A scoping review of the use of ethnographic approaches in implementation research and recommendations for reporting.

BACKGROUND: Researchers have argued for the value of ethnographic approaches to implementation science (IS). The contested meanings of ethnography pose challenges and possibilities to its use in IS. The goal of this study was to identify sources of commonality and variation, and to distill a set of recommendations for reporting ethnographic approaches in IS.
METHODS: We included in our scoping review English-language academic journal articles meeting two criteria: (1) IS articles in the healthcare field and (2) articles that described their approach as ethnographic. In March 2019, we implemented our search criteria in four academic databases and one academic journal. Abstracts were screened for inclusion by at least two authors. We iteratively develop a codebook for full-text analysis and double-coded included articles. We summarized the findings and developed reporting recommendations through discussion.
RESULTS: Of the 210 articles whose abstracts were screened, 73 were included in full-text analysis. The number of articles increased in recent years. Ethnographic approaches were used within a wide variety of theoretical approaches and research designs. Articles primarily described using interviews and observational methods as part of their ethnographic approaches, though numerous other methods were also employed. The most cited rationales for using ethnographic approaches were to capture context-specific phenomena, understand insiders' perspective, and study complex interactions. In reporting on ethnographic approaches, we recommend that researchers provide information on researcher training and position, reflect on researchers' positionality, describe observational methods in detail, and report results from all the methods used.
CONCLUSION: The number of IS studies using ethnography has increased in recent years. Ethnography holds great potential for contributing further to IS, particularly to studying implementation strategy mechanisms and understanding complex adaptive systems.

The field of IS makes use of research methods and frameworks from various scholarly traditions (Bauer et al., 2015; NIH Fogarty International Center, 2018). The field’s inclusive methodological ethos facilitates innovation and allows implementation researchers to draw on a diverse research toolset (Bauer et al., 2015; Proctor et al., 2009). However, this methodological breadth can also present a challenge to the interpretation of research, particularly when there is a lack of agreement on common meanings and standards in the use of approaches (Newhouse et al., 2013).

In recent years, researchers have argued for the value of ethnographic approaches in IS. Ethnography distinguishes itself from other approaches in health research by its deep engagement in a small number of cases to produce highly detailed data, its focus on human interaction in everyday environments to capture data outside of formal research processes, and its emphasis on building relationships with participants to access insiders’ perspectives, referred to as the “emic” viewpoint (Reeves et al., 2008). These features are potentially well suited to delineating complex implementation processes, revealing contextual factors that affect implementation, and incorporating the perspectives of end-users in implementation, among other benefits (Baumbusch et al., 2018; Bunce et al., 2014; Cohen et al., n.d.; Hamilton & Finley, 2020; Tumilowicz et al., 2015). Implementation researchers have also developed approaches based in ethnography for use in implementation research and practice. Palinkas and Zatzick (2019) developed a rapid approach to clinical ethnography that can be used to understand implementation processes and capture contextual details. Finley and colleagues (2018) proposed a process of periodic reflections based in ethnography for timely development of insights into implementation processes in dynamic contexts. Proposals for the use of ethnographic approaches have also been made more broadly in clinical and healthcare research (Greenhalgh & Swinglehurst, 2011; Huby et al., 2007; Morgan-Trimmer & Wood, 2016; Savage, 2000).

Despite the potential value of ethnographic approaches to IS, differing ideas of what ethnography means present a challenge to their use. Because the meaning of ethnography can vary, it can be difficult to interpret what is meant by use of the terms “ethnography” or “ethnographic” in IS work, and thus to evaluate the rigor and reliability of the approach taken. Modern ethnographic approaches have their roots in anthropology, but what is meant by “ethnography” is contested across and within disciplines (Agar, 2006; Brink & Edgecombe, 2003; Hammersley, 2018; O’Reilly, 2012). Ethnography may refer to a method, such as participant observation, or set of methods, sometimes including quantitative methods. However, it may also be understood as a process involving researcher reflexivity, an epistemological orientation emphasizing interlocutors’ perspectives, or a particular research product (Agar, 2006; Clifford & Marcus, 1986; Hammersley, 2018; O’Reilly, 2012).

Given the diverse notions of ethnography, implementation researchers may employ ethnography in different ways. The objective of this scoping review is to examine the uses of ethnography in implementation research. The goal of this study was to identify sources of commonality and variation and distill a set of recommendations for reporting ethnographic approaches in IS.

