Team leadership and patient outcomes in US psychiatric treatment settings

ARTICLE IN PRESS

Social Science & Medicine 62 (2006) 1840–1852 www.elsevier.com/locate/socscimed

Team leadership and patient outcomes in US psychiatric treatment settings Rebecca Wellsa,, Kimberly Jinnettb, Jeffrey Alexanderc, Richard Lichtensteinc, Dawei Liuc, James L. Zazzalid a

Penn State, University Park, PA 16802-6500, USA b The Wallace Foundation, USA c The University of Michigan, USA d The RAND Corporation Available online 7 October 2005

Abstract Previous studies suggest that psychiatric patients mirror the behaviors of the staff members who treat them, but there is little empirical evidence about how staff dynamics affect patients over time. The goals of this study were to examine associations between: (1) team leader discipline and mutual respect among treatment team members; and (2) mutual respect among team members and improvements in patient quality of life. Two models were tested on data from psychiatric treatment teams within the US Veterans Administration. The first examined associations between the discipline of each team’s emergent leader and the level of mutual respect among that team’s members. The second model tested associations between mutual respect among staff and changes over time in patients’ quality of life. The subjects for model 1 were psychiatric staff members (n ¼ 785) whose responses were aggregated for team-level analyses (n ¼ 78). Mutual respect was highest in social worker-led teams and lowest in physician-led teams. The subjects for model 2 were 1638 seriously mentally ill patients in 44 of the units examined in the first model. When mutual respect among staff was greater, patients improved more over time in their satisfaction with the quality of their housing, relations with families, social life, and finances. Together, these analyses imply that mutual respect may improve patient outcomes and that leadership by some disciplines may facilitate such dynamics. In general, leaders may consider learning from other disciplines’ strengths to improve their impact. r 2005 Elsevier Ltd. All rights reserved. Keywords: US; Teams; Leadership; Psychiatric; Respect; Patient outcomes

Introduction Corresponding author. Tel.: +1 814 865 8893;

fax: +1 814 863 2905. E-mail addresses: [email protected] (R. Wells), [email protected] (K. Jinnett), [email protected] (J. Alexander), [email protected] (R. Lichtenstein), [email protected] (D. Liu), [email protected] (J.L. Zazzali). 0277-9536/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2005.08.060

Although considered essential to coordinating complex care, interdisciplinary collaboration is difficult to actualize in practice, and even more challenging to link to demonstrable patient outcomes. On the face of it, we might expect that a key to success would be the level of respect achieved among disciplines (Gianakos, 1997), including both

ARTICLE IN PRESS R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

members’ deference to each other’s competence and interpersonal consideration (David De Cremer, 2002; Webster’s, 1995). There has, however, been virtually no empirical examination of respect among health care team members (Heiselman & Noelker, 1991). This study employs two complementary empirical models to explore how mutual respect among interdisciplinary team members may evolve and affect patient outcomes. The first model builds on role theory and the sociology of disciplines to examine whether greater mutual respect among staff is more likely when certain disciplines are influential. The second model examines associations between staff mutual respect and patient outcomes over time. This two-model approach allows identification of factors directly affecting patient outcomes and some dynamics that may underlie those factors. The combination of results yields actionable findings for leaders in clinical settings.

The role of mutual respect within treatment teams Maintaining respect among diverse personnel can be challenging for interdisciplinary team leaders. Members of different disciplines often bring conflicting philosophies of care as well as disciplinespecific jargon into their interactions (Colombo, Bendelow, Fulford, & Williams, 2003; VinokurKaplan, 1995). Disciplines also claim unique domains within the treatment process, yet there is actually substantial functional overlap (Herrman, Trauer, & Warnock, 2002). For instance, psychiatrists, psychologists, nurses, and social workers all provide patient therapy. Such role blurring may foster competition among team members, manifested through contention around definitions of patient problems (Sims & Sims, 1993) and what constitutes appropriate care (Caudill, 1958; Stanton & Schwartz, 1954). We propose that the respect that members of such diverse groups show each other may be a key to their collective performance. Previous research has examined respect as an individual phenomenon (Baumeister & Leary, 1995; David De Cremer, 2002; Leary & Baumeister, 2000), albeit sometimes in the context of dyadic relationships (Heiselman & Noelker, 1991) or relative to groups (Ellemers, Doosje, & Spears, 2004). In this study we extend this construct to the group level. Thus, we define mutual respect as the extent to which team members

1841

appreciate each other’s competence and show personal consideration for each other. Leadership discipline and mutual respect Fostering mutual respect among team members falls within the relational dimension of team leadership, the complement to task-related direction that involves attending to ‘‘the interpersonal transactions that take place within the group: who is talking with whom (or not doing so), who is fighting with whom, who is pairing up with whom, and so on’’ (Hackman, 1987, p. 321). This function has been termed ‘‘group maintenance’’ in the team literature (Benne & Sheats, 1948) and ‘‘initiating consideration’’ in the leadership literature (Stogdill, 1974). Team leaders can affect relational dynamics among members in a number of ways: by structuring rewards so that people win by collaborating rather than competing; by intervening when conflicts begin to develop; by teaching staff how to manage such conflicts proactively themselves (Kozlowski, Gully, McHugh, Salas, & Cannon-Bowers, 1996); and through the norms they model in their own behavior. A recent study in a primary care setting found the discipline of program managers to be unrelated to the quality of interdisciplinary collaboration (Sicotte, D’Amour, & Moreault, 2002). In this study, our focus was on emergent rather than the designated leadership Sicotte et al. examined, although of course the two can coincide. That is, we were interested in the role played by individuals identified by their colleagues as especially influential in team processes. This focus is particularly appropriate to studies of interdisciplinary teams, which are supposed to suspend rank when appropriate to defer to members’ situationally relevant expertise (Shaw, 1990). Thus, for instance, even when physicians are present, they may not always emerge as providing the strongest team leadership. As leadership becomes increasingly dispersed among disciplines, it is relevant to ask whether some disciplines may facilitate more mutual respect than do others. Because there is now substantial evidence from both medical sociology and social psychology attesting to the differences between disciplines, there are also likely to be systematic differences in leadership behavior across disciplines. Differences between disciplines may begin with self-selection, with medicine appealing to people

