Communication and Relational Ties in Inter-Professional Teams in Norwegian Specialized Health Care: A Multicentre Study of Relational Coordination

Introduction: The delivery of integrated care depends on the quality of communication and relationships among health-care professionals in inter-professional teams. The main aim of this study was to investigate individual and team communication and relational ties of teams in specific care processes within specialized health care. Methods: This cross-sectional multi-centre study used data from six somatic hospitals and six psychiatric units (N = 263 [response rate, 52%], 23 care processes) using a Norwegian version of the Relational Coordination Survey. We employed linear mixed-effect regression models and one-way analyses of variance. Results: The mean (standard deviation) relational coordination total score ranged from 4.5 (0.33) to 2.7 (0.50). The communication and relationship sub-scale scores were significantly higher within similar functional groups than between contrasting functional groups (P < .05). Written clinical procedures were significantly associated with higher communication scores (P < .05). The proportion of women in a team was associated with higher communication and relationship scores (P < .05). Conclusion: The Relational Coordination Survey shows a marked variation in team functions within inter-professional teams in specialized health-care settings. Further research is needed to determine the reasons for these variations.

and a higher level of organized care. Further, the authors also identified a significant increase in the level of individual competence and perceived "teamness" as a result of implementing clinical procedures. However, Deneckere et al. did not observe any association between communication skills or relational ties in inter-professional teams and the implementation of clinical procedures [6].
Research has produced inconsistent results with regard to the effect of team composition and size on teamwork. No investigations have assessed the relationship among age, use and development of clinical procedures, level of experience, and team functioning.
Owing to the lack of reports assessing the association between team function and relevant individual and team factors, we conducted a study on a range of interprofessional teams to determine the associations among age, use and development of clinical procedures, composition, years of experience in team, team size, and team functioning at both the individual and team level. In the present study, we thus undertook the following. First, within specialized health-care settings, we investigated levels of communication and relational ties in interprofessional teams in specific care processes. Second, we assessed the association between individual-and teamlevel characteristics of inter-professional teams and communication and relational ties in those teams.

Relational Coordination
A recent review identified 10 measurement tools measuring teamwork that meet the criteria for psychometric validity [20]. Those tools survey teamwork functions, such as communication, coordination, shared decision making, collaboration, active conflict management, shared objectives and respect. The Relational Coordination Survey was among the 10 recommended measurement tools [8]. This survey is used within health-care services as well as in primary care, community, and hospital settings; it is a useful tool when measuring the quality of communication and relational ties in inter-professional teams in different parts of health care [22][23][24][25][26].
Relational coordination is defined as a "mutually reinforcing process of interaction between communication and relationships carried out for the purpose of task integration" [27]. Rather than examining relationships among particular individuals, the focus of relational coordination is on relationships among professional groups with similar roles (hereafter, "functional groups") [8]. Relational coordination theory has been found to be a sound framework for investigating care coordination in inter-professional teams [28].
The original Relational Coordination Survey has two sub-scales: a communication sub-scale comprising four survey questions (evaluating the frequency, accuracy, timeliness, and problem-solving nature of communication); and a relationship sub-scale with three survey questions (concerning shared goals, shared knowledge, and mutual respect) [25]. The seven items in the Relational Coordination Survey employ a five-point Likert scale. Respondents are asked to complete each item according to their perception of communication or relationships with specific functional groups of health professionals in their team, e.g., physicians, nurses, physiotherapists, and administrative personnel. This creates a matrix with seven Relational Coordination Survey items for each functional group.
The functional groups included in each team vary according to which types of functional groups are considered relevant for the particular care process under assessment. The scores for the two sub-scales are derived by calculating the mean of the four communication and three relationship scores [29]. Higher scores indicate better communication and relational ties within the interprofessional team.
For use in the present study, the Relational Coordination Survey was translated to Norwegian and piloted on 10 health-care professionals within a hospital by Størkson et al. [30]. An authorized translation agency translated the US-English version of the Relational Coordination Survey into Norwegian language. A research team discussed linguistic and cultural aspects. Minor amendments on the Norwegian Relational Coordination Survey were made due to minor difficulties regarding the interpretation of items and contextual issues before a second authorized translator translated the survey back into English language. This version was found comparable to the original version. This was accepted by the author of the original version of Relational Coordination Survey. A psychometric assessment of the Norwegian version of the Relational Coordination Survey constitutes part of the present study.

Design and participants
This cross-sectional multi-centre study used data from six somatic hospitals and six specialist psychiatric units within the Western Norway Regional Health Authority, constituting 27 care processes in total. The team members (N = 503) received information about the project by e-mail, including a link to the Relational Coordination Survey in Corporater Surveyor, version 3.3 (Corporater Inc., Norway) [30]. In all, 301 health-care professionals (60%) responded. All these participants were used in analyses of the psychometric properties of the Norwegian version of the Relational Coordination Survey.
We inspected the data for inconsistencies and missing items. Respondents with missing items were excluded as follows. First, we excluded individuals who had completed less than 40% of the survey response alternatives (there was one response alternative for each functional group of health professionals) among each of the seven items. Second, we excluded participants if they responded to three or fewer of the seven items. Finally, we excluded respondents in teams with fewer than four valid respondents. That left 263 (52%) participants in the final analysis, representing 23 care processes ( Table 1).