Methods

Our research team included scholars with expertise in ethnographic methods and IS, including four trained anthropologists (EPF, JF, ABH, LAP). We chose to undertake a scoping review because of the exploratory nature of our research question. We did not seek to assess the quality of studies, as is typically done in systematic reviews. We followed the approach to scoping reviews delineated by Levac and colleagues (2010). Our primary research questions were as follows: how are ethnographic approaches used in IS, why are they used, and what constitutes ethnographic approaches? In examining what constituted ethnographic approaches, we principally considered what methods were described and how they were used.

Search criteria

We included English-language academic journal articles meeting two criteria: (1) IS articles in the healthcare field and (2) articles that described their approach as ethnographic. For the purposes of conducting searches, we operationalized the first criteria as articles containing the phrases “implementation science,” “implementation research,” “knowledge translation,” or “knowledge to action.” We operationalized the second criteria as articles including variants on “ethnography” (search term: “ethnograph*”) in the title or abstract. In March 2019, we implemented our search criteria in four academic databases: MEDLINE, Web of Science, EMBASE, and ProQuest Health Management. We also conducted a search in the journal Implementation Science for articles that included the term “ethnograph*” anywhere in the article.

Abstract review

Our search produced 210 unique records. Three co-authors (JF, AKG, ADH) screened the abstracts to ensure that they fit our criteria of IS articles in the healthcare field that described their approaches as ethnographic. At this stage, we used Eccles and Mittman’s (2006) definition of implementation research: “the scientific study of methods to promote the systematic uptake of research findings and other evidence-based practices into routine practice” (p. 1). We did not require that articles make explicit reference to the IS field to be considered as implementation research.

We included all articles that described their approach as ethnographic but excluded meta-ethnographies (a method of synthesizing qualitative evidence from multiple studies). We included study protocols in our review. Two co-authors screened every abstract. If there was disagreement between the two co-authors, a third co-author broke the tie. We excluded 92 articles that did not describe their approach as ethnographic, 40 articles that were not considered implementation research, and 5 articles that were not health-related. The excluded articles referenced ethnography and IS, which is why they were captured in our search, but these articles did not characterize their own approaches as ethnographic or did not report on implementation research. After abstract review, 73 unique articles remained for full-text review, of which 21 articles were study protocols (see Appendix 1).

Full text analysis

The 73 included articles were published between 2004 and 2019, with 52 (71%) articles published since 2014. The articles were published across 31 academic journals. The most common journals were Implementation Science, BMC Health Services Research, and Social Science & Medicine.

We iteratively developed a codebook using inductively and deductively developed codes. To understand how ethnographic approaches were used in IS, we coded several article characteristics including characteristics of study design, study objectives, outcome types, and theory, model, or framework (TMF) used (Nilsen, 2015). We coded outcome types as implementation outcomes, service outcomes, and client outcomes following Proctor and colleagues (2011). We developed a simple schema for characterizing articles’ objectives: (1) developing tools for implementation research; (2) assessing implementation determinants (e.g., context, barriers, and facilitators); and (3) evaluating implementation strategies. Articles could be classified as having more than one of these objectives.

To understand why ethnographic approaches were used, we coded articles’ rationales for use of these approaches. We only coded explicit explanations of why ethnographic approaches were used to minimize subjective inferences. To understand what constituted ethnographic approaches, we documented what methods were used and the descriptions of these methods.

Four co-authors (GHC, JF, AKG, IR) coded the same six articles, iteratively revising codes until there was agreement on coding. The final codes were reviewed by the entire study team. These four co-authors then alternatively paired to double-code each article. Coding pairs first independently coded each article and resolved discrepancies by consensus. Any codes that could not be aligned by paired consensus were resolved by the additional coders. Coders also kept unstructured memos noting observations of interest that were not captured by codes. As such, memos did not produce quantitative results but observations on notable practices and distinctions. After coding, the results were summarized and reviewed by the study team to draw out key insights. Based on these results and reviews of existing recommendations for reporting on the qualitative methods, the study team developed recommendations for clear reporting on ethnographic approaches in IS (Tong et al., 2007).

Results

How are ethnographic approaches used in IS?