ARTICLE IN PRESS 1842

R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

who value autonomy highly and nursing and social work drawing those who are more interested in teamwork. Any such initial differences are then likely to be reinforced through the different disciplines’ disparate educational experiences. Social work curricula incorporate sociology, psychology, administration, and law, in addition to medicine (Snelling, 1974). Nursing education stresses a holistic view of patients, including psychological and social factors. By contrast, medical school emphasizes somatic conditions and intervention (McMahon, Hoffman, & McGee, 1994). After training, the experience of care provision continues to differ across disciplines (Liedtka & Whitten, 1998; Parker, 2002; Tjosvold & Fottler, 1996; Willshire, 1999). In inpatient care, nurses manage coordination with other units within hospitals, and are therefore cognizant of care processes’ interdependence (Gibson, 1989). Social workers serve as liaisons between treatment teams and families (Gibson, 1989) and between patients and community resources (Mellor & Lindeman, 1998; Toseland, Palmer-Ganeles, & Chapman, 1986). Thus, social workers’ functions emphasize mutual dependence as well, although in somewhat different ways than nurses’ roles do. In contrast, psychiatrists often feel pressured to focus more narrowly on issues such as diagnosis and medication, especially in light of increasing pressures for efficiency (Andreason, 2001; Willshire, 1999). Differences in personalities, socialization, and functional roles may all contribute to variations across disciplines in their priorities for clinical leadership. In previous research, although physicians, social workers, and nurses all agreed that respect for colleagues was important, social workers (Abramson & Mizrahi, 1996) and nurses were most likely to choose respect and trust as the first priority, whereas physicians were most likely to choose competence (Bates, 1966; Prescott, 1985). Together, these factors imply that treatment team leader discipline can have substantial implications for mutual respect within teams. Nurses and social workers may be more predisposed by temperament toward teamwork; have more emphasis in their training on collaboration; and experience more interdependence on the job than physicians do. For these reasons, nurses and social workers may be more effective than physicians at fostering professional mutual respect within the teams they lead.

This leads to our first hypothesis: H1. Psychiatric treatment teams led by either registered nurses (RN) or social workers will have higher levels of mutual respect than teams led by physicians. Effects of mutual respect among staff on psychiatric patient outcomes When staff members value each others’ competence, they are likely to both volunteer more information and hear more of what each other is saying (Weick & Roberts, 1993). Such communication supports coordination and problem solving (Caudill, 1958), and may thus improve patient care (Knaus, Draper, Wagner, & Zimmerman, 1986; Parker, 2002; Schmidt & Svarstad, 2002). In mental health care, the architect of an early experimental ward attributed its unusually high recovery rates largely to harmonious relationships among staff (Perry, 1982). Conversely, other case studies conducted in psychiatric contexts found that conflicts within treatment staffs led to disturbances among patients (Caudill, 1958; Stanton & Schwartz, 1954). Mutual respect may also affect the social support that staff members experience. This can be one of the most powerful benefits of collaboration, especially when work conditions are discouraging, as in mental health care (Toseland et al., 1986). Research has found that human services staff members who have strong social support are less likely to suffer burnout, a process in which emotional exhaustion can lead to diminished care effectiveness (Pines, 1983). Likewise, Kahn (1993) has argued that in a care giving organization, where staff members can easily be depleted of their emotional energies, the support of colleagues can enable staff members to release these stores of energies to achieve clinical and organizational goals. Thus, mutual respect can be viewed as an indicator of the kind of supportive environment that Kahn and Likert (1961) before him, believed would motivate individuals to achieve both personal and job-related goals. Finally, the level of mutual respect among staff affects the models available to patients for negotiating conflict in their own lives. Severe mental illness distorts people’s perceptions of social situations and undermines their ability to perform interpersonally. Thus, one function of psychiatric care is helping patients develop social skills. Such skills may not only facilitate progress within treatment but also

ARTICLE IN PRESS R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

help patients minimize the impact of disease in their day-to-day lives. The ability to show people respect is essential to negotiating and retaining housing and employment and interacting with family members and others. Thus, by improving both task and relational communication, sustaining treatment staff emotionally, and providing positive behavioral models, mutual respect among staff may affect patients’ ability to improve their quality of life over time. This leads to the second hypothesis: H2. Mutual respect among psychiatric treatment staff will be associated with patients’ improvement in quality of life over time. To explore these two hypotheses we ran two separate models. The first model simply tested for an association between leader discipline and mutual respect, using multivariate regression to control for potential confounders. The second model tested the relationship between team level mutual respect and change in patient outcomes over time; the nested nature of these data necessitated multilevel modeling. These two models are now described. Methods for model 1: treatment team leadership and mutual respect among staff The first hypothesis was that teams led by social workers or nurses would have higher mutual professional respect among staff than those led by physicians. Sample The sample in the first model consisted of 78 Veterans Administration (VA) psychiatric treatment teams operating in units serving the seriously mentally ill (with one team within each unit). The 51 inpatient and 27 outpatient units were selected for inclusion in the study because they met the criterion of at least 50% of their patients having a serious mental illness, had a cumulative length of stay in all veteran medical centers of at least 150 days in the past year and/or had 5 or more admissions to any veteran medical center in the past year. Over one-third of the sample were registered nurses (37%, hereinafter called nurses), followed by nurse aides (16%), social workers (14%), and physicians (8%). The remaining 25% included, in decreasing order of frequency, licensed practical nurses, recreational therapists, and psy-