Individual-specific variables
Respondents were asked to report the following information: professional group (registered nurse [somatic], physician, medical laboratory technician, physiotherapist, social worker/occupational therapist/social educator, or Table 1: Overview of team characteristics in 23 care processes included in the valid sample, team size, fraction of women in team, age and professional group distribution and clinical procedure use in teams (N = 263).

Team-specific variables
Based on the individual variables, we defined team variables to characterize the composition of the team: the proportions of (1) women; (2) team members older than 40 years; and (3) physicians in the team and team size. The team was said to have a clinical procedure if ≥80% of team members answered yes or under development to the related question.

Predictor variables
Individual-specific predictor variables for the survey communication and relationship sub-scales, as reported by each professional respondent, were age, sex, use of clinical procedures, and physician in the team. Team-specific predictor variables for the survey sub-scales (summarized for each team) were proportion of women in team, team members >40 years, use of clinical procedures, proportion of physicians, and team size.

Statistical analysis
We employed confirmatory factor analysis (maximum likelihood estimation with robust standard errors, Satorra-Bentler correction) to test the factor structure. To define a satisfactory model fit, we used the following: a cut-off at 0.95 or higher for the comparative fit index; cut-off at < 0.06 to 0.08 for the root mean square error of approximation; cut-off at 0.8 or lower for the standardized root mean square residual; and cut-off at 0.95 or higher for the Tucker-Lewis index [31]. To assess intra-scale consistency, we computed Cronbach's alpha. A construct validity test could not be performed as there were no comparative instruments available for Norwegian health care settings. We tested differences among functional groups (nurses, physicians, therapists/other) with regard to the communication and relationship sub-scale scores by one-way analysis of variance and illustrated by graphical tools. To assess the association between the predictor variables and the sub-scale scores, we used linear regression models with the communication and relationship sub-scale scores as outcome variables.
For the individual variables (age, sex, profession, and use of clinical procedures), we took into account correlations within each team. Thus, we used a linear mixedeffects model, including the individual variables as fixed factors and team affiliation as random effect.
For the team-specific variables, we used a simple linear regression model with the team mean of the sub-scales as outcome and team-specific variables as predictor. We estimated the univariate model for each predictor as well as the multivariate model for the individual variables and team-specific variables. Tests were two-tailed and the significance level was set to 0.05.
Informed consent to participate was assumed when respondents returned a completed survey. Returned questionnaires were de-identified and data were stored according to appropriate regulations. This study was approved by the Norwegian Social Science Data Services in 2012 (reference no. 29128), which, with this type of material, is the relevant body for approval.

Psychometric properties
Previous research has suggested both a one-factor and two-factor approach for the Relational Coordination Survey [8,35]. However, the factor structure of our sample revealed a better model fit with the two-factor structure than the one-factor model [8,35]. Three estimates of fit-comparative fit index, Tucker-Lewis index, and standardized root mean square residual (the latter is independent of the χ² and sample size [32])-showed: 0.86, 0.79 and 0.09 for the 1-factor solution, respectively. Further, the chi-square from the 1-factor solution was 164.8 (p =< 0.001) with 14 degrees of freedom giving a normed χ² of 11.8. For the 2-factor solution the three estimates of fit showed an acceptable fit: 0.93, 0.89, and 0.06, respectively. Further, the chi-square from the 2-factor solution was 84.2 (p =< 0.001) with 13 degrees of freedom giving a normed χ² of 6.48. A chi-square difference test (χ² diff = 83.6, p =< 0.001) suggested that fit was most favourable the 2-factor solution.
Cronbach's alpha for the communication and relationship sub-scales was 0.93 and 0.80, respectively. This estimated intra-scale consistency supported the internal reliability of the measured items in a two-factor structure [36]. Table 2 lists the reported survey scores in each professional's team by different individual characteristics. Among the mean scores in Table 2, there is a trend for higher scores in the relationship than with the communication sub-scale. There are, however, no clear age or sex-related differences.
Individual team members who reported that they used a written clinical procedure on a daily basis or were in the process of developing procedures reported higher communication sub-scale scores than team members who did not use or were not developing such a procedure ( Table 2). Table 3 indicates that using or developing a clinical procedure was significantly associated with higher communication sub-scale scores in the multivariate model (B = 0.20; 95% confidence interval for B, 0.00-0.41; P = 0.049). There was a marginal non-significant result that being a physician was associated with higher relationship sub-scale scores (B = 0.17; 95% confidence interval, 0.00-0.34; P = 0.051).

