Speaking Up in the Operating Room:

This paper examines learning in interdisciplinary action teams. Research on team effectiveness has focused primarily on single-discipline teams engaged in routine production tasks and, less often, on interdisciplinary teams engaged in discussion and management rather than action. The resulting models are of limited use for explaining differences in learning in interdisciplinary action teams. Members of these teams must coordinate action in uncertain, fast-paced situations, and the extent to which they are comfortable speaking up with observations, questions, and concerns may be critical to team outcomes. To explore what leaders of action teams do to promote speaking up and other learning processes, as well as how organizational context may affect team learning outcomes, I analyze data from 16 operating room teams learning to use a new technology for cardiac surgery. Team leader coaching, speaking up, and boundary spanning combined to facilitate successful technology implementation. The most effective team leaders helped teams learn by communicating a motivating sense of purpose for shifting to the new technology and by framing the learning process as a teamwork rather than technical challenge, thereby minimizing concerns about status differences and promoting speaking up in the service of learning. INTERDISCIPLINARY ACTION TEAMS 3 Introduction Teams come in a staggering variety of forms from self-managed production teams to product development teams, cockpit crews, and sports teams. A technical-rational conception of team effectiveness has dominated research on teams and does not adequately capture or speak to this range of opportunities for teamwork (West, 2000). Most studies have emphasized ongoing teams with welldefined tasks and promoted theories in which team structures such as task design, member composition, and organizational support are designed in advance to achieve specific objectives (e.g., Hackman, 1987; Campion, Medsker, and Higgs, 1993; Cohen and Ledford, 1994; Wageman, 2001). The resulting models are of limited applicability in situations where team goals change over time (McGrath and O'Connor, 1996), are multiple and competing (Brodbeck, 1996), or where task demands require constant responsiveness to unpredictable situations (Sundstrom, De Meuse, & Futrell, 1990). For example, product development team goals can change when market research reveals new information; work teams in high-risk industries face competing goals in the form of executing current tasks flawlessly and incorporating new practices, technologies or techniques that could deliver better performance in the future, and action teams consist of "highly skilled specialists... cooperating in brief performance events that require improvisation in unpredictable circumstances" (Sundstrom, De Meuse, & Futrell, 1990, p. 121). These teams face task variability that eludes a static, structural solution. The role of team leaders has been downplayed in structural models of team effectiveness. Indeed, an ideal of team self-management is to be achieved through effective team design and organizational context support (e.g., Wageman, 2001; Hackman, 1990; 1998). Although a number of recent papers have begun to develop models to describe the team leader role (Kozlowski et al, 1996), the emphasis in empirical research has been on self-managed and leaderless teams. In contrast, in other literatures in which teams are considered – such as new product development (Wheelwright and Clark, 1995) and medicine (Shortell et al, 1994) – the role of team leaders has been viewed as critical to team performance. These teams face relatively higher demands for ongoing learning and problem INTERDISCIPLINARY ACTION TEAMS 4 solving than do self-managed work teams in production, service, and sales. Grappling with the need for adaptation and change in teams, an emerging literature on team learning has adopted a socio-emotional conception of teamwork in which team leader actions and members' subjective perceptions play central roles (Edmondson, 1996; Alderfer, 1987). According to this perspective, when teams face uncertain or changing contexts, their members are interpersonally vulnerable (Edmondson, 1999). If they challenge current practices or norms, raise concerns, or point out errors, others may reject or humiliate them. Moreover, speaking up across identity groups (such as gender or race) or organizational boundaries (such as rank or function) can exaggerate reluctance to communicate and inhibit motivation in a work group (Alderfer, 1987). The interpersonally safe route is to remain silent or continue past practices, but this poses technical risk if the context calls for novel team responses. Thus, not speaking up protects individuals but harms the team. Much research on speaking up has focused on extra-role behavior, studying when people are willing to speak up about aspects of the organization or work context that go beyond the demands of their jobs (Morrison and Phelps, 1999; Van Dyne et al, 1994; Morrison and Milliken, 2000). This paper in contrast investigates speaking up in the context of intra-role behavior, in particular when learning new