1843

chologists, among other occupations. Because day and night shifts tend to have very little interaction, only individuals working during the day were included in this sample. The data source was a survey addressing a variety of job-related attitudes, distributed to all direct patient care providers in sample units in 1992. Items addressed job satisfaction, professional relations, and respondent background, among other factors. There was a useable response rate of 94%. Team membership was defined through individuals’ responses to a survey question about whether or not they belonged to a treatment team, and if so, which was their primary team. Teams with fewer than three members were excluded from the sample. This yielded an initial sample of 94 teams. For interpretive clarity, the regression was then run on the 78 teams whose modal perception of leader was physician, social worker, or nurse. Of the 16 omitted teams, four had selected psychologists as leaders, five had selected ‘‘other,’’ and seven were tied in their perception. There was a much lower % of nurses in social worker-led teams, which raised the possibility that it was having fewer nurses rather than being led by social workers that was associated with better mutual professional respect. The pattern was the same, however, in analyses run omitting all nurses. Dependent variable Mutual respect among staff The scale assessing mutual respect among staff members included four items. The first two relate to recognition of others: ‘In general, the people I work with are excellent mental health professionals’ and ‘My coworkers have a strong interest in helping their patients.’ The second two relate to the extent to which the focal individual perceives that s/he is recognized by others: ‘My coworkers appreciate my efforts’ and ‘The people I work with take a personal interest in me.’ This scale had not been part of the original design of the larger study. Instead, the measure emerged from principal components and maximum likelihood factor analysis of a range of treatment staff satisfaction dimensions. These techniques supported a single-factor solution for the four items listed above, with an alpha coefficient (0.77) indicating sufficient internal consistency to describe them as indicating a common latent construct. The question then was, what did these four items collectively

ARTICLE IN PRESS 1844

R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

indicate? Complementing our empirical analyses with logical structural analysis (Pedhazur & Schmelkin, 1991), we identified this scale as congruent with the task and personal dimensions of respect indicated by both the dictionary definition of this term (Webster’s, 1995) and emergent themes from Heiselman and Noelker’s (1991) qualitative study of nursing assistants’ experiences of respect from patients and their families. A high level of within-team agreement on this factor, indicated by a Rho Within Group index (James, Demare, & Wolf, 1984) of 0.89 (on a scale of 0–1), justified aggregation to the team level. Statistical tests (PROC GENMOD, in SAS), indicate that this attribute was also generally stable: of the three leader disciplines chosen by teams, only agreement on physician leadership changed significantly, and then only between the first and second waves of data collection. Independent variables Discipline of the leader This variable was based on a prompt for each team member: ‘In your view, who provides the strongest leadership on your team?’ Options included physician, head nurse, and social worker, as well as psychologist and ‘it depends on the type of patient.’ The team level variable was constructed as the modal perception within each team. This measure captured the emergent nature of leadership by measuring shared perceptions of influence within teams rather than formally designated roles. In addition to our theoretical focus on leader discipline, we included two control variables because of their potentially confounding effects on mutual respect within treatment teams: Team size Team size was measured as the number of patient care staff assigned to a unit who indicated that they belonged to the treatment team. This was often a subset of the unit size because, for instance, nurse aides often did not belong to the treatment team. Previous research has linked larger team size to greater problems in interpersonal dynamics (Mullen, Symons, Hu, & Salas, 1989; Thomas & Fink, 1963). Median tenure in team This variable indicated the median tenure of team members in their positions. This was intended to

control for the potentially confounding effect of team maturity on team dynamics. Analyses The first model was a cross-sectional ordinary least squares regression testing the association between the modal perception among team members of which discipline ‘provides the strongest leadership on your team’ and their average reported level of mutual respect. Power analyses using PASS 6.0 software indicated a 92% chance of detecting a covariate explaining 10% of the variance in the dependent variable at alpha ¼ 0.05. Results: discipline of the leader and mutual respect among staff Physicians were the most frequently perceived as team leaders (in 56% of all 94 teams) even though they were only present in 70% of the teams in the sample: 75% of those teams that included a physician chose that discipline as providing the strongest leadership. In contrast, 80 teams included social workers but only 17 of them (21%) chose that discipline as the most influential, and 12 out of 83 teams (15%) with nurses chose nursing as providing the strongest leadership (Table 1). After restricting the sample to those in teams that had chosen physicians, nurses, or social workers as most influential, there were 785 individuals within 78 teams. On average, physicians were older (58 years) than social workers (48) and nurses (46). Physicians were also more frequently male (75%) and white (75%) than were social workers (55% male and 66% white) and nurses (15% male and 45% white). The mean position tenure was between 512 and 634 years for all three disciplines. Table 1 Descriptive statistics for model 1: team leader discipline and mutual respect among staff (N ¼ 78) Variable

Mean

Range

Std. dev.

Mutual respect among staff Team size Median tenure in team (in years) Physician led Social worker led Nurse led

5.5 10.1 3.5

0–7 3–32 0–17

0.5 5.2 2.5

63% of units 22% of units 15% of units

NA NA NA

0.5 0.4 0.4

ARTICLE IN PRESS R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852 Table 2 Cross-sectional regression model: team leader discipline and mutual respect among staff (N ¼ 78Þ Variable

Coefficient

Std. error

Constant Team size Median tenure in team RN led Social worker led

5.56***
0.04** 0.03 0.28+ 0.53***

0.15 0.01 0.02 0.14 0.15

Adjusted R2 ¼ 0:352. +po0.10; *po0.05; **po0.01; ***po0.001.

