Our scoping review followed a predetermined (but unregistered) protocol that was developed in accordance with the PRISMA-ScR (Preferred Reporting Items of Systematic Review and Meta-Analyses Extension for Scoping Reviews) [14,15] and followed methods used in published peer-reviewed scoping reviews [16]. An exploratory search of 1 database over a 2-year period, conducted in consultation with a medical librarian and mental health experts on our team (GS and CT), confirmed that the studies on EMRs implemented in mental health settings were limited. Therefore, in our scoping review, we also conducted a review of reviews to capture implementation literature across EMRs in all health settings and not just mental health, given that the broad issues around the usability of EMRs in general health settings are potentially relevant. The results of both search strategies were analyzed; for synthesis, we used a combination of results from primary studies and review papers.

In our scoping review, we defined mental health professionals as psychiatrists, psychologists, nurses, and any other health professional involved in treating people with mental health disorders in health service settings, including allied health professionals. These settings could be mental health clinics, or general inpatient or outpatient clinics but needed to be in high-income countries. High-income countries were classified as category 1 countries by the Organisation for Economic Co-operation and Development (OECD) [17]. This criterion was used to maintain relevance to the local policy setting context. Implementation determinants were defined as barriers and enablers that may prevent or facilitate improvements in practice [18], as reported in the included studies. The Proctor framework provides a systematic taxonomy of implementation outcomes (ie, acceptability, adoption, appropriateness, feasibility, fidelity, implementation cost, penetration, and sustainability), distinguishing these from service and patient outcomes [13].

Our scoping review combined two systematic searches; the first captured published studies reporting primary data on the use and implementation of clinician-facing EMRs specifically in mental health settings (henceforth termed “primary studies”). The second search captured published reviews on the use of clinician-facing EMRs as implemented in all health settings irrespective of their relationship to mental health (henceforth termed “reviews”). The searches were conducted in three academic databases (MEDLINE via the PubMed Interface, EMBASE, and PsycINFO) and used the terms outlined in Table 1. Additionally, we manually searched the reference lists of the included studies (primary studies and reviews) for other relevant publications. All searches were limited to studies and reviews published between January 2010 and June 2020. The search strategies were devised by the review team with the assistance of an experienced medical librarian.

In both searches, articles were included if they met the following inclusion criteria: investigated implemented clinician-facing EMRs; conducted in high-income countries (countries classified as category 1 by the OECD [17]); assessed and reported implementation outcomes and contextual determinants of implementation (ie, barriers and facilitators); and published between January 1, 2010, and June 30, 2020. The population and study type inclusion criteria differed between the searches. The review of primary studies included studies related to mental health clinicians, whereas the review of reviews included literature reviews of studies about any health professionals in any health setting.

In both searches, articles were excluded if the implemented EMRs were exclusively patient-facing ones, they did not report on implementation processes or outcomes, or they were not published in English. The complete list of the inclusion and exclusion criteria is available in Multimedia Appendix 1.

Reference details, including abstracts, were downloaded into the reference management software EndNote X8 (Clarivate) [19]; duplicates were removed and the deduplicated list was exported to Rayyan QCRI [20], a systematic reviews web app, for title and abstract screening. Five investigators (HLT, LT, LAE, AG, and IM) independently conducted the two-phase screening process: (1) title and abstract screening and (2) full-text screening. Two investigators (HLT and IM) cross-checked 50% of the records to ensure that article screening was consistent in accordance with accepted practices [21]. Interrater reliability (Cohen kappa coefficient) in this cross-checking indicated strong agreement (>0.8) [22]. A custom data extraction workbook in Excel (Microsoft Corporation) was developed and tested. Data were systematically extracted by six investigators (HLT, LL, LT, AG, LAE, and IM). Four investigators (LT, LAE, YZ, and IM) examined the data for consistency and cross-checked the extracted data against original articles.

Key information extracted included the study publication details (authors, date of publication, country of study, and number of studies in reviews), health settings, study methods (quantitative, qualitative, and mixed methods), design features of EMRs, and implementation barriers, enablers, and outcomes. To ensure consistency in our review, the Proctor framework of implementation outcomes presented in Table 2 was used as the guiding structure, with definitions tailored to suit the EMR implementation context [13].

Primary studies were appraised for quality using a modified Hawker appraisal tool and scoring system [23,24]. This tool was selected as it is designed to review evidence from a variety of methods [23]. The Critical Appraisal Checklist for Systematic Reviews and Research Syntheses developed by the Joanna Briggs Institute was used to appraise the systematic review studies [25]. Two investigators (HLT and IM) appraised 10% of the articles independently to ensure consistency. Quality assessment results were reported to reflect the quality of the studies and reviews included in our scoping review. We did not exclude studies based on quality assessment.