tasks or coordination routines. I explore organizational and group influences on speaking up in a study of 16 operating room teams learning to use a new technology, and show that speaking up is a vital aspect of the team learning process in this context. Data analysis identifies aspects of the learning process that vary across teams and sheds light on the relationship between these variables and technology implementation success. Further exploration of the qualitative data examines how leaders promote speaking up in teams. Speaking Up and Learning in Interdisciplinary Action Teams Interdisciplinary action teams Action teams are defined as teams in which members with specialized skills must improvise and coordinate their actions in intense, unpredictable situations. Some sports teams, such as those playing basketball or hockey, exemplify this definition, as do emergency medical teams, operating INTERDISCIPLINARY ACTION TEAMS 5 room teams, and cockpit crews (Sundstrom et al, 1990). Although most of these groups are composed of individuals with similar training such as pilots or basketball players, some action teams, notably in the operating room, are interdisciplinary. Action teams, by definition, must respond to unexpected events in a coordinated way, often requiring a free and open transfer of information to coordinate action and interaction. This openness can be particularly challenging for teams that include different disciplines, because associated differences in status, training, language, and norms can impede communication and shared understanding. Disciplinary differences lead to communication problems, due to specialized training and terminology and to differences in what is taken for granted by individuals in a given specialty (Dougherty, 1992). In interdisciplinary teams, it is often the case that the team leader is in the unique position of seeing the whole picture, or understanding how different sources of expertise fit together in the project (Wheelwright and Clark, 1995). Team leaders in such teams can help their teams create shared meaning about the situations they face. Research on high-reliability organizations has emphasized notions of shared cognition to facilitate coordination in action contexts by allowing team members to anticipate each other's moves to promote responsiveness in unexpected situations (Weick, 1993; Weick and Roberts, 1993). However, this work has not investigated the challenge such teams face when they must alter their coordination routines to accommodate major changes in procedures, equipment, context, or members – that is, when they must develop or modify collective mental models. Altering team routines may be particularly difficult in the action context, because routines take on a habitual nature (Gersick and Hackman, 1992). This paper explores the role of speaking up in a team to help create new routines. Speaking up and team learning The challenges faced by interdisciplinary action teams (IATs) are intensified when coping with substantial change in addition to their usual need to coordinate and improvise; coordination is likely to be difficult during a transition period between old and new routines. The introduction of a new technology, for example, may require both new individual skills and new interpersonal routines. INTERDISCIPLINARY ACTION TEAMS 6 Successful implementation (Yin, 1977) of new equipment, new members, or new practices in a team can be conceptualized as the outcome of a learning process. This learning process involves experimentation with new actions and behaviors as well as explicit discussion about the changes being made (Edmondson et al, 2001), during which team members share both technical knowledge and social knowledge about who knows what (Moreland, 1999). In teams in which members are more comfortable speaking up with observations, concerns, and questions, the learning process is likely to be more effective than those in which they are not. The process may be characterized by more trial and error and by effectively building a repertoire of shared experiences of what works and what doesn't. Ease of speaking up should facilitate regaining smooth coordination of actions, following the transition period, building confidence in and commitment to the new technology. In sum, successful implementation – defined as ongoing incorporation of new practices into the team's repertoire of capabilities (Yin, 1977) – should be aided by speaking up in IATs. Proposition 1: Speaking up in IATs facilitates implementation of new practices. The belief that speaking up is desirable and possible in a team is likely to be influenced by both organizationaland group-level influences. For example, organizational culture includes norms for communication, making some organizations more informal, direct, or confrontational, and others more formal, indirect, or polite (Schein, 1985). Organizational variables that are especially relevant for implementing change include top management support and resources (e.g., Yin, 1977). At the same time, groups in an organization vary widely in psychological safety, suggesting that group-level influences also may explain variance in speaking up behavior (Edmondson, 1996; 1999; 2002). Organizational context Most teams exist within organizational contexts, such as hospitals, airlines, or athletic associations (Hackman, 1987). The nature of the organizational context is likely to influence a team facing a significant learning challenge. First, if a team has resources and slack in its schedule to allow practice and debriefing, it may be able to learn more easily because it has time for INTERDISCIPLINARY ACTION TEAMS 7 experimentation. Second, an organization can provide support for learning in the form of information systems that simplify record keeping and access (Davenport and Prusak, 1998). Third, senior management can signal that they support innovation and change, through words and resources. Management support also can make a change project visible and energize others in the organization to provide assistance when needed. Finally, having a history of innovation in an organization may facilitate acceptance of new technology (Cohen and Levinthal, 1990; Iansiti and Clark, 1994). These factors together constitute a supportive organizational context for implementing new team practices. Proposition 2: A supportive organizational context facilitates successful implementation of new practices in IATs. Organizational experience, management support, information systems, and other resources are not likely to help team members speak up. Organizational support cannot shield someone from the psychological harm of interpersonal rejection in the face-to-face context of a team. In sum, a supportive organizational context may enable team learning by removing external barriers (such as insufficient resources), while other, group level influences, more directly affect speaking up. Team design Important features of team design include task, role composition, and membership stability (Hackman, 1987; Wageman, 2001; Moreland, 1999). An interdependent team task gives rise to a need for speaking up to coordinate actions; likewise, when a team is composed of members with appropriately diverse skills and experiences speaking up may be perceived as worthwhile because people can benefit from others' perspectives. Action teams in health care, high reliability organizations, or sports, face some degree of membership instability. Such teams tend to draw from a larger pool of members to put together a subset of members at a given point in time, due to the need for around the clock operations or the potential for exhaustion that action teams face. Research on transactive memory in teams suggests that members who work together longer develop an understanding of who knows what, enabling them to coordinate actions while learning a new task (Moreland, 1999). Similarly, in a laboratory study of INTERDISCIPLINARY ACTION TEAMS 8 team adaptation, Okhuysen (2001) found that familiar groups (those consisting of members who know each other well) were more able to initiate opportunities to reflect out loud on the task and change direction. Thus, if a stable subset of team members is working closely together, this may enable more rapid development of new coordination routines. At the same time, overall implementation success in the organization may suffer if other, alternate members are not being exposed to the new routines. Therefore, the effect of team stability on implementation of new practices for IATs may be mixed. Team leadership When action teams must learn new routines, members may feel anxious about changing (Schein, 1985), reducing their willingness to speak up openly with observations, concerns and questions. In addition to helping articulate a meaningful rationale for change, team leaders are in a position to help to reduce anxiety, creating psychological safety for speaking up. Coping with change. In situations of uncertainty or change, the meaning of various stimuli and events is often unclear. Literature on change emphasizes the importance of leadership (Kotter, 1990; Schein, 1985). Likewise models of leadership emphasize the crucial role of leaders in managing change, uncertainty, and crisis (Bennis, 2001; Denis et al, 2001). In these related streams of work, the notion of leadership (at the top of an organization rather than in a team) and the experience of change are tightly linked. In contrast, existing models of team leadership emphasizes a managerial role, in which team leaders assess the context, the task and the team's skills, and then structure and guide ongoing activities accordingly. Although recent models of team leadership include dynamic considerations – noting that team tasks are not always static and so leaders must play a key role in developing their teams and creating focused training opportunities as needed (Kozlowski et al, 1996; Zaccaro and Marks, 1999) – in the organizational leadership literature, these activities resemble those ascribed to managers in contrast to leaders (Zaleznick, 1992; Kotter, 1990). Whereas leaders attend to meaning and the emotional experience, in part by calling attention to their vision of a desired future state, managers attend to the pragmatic and to the here and now – organizing INTERDISCIPLINARY ACTION TEAMS 9 complexity rather than interpreting reality. The