As shown in Table 2, team leadership by nurses, relative to physicians, was positively related to the degree of mutual respect team members reported having with coworkers, at a marginal level of statistical significance (p ¼ 0:05). Leadership by social workers was strongly associated with mutual respect, relative to physician leadership (po0:001). Mutual respect among staff was also lower in larger teams (po0:01), in keeping with previous research indicating that interpersonal relations suffer in larger groups (Thomas & Fink, 1963). This pattern of results indicates that mutual respect within psychiatric treatment teams did vary with the discipline of teams’ emergent leaders. Specifically, there was suggestive evidence that leadership by nurses rather than physicians was associated with higher mutual respect among staff, and leadership by social workers was significantly related to mutual respect. Unfortunately, we did not have data about the demographics of leaders as perceived by all team members (our operationalization of emergent leadership). However, the survey did ask each person how much influence s/he had over team decisions. Viewing self-perception of leadership as a reasonable proxy for leadership as perceived by others, we regressed individuals’ perceived influence in the team on discipline, sex, age, and race. The variables for discipline were the only significant predictors. This supports our contention that discipline per se, rather than its demographic correlates, was associated with mutual respect among team members. Methods for model 2: mutual respect among staff and patient outcome improvements The second hypothesis was that mutual respect among staff would affect patients’ abilities to improve their quality of life over time.

1845

Sample The sample for this model consisted of 1638 seriously mentally ill patients who received treatment between 1992 and 1999 within the 44 units from the first sample for which patient outcome data were available. Treatment staff identified patients belonging to each unit. The primary diagnosis for the majority (82%) of patients was a schizoaffective or schizophrenic disorder; for 14% it was a bipolar or major depression; and for 3% the diagnosis was dementia or alcohol-related disorder. The measure of mutual respect came from the 1992 job-related attitudes survey described previously. Although there were two later waves of this survey (in 1994 and 1995), repeated measures analyses revealed no significant change in mutual respect over time. Thus, only 1992 data were included in the analysis. The stability of mutual respect in the sample may have been partially attributable to the teams’ stage of development of the teams, which averaged 3.5 years old at the beginning of our study. Research on work groups has found that, after an initial period of negotiation, members quickly establish norms that are then generally very stable (Gersick, 1989; Tuckman & Jensen, 1977). Power analyses conducted in STATA indicated an 80% chance of detecting a moderate (0.46) effect size. Patient data were collected as part of the US Department of Veterans’ Affairs’ Long-Term Mental Health Enhancement Program in its 29 longterm neuropsychiatric hospitals. As part of this longitudinal evaluation, trained clinicians collected patient data using the Quality of Life Scales described below. A survey of unit directors provided data on the mean psychosocial functioning of patients on the unit (through the Global Functioning score, also described more below) and team size. Dependent variables Quality of Life Scales Dependent variables in the patient outcomes models are based on Lehman’s Psychiatric Quality of Life Interview, specifically designed to be used with seriously mentally ill patients (Lehman, 1988, 1991). The Lehman Quality of Life Scales measure dimensions of individual quality of life across a variety of areas, including housing, leisure activities, family and social relations, health, safety, and finances. For instance, the housing scale includes

ARTICLE IN PRESS 1846

R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

how the patient feels about ‘the people who live (in the houses, apartments, near yours/on your unit or ward)’; ‘people who live in this community or town’; ‘the outdoor space there is for you to use’; ‘the particular (neighborhood/hospital) as a place to live’; ‘this community/town as a place to live’; and ‘how safe you feel in this (neighborhood/hospital)?’ Each of the seven domains involved in the current study was assessed through 4–6 items, each of which was measured on a 1-to-7-point ‘‘feeling’’ scale, where 1 indicates ‘‘terrible’’ and 7 indicates ‘‘delighted.’’ The psychometric properties of the Quality of Life interview have been extensively studied (Becker, 1995; Lehman, 1991). Cronbach alphas range from 0.79 to 0.88; test–retest reliability ranges from r ¼ 0:41 to 0.95; factor analyses have effectively discriminated between Quality-of-Life and Mental-Health Scales (Lehman, 1988). The Lehman Quality of Life Scales are more reliable with the Schizoaffective/Schizophrenia and Bipolar/ Major Depression groups than those patients with dementia and alcohol-related diagnoses. The latter two categories, however, represented only 3.4% of the current sample (Atkinson & Zibin, 1996, Chap. 5). Trained clinical staff assisted patients in completing the Quality of Life instrument at program entry, every 6 months for 2 years, and every year thereafter for the remainder of the client’s enrollment. Therefore, quality of life is a time-varying dependent variable and, as such, permits assessment of change in patient quality of life over time. The days variable described later represents the time basis for the quality of life assessments.

illness. High score ranges indicate minimal symptoms and good functioning; low ranges may indicate persistent danger of hurting or killing someone, although they may also simply indicate very low ability to function or communicate. The composite patient severity measure for the team was calculated as a weighted mean of the proportion of patients on each unit falling in each functioning score range. Unit size. Total number of patient care staff assigned to a unit. This is different than the team size control used in model 1. In that model, we were interested in the possible confounding effects of team size on team dynamics. In model 2, we wanted to control for the possible effect of larger unit size on patient outcomes. Thus the total number of staff, rather than just in the official treatment team, was the relevant control variable in model 2. Individual level Prior inpatient days. Number of days hospitalized in a VA Medical Center for a mental health condition in year prior to program enrollment. This controlled for individual patients’ severity of illness. Age. Measured as patient age in years. This controlled for the potentially confounding effect of age on quality of life. Diagnosis. Defined as the patient’s primary diagnosis upon program entry. Dementia and alcoholrelated disorders were used as the referent group, with schizoaffective disorders and schizophrenia and bipolar and major depression as the two comparison categories.

