Factors influencing national implementation of innovations within community pharmacy: a systematic review applying the Consolidated Framework for Implementation Research

Eligible studies sought pharmacy staff’s perspectives on barriers and facilitators to implementing national innovations. An innovation was considered a practice, object, or idea perceived to be new to the setting in which it was implemented [18]. Implementation was considered the process by which an innovation was introduced and applied within the pharmacy setting [26, 27]. Studies solely focusing on views of adoption prior to the implementation of an innovation were outwith the scope of this review as this was considered the preliminary decision-making process of the pharmacy staff to use an innovation [18], which was considered conceptually dissimilar to implementation. Studies involving participants from mixed disciplines (e.g. general practitioners as well as community pharmacy staff) were included if it was possible to extract the data solely pertaining to community pharmacy staff perspectives. Studies from any country were considered eligible for inclusion. Qualitative, quantitative, and mixed method studies were included from peer-reviewed journal articles, conference proceedings, poster presentations, and unpublished literature. The exclusion criteria are presented in Table 1.

The databases Medline, EMBASE, PsycINFO, and the Cumulative Index of Nursing and Allied Health Literature (CINAHL) were searched from their inception on 17 December 2015. Unpublished literature was searched within the Open Grey database [28]. See Additional file 1 for the full Medline search strategy. The search was limited to the English language and covered all studies available up until the search date. Supplementary searches were applied from December 2015 onwards until data analysis concluded in March 2017 (Table 2).

Titles and abstracts were screened within Covidence [29], with potentially relevant studies progressing onto full-text screening. The primary reviewer (NW) completed the study selection, with a 20% subset of the title/abstracts and full-texts screened independently. A percentage of agreement was calculated and categorised using the following thresholds: < 70%, poor; 70–79%, fair; 80–89%, good; and > 90%, excellent [30]. A percentage of agreement > 80% was considered adequate [31]. Where the data were published in more than one format, the format which underwent the most extensive peer-review process was included (e.g. a journal article would be selected for data extraction over a conference proceeding).

A data extraction table was devised [24] and piloted in approximately 10% of studies. Piloting identified that delineating the data to barriers and facilitators was oversimplistic as the studies also reported on suggestions of what would have facilitated implementation. These were termed “hypothetical facilitators” and were extracted separately.

Quality assessment tools were used specific to the method(s) employed. This was to ensure that existing purposefully designed tools were used to assess the quality of either qualitative, quantitative, or mixed method studies to comparable depth. They consisted of a series of questions exploring aspects such as the clarity of the aim, appropriateness of the methodology, recruitment of participants, and data analysis. The 34-item Critical Appraisal Skills Programme (CASP) tool was used to appraise qualitative studies [32]. For questionnaire design studies, the Boynton and Greenhalgh Quality Checklist (BGQC) tool was used [26]. The mixed method studies all applied interviews alongside a questionnaire. Existing mixed method quality assessment tools did not assess questionnaires to the same depth as the BGQC tool [33] and would not offer comparable quality assessment. Therefore, the mixed method studies were assessed using the initial screening questions within the Mixed Methods Appraisal Tool (MMAT) [34], which explores the appropriateness of the mixed method approach, with each method then assessed by the CASP or BGQC tool [26, 32].

The quality assessment tools each have screening questions on the clarity of the aim and appropriateness of the research design. Studies were excluded if these initial criteria were not met. Questionnaire studies which used only closed-ended questions were excluded unless based on previous qualitative work or wider literature as the researchers would have introduced bias based on their a priori assumptions of influential factors [35]. To generate the quality assessment result for each study, each question within the quality assessment tool was attributed a score of 2 if the study fully met the criteria, 1 if partially met, and 0 if not met or unclear. The quality assessment results are presented as percentages as not all questions were applicable to every study [32, 36]. The quality assessment results for the mixed method studies were calculated from the lowest scoring method to ensure the final result did not exceed the quality of the studies weakest component [34]. The quality assessment was conducted by the primary reviewer (NW), with clarification from a mediator when required (ED).

The Consolidated Framework for Implementation Research (CFIR) was selected to synthesise the data [37]. The CFIR is a determinant implementation framework of factors influencing implementation amongst five domains: intervention characteristics, the inner setting, the outer setting, characteristics of the individual, and the implementation process [38]. It is widely applied [39], commonly to explore healthcare practitioners’ experiences of implementing an innovation [40], which facilitates cross comparison of results [41]. Each CFIR domain has a number of constructs which are defined in Additional file 2.