To understand how ethnography is deployed in IS, we documented the characteristics of IS articles that used ethnographic approaches. Of the 73 articles included, 32 (44%) articles did not report using a specific TMF (Table 1). The remaining articles cited 27 distinct TMFs, with some using multiple approaches. Prominent IS frameworks were the most frequently employed. The most cited TMFs were the Consolidated Framework for Implementation Research (CFIR; Damschroder et al., 2009); Promoting Action on Research Implementation in Health Services (PARiHS; Kitson et al., 2008); Reach, Effectiveness, Adoption, Implementation, Maintenance (RE-AIM; Glasgow et al., 1999); and Diffusion of Innovations (Rogers, 2003).

Table 1.

Characteristics of articles using ethnographic methods in IS.

Characteristics	Number of studies (%)
Theory, model, or framework
None	32 (44%)
CFIR	6 (8.2%)
PARiHS	5 (6.8%)
RE-AIM	5 (6.8%)
Diffusion of innovation	4 (5.5%)
Study type
Assessing determinants	64 (88%)
Testing strategies	20 (27%)
Developing tools/methods	11 (15%)
Outcome type
Implementation	72 (99%)
Service	14 (19%)
Client	14 (19%)
Approach
Multi- or mixed methods	43 (59%)
Hybrid	10 (14%)
Non-hybrid, qual only	30 (41%)
Setting
Healthcare	62 (85%)
Community	16 (22%)
Other	3 (4.1%)
Participants
Providers	66 (90%)
Administrators	40 (55%)
Staff	35 (48%)
Patients and families	30 (41%)
Other	11 (15%)

IS: implementation science; CFIR: Consolidated Framework for Implementation Research; PARiHS: Promoting Action on Research Implementation in Health Services; RE-AIM: Reach, Effectiveness, Adoption, Implementation, Maintenance.

When it came to articles’ objectives, 64 (88%) articles were focused on assessing the determinants of implementation. Some of these articles aimed to better understand contexts in which implementation efforts could take place without seeking to assess a specific intervention. For instance, Charani and colleagues (2017) used observations and interviews to understand how surgical teams made decisions about antibiotic use. Twenty (27%) articles evaluated implementation strategies, nearly all using multi- or mixed-methods approaches. Sax and colleagues (2013) described a protocol for a mixed-methods evaluation of a “work package” of interventions to improve infection control in intensive care units. Eleven (15%) articles developed tools for implementation research. Some of these developed novel methods, as the previously discussed examples led by Palinkas and Zatzick (2019) and Finley and colleagues (2018). Other articles used ethnographic approaches to elaborate theoretical approaches. For example, Jenkins and colleagues (2016) used an approach involving site visits, meetings, and written communication to develop a framework for guiding community-based knowledge translation. Bardosh (2018) used three rapid ethnographic studies to inductively develop a framework for assessing the effectiveness of interventions aimed at neglected tropical diseases.

Seventy-two (99%) articles included implementation outcomes. Many of these sought to identify barriers and facilitators to implementation of interventions in specific contexts. Some articles identified specific implementation outcomes of interest to them. In their study of an innovation to improve dysphagia care, which employed a mixed-methods approach that included use of an “ethnographic field journal,” Ilott and colleagues (2013) clearly identified “adoption and adaptation” as their primary outcomes of interest. Other articles framed their objectives in terms of capturing a particular point of view or aspect of practice. Dainty and colleagues (2016) used interviews and observation to describe the implementation of a post-cardiac arrest consult team “from the participant perspective.” Boaz and colleagues (2016) examined the roles of patients across several quality improvement projects using interviews, observations, reflective diaries, and document analysis (Boaz et al., 2016). Fifteen (21%) articles included service outcomes. Jacobsen and colleagues (2017) used survey instruments to track person-centeredness of care in nursing homes before and after an educational initiative. Fifteen (21%) articles included client outcomes. Mumtaz and colleagues (2016) described a protocol for evaluating the quality of maternal health services that included interviewing women about their care experiences.

Forty-three articles (59%) reported on multi- or mixed-methods research with 10 (14%) of these reporting on hybrid implementation-effectiveness studies (Anguera et al., 2018; Curran et al., 2012). In memos, coders noted that these studies at times employed ethnographic approaches sequentially and at times concurrently with quantitative methods. Some articles reported qualitative and quantitative findings together, while others referenced quantitative findings published separately. Dorsey and colleagues (2015) used a “rapid ethnographic assessment approach” to inform measurements in a subsequent randomized controlled trial of a psychosocial intervention. Clarke and colleagues (2013) drew upon interviews, observations, and document review for process evaluation of a stroke care training intervention. Both articles reported on results of ethnographic approaches separately from quantitative trial results.