The extracted data were analyzed for common features and summarized into tables. Implementation outcomes were grouped by outcomes (eg, satisfaction), and barriers and enablers were grouped by themes (eg, technical factors). Barriers, enablers, and implementation outcomes were also categorized using the Proctor framework (Table 2), while recognizing some degree of overlap among constructs as suggested by other publications [13,26]. Assignments to the constructs were based on the definitions applied by the review team (Table 2) rather than the assignment made by the authors of the included articles owing to inconsistencies in the manner of defining, measuring, and reporting implementation outcomes [13]. Three investigators (YZ, LAE, and IM) reviewed the assignment of all the results, and any discrepancies were discussed among the three authors until a consensus was reached. Summary statistics (frequencies and proportions) were calculated for the final assignment.

The search for primary studies yielded 1546 results relevant to mental health professionals or settings (Medline: 606; EMBASE: 620; PsycINFO: 320). Manually searching the article reference lists yielded 2 more papers. Among these, 271 duplicates were removed; after title/abstract screening, 1209 papers were excluded because they did not meet the inclusion and exclusion criteria. Furthermore, 68 studies underwent full-text review, and another 56 papers were excluded. We included 12 primary studies for data extraction and synthesis, as shown in Figure 1.

The search for reviews yielded 484 results (Medline: 175; EMBASE: 297; PsycINFO: 12). We identified 4 additional papers by manually searching article reference lists. Then, 73 duplicates were removed, and after title and abstract screening, 377 papers were excluded. Another 27 were excluded after full-text review, and 11 were included for data extraction and synthesis (Figure 1). A total of 23 studies were included for data extraction and synthesis from the 2 searches.

Half of the primary studies (6/12, 50%) were from the United States of America. The remaining primary studies were from Canada (2/12, 16.7%), the United Kingdom (2/12, 16.7%), France (1/12, 8.3%), and Sweden (1/12, 8.3%), as shown in Table 3. Most primary studies were conducted using quantitative methodologies (5/12, 42%), and fewer studies were conducted using qualitative (3/12, 25%) or mixed (4/12, 33%) methodologies. Each of the review studies included publications from several countries; however, in each review, at least more than 50% of the countries were OECD nations, as observed in Table 4. Among these 11 review studies, 1 (9%) focused on mental health settings [27], and the remaining 10 (91%) involved general health settings.

The primary studies scored highly on the modified Hawker appraisal tool [23], with an average score of 30.3 (SD 3.81) out of a possible 36. The reviews scored an average of 7.6 (SD 1.45) out of a possible score of 11 on the Joanna Briggs Institute Critical Appraisal Checklist for Systematic Reviews and Research Syntheses [25] (see Table S1 of Multimedia Appendix 2 for details).

The features of the EMRs were described in 8 of the 12 (66.7%) primary studies in mental health settings. Features ranged from the simple electronic storage of personal and health information documentation [28], e-ordering of consultations [31], and capability to enter free-text notes [30] to features that aimed to improve care quality including embedded assessment tools [32,35,36], and care coordination plans [29,32,38]. Specific examples included implementing automated alerts to develop safety plans for children and youth with suicidal ideations [35] and embedding an e-consultation pathway prompt linking primary health providers with a psychiatrist [31]. Overall, the description of EMR features was limited among the included studies. Four studies did not report on specific features; instead, they simply described the EMRs as storage of clinical notes and test results to improve the accuracy and completeness of clinical information [33,34,37,39]. Only 1 of the 11 (9.1%) reviews provided a comprehensive description of EMRs among the included studies [46] (Table 4).

In mental health settings, the adoption of EMRs is seemingly impacted by technical and organizational factors, as well as by clinician perceptions of appropriateness and acceptability. Clinicians perceived EMRs as acceptable and appropriate if they improved documentation completeness without interrupting workflow. Clinicians tend to value EMRs that support quality of care, are fit for purpose, have readily available technical support, and do not interfere with the clinician-patient relationship. Overall, the body of evidence specific to mental health was small. The implementation determinants and outcomes identified in general health settings aligned with and expanded on the mental health–specific findings. For example, the cost of implementation was identified as an additional barrier to adoption, apart from the technical and organizational factors identified in the mental health literature. However, evidence from general health settings did not consider the unique challenges of implementing EMRs in mental health settings. We have drawn on the evidence from general and mental health settings to make three recommendations for future implementation of EMRs in mental health settings.

