This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
S1 Table: Summaries of Interventions. Summaries of each study and intervention included in the meta-analysis is provided in the S1 Table.
GUID: 79D00175-E12C-4E2E-9125-1D80ED4C8F10S1 File: PRISMA Checklist. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Checklist [82] for this review is presented in the S1 File.
GUID: 45FE14FC-D821-4D53-9BE1-BBEB5FB136D9All relevant data are within the paper and its Supporting Information files. Raw data (taken from the studies in our meta-analysis) are available upon request from the corresponding author.
The objective of this study was to conduct a systematic review and meta-analysis of teamwork interventions that were carried out with the purpose of improving teamwork and team performance, using controlled experimental designs. A literature search returned 16,849 unique articles. The meta-analysis was ultimately conducted on 51 articles, comprising 72 (k) unique interventions, 194 effect sizes, and 8439 participants, using a random effects model. Positive and significant medium-sized effects were found for teamwork interventions on both teamwork and team performance. Moderator analyses were also conducted, which generally revealed positive and significant effects with respect to several sample, intervention, and measurement characteristics. Implications for effective teamwork interventions as well as considerations for future research are discussed.
From road construction crews and professional soccer squads to political parties and special operations corps, teams have become a ubiquitous part of today’s world. Bringing a group of highly-skilled individuals together is not sufficient for teams to be effective. Rather, team members need to be able to work well together in order for the team to successfully achieve its purposes [1, 2]. As a result, there has been a proliferation of research assessing whether, and how, teams can be improved through teamwork training. A wide range of studies have shown positive effects of teamwork interventions for improving team effectiveness across several contexts such as health care (e.g., [3]), military (e.g., [4]), aviation (e.g., [5]), and academic (e.g., [6]) settings. Similarly, improvements in teamwork have been observed as a result of training with a variety of team types including new teams (e.g., [7]), intact teams (e.g., [8]), and those created for laboratory-based experiments (e.g., [9]). In sum, the extant empirical evidence to date appears to suggest that teams can be improved via teamwork training.
Within teams, members’ behaviors can be categorized in terms of both taskwork and teamwork processes [2]. Marks et al. [10] differentiated between the two by suggesting that “taskwork represents what it is that teams are doing, whereas teamwork describes how they are doing it with each other” (p. 357). Specifically, while taskwork involves the execution of core technical competencies within a given domain, teamwork refers to the range of interactive and interdependent behavioral processes among team members that convert team inputs (e.g., member characteristics, organizational funding, team member composition) into outcomes (e.g., team performance, team member satisfaction) [2, 10]. Some examples of teamwork (and respective comparisons to taskwork) include: the seamless communication between a surgeon, nurse, and anaesthesiologist, rather than the technical competencies of these practitioners; the synergy between a quarterback and receiver to complete a passing play, rather than their respective skill sets related to throwing or catching a football; the collaborative adjustments a flight crew makes in response to adverse weather or system problems, rather than each individual’s aviation skills; and so forth. Research from an assortment of studies indicates that teamwork—the focus of the current paper—is positively related to important team effectiveness variables, including team performance, group cohesion, collective efficacy, and member satisfaction [1].
Teamwork has been conceptualized within several theoretical models. For example, in their review, Rousseau et al. [2] reported that 29 frameworks related to teamwork have been published. Although there is much overlap across these models, there are also some notable differences. These relate to the number of dimensions of teamwork being conceptualized as well as the specific labelling of these dimensions. One thing that is generally agreed upon, however, is that teamwork is comprised of multiple observable and measurable behaviors. For instance, two highly cited frameworks by Marks et al. [10] and Rousseau et al. [2] consist of 10 and 14 dimensions of teamwork, respectively. In general, teamwork models focus on behaviors that function to (a) regulate a team’s performance and/or (b) keep the team together. These two components coincide with the two respective processes that Kurt Lewin, the widely recognized father of group dynamics, originally proposed all groups to be involved in: locomotion and maintenance [11].