Sixty-two (85%) articles described studies in healthcare settings and 22 (30%) in other settings, including homes, places of worship, and social service offices. Sixty-six (90%) articles included clinical providers as participants. Of these, 55 (75%) also included other types of participants, most often administrators and non-clinical staff. Thirty (41%) articles involved patients or their families. In memos, coders noted that articles generally presented settings and participants as planned prior to data collection. By contrast, Shaw and colleagues (2017) described beginning their study of transitional care in a hospital but then following patients into other settings, including patients’ homes, as the study progressed.

Why are ethnographic approaches used?

Thirteen (18%) articles did not provide an explicit rationale for the use of ethnographic approaches (Table 2). For the remaining articles, we grouped rationales for ethnographic approaches into six inductively developed categories, which reflected authors’ stated intent to understand context-specific phenomena (52 articles, 71%); clarify emic (insiders’) perspectives (31 articles, 42%); study complex interactions (28 articles, 38%); examine sociocultural factors (21 articles, 29%); strengthen reliability of findings through use of triangulation (16 articles, 22%); and integrate theoretical models or frameworks (7 articles, 9.6%).

Table 2.

Description of ethnographic methods in IS.

Outcome	Number of studies (%)
Ethnographic rationale
None provided	13 (18%)
Context-specific knowledge	52 (71%)
Emic perspective	31 (42%)
Complex system	28 (38%)
Social norms	21 (29%)
Triangulation	16 (22%)
Theory	7 (9.6%)
Researcher description
No description	40 (55%)
Research position	23 (32%)
Researcher training	22 (30%)
Training and position	12 (16%)
Interviews
Not used	5 (6.8%)
Semi-structured	45 (62%)
Unspecified	17 (23%)
Formal	16 (22%)
Informal	14 (19%)
Conversation	11 (15%)
Structured	4 (5.5%)
Unstructured	1 (1.4%)
Observation
Not used	11 (15%)
Participant	20 (27%)
Non-participant	20 (27%)
Unspecified	24 (33%)
Fieldnotes
Not used	13 (18%)
Unspecified format	55 (75%)
Structured format	5 (6.8%)
Document review
Not used	30 (41%)
Existing	39 (53%)
Study	15 (21%)
Focus group
Not used	48 (66%)
Unspecified type	20 (27%)
Semi-structured	4 (5.5%)
Unstructured	1 (1.4%)

IS: implementation science

Several studies drew on more than one of these rationales. Bunce and colleagues (2014) discussed the value of an ethnographic approach for process evaluation as “placing the intervention in its historical and social context, ‘being there’ to document the process as it unfolds and as interpreted by its participants” (p. 2). In their study of knowledge translation strategies to improve spinal care, Webster and colleagues (2014) explained that, through an ethnographic approach, “the cultural norms, local context and specific needs of various professionals can be explicated when building an account of how policymakers, clinicians, and hospital administrators interact” (p. 2). Dixon-Woods and colleagues (2012) succinctly explained multiple rationales for their ethnographic approach to studying a patient safety program in British hospitals:

Ethnography enables detailed, contextualized descriptions of behavior and of how people make sense of the situations in which they live and work and, consequently, why their own actions make sense (Hammersley & Atkinson, 2007). Ethnography is an especially useful approach to studying patient safety, as it provides an opportunity to observe firsthand how events are classified and communicated in particular ways, as well as the social, cultural, and organizational influences on such classificatory work (Bosk, 2003; Waring, 2009). In particular, ethnography enables insights into how professionals in health care settings account for patient safety issues (Bosk, 2003) in ways that other methods may not detect. (p. 554–555)

Several studies applied ethnographic approaches to reveal processes by which implementation strategies worked or failed to work at specific sites. These studies emphasized the value of ethnography for the study of context and complex systems. Drew and colleagues (2019) used an ethnographic approach to understand divergences in client and implementation outcomes across different hospitals as part of an initiative improve hip and knee replacement surgery in the United Kingdom. McCullough and colleagues (2015) used semi-structured interviews and observations to understand how contextual factors affected differential uptake of evidence-based anticoagulation practices across clinics in a network.

