To identify all existing descriptions of CoP groups in the health and business literature, we used the following strategy to search for studies published between 1991 and 2005:

1. Searching electronic bibliographic databases, including Medline, CINHAL, HealthSTAR, EMBASE, ERIC, ECONLIT, AMED, and ProQuest. The search strings for Medline (Additional file 1) were adapted for other databases.

2. Hand-searching key journals, including Journal of Continuing Education in the Health Professions, Medical Education, and, Harvard Business Review.

3. Examining the reference lists of the included articles and books for additional literature.

In addition, we consulted with members of CP Square http://www.cpsquare.org about the search strategy and the review methodology through two teleconferences on 19 and 23 November 2004. CP Square is a 'CoP of CoP' hosted by Wenger and colleagues.

The literature search was conducted in September 2005 by one of the researchers (LL) and a librarian/information scientist (JM). To examined how CoP groups were defined and used, we restricted our search to primary studies involving groups that were either labelled as CoPs or were developed using CoP and/or other related theories (e.g. situated learning, legitimate peripheral learning) as the guiding framework. To examine the evidence on the effectiveness of CoPs in the health sector, we identified articles that: evaluated CoPs for improving health professional performance, health care organizational performance, professional mentoring, and/or patient outcome; and used experimental, quasi-experimental (controlled clinical trials (CCT), interrupted time series (ITS), controlled before-and-after (CBA), or observational designs (before-and-after studies, cross-sectional studies).

The article selection involved a two-phase review. In the first phase, two reviewers (LL and CN) screened the titles and abstracts to identify primary studies that described or evaluated a CoP group. In the second phase, two reviewers, LL and CD (a research coordinator), categorized the included articles into one of five sectors: health care, business, education, information science, and other. All disagreements were discussed, and a third reviewer (MJ) was involved if no consensus was reached.

To understand how CoPs were defined and used in the business and health sectors, our literature review was guided by the meta-narrative technique 78. It began by studying the key theoretical publications, reviews, and critique papers; analyzing the key components of a CoP; and using the information to develop a data extraction form. The form was tested on three health sector articles by four research team members (LL, CD, CM, and MJ) and a collaborator from a research funding agency (PM). The content was subsequently modified to capture the interpretation of the CoP concept, and the development, organization, and activities within CoP groups. The final version included the following categories: the study authors' definition of a CoP; duration of the CoP group; members and their disciplinary backgrounds; methods and frequencies of communication; administrative structure; and statements that described 'community,' 'domain,' and 'practice' as defined by Wenger et al. 3. Data extraction of all health sector studies was done independently by two reviewers (LL and CD). The remaining papers were reviewed by CD, and the data were verified by LL.

We conducted concept analysis to explore the interpretation of the CoP concept and the characteristics of CoP groups 9. The analysis aimed to highlight the similarities and differences in findings across sectors. In this review we focused on three areas: the authors' interpretations of the CoP concept, the key characteristics of CoP groups in primary studies, and the common elements of CoP groups. The characteristics of these groups, reported in primary studies, were summarized in five categories:

1. Why was the group formed?

2. Who was included in the group?

3. How did members communicate?

4. What did the members do or produce, individually or collectively?

5. Where did members interact with each other?

Each sector was reviewed separately, and codes were inductively developed by LL. These codes were uncovered by identifying similarities or differences in phrases, as well as meaningful patterns and processes between and within the different sectors. They were then merged into broader themes. Key reviews and critiques were used to verify the analysis. Throughout the process, the reviewers had frequent discussions and sought input from other team members to identify additional codes and themes. We subsequently discussed the analysis with other researchers with an interest in CoPs for further feedback.

To assess the effectiveness of CoPs, a separate data extraction form was developed to record the following information: number and type of participants, sex, age, the description of CoPs (intervention groups) such as settings and organizational structures, and the description of interventions received by the control group. For each continuous measure, the baseline value and standard deviation were extracted. Also, mean changes from baseline with standard deviations in outcome measures assessed at the end of the treatment period and at the follow-up period were recorded, if available. For dichotomous data, medians and interquartile ranges were recorded at baseline and the subsequent assessments.