With regard to regulating team performance (i.e., locomotion), teamwork behaviors include those that occur (a) before/in preparation for team task performance, (b) during the execution of team performance, and (c) after completing the team task [2]. First, with regard to teamwork behaviors that occur before/in preparation for team task performance, these include the active process of defining the team’s overall purpose/mission, setting team goals, and formulating action plans/strategies for how goals and broader purposes will be achieved. These behaviors help ensure that all team members are clear in terms of what is required of them in order for the team to function effectively. Second, teamwork behaviors that occur during the execution of team tasks include actions that correspond to members’ communication, coordination, and cooperation with each other. At this stage, team members translate what they have previously planned (during the preparation phase) into action. Third, in terms of teamwork behaviors that occur after completing the team task (i.e., reflection), these include monitoring important situations and conducting post-task appraisals of the team’s performance and system variables (e.g., internal team resources, broader environmental conditions), solving problems that are precluding team goal attainment, making innovative adjustments to the team’s strategy, and providing/receiving verbal and behavioral assistance to/from teammates. Hence, team members determine whether their actions have moved them closer towards accomplishing the team goals and objectives, and whether any modifications are required in order to facilitate future success. In addition to these three dimensions concerned with the regulation of team performance, a fourth dimension of teamwork involves behaviors that function to keep the team together (i.e., maintenance). These behaviors focus on the team’s interpersonal dynamics, and include the management of interpersonal conflict between members and the provision of social support for members experiencing personal difficulties. Managing interpersonal dynamics is critical as it is theorized that teams cannot operate effectively when these issues are present [2].
Teamwork interventions have utilized a number of training methods in order to target the regulation of team performance (i.e., preparation, execution, reflection) and management of team maintenance (i.e., interpersonal dynamics) dimensions. These intervention strategies generally fall under one of four categories. First, the most basic approach to training and developing teamwork involves providing didactic education to team members in a classroom-type setting, such as lecturing about the importance of providing social support within the team or promoting ways to manage interpersonal conflict among teammates. This type of training has been found to be useful for enhancing team effectiveness (e.g., [12]). A second category of team training involves utilizing a more interactive workshop-style format, wherein team members take part in various group activities, such as having discussions about the team’s purposes and goals (e.g., [13]) or working through case studies together (e.g. [14]). The third broad category of team training involves simulation training, wherein teams experientially enact various teamwork skills, such as interpersonal communication and coordination, in an environment that mimics upcoming team tasks (e.g., airline simulators or medical patient manikins). Although often used as a means of fostering taskwork competencies (e.g., teaching new surgeons how to perform the technical skills of a medical operation), simulation training has been found to be an efficacious approach to teamwork intervention (e.g., [15]). In addition to these three training approaches that occur outside of the team task environment (i.e., training within classroom and simulation settings), teamwork can also be fostered by incorporating team reviews in-situ (i.e., where the team actually performs its tasks), which allows teams to monitor/review their quality of teamwork on an ongoing basis. These team reviews involve some form of team briefs before (e.g., creating action plans), during (e.g., monitoring team members’ actions), and/or after (e.g., assessing the team’s performance) team task execution, and have also been shown to be efficacious in previous studies (e.g., [16]).
The effectiveness of teamwork interventions can be determined with an assortment of criteria, including team- and individually-based behaviors, cognitions, and affective states. Hackman and Katz 2010 [17] posit that team effectiveness can be determined by examining the extent to which the team has achieved its a priori objectives. Since the broad purpose of forming a team is to produce something of value, it is perhaps unsurprising that the most widely tested criterion of team effectiveness has been team performance [18–20]. Thus, although teams come from an array of settings and are idiosyncratic in their own ways, one question that essentially all teams address at some point during their tenure is whether they are performing well. For example, is that road construction crew fixing potholes adequately? Does the local soccer squad have a respectable winning percentage? Has an elected political party successfully completed the tasks for which they campaigned? Did a special operations corps achieve the mission it set out to accomplish? When taken in concert, questions related to team performance are often of central interest when characterizing a team’s effectiveness.