Some studies used ethnographic approaches to mobilize, legitimize, or formalize experiences of researchers and practitioners working in implementation efforts. These studies emphasized the value of ethnography for the study of complex systems and emic perspectives. Reflecting on their role as applied anthropologists working in a program to provide mental health crisis services in New York City, Pope and colleagues (2016) argued that “ethnographically based qualitative research might be used to convert the ‘noise’ of actual implementation process into information with instructive power (Hohmann & Shear, 2002; Rapkin & Trickett, 2005)” (p. 508). English and colleagues (2011) made a similar proposal as they drew on their own experiences to reflect on why an implementation strategy to improve pediatric hospital services in Kenya produced variable effects. Finley and colleagues (2018) formalized this potential of ethnographic approaches to provide knowledge and insights on implementation efforts as they unfold by developing a procedure for guided discussion.

What constitutes ethnographic approaches?

Articles generally described using a distinct set of methods that constituted their ethnographic approaches, primarily interviews and observation. However, we noted substantial diversity in how methods were described. In particular, the meaning of “ethnographic” was not always articulated.

Sixty-eight (93%) articles used interviews as a part of their methods (Table 2). Forty-five (62%) articles described interviews as semi-structured, four (5.5%) as structured, and one (1.3%) as unstructured. Sixteen (22%) articles described interviews as formal and 14 (19%) as informal. Eleven (15%) articles described using conversations as part of their methods. Seventeen (23%) articles used interviews without describing the type of interview conducted.

In memos, coders noted variation in the level of detail with which articles described interviews or explained how interviews contributed to an ethnographic approach. Sobo and colleagues (2008) used semi-structured interviews as a part of their “focused ethnographic assessment” of an HIV testing intervention for US veterans. They explicitly described what was meant by their “ethnographic” interview approach:

Interviewers adopted techniques designed to elicit information that interviewees themselves deemed important and to expose understandings existing below any official discourse (Campbell & Gregor, 2004). For example, interviewers adopted an “interested listener” rather than a dictatorial role (Quinn, 2005, p. 41). They sought to avoid collusive conversational turn-taking and gapfilling in which implicit meanings are assumed to be shared (Campbell & Gregor, 2004). (p. 445)

Sixty-two (85%) articles described using observation as part of their methods. Of these, 20 (27%) described observation as participant, 20 (27%) as non-participant, and 24 (33%) did not specify. As with interviews, coders noted in memos variation in the level of detail with which observational methods were described. In several articles, the only mention of an ethnographic approach was to describe the use of “ethnographic observations” without elaborating on what this meant. By contrast, Latif and colleagues specified the constructs and variables researchers attempted to document through observations in their study of an intervention to improve patient adherence across pharmacies in England. They also provided an explanation of how observation contributed to their ethnographic approach: “Through firsthand observation and direct engagement, organizational ethnography offered an in-depth or ‘thick’ description of the social, through which analysis of the social-cultural and organizing context was possible” (p. 969).

Many articles described use of fieldnotes in conjunction with observation. In all, 60 (82%) articles described using fieldnotes. Some articles merely noted that fieldnotes were used, while others described the contents of fieldnotes and how they were used. In a protocol for the study of the implementation of a federal Canadian mental health strategy, Park and colleagues (2015) described how trained observers would keep fieldnotes of events such as team meetings that would capture contextual factors and personal interactions. They also described how observers and other research team members would reflect on fieldnotes as part of their planned data analysis.

A variety of other methods were also used as part of ethnographic approaches. Forty-three (59%) articles described analyzing documents as part of their methods. These primarily involved reviewing existing documents from study sites, such as written communications and protocols. Twenty-five (34%) studies used focus groups. Other methods were used less frequently. Zobrist and colleagues (2017) used cognitive mapping in their study of young child feeding interventions in Senegal. Patel and colleagues (2014) described a plan to use video ethnography in their study protocol for an initiative to improve cardiovascular risk management in primary care clinics.

Articles generally provided little information on researchers’ training and positionality (role and social position in relation to participants) as part ethnographic approaches. Forty (55%) articles did not provide any description of the researchers involved in ethnographic data collection. Only 12 (16%) of articles mentioned researchers’ training in use of ethnographic methods and relationship to the study’s setting. When descriptions were provided, they were often minimal. By contrast, Dlamini-Simelane and Moyer (2017) detailed Dlamini-Simelane’s participation as an HIV counselor as part of data collection for their study of nurse-led HIV treatment in Swaziland, reflecting on the insights that could be gained in this role.

There was lack of clarity in how long researchers were engaged in data collection. Fifty-six (77%) articles provided some information on study duration. However, there was substantial variation in how duration was reported. Some studies reported the start and end dates of data collection, while others reported number of site visits or hours of observations performed. Of the studies reporting on length of data collection, the shortest period was 2 weeks and the longest was 5 years (Bunce et al., 2014; Dorsey et al., 2015).