Independent variables Team level Mutual respect among staff. Measured as indicated in the first methods section. Mean patient psychosocial functioning in team. To control for the functional status of patients assigned to a treatment team, unit heads were asked to provide the percentage of patients on their unit who fell into each score range of the Global Assessment of Functioning Scale (GAF). The higher the mean, the higher the average level of patient function on the team. The Functioning Scale was developed by the American Psychiatric Association to measure the functional status of persons with serious mental

Time level Days. Number of days from baseline (program entry) on which a patient has been assessed for quality of life. Days represent the time duration between assessments of client’s quality of life, the dependent variable. This allowed us to examine associations between mutual respect among staff members at baseline and changes in patient outcomes over time. Analyses The second set of regressions modeled each dimension of individual patients’ quality of life on independent variables at three levels of analysis:

ARTICLE IN PRESS R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

1847

time (level 1), individual (level 2), and team (level 3), using SAS PROC MIXED. Multilevel analysis (also known as hierarchical linear modeling and random coefficients or random effects modeling) accommodates nested data structures such as these by separating within-team and within-person variance from between-team and between-person variance (Bryk & Raudenbush, 1992). This corrects for disaggregation and aggregation biases by estimating parameters at different levels of analysis, thus explicitly accounting for the non-independence of individuals within the same group (Hofmann, 1997). Multilevel analysis also allows for variable assessment schedules and missing assessments within clients.

portion of the 0–90 range. The average unit had 17.3 staff members and the mean rating of mutual respect among staff was 5.5, the same as the mean in the larger sample examined in the first model. Intraclass correlation coefficients show how much variance in dependent variables is attributable to each level of analysis relevant to the model. On average across the seven Quality of Life Scales, 41% of the variance occurred at the time level (level 1), 47% of the variance was at the patient level (level 2), and 12% was at the unit level (level 3). The significant amount of variance at each level indicates the need for multilevel methods to partition variance appropriately.

Results: staff mutual respect and improvements in patient quality of life

Final multilevel model

The means for patient Quality of Life Scales varied from 4.3 for finances to 5.0 for safety on a 1–7 scale (Table 3). The average patient was 60 years old and had been in a veterans’ medical facility more days in the past year (264) than he had been out. The average number of days in treatment during the study was 618. The mean team level functioning score was 37.2, at the low-middle

Table 4 lists the results of the final full multilevel models of patient quality of life. These are common, saturated models for the purpose of comparing the effects of a single set of independent variables across seven dimensions of patient quality of life. The row showing associations between mutual respect among staff and changes in patient quality of life over time is shaded.

Table 3 Descriptive statistics for model 2: mutual respect among staff and patient outcomes Individual level variables (N ¼ 1638Þ

Mean

Range

Std. dev.

Quality of life Housing Leisure Family Social Health Safety Finance Patient age Prior inpatient stay: number of days in VAMC in past year Number of days in study

4.7 4.7 4.7 4.8 4.8 5.0 4.3 60.6 263.6 617.7

1–7 1–7 1–7 1–7 1–7 1–7 1–7 31–96 0–365 0–2675

1.1 1.0 1.2 1.0 1.0 1.0 1.5 12.8 107.3 522.0

Primary diagnosis

Frequency (%)

Range

Std. dev.

Schizoaffective & schizophrenic Bipolar & major depression Dementia & alcohol-related Unit level variables (N ¼ 44) Mean patient functioning: GAF score Unit size Mutual respect among staff

82.5 14.1 3.4 Mean 37.2 17.3 5.5

11.8–59.0 4–39 4.4–6.8

11.1 10.1 0.6

*po0.05; **po0.01; ***po0.001.

Bipolar

Psychiatric diagnosis  time Dementia/alcohol

Patient’s age  time

Individual level effects on slope Prior inpatient stay  time

Mutual respect among staff  time

Team size  time

Team level effects on slope Mean patient functioning  time

Overall slope (change over time)

Bipolar diagnosis

Psychiatric diagnosis Dementia/alcohol

Patient age

Individual level effects on intercept Prior inpatient stay

Mutual respect among staff

Unit size

Team level effects on intercept Patient functioning on the unit (mean GAF)

Overall intercept (at program entry)

Variables


0.0002 (0.0001)
0.0000 (0.0000)

9.128E
8*** (0)
4.75E
6** (2.333E
6)


5.6E
6* (3.295E
6) 7.371E
6** (3.544E
6) 0.0001119* (0.0000)


0.0001 (0.0004)


0.2489 (0.1604) 0.0966 (0.0750)


0.0006** (0.0002) 0.0042* (0.0022)

0.0134*** (0.0042)
0.0026 (0.0048)
0.0044 (0.0902)

3.9617*** (0.5935)

Housing


0.0001 (0.0001)
0.0000 (0.0000)

1.291E
7*** (0)
6.11E
6*** (2.107E
6)

1.259E
6 (4.67E
6) 2.99E
6 (5.436E
6) 0.000076 (0.0001)


0.0000 (0.0006)

0.0352 (0.1559) 0.3078*** (0.0713)

0.0001 (0.0002) 0.0059*** (0.0022)

0.0055 (0.0047)
0.0041 (0.0055)
0.0787 (0.1010)

4.3200*** (0.6545)

Leisure

Table 4 Longitudinal multi-level model: mutual respect among staff and patient outcomes


0.0002 (0.0001) 0.0000 (0.0000)


1.62E
7*** (0)
5.22E
6** (2.653E
6)

1.075E
6 (3.865E
6) 4.219E
6 (4.215E
6) 0.000158** (0.0000)


0.0007 (0.0005)