In addition to assessing the outcome variable of team performance, researchers have also been interested in whether teamwork training actually improves teamwork itself. The efficacy of these interventions can be determined with a number of objective (e.g., products produced by an industry team), self-report (e.g., questionnaires regarding perceived social support amongst team members), and third-party assessments (e.g., expert ratings of team behaviors). Both general/omnibus measures of teamwork (e.g., [21]) as well as those assessing specific dimensions of teamwork (e.g., communication [22]) have been operationalized to examine the effectiveness of these interventions. For example, do team goal setting activities actually result in members creating and pursuing effective team goals? Does simulation training improve the requisite coordination processes among aviation cockpit crews? Has a didactic lecture contributed to improved conflict management among team members? Answering these types of questions is important for determining whether an intervention is actually efficacious in changing the variable that is targeted for improvement (i.e., teamwork behaviors).
Prior to outlining the purposes of this systematic review, it is important to recognize that previous quantitative reviews have been conducted that addressed—to some degree—teamwork training. In preparation for this systematic review, we conducted a scoping review which revealed that eight previous meta-analyses have assessed teamwork intervention studies in some way. However, these reviews were delimited based on various sample and/or intervention characteristics. For example, some reviews included studies that were only conducted with certain team types (e.g., intact teams [23]) or within a particular context (e.g., sports [24]; medical teams [25]). Others were delimited to specific training programs/strategies that were restricted to a narrow range of teamwork strategies (e.g., [23, 25–29]). Finally, studies that used a combination of teamwork and taskwork intervention components have been systematically reviewed [30]; however, these types of interventions result in a limited ability to determine the extent to which the resulting effects were due to teamwork training versus taskwork training.
It should also be noted that all but one [23] of these previous reviews pooled together studies that included a control condition (i.e., wherein teams do not receive any type of teamwork training) and those that did not (as mentioned above, that study only analyzed the effects of certain teamwork strategies). This is an important consideration, as it has been suggested that controlled and uncontrolled studies should not be combined into the same meta-analysis due to differences in study quality (which is a major source of heterogeneity) and since stronger conclusions can be derived from controlled interventions compared to uncontrolled interventions (e.g., [31]). Therefore, while previous systematic reviews have provided valuable contributions to the teamwork literature, a systematic review that assesses the effects of controlled teamwork interventions across a range of contexts, team types, and involving those that targeted diverse dimensions of teamwork appears warranted. In doing so, a more comprehensive assessment of the efficacy of these teamwork interventions is provided, while also having the capacity to look at the potential moderating effects of various sample, intervention, and measurement characteristics. Moreover, by including only controlled studies, one is able to make stronger conclusions regarding the observed effects.
The overall purpose of this study was to better understand the utility of teamwork training for enhancing team effectiveness. Specifically, a meta-analysis was conducted on controlled studies (i.e., comparing teams who have received teamwork training with those who have not) that have examined the effects of teamwork interventions on teamwork processes and/or team performance. To better disentangle the effectiveness of these studies, we also sought to assess potential moderators of these main effects; that is, to determine whether there are certain conditions under which the independent variable of teamwork training more strongly (or weakly) causally influences the dependent variables of teamwork behaviors or team performance [32]. The specific moderators that we assessed included: (a) the team context/field of study, (b) the type of teams that were trained, (c) the primary type of intervention method employed, (d) the dimensions of teamwork that were targeted in the intervention, (e) the number of dimensions targeted, (f) the types of measures used to quantify the training effects, and (g) in studies where teamwork was assessed as an outcome variable, the dimensions of teamwork that were measured. It was hypothesized that teamwork training would have a positive and significant effect on both teamwork and team performance and that these effects would be evident across a range of the aforementioned sample, intervention, and measurement characteristics/conditions.
Searches for potential articles were conducted in the following databases: PsycInfo, Medline, Cochrane Central Register of Controlled Trials, SportDiscus, and ProQuest Dissertations and Theses. Hand searches were also conducted across thirteen journals that typically publish articles on group dynamics (e.g., Group Dynamics: Theory, Research, and Practice; Small Group Research, Journal of Applied Psychology; Personnel Psychology, Human Factors; Academy of Management Journal, Journal of Sport & Exercise Psychology). In each database and journal search, the following combination of search terms were used: (team OR interprofessional OR interdisciplinary) AND (intervention OR training OR building OR simulation) AND (teamwork OR mission analysis OR goal specification OR goal setting OR planning OR strategy OR coordination OR cooperation OR communication OR information exchange OR information sharing OR monitoring OR problem solving OR backing up OR coaching OR innovation OR adaptability OR feedback OR support OR conflict management OR situation awareness OR confidence building OR affect management). These terms were based on various models of teamwork that exist within the literature (see Rousseau et al. [2] for an overview of these models). An additional search was conducted within these databases and journals using the search terms (TeamSTEPPS OR Crew Resource Management OR SBAR [Situation-Background-Assessment-Recommendation]), as several articles in the initial search used these specific training programs. We also searched the reference sections of the articles from past teamwork training review papers as well as from articles that initially met inclusion criteria to determine if any additional articles could be retrieved. The searches were conducted in September 2015 and no time limits were placed on the search strategy. Each article was first subjected to title elimination, then abstract elimination, and finally full-text elimination.