management function helps people cope with complexity, helping to structure, assign and guide task execution (Kotter, 1990; Zaleznik, 1992); the leadership function helps people cope with change. Authority dynamics. In some teams, team leaders have unusual degree of power or authority relative to team members; for example, leaders of "heavyweight" product development teams have both positional authority and the ear of senior managers (Wheelwright and Clark, 1995). Similarly, in operating room (OR) teams, surgeons have enormous status and power relative to other team members. Those lacking power in a team are more dependent on those with power than the other way around, and thus are motivated to conform rather than take risks (Lee et al, 2001; Depret & Fiske, 1993). Lacking formal power, individuals in groups tend to be reluctant to take risks, deviate from norms, or try novel actions that may not work (Estrada, Brown, & Lee, 1995). Power and status are likely to be highly salient when individuals make decisions about whether or not to speak up in the context of doing work. Team leader coaching refers to direct interaction with the team that is intended to shape individual and team activities to promote desired outcomes (Wageman, 2001). Coaching thus includes providing clarification and feedback, seeking members' input, listening to concerns, and being accessible and receptive to others' ideas and questions. These include both managerial activities (such as coordinating members' actions) and leadership activities (such as inspiring commitment, effort and openness). One way that leaders can communicate openness and fallibility to mitigate power imbalances is through self-disclosure (Gabarro, 1987). Creating psychological safety. Psychological safety describes the degree to which people believe they will not be penalized, ridiculed, or rejected for speaking up, or for asking questions, seeking help, or admitting error (Edmondson, 1999). Team leaders can influence members' sense of psychological safety by whether they communicate openness to and interest in different views, such 1 Psychological safety and beliefs about how easy it is to speak up on the job are highly related concepts; psychological safety is the belief that it is easy it is to speak up; speaking up is a behavioral manifestation of that belief. INTERDISCIPLINARY ACTION TEAMS 10 as by inviting input from multiple team members on a specific topic. This enables different voices to be heard and also is likely to lower the hurdle for subsequent communication. More generally, to encourage speaking up under conditions of uncertainty, team leaders must find away to minimize concerns others may have about being humiliated or rejected in any way by higher status, more powerful others. Proposition 3: Coaching behavior by the team leader, including efforts to downplay status differences, creates psychological safety for speaking up in IATs. Boundary spanning as a team learning process Confronting novelty and change, teams face both within-team and external coordination challenges. Members of action teams in complex organizations often must coordinate objectives, schedules, and resources with those outside team boundaries. Referred to as boundary spanning (Ancona, 1990), these activities allow teams to obtain and communicate information relevant to new practices. Without effective boundary management, teams are in danger of making decisions that are inconsistent with other organization goals or constraints or of failing to take advantage of available support and resources. When new team practices have implications for others in the organization, effective boundary spanning by a team is likely to promote implementation success. Proposition 4: Boundary spanning in IATs facilitates successful implementation of new practices that affect or require input from others in the organization. Interdisciplinary Action Teams in Health Care I explore the theoretical relationships discussed above in the health care setting because of the prevalence and importance of IATs in this context. Interdisciplinary teams exist in the operating room, in intensive care units (ICUs), the emergency room, and in clinics that treat chronic diseases, which benefit from interdisciplinary team-based care (Horbar, 1999; Shortell et al, 1994; Klein et al, 2002). These teams face a series of unique patient situations (requiring effective event-based learning), coupled with more gradual changes in techniques and technologies supporting the work (requiring periodic learning of new approaches for patient care). Identifying factors associated with INTERDISCIPLINARY ACTION TEAMS 11 effective learning and process improvement is of increasing importance for health care researchers and practitioners because of pressures to reduce cost and improve quality. To study team learning in this context, it was important to find multiple teams facing a learning challenge at the same time, where many other features of the context could be held relatively constant. A new cardiac surgery technology introduced in many U.S. hospitals in the late 1990s met these criteria. Operating room teams in cardiac surgery Four disciplines come