0.2965 (0.2007) 0.1356 (0.089)

0.0004 (0.0003)
0.0003 (0.0027)

0.0010 (0.0040) 0.0029 (0.0042)
0.1003 (0.0799)

4.9726*** (0.5523)

Family


0.0001 (0.0001)
0.0000 (0.0000)


1.72E
8*** (0)
3.32E
6 (2.026E
6)

1.062E
6 (4.283E
6) 3.774E
6 (4.969E
6) 0.000154* (0.0000)


0.0006 (0.0005)

0.1222 (0.1495) 0.2484*** (0.0682)

0.0002 (0.0002) 0.0046** (0.0021)

0.0059 (0.0042)
0.0019 (0.0049)
0.1208 (0.0908)

4.6608*** (0.5925)

Social


0.0000 (0.0001)
0.0000 (0.0000)


1.34E
7*** (0)
6.47E
6*** (1.903E
6)

3.004E
6 (3.489E
6) 2.251E
6 (3.969E
6) 0.000062 (0.0000)


0.0000 (0.0004)

0.0788 (0.1503) 0.2876*** (0.0694)

0.0001 (0.0002)
0.0000 (0.0021)


0.0019 (0.0041) 0.0032 (0.0047) 0.0253 (0.0876)

4.3695*** (0.5737)

Health


0.0000 (0.0001)
0.0000 (0.0000)

1.49E
7*** (0)
1.04E
6 (2.132E
6)

7.412E
7 (4.217E
6) 4.345E
6 (4.837E
6) 0.00007 (0.0000)


0.0004 (0.0005)

0.0710 (0.1597) 0.0755 (0.0741)


0.0001 (0.0002) 0.0024 (0.0023)

0.0037 (0.0053) 0.0001 (0.0062) 0.0225 (0.1138)

4.5238*** (0.7329)

Safety


0.0002 (0.0001)
0.0000 (0.0000)

8.97E
7*** (0) 8.129E
7 (3.097E
6)

6.143E
6 (6.56E
6) 3.461E
6 (7.557E
6) 0.000236* (0.0001)


0.0012 (0.0009)

0.0746 (0.2319) 0.4032*** (0.1055)


0.0006* (0.0003)
0.0010 (0.0032)


0.0111* (0.0064) 0.0031 (0.0073)
0.1792 (0.1345)

5.2273*** (0.8784)

Finance

1848 R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

ARTICLE IN PRESS

ARTICLE IN PRESS R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

Results: effects of mutual respect among staff and quality of life There were no associations between mutual respect among staff and patients’ quality of life at the time of unit entry. However, patients in units with greater mutual respect among staff did improve more over time in four quality of life dimensions: housing (po0:05), relations with family (po0:01), social life (po0:05), and finances (po0:05). Recalling the modest statistical power available with 44 units, we can consider these associations to have met a very conservative test of significance. Fig. 1 uses housing to illustrate the effect of staff mutual respect on the client’s satisfaction on quality of life over time. The figure reveals that patients in units with high levels of mutual respect among team members improve in quality of life over time more rapidly than patients treated in units with lower levels of staff mutual respect (improving on average by a full point within the 7-point scale over 3 years, versus virtually no improvement for patients in low mutual respect units). These effects suggest that high levels of staff mutual respect may facilitate patient progress on quality of life.

1849

Unit level effects Average patient functioning within individual units was positively associated with patient satisfaction with the quality of housing situations reported at the time of unit entry as well as with decreases in their satisfaction with housing reported over time. Average patient functioning in the unit was negatively associated with patients’ financial conditions upon unit entry. Patients in larger units on average improved more over time in housing status. Individual level effects The greater the number of days patients had been in inpatient units prior to program enrollment (prior inpatient stay) the lower their ratings of the quality of their housing and financial situations at the time of program entry. Longer prior inpatient stay of clients was also associated with greater improvements in satisfaction with housing, leisure, safety, and finance, and decreases in satisfaction with family relations, social life, and health over time. Older patients reported better housing, leisure, and social situations upon program entry. Age was negatively associated with changes over time, however, in housing, leisure, family relations, and health. Discussion

Effects of control variables on patient quality of life In addition to the hypothesized effects, there were some significant associations between controls and outcomes.

Quality of Life - Housing

Low mutual respect team High mutual respect team 7.00 6.00 5.00 4.00 3.00 2.00 1.00 entry

1 year

3 years

Time in Program Fig. 1. Effects of mutual respect among staff on quality of life over time. Using housing as an example. Note: For graphical display purposes quality of life was predicted at three time points (0 days, 365 days, 1095 days) and two team mutual respect levels (1 ¼ low and 7 ¼ high). All other variables were held constant as follows: Prior Inpatient Stay ¼ 500, Age ¼ 75, Diagnosis ¼ 1, Mean Patient Functioning ¼ 30, Team size ¼ 10.

Our study yields two central findings. First, in the psychiatric treatment teams studied, leadership by different disciplines was associated with varying levels of mutual respect among staff members. In developing our hypotheses, we speculated that nurses and social workers might self-select into these professions as well as be trained and socialized in ways that facilitated positive interdisciplinary relations. The data supported our prediction, with nurse relative to physician leadership associated with higher mutual respect at a marginal level of statistical significance and social worker leadership associated with significantly higher mutual respect. This does not necessarily mean that nurses and social workers are more effective leaders in a global sense. Leadership is a multidimensional function, involving many aspects of both task and group maintenance. Nor are such patterns immutable. Thus, we would not suggest that teams choose different disciplines to lead them (even if we thought that someone would follow that advice!). Instead, our recommendation is that each discipline learns from each other’s strengths to optimize its own leadership performance. Future, more fine-grained