To be included in the meta-analysis, a study needed to examine the effects of teamwork training by comparing teams in an experimental condition (i.e., those who received teamwork training) with those in a control condition (i.e., where teams did not receive teamwork training). Cross-sectional/non-experimental studies were excluded, as were intervention studies that did not include a control condition. As this review was only concerned with teamwork interventions, studies that focused on training taskwork—whether independent of, or in addition to, a teamwork intervention—were excluded. For example, as previously mentioned, simulation-based training (SBT) has been used as a means of training individuals to perform technical skills and also to enhance teamwork. In order for a SBT intervention to be included in this meta-analysis, it had to be clear that only teamwork (not technical skills) was being targeted during training. In order to address our primary research question, the study had to provide data on at least one teamwork dimension and/or team performance. The study also needed to provide sufficient statistics to compute an effect size. In cases of insufficient data, corresponding authors were contacted for this information. The articles were delimited to those published in the English language.
Articles that met the aforementioned eligibility criteria were extracted for effect sizes and coded independently with respect to seven moderators by two of the authors (DM and GR). Interrater reliability for the coding of these moderators was over 90%, kappa (SE) = 0.80 (0.01). The moderators examined were based on a scoping review (the purpose of which included identifying pertinent characteristics that were commonly reported in previous teamwork intervention research), which was conducted in preparation for this systematic review. The moderators that were examined in this review included (1) the context within which an intervention was conducted (health care, aviation, military, academia, industry, or laboratory experiment), (2) the type of team targeted (intact or new), (3) the primary training method applied to conduct the intervention (didactic education, workshop, simulation, or team reviews), (4) the dimension(s) of teamwork (preparation, execution, reflection, and/or interpersonal dynamics) targeted in the intervention as well as (5) the number of dimensions targeted (between one and four), (6) the type of measure used to derive effect sizes (self-report, third party, or objective measures), and—when teamwork was assessed as the criterion variable—(7) the specific dimension(s) of teamwork that were measured (general, preparation, execution, reflection, and interpersonal dynamics).
Once coded, data were entered into the software Comprehensive Meta-Analysis, Version 2 [33] and analyzed as a random-effects model (DerSimonian and Laird approach). This type of model assumes that there is heterogeneity in the effect sizes across the included studies and is the appropriate model to use in social science research, as opposed to a fixed-effects model (which assumes that effect sizes do not vary from study to study) [34, 35]. Where possible, effect sizes for each study were derived from means, standard deviations, and sample sizes at baseline and post-intervention [34, 36]. If these statistics were not fully provided, they were supplemented with F-statistics, t scores, correlations, and p-values to compute the effect size. Each study was given a relative weight based on its precision, which is determined by the study’s sample size, standard error, and confidence interval (i.e., the more precise the data, the larger the relative study weight) [34].
In instances where a study provided data to calculate multiple effect sizes (such as when several measures of the criterion variable—teamwork or team performance—were examined), these effects were combined into one overall effect size statistic (i.e., a weighted average) for that study. This was done to ensure that those studies that had multiple measures of teamwork or team performance were not given greater weight compared to studies that only provided one effect size (i.e., only had one measure of performance or teamwork), which could potentially skew the overall results [34]. The exception to this was when articles reported the effects of more than one intervention (i.e., had multiple experimental conditions), each of which had a unique teamwork training protocol. In these cases, an effect size from each intervention was computed. Thus, these articles would contribute multiple effect sizes to the total number of comparisons within the meta-analysis. To correct for potential unit-of-analysis errors in these particular articles, the sample size of the control condition was divided by the number of within-study comparisons [31]. For example, if three different types of teamwork interventions were compared to one control condition (e.g., which had a sample size of 30 participants), the n of the control condition was divided by 3 (i.e., 30/3 = 10) when calculating the effect sizes of those interventions. Cohen’s d was used as the effect size metric to represent the standardized effect (i.e., the average magnitude of effectiveness) of teamwork interventions on teamwork and team performance [37]. Standard errors and 95% confidence intervals were computed to test for the accuracy of the standardized effects obtained.