Team-level associations
With the survey scores for different teams ( Table 4), we found the mean (standard deviation) for communication and relationship sub-scale scores ranged from 4.3 (0.52) to 2.7 (0.34) and 4.5 (0.33) to 3.2 (0.71), respectively. Notably, the communication and relationship sub-scale means were among the highest in teams responsible for stroke patients. We found no clear differences concerning in-or outpatient or somatic or psychiatric care processes.

Discussion
Based on the normed χ², comparative fit index, Tucker-Lewis index, and standard root mean square residual estimates of fit from the confirmatory factor analysis; we conclude that the Norwegian version of the Relational Coordination Survey is acceptable for use in specialized health-care settings employing the two suggested sub-scales of communication and relationship. The chi-square test is perceived inappropriate as it is sensitive to large study populations (above 200) and therefore tends to reject models too often [37]. This conclusion is supported by earlier investigations of the factor structure of the survey employing exploratory factor analyses [8,35].
The use of the Relational Coordination Survey in the included care processes revealed relatively large differences in the quality of teamwork through the survey sub-scales ( Table 4). The better communication and relational ties in these inter-professional teams may reflect an increased effort to improve integrated care for these patient groups. Previous research has shown that implementation of specific inter-professional teams and specific guidelines within stroke rehabilitation have improved patient outcomes [38,39].
At the level of the individual respondent, we observed that being a physician was associated with higher relationship sub-scale scores within teams. This may reflect physicians typically having a central, coordinating role in inter-professional teams in specialized health-care settings in Norway. However, this result is contrary to that of Hartgerink et al. [23]; they found that being a physician  was associated with lower perceived team communication and relational ties. The authors explained this negative association as the result of medical specialists often making their treatment decisions independently of othersand consequently not interacting frequently with other team members.
In the present study, team members in the same profession communicated better with others in the functional group to which they belonged than with members of other functional groups. Inter-professional teamwork has received much attention lately; however, this result may reflect a lack of understanding of different roles and poor communication skills across contrasting functional groups. Furthermore, inter-professional education that includes hands-on inter-professional teamwork practice is not yet fully implemented in all education programmes within health care [40].
Individual team members' development or daily use of a written clinical procedure was associated with significantly higher communication sub-scale scores ( Table 4). This finding may reflect the fact that clinical procedures serve as a coordinating mechanism, assuring necessary levels of communication in inter-professional teamwork [41]. Moreover, Deneckere et al. [6] found that coordination of care and communication improved in interprofessional teams developing clinical procedures. In our study, each team member stated whether or not a written clinical procedure was in daily use. However, the team members reported this information inconsistently. This discrepancy may be explained by respondents' interpretation of the term "clinical procedure": it may be understood differently from one respondent to another [30]; it may also reflect individual respondents being unaware of the existence of a particular procedure. Further, respondents who were aware of the existence of the clinical procedure may not actually have used it in the care process.
A team with a greater proportion of female members was associated with higher communication sub-scale scores. One explanation for this finding may be that women tend to be more oriented towards interpersonal relations and social interactions-and therefore provide higher communication sub-scale scores-than men [42]. Research has shown that nurses were more positive towards collaborating in a team environment than physicians, who traditionally learn to make more independent decisions [23,43]. Another study found a positive relationship between women and the degree of relational coordination [23].

Strengths and limitations
One strength of this study is the collection of data on a wide array of care processes typical for specialized healthcare settings. The inclusion of inter-professional teams from a broad range of clinical areas probably reduced the risk of selection bias. Furthermore, this inclusion increased the reliability and generalizability of the findings. The hierarchical statistical approach (which is appropriate when investigating associations of individual characteristics clustered at the team level) made false-positive findings (type I findings) less likely. However, this study has several limitations. The crosssectional design allowed us to identify associations and characteristics of inter-professional teams in specialized health-care settings but not determine causality.
The median response rate for surveys has declined slightly since 1975 [44]. In the present study, the response rate was acceptable (52%); however, we had limited information on individuals who did not return the survey, for example whether the majority were men or women. Consequently, an inclusion bias cannot be excluded. Further, the number of respondents in each care process is relatively low; results may therefore reflect a coincidental expression of the individual teams' performance rather than cultural differences. However, more studies are needed to clarify these findings further.
Communication and relationships are believed to be different within and between professional groups [45].
By merging specific categories of professional groups in some analyses, we lost the possibility of identifying patterns or levels of responses specific to each of those groups. Further, although the Relational Coordination Survey showed satisfactory psychometric properties in earlier investigations [8,46], we cannot rule out measurement error or issues related to construct validity in the present study.

Conclusion
This study represents the first exploration of interprofessional teamwork using the Relational Coordination Survey in a Norwegian context. The communication and relationship sub-scale scores were significantly higher within unique functional groups than between contrasting groups; this implies there is a need for inter-professional education programmes to enhance the understanding of health professionals' roles and communication skills among team members. Our findings indicate that Future studies should be designed as longitudinal investigations. They should include outcomes at the patient and system level. They should also examine causal aspects of the communication and relationship skills of the Relational Coordination Survey to determine the quality of health-care delivery.