ARTICLE IN PRESS 1850

R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

studies may determine what factors produced the pattern of associations reported here. If some disciplines are more effective at empathic listening, validation, or mediation, these skills could be more systematically incorporated into the curricula of other disciplines. Such training could also be provided through continuing medical education. Even in the absence of this level of resource investment, teams may benefit from occasional, brief discussions about process. For instance, leaders might ask members what they thought the team was doing well, what it might do better, and how. Such simple interventions could improve mutual respect both by the very act of asking for feedback and acting on constructive input. Previous literature suggests that differences in both training and work experience foster a distinct set of behaviors among social workers (Abramson & Mizrahi, 1996; Snelling, 1974), especially in contrast to physicians. Social workers’ emphases on more holistic psychosocial approaches to treatment and systems dynamics, however, are shared with nurses (Bates, 1966; Gibson, 1989; McMahon et al., 1994; Prescott, 1985). Thus it was somewhat surprising that nurse (vs. physician) leadership was not more significantly associated with the level of mutual respect found among staff. Perhaps because of social workers’ boundary spanning roles, they are uniquely qualified to bridge disciplinary differences in their team facilitation. The second major finding is that mutual respect within treatment teams was associated with differential patient improvement over time in some areas of quality of life. This finding as well is open to more than one interpretation. Because we did not test the mediating roles of enhanced coordination and social modeling within treatment teams, we do not know how much each may have contributed to improved patient outcomes. Relative to patients’ financial status, we might infer that coordination plays a central role: perhaps more mutually respectful treatment teams were more successful at working together to help patients manage the complex transition into stable employment and/or avoid economically costly life disruptions and were more respectful in their interactions with individual patients. Social modeling seems like a more likely candidate for helping patients improve their relationships with their families: one can imagine, for instance, that patients who saw more respectful interactions among treatment staff might mirror those behaviors in interactions with family mem-

bers. We could imagine both coordination and social modeling affecting patient housing status, given that both logistical and interpersonal factors may affect the ability to secure and retain, for instance, stable rental situations. The lack of effect of mutual respect on the leisure, health and safety dimensions was puzzling. Because the Veterans Health Administration is focused on the provision of health care to veterans and has engaged in safety-related initiatives (VA, 2005), there may be constant press on all teams regardless of mutual respect levels to be accountable for health and safety issues. Similarly, perhaps satisfaction with leisure activities is more responsive to organizational budgets and community offerings than to the degree of mutual respect among staff members on teams. It is possible that the respect staff members show each other affects quality of life more than other patient outcomes. This aspect of generality is difficult to speak to without knowing more about which mediating mechanisms led to the effects seen here. In settings in which mutual respect among staff members does affect patient outcomes, however, managers should consider these interpersonal dynamics a patient quality issue. Although our results suggest a number of potentially important associations between team leadership, team dynamics, and patient outcomes, this was an observational study and thus may have suffered from reverse causality and/or omitted variable bias. It is possible that the association between leader discipline and mutual respect is caused by a common, unmeasured factor. For instance, certain belief structures within teams (such as a greater emphasis on systems factors) may lead members to both see social workers as providing leadership and value each other’s contributions to the team. Conversely, physicians may more frequently lead teams in settings characterized by other unmeasured factors that affect mutual respect among team members (such as settings where medical training occurs). Other factors that are correlated but not causally linked with mutual staff respect, such as levels of interaction among team members, may also affect patient outcomes. Future longitudinal and qualitative studies may disentangle causality in relation to disciplinary background and team dynamics. One might also ask how relevant these findings are, given the continuing trend toward shorter patient lengths of stay. Two elements of the analysis

ARTICLE IN PRESS R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

partially account for this. First, the model shown on Table 4 accounted for the duration of prior exposure to care; thus, the patterns of association reflect additional increments rather than cumulative totals of care. Second, the sample included both inpatient and outpatient units. Thus, these findings are not only applicable to patients with extended inpatient stays. Finally, the patient population in this study was the seriously mentally ill, operationalized in terms of high prior utilization. This raises questions about the relevance of this study to other care settings. We believe the relevance is quite high, given the increasing emphasis on long-term management of chronic conditions. Any lessons we can derive from previous generations of mentally ill veterans may help us better serve future survivors. This study does support the general contention that interpersonal dynamics among treatment team members can affect patient outcomes. This is a powerful insight that merits both further empirical examination and managerial action. Managers and team leaders should consider sanctions and rewards that discourage competition and encourage mutual respect. They should intervene in constructive ways when staff conflicts begin to develop and support staff in learning how to manage such conflict themselves and model the respect they seek to foster in their team. Finally, even in the absence of structured training programs, clinicians who are not initially successful in facilitating mutual respect among other staff may learn from members of other disciplines, such as leaders of other units. In so doing, they may not only help improve team coordination and communication, but patient outcomes as well. Acknowledgements Data collection was supported by the US Veterans Health Administration under the auspices of its Serious Mental Illness Treatment Research and Evaluation Center (SMITREC). The authors wish to thank Frederic C. Blow and Stephanie Visnic for their contributions to this study, as well as all the clinicians and patients who participated in data collection. References Abramson, J. S., & Mizrahi, T. (1996). When social workers and physicians collaborate: Positive and negative experiences. Social Work, 4(1), 270–281.