To reduce heterogeneity and improve the interpretability of the results, we pooled studies into those that measured teamwork as its criterion variable and those that measured team performance. Pooling studies in this manner not only reduces heterogeneity but also allowed us to identify the extent to which teamwork interventions impact team performance and, separately, the extent to which they affect teamwork processes. Heterogeneity within the meta-analysis was also assessed by computing a Q value—which estimates the variability in the observed effect sizes across studies—and an I 2 statistic—which estimates the ratio of the true heterogeneity to the total observed variation across studies. High Q and I 2 statistics can be problematic for interpreting the results of a meta-analysis and can also indicate that the meta-analysis includes outlier studies. We also planned to identify and exclude outliers from subsequent moderator analyses in two ways. First, sensitivity analyses were carried out by removing a single intervention from the meta-analysis and noting the resulting effect size—this estimates the impact that each individual intervention has on the overall effect size of teamwork or team performance. If the resulting effect size with an intervention removed (i.e., K– 1) is substantially different than the effect size with that intervention present, this may suggest that it is an outlier and needs to be removed [34]. Second, we noted any studies that had abnormally high effect sizes and standardized residuals (above 3.0), especially when these values were accompanied by narrow confidence intervals. If heterogeneity (Q and I 2 ) is substantially reduced upon removal of a study, this further confirms that the study is an outlier and should be omitted from subsequent subgroup/moderator analyses.
Once the two pools of studies were produced, bias within each pool was assessed. First, publication bias was examined by calculating a fail-safe N statistic, which estimates the number of unpublished studies with null findings that would have to exist to reduce the obtained effect size to zero [38]. If this number is sufficiently large—Rosenberg [39] recommends a critical value of 5N+10—then the probability of such a number of studies existing is considered to be low. For example, if 20 studies were included in a meta-analysis, then the resulting fail-safe N should be larger than 110 (i.e., 5*20 + 10); if this value was not larger than 110, then publication bias is likely within this pool of studies. We also obtained two funnel plots (one for studies where teamwork was the outcome variable and one for team performance as the outcome) to provide a visual depiction of potential publication bias. We then conducted an Egger’s test as a measure of symmetry for these two funnel plots. If this test statistic is significant (p < 0.05), this denotes that the distribution around the effect size is asymmetric and publication bias is likely present [34].
The literature search from the five databases returned 22,066 articles, while the hand searches of the 13 journals returned 3797 articles, vetting of studies from previous team training reviews returned 191 articles, and the ancestry search of reference lists returned 471 articles (see Fig 1 ). After removing duplicates, 16,849 articles were subject to title and abstract screening, where they were dichotomously coded as ‘potentially relevant’ or ‘clearly not relevant’. 1517 potentially relevant articles were then full-text reviewed and coded as meeting eligibility criteria or as ineligible for the following reasons: (1) not a teamwork intervention; (2) teamwork-plus-taskwork intervention; (3) insufficient statistics to compute an effect size; (4) not including a measure of teamwork or team performance; or (5) not including a control group. As a result of this eligibility coding, 51 articles were included in the meta-analysis. 13 of these studies reported results on two or more interventions, bringing the total number of comparisons (k) to 72 with 8439 participants (4966 experimental, 3473 control). See S1 Table for descriptions of each study with regard to study context, type of team and participants, targeted teamwork dimensions of the intervention, number of effect sizes, the criteria measured, and an overview of the intervention.