Coders noted in memos that studies primarily described results from interviews, less often describing results from observational methods, such as descriptions of settings or events. Coders also noted that studies often described results without referencing larger sociohistorical context or reflecting on the positionality of researcher regarding the research subject, in contrast to how ethnographic results are often presented in social science fields. One exception was Dixon-Woods and colleagues’ (2012) study of a program to reduce catheter-related infections in British hospitals. These researchers provided detailed descriptions of how staff responded to the program, placing these reactions within historical context of the British healthcare system.

With regard to analyzing study results, 64 (88%) articles used some form of data coding to identify themes, at times using formal named approaches (e.g., qualitative content analysis, framework analysis, matrix analysis) and at times describing more ad hoc approaches (i.e., developing themes through reading and discussion without a specific or formal process). Studies that did not use data coding either did not explicitly describe an analytic approach or drew on other approaches to the analysis of ethnographic data. In their study of how nurses use research in a pediatric critical care unit, Scott and colleagues employed Fetterman’s two-phase approach to ethnographic analysis involving “making order of data” followed by interpretation (Fetterman, 1998; Scott et al., 2008).

Discussion

Our findings suggest that ethnographic approaches are well-accepted in IS. The number of implementation research articles reporting use of ethnographic approaches has increased substantially in recent years, with many articles published in prominent journals. In addition, the use of ethnographic approaches appears compatible with a wide array of TMFs in IS. Ethnographic approaches can be deployed in a variety of mixed and multi-methods designs, including as a part of hybrid implementation-effectiveness studies.

We found that interviews and observations were the most frequently used methods in the articles we reviewed. However, we noted substantial variation in the detail with which these approaches were described. In some studies, the only mention of ethnography was to describe observations as “ethnographic.” This finding suggests “ethnography” is at times used to legitimize observational methods in IS. We also found a few articles reported using informal interviews and conversations. Such informal interactions could reveal valuable insights that are not as easily appreciated in formal interviews.

Ethnographic approaches were primarily used to assess implementation outcomes in healthcare settings. A minority of articles reported data collection with patients or in community settings. This finding is not surprising given that implementation efforts are typically aimed at clinical professionals. That said, ethnographic approaches are well suited for multisite research and capturing perspectives from diverse actors. Ethnographic approaches also allow for incorporation of new sites and participants as studies progress. These may represent underutilized possibilities of ethnographic approaches in IS.

The leading reasons given for using ethnographic approaches among the included articles were studying context-specific phenomena, capturing emic perspectives, and studying complex interactions. These rationales are consistent with proposals for the use of ethnographic methods in healthcare research (Huby et al., 2007; Savage, 2000). In particular, we noted that ethnographic approaches were used to detail implementation processes and formalize the experience of individuals involved in implementation as research results.

The rationales for use of ethnographic approaches in IS suggest that collaborative approaches to ethnography may be especially valuable for studying adaptations to interventions and implementation strategies (Stirman et al., 2019), “mechanisms” by which implementation strategies succeed or fail (Lewis et al., 2018, 2020; Powell et al., 2019; Zuckerman et al., 2017), and complex adaptive systems (Braithwaite et al., 2018; Reed et al., 2018). The iterative nature of ethnography makes it well-suited to investigating emergent phenomena that are a feature of such systems, especially as interventions and strategies interact with contextual factors.

An additional possible contribution of ethnography to implementation is increased reflexivity and awareness of positionality (Barry et al., 1999; Bikker et al., 2017). There is growing recognition in academic healthcare for the ways systems of oppression (e.g., racism, sexism, and colonialism) operate within and through healthcare institutions (Evans et al., 2020; Horton, 2019; National Academies of Science Engineering and Medicine, 2018). Especially, given that underrepresented minorities are often excluded from academic positions, researchers must critically address their positionality in relation to systems of oppression when conducting research (Association of American Medical Colleges, 2019; Blackstock, 2020; McKay, 2018; Sufrin, 2015).

Some scholars have argued that research is only “ethnographic” if it employs a particular theoretical orientation, not merely a distinct set of methods (Jowsey, 2016; Lambert, 2002; Waring & Jones, 2016). In conducting this review, we did not seek to reconcile tensions between traditional ethnographic approaches and more structured or abbreviated approaches used in health fields. Rather, our intention was to promote clear communication and thoughtful application of ethnographic methods, so that such tensions may be productive sources of critique and innovation, rather than merely sites of discord or misjudgment. Toward that end, we developed reporting recommendations.