Although we were unaware of the number of articles that would meet the eligible criteria, we anticipated extreme heterogeneity among the included studies. Hence, our analysis plan for RCTs, CCTs, and CBAs included calculating standardized effect sizes for the continuous measures, and calculating the number of comparisons showing a positive direction of effect, median effect sizes, and number of comparisons showing statistically significant effects for the dichotomous measures. For the ITS comparisons, the significance of changes in level and slope would be reported. For observational studies, a descriptive summary would be presented.

The term CoP emerged in business literature in the mid-1990s, but articles about social learning and knowledge management had already appeared in journals such as Harvard Business Review as early as the early 1990s 4344. Most (77.8%) of the primary studies were conducted in the US (Additional file 3). The earlier studies focused on apprentice training, but the term was later used to describe a variety of groups, including formal training sessions 13, informal learning groups 141516, multidisciplinary teams 171819, and virtual communities 202122. Most studies cited Wenger 123, but one 23 referred only to Brown and Duguid 45, and one did not cite any of the seminal work 10.

In 1996, Henning studied refrigeration service technicians and documented the information exchange and mentoring that took place during informal gatherings 14. In addition to learning and building a professional identity, new workers gained confidence in making work-related decisions. Similar findings were reported by Attwell on the experience of an apprentice in the train re-servicing industry 10. Harris et al. also highlighted the importance of the interaction with mentors in the workplace, which helped new tradespersons make sense of contradictory information that they learned in the classroom 11.

A prominent characteristic of the business CoPs is a willingness to invest time and resources to facilitate activities for members to socialize. While some groups were encouraged explicitly by employers to connect with others on and off the job 1416172022, others were provided with communication equipment to enable networking 2122. Also, these groups tended to use a range of formal and informal activities. For example, Henning documented the on-the-job meetings and after-work telephone calls among refrigeration technicians 14, Robey illustrated the mix of formal face-to-face meetings and after-work social activities for workers of a soft goods manufacturing company who worked at different sites 22, and Benner described the organized monthly social outing of women working in the information technology companies 16.

Most (92.3%) primary studies of health sector CoPs were from the UK or the US. The term 'community of practice' began to surface in this field in the mid-1990s and was often used as a label for groups and teams, rather than a social learning concept. Learning, sharing information, and identity-building were the major focus of these groups, with situated learning and/or legitimate peripheral participation being the guiding concepts.

In 1995, Jenkins and Brotherton published a series of papers on the use of situated learning in the occupational therapy curriculum 464748. They argued that occupational therapists consolidate their knowledge and skills most effectively while practising in the clinical setting (i.e. a CoP), and recommended early clinical placements as part of the professional training 464748. In a later case study, Lindsay documented the growth of occupational therapy students as they practised applying clinical reasoning skills acquired from a seminar through working with mentors and patients 30. As students gained experience and confidence in the clinical setting, they were advanced to more complex cases. This process, described by Lave and Wenger as legitimate peripheral participation 1, helped to shape students' career goals and identities as occupational therapists.

In nursing, Cope et al. also promoted the use of legitimate peripheral participation as a theory for students to gain skills and professional identity in their clinical placements 5. The term 'communities of practice' began to appear in the medical literature around 2002 when Parboosing published an opinion article discussing the use of CoP groups to facilitate continuing professional development for physicians 49. Also, Winkelman and Choo envisioned a CoP as an intervention for patient empowerment 50.

All the primary studies were published in 2000 or later, and the term CoP was used as a synonym for a group of health professionals who are working together. Some authors even argued that a cohesive multidisciplinary team with a clear sense of identity was a CoP 51. We found that 12 of the 13 primary studies cited Wenger and colleagues' definition of a CoP; however, the actual structure and function of these groups varied greatly. Examples of CoP groups include (Additional file 4): clinical placements where students interacted with and learned from expert practitioners 53052, informal learning groups (e.g. journal clubs 32), health care agency collaboratives that aimed to achieve a common goal (e.g. to improve primary care for older people 33), and virtual communities where practitioners from different sites discussed work-related issues 35363738. Grounded in situated learning and legitimate peripheral participation, studies on clinical placements and apprenticeship tended to focus on students' acquisition of knowledge, skills, and professional identities. However, in groups that focused on information-sharing/-creation, CoP was primarily used as a managerial tool for continuing professional development and improving quality of care, rather than identity development.