Results of the overall effect of teamwork interventions on teamwork processes along with summary statistics and sensitivity analyses (i.e., the final column marked ‘ES with study removed’) for this pool of studies are presented in Table 1 . This pool included a total of 39 interventions from 33 studies. The results revealed that teamwork interventions had a significant, medium-to-large effect on teamwork, d (SE) = 0.683 (0.13), 95% CI = 0.43–0.94, Z = 5.23, p < 0.001; Q (df) = 660.7 (38), I 2 = 94.2. The funnel plot for this pool of studies is shown in Fig 2 . The fail-safe N was 3598, which is sufficiently large, as it exceeds the critical value of 205 (5*39+10). The funnel plot for this pool of studies is presented in Fig 2 . Egger’s value for this funnel plot was not significant (B = 0.364, SE = 1.30, 95% CI = -2.26–2.99, t = 0.28, p = 0.78), which also suggests that bias was not present. Two studies were identified as outliers within this pool of studies: Morey et al. [3] and Marshall et al. [22]. The resulting effect size when these studies were excluded was d (SE) = 0.550 (0.08), 95% CI = 0.39–0.71, Z = 6.73, p < 0.001; Q (df) = 187.53 (36), I 2 = 80.8. Subsequent moderator analyses were conducted with these two outlier studies being omitted.
Circles filled with black indicate outlier studies.
| Study | Relative Weight | Effect Size (SE) | 95% CI (lower, upper) | Z-value | p-value | ES with intervetion removed |
|---|---|---|---|---|---|---|
| Aaron 2014 [13] a | 2.43 | 1.432 (.35) | .74, 2.13 | 4.04 | < .001 | 0.67 |
| b | 2.48 | .869 (.33) | .22, 1.52 | 2.61 | .009 | 0.68 |
| Becker 2005 [40] | 2.75 | .635 (.21) | .22, 1.05 | 3.02 | .003 | 0.69 |
| Beck-Jones 2004 [41] a | 2.70 | -.030 (.24) | -.50, .44 | -0.13 | .898 | 0.70 |
| b | 2.69 | -.003 (.24) | -.47, .47 | -0.01 | .990 | 0.70 |
| Beranek 2005 [42] | 2.67 | .649 (.25) | .16, 1.13 | 2.62 | .009 | 0.68 |
| Bjornberg 2014 [9] | 2.83 | .080 (.16) | -.23, .39 | 0.50 | .615 | 0.69 |
| Brannick 2005 [5] | 2.72 | 1.229 (.23) | .79, 1.67 | 5.47 | < .001 | 0.69 |
| Bushe 1995 [43] a | 2.53 | .405 (.31) | -.20, 1.01 | 1.31 | .192 | 0.69 |
| b | 2.53 | .534 (.31) | -.08, 1.14 | 1.71 | .086 | 0.69 |
| Cheater 2005 [12] | 2.82 | .336 (.17) | .00, .67 | 1.97 | .049 | 0.69 |
| Clay-Willaims 2013 [44] a | 2.04 | .531 (.51) | -.46, 1.53 | 1.05 | .296 | 0.69 |
| b | 2.06 | -.213 (.50) | -1.20, .77 | -0.43 | .671 | 0.70 |
| c | 2.12 | 0.000 (.48) | -.94, .94 | 0.00 | 1.00 | 0.70 |
| Dalenberg 2009 [45] | 2.82 | 1.001 (.17) | .68, 1.33 | 6.02 | < .001 | 0.67 |
| Deneckere 2013 [46] | 2.92 | .129 (.09) | -.04, .29 | 1.52 | .129 | 0.70 |
| Dibble 2010 [47] | 2.92 | -.242 (.09) | -.42, -.07 | -2.72 | .007 | 0.71 |
| Eden 1986 [48] | 2.92 | .427 (.09) | .07, .42 | 2.73 | .006 | 0.70 |
| Ellis 2005 [14] | 2.88 | .792 (.13) | .54, 1.05 | 6.14 | < .001 | 0.68 |
| Emmert 2011 [49] | 2.54 | .763 (.31) | .16, 1.36 | 2.48 | .013 | 0.68 |
| Entin 1999 [50] | 2.32 | .771 (.40) | -.01, 1.55 | 1.93 | .054 | 0.68 |
| Friedlander 1967 [51] | 2.72 | .495 (.22) | .06, .94 | 2.21 | .027 | 0.69 |
| Green 1994 [52] a | 1.91 | .665 (.56) | -.44, 1.76 | 1.19 | .236 | 0.68 |