Firstly, EMR implementation requires embedded long-term evaluation. In this review, we identified that approximately half of the studies focused on the early‐to-middle stage implementation outcomes (ie, adoption, acceptability, appropriateness, and feasibility) [13], whereas later‐stage implementation outcomes (ie, penetration and sustainability) [13] and implementation costs were rarely evaluated. Fidelity was assessed in only one primary study in a mental health setting. This is in contrast with the implementation research outside of research on EMRs, where implementation fidelity has more often been assessed compared with other outcomes [13]. This may be related to the nature of EMR technologies, which can be tailored and used flexibly to suit particular practices or service needs [50]. Sustainability was also not reported in any of the included studies, a finding that is consistent with implementation research outside of the EMR field where the assessment of program sustainability has been identified as a neglected area [13,51,52]. Limited research on the cost, fidelity, penetration, and sustainability of EMRs suggests limited evaluation and impact assessment, and the lack of long-term goal setting, particularly at the organizational level. We recommend that future implementation of EMRs in mental health settings must include continuous and embedded evaluation to explore long-term outcomes and impacts for health professionals and patients while identifying the determinants of cost, fidelity, penetration, and sustainability. Findings from thorough evaluations are needed to inform the future design, policies, and uptake of EMRs in mental health and other health settings.

Secondly, implementation of EMRs needs to adopt co-design principles and a human factors approach, including clinician participation in formative and summative usability testing prior to and during implementation [53,54]. The successful uptake of EMRs is influenced by clinicians’ perceptions of appropriateness and acceptability. In mental health settings, this was negatively impacted when EMRs misaligned with established workflows. It was also affected by organizational factors such as high staff turnover, low staff buy-in, and low capacity or willingness to change shown by clinicians. Evidence from general health settings suggests that these determinants can be modified by specific facilitating features such as staff training, clinical champions, buy-in from clinicians and leaders, IT support, and, above all, good fit for purpose with minimal disruption to clinical workflows. However, EMRs are commonly designed by IT professionals; although well intentioned, the software is often insufficiently flexible to meet the needs of clinicians at the frontlines of care [55]. Good fit with clinical workflows and local clinical contexts can be achieved through user-centered design processes and collaboration between clinicians and IT professionals [56,57]. Outside of this review and in general health settings, authors have recommended routine use of co-design principles and frameworks, formative evaluations in consultation with clinicians, and frameworks to assess the fit of off-the-shelf EMRs [58]. In the mental health–specific literature covered in this review, co-designing was not analyzed. In future, to enhance the fit of EMRs to the unique and sensitive clinical work undertaken in mental health, we recommend that the development and implementation of EMRs include co-design and formative evaluations to achieve an optimal fit to support usability for clinicians and patient centeredness.

Lastly, the implementation of EMRs needs to be guided by theories and frameworks to successfully navigate behavior change, and interactions between people and technology. In an environment where sensitive issues are addressed and building rapport and trust with patients is especially important, simply “injecting technology” is unlikely to yield better care, experience for health professionals, or successful implementation. For example, in this review, organizational factors such as leadership and culture were the common determinants of EMR implementation [32,41,42]. It is also likely that external factors (eg, health system structure, funding, and governance) impact EMR implementation as seen in other areas of mental health [59]; however, this was not addressed in the included studies. Successful implementation of EMRs requires structured methodology and careful planning, as changes in a social environment often require new skills and can have unpredictable impacts [60]. We recommend that the development, planning, implementation, and evaluation of EMRs could be improved by applying appropriately structured guiding theories and frameworks (eg, behavior change theory [61] or the normalization process theory [62]).

Despite a rigorous search strategy, it is possible that some potentially relevant studies were missed owing to a wide range of terms used to describe EMRs (eg, health information systems and electronic health records). Nevertheless, our search strategy identified over 2000 potential publications across the two search strategies, reflecting its high level of comprehensiveness.

Further, the inconsistent use of implementation outcome terminologies across the literature and some degree of overlap among constructs as suggested by other publications [13,26] made it challenging at times to classify outcomes into the Proctor categories. Although this may have resulted in the misclassification of some findings, they were applied as closely as possible to the Proctor definitions. A robust process where classifications were reviewed by three of the authors (YZ, LAE, and IM) and any discrepancies were discussed until a consensus was reached is a methodological strength supporting our synthesis.

Lastly, although we assessed study quality using validated tools, owing to the limited evidence available, it was not feasible to exclude studies or distinguish findings based on quality. Quality assessment results are described in Table S2 of Multimedia Appendix 2.

The body of evidence about the implementation of EMRs in mental health settings is currently limited. Key enablers of the adoption of EMRs by clinicians in all health settings included clinician buy-in, staff training, IT support, and appropriate fit with the clinical context and workflows. Specific issues identified in mental health settings included limited suitability of the drop-down or checklist options and their impact on clinical workflows and patient-clinician interactions. Future implementation of EMRs could be facilitated through co-design with clinician end users, embedding routine implementation process evaluations, and including routine feedback from clinicians to facilitate adjustments and ensure usability and the best fit with the clinical context and person-centered care. Additionally, it is imperative that future implementations include embedded evaluations to assess long-term impacts on organizations, clinicians, and patients in mental health settings to inform future design, implementation, policy, and funding decisions. Lastly, the implementation of EMRs needs to recognize and address the interplay between the social factors and technical aspects of EMRs as a sociotechnical system to support successful uptake. Future research should consider the application of guiding social theories, implementation frameworks, and consistent use of terminology.