together in the cardiac surgery operating room. Cardiac surgeons, the team leaders, work with anesthesiologists, nurses (scrub and assisting) and technicians called perfusionists who run the heart-lung bypass machine. Teamwork in this setting involves seamless coordination of many small tasks that together constitute the surgical procedure. The most frequent cardiac procedure is coronary artery bypass graft (CABG); another common procedure is mitral valve repair (MVR). The role of each team member in these procedures is clearly defined, and everyone is able to monitor the progress of the operation and to anticipate what actions will be needed by simply looking at the heart itself and at what the surgeon is doing. Minimally invasive cardiac surgery (MICS), an innovation developed and manufactured by a device company called Minimally Invasive Surgical Associates (MISA), differed from the conventional approach to CABG and MVR in that the patient’s breastbone was not split apart. This reduced the extent of pain and recovery time for patients, such that they were able to resume normal activities more quickly than after conventional cardiac surgery. Using special new equipment, the heart was accessed through small incisions between the ribs, and the patient was connected to the bypass machine through the artery and vein in the groin. A tiny deflated balloon, threaded into the aorta and then inflated to prevent blood from flowing backwards into the stopped heart, replaced the traditional clamp inserted directly into the chest. MICS increased the coordination challenge on the OR team dramatically, and all hospitals purchasing the technology were required to send a team to a three-day training program run by MISA 2 All company, product, hospital, and individual names are pseudonyms. INTERDISCIPLINARY ACTION TEAMS 12 before using the technology on patients. During the surgical procedure the tiny balloon-clamp had to be constantly monitored by the team with specialized ultrasound technology. Team members had to coordinate their actions to guide the balloon into place and then monitor it for hours to make sure it stayed in place. Unlike conventional surgery in which surgeons relied on direct visual and tactile data, MICS called for team members to supply the surgeon with information displayed on monitors. The new technology thus not only changed individual team members' tasks, it blurred role boundaries and increased the need for speaking up, to coordinate action that had previously been sufficiently routine to be coordinated without speaking. Method Research design This study used a multiple case study design (Eisenhardt, 1989; Yin, 1989) to advance theory on learning in interdisciplinary action teams. This design combined strengths of qualitative data for better understanding a process with the opportunity to learn from patterns of variance across cases. Multiple case comparisons encourage researchers to look beyond initial impressions from a single site and further inform theory building efforts by suggesting how different constructs relate to each other (Eisenhardt, 1989). This design thus allowed me to explore team processes qualitatively and to examine quantitative correlations between constructs as preliminary tests of the research propositions. Sample. The sample consisted of 16 cardiac surgery teams selected to ensure differences in geography and innovation history from the approximately 150 U.S. hospitals that had purchased the technology. Each hospital had only one OR team learning MICS, although the teams varied in size. All were learning the new technology within the first year of its approval by the FDA, and all were in well-respected cardiac surgery departments. Data Collection Preliminary fieldwork. To understand the technology and the surgical process, I attended MISA’s three-day training program with three other management researchers. We attended lectures, observed an OR team in hands-on laboratory sessions, and interviewed members about how they INTERDISCIPLINARY ACTION TEAMS 13 perceived the implementation challenge. We then developed an interview protocol to assess organizational and team characteristics related to MICS implementation. Because the researchers spanned three academic disciplines, the protocol included questions addressing a range of issues related new technology adoption in health care, creating a shared database for several distinct sets of analyses. This paper reports on data related to team process and outcomes. Interviews. The primary data source is 165 interviews conducted at the 16 hospitals. Four interviewers participated in the first four site visits to promote consistency in asking questions and recording data; a team of two or three of us visited each remaining site. At each site, we interviewed members of the operating room team and other hospital personnel who interacted with team members or were knowledgeable about the technology. Team members interviewed consisted of one to three people in each of four roles: surgeons, anesthesiologists, OR nurses and perfusionists. Others interviewed included cardiologists, intensive care unit (ICU) nurses, general care unit (or floor) nurses, and hospital