing a coding tool. Translational Behavioral Medicine. 2018.

38. Higgins JPT, et al. A revised tool for assessing risk of bias in randomized trials (RoB 2.0). Cochrane Database of Systematic Reviews. 2016;10:1-52.

39. Egger M, et al. Bias in meta-analysis detected by a simple, graphical test. Bmj. 1997;315(7109):629–34.

40. The Cochrane Collaboration. Review Manager [Computer program]. Copenhagen: The Cochrane Collaboration; 2014.

41. Cohen J. A power primer. Psychological Bulletin. 1992;112(1):155–9. 42. IBM Corporation. IBM Statistical Package for the Social Sciences (SPSS)

[Computer program]. New York City: IBM Corporation; 2017. 43. Lipsey MW, Wilson DB. Practical meta-analysis. Thousand Oaks: Sage

Publications; 2001. 44. Wilson, DB. Meta-analysis macros for SAS, SPSS, and Stata. 2005 [April 6

2019]; Available from: http://mason.gmu.edu/~dwilsonb/ma.html. 45. Guyatt G, et al. GRADE guidelines: 1. Introduction—GRADE evidence profiles

and summary of findings tables. Journal of Clinical Epidemiology. 2011; 64(4):383–94.

46. GRADE Working Group - McMaster University. GRADEpro 2015 [Computer program]. Halmilton: GRADE Working Group; 2015.

47. Egger M, et al. How important are comprehensive literature searches and the assessment of trial quality in systematic reviews? Health Technology Assessment. 2003;7(1):1–76.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Fontaine et al. Systematic Reviews (2019) 8:305 Page 10 of 10