A directed content analysis approach [42] was applied where data extraction was conducted inductively, with the synthesis afterwards deductively aligned to the CFIR [43, 44]. This allowed data capture of barriers and facilitators not within the CFIR to test its applicability within the community pharmacy context. As the CFIR constructs are conceptually broad (e.g. one construct is “Knowledge and Beliefs”), data within each CFIR construct was explored for emergent subconstructs [45]. A table quantifying the emergence of barriers, facilitators, and hypothetical facilitators within each CFIR construct was developed, with overarching thematic areas identified from visual analysis [45]. A descriptive narrative synthesis method was chosen to present commonly reported CFIR constructs to facilitate integration of qualitative and quantitative results [46‐48].

To examine the robustness of the results, a sensitivity analysis involved removal of studies with a quality assessment result of < 50% to observe what effect this had on the reporting frequency of the barriers, facilitators, and hypothetical facilitators [49]. As different studies evaluated the same innovation, the results were categorised both by study and by innovation to see how this affected reporting frequency.

Thirty-nine studies were included from the 5874 studies which had titles and abstracts screened (Fig. 1) [6, 50‐87]. The percentage of agreement of the titles and abstracts independently screened was 94% (excellent), and for full-texts was 88% (good).

All studies were published from 2002 onwards, and most since 2010 (n = 28, 72%). Approximately half (n = 20, 51%) originated from the UK. Ten studies originated from other European countries, and the other nine were from Australia (n = 3), Malaysia (n = 2), New Zealand (n = 2), Saudi Arabia (n = 1), and Cambodia (n = 1). The innovation types can be categorised into four subtypes: clinical service (n = 21); pharmacovigilance (n = 6); e-technology (n = 2); and legislative change (n = 10) such as policy changes and reclassification of medicines. Some studies evaluated the same innovation: the UK “Healthy Living Pharmacy” framework (n = 5), the UK “New Medicines Service” (n = 3), the “Danish’ Inhaler Technique Assessment Service” (n = 2), the UK “Medicines Use Review” (n = 3), the Malaysian spontaneous adverse drug reporting system “MADRAC” (n = 2), and the Swedish implementation of ePrescribing (n = 2). Resultantly, the 39 studies report on 28 innovations. Excluding one study which did not provide participant numbers [69], the total number of participants included from the studies is 12,172. Only ten studies (26%) explored perspectives of community pharmacy support staff [52, 54, 57, 61, 63, 69, 70, 75, 82, 83], and two did so exclusively [57, 82]. The full characteristics of included studies are within Additional file 3.

An overview of the quality assessment results is within Table 3, with the full quality assessment in Additional file 4. Five studies were excluded as they applied questionnaires with only closed-ended questions which were not reported to be developed from reference to literature or previous qualitative findings [88‐92].

The reporting frequency of the barriers, facilitators, and hypothetical facilitators aligned to the CFIR constructs is shown in Table 4. A full presentation of constructs and subconstructs is in Additional file 5. When removing studies with a quality assessment score of < 50% (n = 12, 41%), no changes were identified to the most commonly reported CFIR constructs amongst the barriers, facilitators, and hypothetical facilitators. Likewise, when categorising results by innovation and not by study, the most commonly reported CFIR constructs remained the same. Therefore, for completeness, all studies were retained within the analysis.

Thirteen (33%) CFIR constructs were cited by at least ten studies (25%) as a barrier, facilitator, or hypothetical facilitator. Overarching thematic areas spanning across the CFIR domains were identified: pharmacy staff engagement, operationalisation of the innovation, and external engagement (Table 5).

A critical output of this systematic review was the identification of the three overarching thematic areas (Table 5). The identification of these overarching thematic areas allowed for a preliminary causal theory to be developed (Fig. 2). This was not an initial objective of this review as per the initial PROSPERO record (Registration number: CRD42016038876). This preliminary theory proposes that successful national implementation of an innovation requires (1) the pharmacy staff to positively engage with an innovation, (2) the innovation to be easily operational within community pharmacies, and (3) the positive external engagement of patients and other healthcare professionals. The authors view a community pharmacy-specific theory to be necessary due to its distinctiveness when compared to other primary care settings—for example, patients are able to consult with a healthcare professional within community pharmacies without the need for an appointment. Furthermore, the theory presents the high-level influencers of successful implementation within this setting, which may facilitate evaluations if exploration of all 39 CFIR constructs is not feasible.

Further work is needed to understand the interaction between the three overarching thematic areas and to consider how the influence of each may differ within varying contexts and settings. Once tested further, applying this theory at an early stage of an innovation’s implementation process could inform potential refinements to an innovation. As “External Engagement” was an emergent theme, this exemplifies the impact of the surrounding outer context and indicates that the community pharmacy setting is not detached from the wider primary care setting. Wider exploration of the perspectives of other healthcare professionals and the public may strengthen this understanding.