senior management. Multiple informants at each site allowed cross checking of responses about factual issues and reduced the chances of obtaining unrepresentative responses from individuals. To minimize the effects knowledge of implementation success would have on our perceptions of sites, interviewers were blind to the outcome data throughout site visits. Each interview began an open-ended question ("Can you tell us how MICS got started at this hospital and how it's going?") to obtain views about what mattered before limiting informants' responses with specific questions. The questions that followed addressed more specific team issues, such as how the team worked together, what they did to prepare (following training but before the first case), and who was involved at various stages of the learning process. Questions about the hospital as an organization included both broad ("How is this hospital different from other hospitals in which you've worked?") and specific topics ("What was the role of management in MICS?" and 3 Together we represented the fields of economics and operations management, medicine, quality improvement and statistics, and the author's field of social psychology and organizational behavior. The interview protocol is available from the author. 4 Other papers examine differences in the slope of the learning curve across sites, technology implementation, and managing innovation in health care. INTERDISCIPLINARY ACTION TEAMS 14 "What kinds of data/information are kept in this department? Who uses this information? For what purposes?"). Consistent with the aim of a multiple case study design, each new visit was used to check an emerging understanding of salient factors in the team learning process. This iterative process clarified not only which aspects of the team learning process and organizational context varied but also the nature and range of variance. As a result, we developed a set of categories that corresponded to each of the consistently asked questions. For example, to assess psychological safety for speaking up, we asked informants about what they would do (or had done) if they noticed a marginally adverse indicator in one of the visual or numerical monitors – one that presented a situation in which the patient was in no immediate danger but could indicate a problematic trend. We were able to code responses as representing one of three basic options: the atmosphere and interaction in this team is characterized by (3) open reciprocal communication (very free and effortless), or (2) respectful but guarded communication (picking the right moment to speak, pronounced awareness of status differences), or (1) communication that is quite limited, with some members extremely hesitant to speak up. 5 This process generated a database with numerical codes for each variable for each hospital. We also took extensive notes to justify the coding and to capture the additional detail and anecdotes provided in these responses. Data Analyses Data analysis for this paper occurred in four stages. First, I conducted preparatory analyses of the interview data to aggregate the coded questions into a set of meaningful independent variables. Second, a research assistant categorized and coded the free-form interview notes. Third, I analyzed archival data to create a measure of implementation success (the outcome variable in this study), which assessed how effectively a team implemented the new technology into the hospital as an ongoing practice. Fourth, I examined qualitative and quantitative relationships between the independent and dependent variables and then further analyzed the qualitative data to explore team 5 In almost all cases, interviewers selected a common code for all constructs discussed in this paper. The few discrepancies were resolved by discussion, citing evidence from multiple interviews capturing responses to the same question. INTERDISCIPLINARY ACTION TEAMS 15 leader behaviors associated with effective team learning. The strength of these data lie in their opportunity to provide qualitative insight; the derived (quantitative) measures were used as a check against researcher bias and as a way to leverage and extend intuition to learn more from the data than would otherwise be possible.. Stage 1. I analyzed all coded interview data related to the team learning process and the organizational context to create composite variables. Following the logic of within-method triangulation (Jick, 1979), I examined correlations between coded responses to questions assessing the same construct (Campbell and Fiske, 1959). This provided measures of internal consistency reliability and discriminant validity for each composite variable, as shown in Table 1; non-composite (or single-measure) variables are shown in Table 2. My aim in these analyses, described in more detail in the Appendix, was to assess the similarity of responses to related questions. The result is a set of numerical ratings for five team constructs (team stability, team preparation, procedure innovation, boundary spanning, and psychological safety) and four organizational constructs (history of innovation, resource constraints, management support, and