Recommendations for reporting

Rather than develop criteria for reporting on ethnographic methods de novo, we sought to build on existing recommendations. The Consolidated Criteria for Reporting Qualitative Research (COREQ) is a checklist intended to improve reporting of qualitative research in health sciences (Tong et al., 2007). It is the most cited of the EQUATOR Network’s key reporting guidelines for qualitative research (Centre for Statistics in Medicine, n.d.). COREQ consists of 32 items separated into three domains: (1) research team and reflexivity, (2) study design, and (3) analysis and findings. Our study team agreed that these criteria were useful for the reporting on ethnographic methods. However, we noted several areas where additional information and emphasis could improve clarity in reporting (Table 3).

Table 3.

Recommendation for reporting on ethnographic approaches in implementation research.

COREQ domains	Recommendations for ethnographic approaches
Research team and reflexivity	• Provide Information on researchers’ experience and training in ethnographic methods• Clarify which researchers participated in data collection and to what extent• Provide detailed information on researcher’ position with respect to research site and participants• Report on how researchers presented themselves to participants. including measures taken to earn participants’ trust
Study design	• Consider whether describing approach as ethnographic adds clarity or confusion• Explain what is meant by ethnographic approach and why it Is used• Consider employing and referencing specific approaches to rapid or focused ethnography if appropriate• Provide detail on how observational methods were employed. including their goals and who employed them• Report on relevant informal and unstructured activities• Report on how approach evolved during course of study
Analysis and findings	• Describe results from all methods used including observational• Explicit data coding is not required but an approach to theme identification should be described• Explicitly incorporate reflexivity in analysis• Consider sociohistorical context in results and analysis if appropriate

COREQ: Consolidated Criteria for Reporting Qualitative Research.

Domain 1: research team and reflexivity

COREQ’s first domain includes criteria on researcher credentials (Item 2), occupation (Item 3), gender (Item 4), experience and training (Item 5), and relationship with participants (Items 6–8). Reflecting on researchers’ characteristics and relationship to studies can add substantially to ethnographic approaches (Barry et al., 1999; Bikker et al., 2017). As Dixon-Woods (2003) has argued, “researchers are required to be reflexive—that is, to reflect on and be able to give an account of how they produced their interpretations—and to be able to show that their interpretation is warranted by the data” (p. 326). Information on researchers’ training can improve readers’ confidence in the execution of methods. Explanations of researchers’ positionality can improve readers’ understandings of how studies were conducted and what insights were available to researchers.

We found that a few articles in our review included information on researchers’ training and relationship to study sites and participants. We recommend that articles delineate researchers’ training and experience in ethnographic methods, including any training provided for completion of the study. We also recommend providing detailed information on researchers’ relationship to the study site and participants, which may go beyond the items explicitly listed in COREQ. Discussing researchers’ social position and characteristics in relation to study sites and participants may be relevant depending on the study’s context and aims.

In the IS field, clarifying how researchers were perceived by participants and their existing roles in organizations is especially important. Implementation efforts may be perceived as burdensome top-down initiatives by some staff. In these situations, staff may withhold their honest views if they perceive researchers as surveillance agents for organizational leadership. Researchers can contextualize these dynamics by clarifying their position within organizations and explicating their relationship with participants. Delineating any strategies that were used to gain participants’ trust is recommended.

Domain 2: study design

COREQ’s second domain includes criteria on methodological orientation and theory (Item 9), participant selection (Items 10–13), setting (Items 14–16), and data collection (Items 17–23). In our review, we found substantial diversity in articles’ study designs, with many using mixed- or multi-method designs. This diversity makes comprehensive reporting of study designs crucial.

Beyond the existing items in COREQ’s Domain 2, we recommend that researchers explain why an ethnographic approach is appropriate for their research question, clarifying what they mean by “ethnographic” and, if relevant, the relationship between their theoretical approach and their use of ethnographic methods. If appropriate, we recommend that researchers consider applying and referencing approaches to focused or rapid ethnography when planning and reporting studies (Higginbottom et al., 2013; Knoblauch, 2005; Palinkas & Zatzick, 2019; Wall, 2015). Such approaches often provide clear guidelines for their use and are typically understood to be distinct from conventional ethnography, which may help avoid ambiguity (Knoblauch, 2005; Wall, 2015).