Compared to the business sector CoPs, the health care CoPs focus mainly on fostering social interactions at the workplace or during task-oriented activities (e.g. a journal club). Four studies described the use of information technology for members to hold informal discussions and formal meetings 35363853, but we did not find any study that supported off-the-job social outings.

The structures of CoP groups in business and health sectors are summarized in Additional files 5 and 6 respectively. Learning and sharing information through socialization appeared to be the central characteristic of CoP groups. We found all groups demonstrated, to varying degrees, the following characteristics:

1. Social interaction – Interaction of individuals in formal or informal settings, in person or through the use of communication technologies.

2. Knowledge-sharing – The process of sharing information that is relevant to the individuals involved.

3. Knowledge-creation – The processes of developing new ways to perform duties, complete a task, or solve a problem.

4. Identity-building – The process of acquiring a professional identity, or an identity of being an expert in the field.

The knowledge-sharing/-creation CoPs and apprenticeship CoPs emphasized different points, with the latter being focused more on identity-building (e.g. student nurses learning to be a nurse, or new technicians learning to be an expert). Also, it appeared that the mature and cohesive groups tended to include processes that address all four characteristics 14353754, while the newer groups tended to invest more in activities that encourage social interaction and knowledge-sharing, but less in identity-development or knowledge-creation activities. Also, knowledge-creation was rarely a focus in the apprentice training because the goal was to learn existing skills rather than to develop new ones. While the process of knowledge-sharing could be observed in all CoP groups, the benchmarks for the other three characteristics were less clear in the emergent and maturing CoPs.

A number of studies from both sectors highlighted the importance of facilitators, and some linked the success and failure of the CoP to this role 151620243032333435363855. However, the actual responsibilities of facilitators and the organizational support required for this role were less clear in the literature. For example, some facilitators played a distinct role from that of the leader and conducted their activities under the direction of the group and/or the leader 343538, while other groups merged the role of the leader and facilitator 3255. The choice of management structure appeared to depend on the size of the group and the availability of human resources. Which model best suited which type of organization was unclear, but facilitator fatigue was mentioned as something that could lead to the downfall of CoP groups 32.

Ambiguity was observed in the power relationships among CoP members. In the apprenticeship CoPs, the hierarchy of power was usually clearly defined by the roles of mentor-mentee or expert-novice. New practitioners moved from the periphery to a position of full participation as they developed their knowledge and skills by learning from skilled practitioners. Those with full participation would play a greater role, and subsequently had more power to direct the group's activities. In contrast, the power relationship was less clear in the non-apprenticeship CoPs. The inherent assumption was that members of a CoP are naturally collegial, honest, and respectful of each other, and that they put aside their personal agendas for the common good. However, in the non-apprenticeship CoPs, members may not necessarily develop beyond a position of peripheral participation (i.e. they remain as learners/observers rather than contributors), and so learning and negotiation of meaning may continue to be only a reflection of the dominant source of power. This could therefore affect the effectiveness of the group when completing a task or achieving a goal.

One example of people remaining in peripheral participation over the evolution of a CoP group, and therefore of power imbalance, was the multi-stakeholder collaborative in the health sector reported by Gabbay et al. 34. This group was formed to develop health care policies for elder care. Group members participated in scheduled meetings that were organized and facilitated by an experienced librarian. However, despite the facilitator's best efforts, the discussion was often dominated by the opinion and agenda of only a few members. As the group evolved, members like physicians, experienced nurses, and representatives from the health authority were entrusted with more power, and their opinions were valued more by the rest of the group. This subsequently affected the policy development, and some key decisions were based on individuals' experience and preferences rather than the evidence.

CoP research in the health sector focused mainly on the exploration of how people shared information, created knowledge, and built a professional identity in a social setting. Researchers predominantly used in-depth interviews and participant observations (Additional files 3 and 4). Action research methods, in which participants were involved in the development, growth, and evaluation of the group, were also used 333437. In this review, we did not find any paper in the health sector that met the eligibility criteria for the quantitative analysis (Additional files 3 and 4); and so the effectiveness of CoP in this sector remained unclear.