information infrastructure) for each hospital. These nine variables represent a set of factors that emerged as varying noticeably across hospitals. Team leader coaching was assessed in the next stage of analysis, using free form qualitative data. Tables 1 and 2 about here Stage 2. A research assistant who had not participated in site visits coded the transcribed verbatim responses using categories that corresponded closely to the interview protocol. Data that fit into more than one category were given multiple codes. The research assistant then developed subcategories by identifying recurring themes within each category, and finally coded each data unit (ranging from one to several sentences) according to major and minor categories, speaker profession, and hospital. The coded data set allowed me to quickly compare particular features across hospitals 6 Other initially interesting variables were dropped as they either did not vary, or were not measurable because informants did not recognize them as salient. INTERDISCIPLINARY ACTION TEAMS 16 by excerpting all data in the category of interest, sorted by hospital, facilitating identification of recurring themes and cross-case analyses. To obtain a measure of leader coaching and to triangulate the measure of psychological safety described above, I used the free-form qualitative data to create measures of psychological safety for speaking up and team leader coaching. Two research assistants rated all quotes identified as related to team leader behavior on a three-point scale from high coaching (creates an open environment, leads discussion, advocates teamwork) to low coaching (doesn't coach, doesn't listen to others, disregards importance of teamwork). Quotations were scrambled so that raters did not confront a cluster of quotes from the same site, creating a bias for similarity of ratings within sites. Using the same approach, they rated quotes previously coded as relevant to speaking up on a three-point scale from high (easy to speak up about anything on one's mind) to low (people appear to be very uncomfortable speaking up and only do it under duress). These two measures of psychological safety for speaking up were correlated (Table 3), providing triangulation of interviewers' assessments of this construct. I subsequently combed through the 628 quotes coded as related to team leader behavior to seek recurring themes and specific behaviors used by the team leaders, for insight into how leaders affect speaking up and learning outcomes. Stage 3. Data on the annual number of cardiac surgery operations in each hospital and on the number of operations conducted in the first six months of using MICS were obtained from each hospital. I calculated implementation success—following Iansiti and Clark (1994)—as the sum of the ranks of three variables: (1) the number of MICS cases conducted in the first six months at each site, (2) the percent of heart operations conducted using MICS in the same period, and (3) whether a site 7 To assess within-site consistency, I used analysis of variance, with the numerical codes as the dependent variable and hospital identity and the independent variable; results verified within-site similarities and between-site differences for both variables (see Appendix for details). 8 The research assistants were rating interviewers' notes and thus – in this sense – testing whether their ratings would replicate interviewers' ratings; however, raters did not know which hospital a given quote came from – preventing bias about the sites and precluding a cognitive tendency to impose within-site consistency. Further, the raters assessed additional data because they had all quotations related to speaking up, not just those in response to a specific question targeting this issue. INTERDISCIPLINARY ACTION TEAMS 17 was increasing, decreasing, or remaining steady in its use of MICS. The measure considered absolute volume, penetration levels and trend, thereby giving credit to several dimensions of adoption success and not unduly penalizing small centers for carrying out fewer MICS operations. I planned this index of team learning outcomes in advance of analysis of the interview data. Stage 4. Because of the small sample size and the nature of our rating systems, I utilized a nonparametric statistical test, Spearman’s rho, to conduct tests of relationships between variables. Nonparametric statistics are inferential tests that do not require assumptions about the distribution of the population from which the samples were taken, and Spearman’s rho is particularly well suited to small samples and to interval-scaled data for which the distribution is not necessarily normal (Saslow, 1982). Although Pearson’s r can inflate the degree of association between two variables of this kind if there are a few extreme values, Spearman’s rho solves this problem by correlating the rank order between two variables. Given the small N and novel, emergent measures, these analyses can be considered tests of the plausibility of the research propositions – that provide insight into relationships and avenues for future research – rather than as conclusive tests of them. Table 3 displays all correlations.