| b | 1.87 | 1.058 (.58) | -.08, 2.20 | 1.82 | .069 | 0.68 |
| Jankouskas 2010 [7] | 2.22 | .778 (.44) | -.08, 1.64 | 1.77 | .077 | 0.68 |
| Kim 2014 [53] | 2.65 | .062 (.26) | -.45, .57 | 0.24 | .813 | 0.70 |
| Marshall 2009 [22] * | 2.70 | 3.277 (.33) | 2.65, 3.95 | 9.90 | < .001 | 0.61 |
| Martinez-Moreno 2015 [54] | 2.86 | .503 (.14) | .23, .78 | 3.63 | < .001 | 0.69 |
| Morey 2002 [3] * | 2.93 | 1.896 (.08) | 1.75, 2.05 | 24.83 | < .001 | 0.64 |
| O’Leary 2011 [21] | 2.82 | .426 (.17) | .10, .76 | 2.54 | .011 | 0.69 |
| Padmo Putri 2012 [6] | 2.82 | -.097 (.17) | -.42, .23 | -0.58 | .561 | 0.71 |
| Prichard 2007 [55] | 2.40 | 1.981 (.37) | 1.26, 2.70 | 5.381 | < .001 | 0.65 |
| Rapp 2007 [56] | 2.61 | .535 (.28) | -.01, 1.08 | 1.93 | .053 | 0.69 |
| Shapiro 2004 [57] | 2.03 | .689 (.52) | -.32, 1.70 | 1.34 | .181 | 0.68 |
| Smith-Jentsch 2008 [4] | 2.63 | 1.103 (.27) | .58, 1.63 | 4.13 | < .001 | 0.67 |
| Thomas 2007 [58] | 2.39 | .891 (.37) | .16, 1.62 | 2.40 | .016 | 0.68 |
| Volpe 1996 [59] | 2.71 | .450 (.23) | .00, .90 | 1.97 | .049 | 0.69 |
| Weaver 2010 [60] | 2.41 | .580 (.36) | -.13, 1.29 | 1.61 | .109 | 0.69 |
| Weller 2014 [61] | 2.64 | 1.563 (.26) | 1.05, 2.08 | 5.92 | < .001 | 0.66 |
| OVERALL | 100 | .683 (0.13) | 0.43, 0.94 | 5.23 |
Note. a, b, c = intervention groups within study; SE = standard error; CI = confidence interval; ES = effect size.
* = Study identified as an outlier and removed from subsequent moderator analyses.
The final column marked ‘ES with study removed’ indicates the results of the sensitivity analysis for each respective intervention.
Results of the overall effect of teamwork interventions on team performance as well as summary statistics and sensitivity analyses (i.e., the final column marked ‘ES with intervention removed’) for this pool of studies are presented in Table 2 . This pool of studies included a total of 50 interventions from 32 studies. It was shown that teamwork interventions had a significant, large effect on team performance—d (SE) = 0.919 (0.14), 95% CI = 0.65–1.19, Z = 6.72, p < 0.001; Q (df) = 851.3 (49), I 2 = 94.2. The funnel plot for this pool of studies is shown in Fig 3 . The fail-safe N was 6692, which is sufficiently large, as it exceeds the critical value of 260 (5*50+10). The funnel plot for this pool of studies is presented in Fig 3 . Egger’s value for this funnel plot was not significant (B = 0.131, SE = 1.19, 95% CI = -2.26–2.54, t = 0.11, p = 0.91), which also implies that bias was not present. There were five outlier interventions (from four studies) in this pool of studies that assessed team performance: Morey et al. [3], Smith-Jentsch et al. [4], one of the interventions from Buller and Bell [63]; teambuilding condition), and both interventions from Bushe and Coetzer [43]. When these outliers were removed, the resulting effect size was d (SE) = 0.582 (0.06), 95% CI = 0.47–0.69, Z = 10.30, p < 0.001; Q (df) = 101.1 (44), I 2 = 56.5. Subsequent moderator analyses were conducted with these five interventions omitted.
Circles filled with black indicate outlier studies.