Given the contested meanings of ethnography, we recommend that researchers consider whether referencing ethnography adds or detracts from understanding their approaches. We propose that the invocation of ethnography merely to justify the use of observation in methods is unnecessary. Describing observational methods as ethnographic is rarely clarifying, given the breadth of meanings associated with ethnography that we observed. Rather, researchers may focus on detailing how observational methods were employed, by whom, and what their goals were. Similarly, researchers merely seeking to convey that their studies were conducted in “real world” settings or during routine activities may describe their work as in situ rather than as ethnographic.

Ethnographic approaches often involve informal interactions in mundane activities that can be difficult to convey in a formal description of methods. Clifford Geertz (1998) famously theorized ethnographic work as “deep hanging out.” Few articles in our review conveyed such unstructured immersive approaches, possibly because they were not used or because researchers were concerned that these would not be perceived as rigorous. We encourage researchers to describe informal, unstructured, and unplanned interactions that contributed meaningfully to data collection and provide important insights. Phenomenological approaches may prove useful to implementation researchers who wish to engage with such unplanned or unstructured research experiences (Cohen et al., n.d.; van Manen, 2016).

An advantage of ethnographic approaches is that they are commonly iterative. Study settings and participants can evolve as researchers gain knowledge and insights (Agar, 2006). We recommend that researchers explicitly delineate how their planned approaches evolved over the course of the study and why. Such descriptions can provide valuable information for readers. Clear descriptions of such iterative approaches may also contribute to evolving ethnographic methodologies and study designs in IS.

Domain 3: analysis and findings

COREQ’s Domain 3 includes criteria related to data analysis (Items 24–28) and reporting (Items 29–32). COREQ presumes that researchers will use some approach to data coding leading to theme derivation in analyses. While ethnographic analyses do generally involve identification of themes, they may not involve explicit data coding (Agar, 2006). Regardless of how themes are derived, we recommend that researchers’ reflexivity be explicitly considered in analyses, given its central importance to ethnographic approaches (Barry et al., 1999; Bikker et al., 2017). We similarly suggest that researchers present results within their sociohistorical context if it will improve the understanding of their findings.

Given the diversity of study designs within which ethnographic approaches are used in IS, differing approaches to presenting results are to be expected. COREQ emphasizes the importance of presenting quotations in results (Item 29). We additionally recommend that researchers present results from all methods used as a part of ethnographic approaches, including observational methods. Doing so may include descriptions of settings or events that were noteworthy. Use of digital media, including audio, photographs, and video, may also be used to convey researchers’ observations (Underberg & Zorn, 2013).

Researchers seeking to follow our recommendations may encounter structural barriers from journals in the form of word limits or formatting requirements. Notably, sociology and anthropology journals, where ethnographic research has traditionally been published, have word limits in the range of 9,000–15,000, rather than 2,000–4,000 that is typical of health journals, and do not dictate the article structure. We encourage health journal editors to allow longer article lengths for qualitative and mixed-methods work. Editors may consider allowing longer versions of methods and results for online publication with abbreviated print versions.

In some instances, our recommendations may suggest a blurring of traditional article structures and headings. Researchers may need to preview results if they describe how methods iteratively evolved or preview discussion topics if they present their results within a sociohistorical context. We encourage researchers and editors to be open to flexible presentations of studies if these convey more faithfully iterative, recursive, and reflexive approaches. We recognize that article length and structure guidelines are unlikely to change in the near term. We encourage researchers to seek creative ways in which to communicate their ethnographic and mixed-methods approaches within existing parameters. This may include using supplemental materials sections or publishing separate study protocols that provide additional methodological detail.

Limitations

As a scoping review, our study provides an exploratory rather than a comprehensive view of ethnography in IS. Our analyses are likely influenced by our disciplinary backgrounds, though we sought to recruit a multidisciplinary team and critically challenge our implicit notions through iterative coding and discussions. Our recommendations are aimed at improving clarity and comprehension in communicating ethnographic approaches, but we did not make suggestions for what ethnography is or how it should be done in IS.

Conclusion

Ethnographic approaches to IS may allow researchers to gain insights that would not be available through other methods, particularly on the interactions between implementation processes and context as well as insiders’ views of implementations efforts. Comprehensive and harmonized reporting approaches could improve the understanding of ethnographic approaches, enhancing its value in IS.