In line with previous work, this review identified that adopted implementation strategies are poorly reported in the literature [14, 21], and the development of innovations was also poorly described. Future studies should explicitly report the implementation strategies being adopted and greater details of the innovation being implemented, including its development, to allow for consideration of how these aspects may influence successful implementation [116, 117].

The emergent CFIR constructs identified in this review complemented those from related reviews [20, 21], which suggests our findings are valid. However, given the methodological differences in the review approaches, this should be interpreted with a degree of caution. Two CFIR constructs cited by Shoemaker et al. and Roberts et al. were not emergent in this review: “Culture” [20, 21] and “Trialability” [21]. As Roberts et al. only reported “culture in pharmacy” and did not elaborate, it was not possible to consider why culture was reported in this previous review and not in the current review. Shoemaker et al. coded “alignment with missions/priorities” and “commitment to providing preventive services” within “Culture”, which the current authors would have coded within “Compatibility” (i.e. how the innovation “aligns with individuals’ own norms and values”) [37]. With regards to “Trialability”, Shoemaker et al. identified that slow expansion of immunisation services facilitated implementation, which would unlikely be salient within scaled-up national services [21]. It was more common for differences to emerge in the valence of the reported CFIR constructs. For example, the current review identified negative patient perceptions (patient needs and resources) and lack of engagement of external healthcare professional (cosmopolitanism), whereas Shoemaker et al. only reported facilitators within these CFIR constructs [21]. This may suggest that emergence of barriers and facilitators at national level is comparable to those in small-scale pilot stages, yet the strength of influence they have on implementation may differ, which would be an interesting area for further research.

From these findings, future implementation strategies (Table 6) for national community pharmacy innovations should involve the use of piloting strategies, promotion of whole-team involvement, pre-implementation exploration, and stakeholder engagement strategies.

Alignment of the results to the CFIR positions the research within the wider implementation science literature. However, the inclusion of qualitative and quantitative studies and the varying level of reporting detail meant it was not possible to weight identified barriers and facilitators and deduce relative influences. The results instead reflected the number of reporting studies, which is not uncommon for reviews of this nature [20, 21, 118‐120]. Primary studies would benefit from applying the CFIR “rating rules” which explore the valence (i.e. positive or negative impact) and strength of influence of emergent CFIR constructs [93]. Nevertheless, tabulation of the results facilitated consideration of the relationship between constructs and cross comparison [45].

The inclusion of qualitative, quantitative, and mixed method studies ensured wide exploration of the barriers and facilitators in this context, which meant that different quality assessment tools were applied. Although tools of similar depth were selected, it is notable that the CASP and BGQC tools covered different aspects of quality [26, 32]. For example, the BGQC tool does not explore the relationship between participants and researchers unlike the CASP tool. Therefore, to what extent the quality assessment results are truly comparable between study types is unknown.

To the authors’ knowledge, using a directed content analysis approach when applying the CFIR is novel and helped assess its applicability to the community pharmacy setting. The CFIR constructs captured most emergent data except for two instances within the CFIR outer setting domain. Patients’ perceptions, awareness, and engagement were not encompassed within the “Patient Needs and Resources” construct as the construct focuses on the ability of an organisation to identify and prioritise patient’s needs (which has been criticised elsewhere [121]). The “Cosmopolitanism” construct overlooks the impact of external healthcare professional’s engagement as it centres on networking with external organisations, and further exploration of how cosmopolitanism may be influenced by the infrastructure of healthcare systems is required. Both of these factors would not appear relevant in any other CFIR domain or construct, and we suggest widening the scope of these CFIR constructs accordingly.

Limiting the search to the English language compromised the ability to identify implementation facilitators and barriers internationally. However, 41% (n = 16) of included studies were from non-English-speaking nations. During full-text screening, 35 studies could not be accessed. As only 8% (n = 29) of the 358 studies screened were included, we estimate only three of these would be eligible for inclusion. Although the study selection process underwent independent validation, alignment of the results to the CFIR was not conducted independently. However, a dedicated CFIR codebook and technical assistance guide was routinely accessed to ensure appropriate alignment to CFIR constructs [93]. Lower quality studies were retained to allow for broad data capture, and removal of the lower quality studies did not affect the representation of the most commonly cited CFIR constructs. Conference abstracts obtained the lowest quality scores—likely to be reflective of their reporting depth—although retaining these ensured representation of the latest research [120].