1851

Andreason, N. C. (2001). Diversity in psychiatry: Or, why did we become psychiatrists? The American Journal of Psychiatry, 158(5), 673–675. Atkinson, M. J., & Zibin, S. (1996). Evaluative review of quality of life instruments. In Quality of life measurement among persons with chronic mental illness: A critique of measures and methods. Retrieved April 17, 2005, from http://www.phacaspc.gc.ca/mh-sm/mentalhealth/pubs/quality_of_life/. Bates, B. (1966). Nurse–physician teamwork. Medical Care, 4, 68–80. Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117(3), 497–529. Becker, M. (1995). Quality-of-life instruments for severe chronic mental illness. Implications for pharmacotherapy. Pharmacoeconomics, 7(3), 229–237. Benne, K. D., & Sheats, P. (1948). Functional roles of group members. Journal of Social Issues, 4(2), 41–49. Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models. Newbury Park, CA: Sage. Caudill, W. (1958). The psychiatric hospital as a small society. Cambridge, MA: Harvard University Press. Colombo, A., Bendelow, G., Fulford, B., & Williams, S. (2003). Evaluating the influence of implicit models of mental disorder on processes of shared decision making within communitybased multi-disciplinary teams. Social Science & Medicine, 56(7), 1557. David De Cremer. (2002). Respect and cooperation in social dilemmas: The importance of feeling included. Personality and Social Psychology Bulletin, 28(10), 1335–1341. Ellemers, N., Doosje, B., & Spears, R. (2004). Sources of respect: The effects of being liked by ingroups and outgroups. European Journal of Social Psychology, 34, 155–172. Gersick, C. J. G. (1989). Marking time: Predictable transitions in task groups. Academy of Management Journal, 32(2), 274–309. Gianakos, D. (1997). Physicians, nurses and collegiality. Nursing Outlook, 45, 57–58. Gibson, D. (1989). Requisites for excellence: Structure and process in delivering psychiatric care. Occupational Therapy in Mental Health, 9(2), 27–52. Hackman, J. R. (1987). The design of work teams. In J. W. Lorsch (Ed.), Handbook of organizational behavior (pp. 315–342). Englewood Cliffs, NJ: Prentice-Hall. Heiselman, T., & Noelker, L. S. (1991). Enhancing mutual respect among nursing assistants, residents, and residents’ families. Gerontologist, 31(4), 552–555. Herrman, H., Trauer, T., & Warnock, J. (2002). The roles and relationships of psychiatrists and other service providers in mental health services. Australian and New Zealand Journal of Psychiatry, 36, 75–80. Hofmann, D. (1997). An overview of the logic and rationale of hierarchical linear models. Journal of Management, 23, 723–744. James, L. R., Demare, R. G., & Wolf, G. (1984). Estimating within-group interrater reliability with and without response bias. Journal of Applied Psychology, 69(1), 85–98. Kahn, W. A. (1993). Caring for the caregivers: Patterns of organizational caregiving. Administrative Science Quarterly, 38(4), 539–563. Knaus, W. A., Draper, E. A., Wagner, D. P., & Zimmerman, J. E. (1986). An evaluation of outcome from intensive care in major medical centers. Annals of Internal Medicine, 104, 410–418.

ARTICLE IN PRESS 1852

R. Wells et al. / Social Science & Medicine 62 (2006) 1840–1852

Kozlowski, S. W. J., Gully, S. M., McHugh, P. P., Salas, E., & Cannon-Bowers, J. A. (1996). A dynamic theory of leadership and team effectiveness: Developmental and task contingent leader roles. In G. R. Ferris (Ed.), Research in personnel and human resources management. Greenwich, CT: JAI Press. Leary, M. R., & Baumeister, R. F. (2000). The nature and function of self-esteem: Sociometer theory. Advances in Experimental Social Psychology, 32, 1–62. Lehman, A. (1988). A quality of life interview for the chronically mentally ill. Evaluation and Program Planning, 11, 51–62. Lehman, A. (1991). Quality of life interview: Core version manual. Unpublished manuscript. Liedtka, J. M., & Whitten, E. (1998). Enhancing care delivery through cross-disciplinary collaboration: A case study. Hospital & Health Services Administration, 185–205. Likert, R. (1961). New patterns of management. New York: McGraw-Hill. McMahon, E. M., Hoffman, K., & McGee, G. W. (1994). Physician–nurse relationships in clinical settings: A review and critique of the literature, 1966–1992. Medical Care Review, 51(1), 83–112. Mellor, J. M., & Lindeman, D. (1998). The role of the social worker in interdisciplinary geriatric teams. Journal of Gerontological Social Work, 30(3–4), 3–7. Mullen, B., Symons, C., Hu, L.-T., & Salas, E. (1989). Group size, leadership behavior, and subordinate satisfaction. The Journal of General Psychology, 116(2), 155–169. Parker, V. A. (2002). Connecting relational work and workgroup context in caregiving organizations. The Journal of Applied Behavioral Science, 38(3), 276–297. Pedhazur, E. J., & Schmelkin, L. P. (1991). Measurement, design, and analysis: An integrated approach. Hillsdale, NJ: Lawrence Erlbaum Associates. Perry, H. S. (1982). Psychiatrist of America: The life of Henry Stack Sullivan. Cambridge, MA: The Belknap Press. Pines, A. (1983). On burnout and the buffering effects of social support. In B. A. Farber (Ed.), Stress and burnout in the human service professions (pp. 155–174). Toronto: Pergamon Press. Prescott, P. A. (1985). Physician–nurse relationships. Annals of Internal Medicine, 103, 127–133. Schmidt, I. K., & Svarstad, B. L. (2002). Nurse–physician communication and quality of drug use in Swedish nursing homes. Social Science & Medicine, 54, 1767–1777.

Shaw, R. B. (1990). Mental health treatment teams. In J. R. Hackman (Ed.), Groups that work (and those that don’t): Creating conditions for effective teamwork (pp. 330–346). San Francisco, CA: Jossey-Bass. Sicotte, C., D’Amour, D., & Moreault, M.-P. (2002). Interdisciplinary within Quebec community health care centers, 55(6), 991. Sims, D., & Sims, A. (1993). Managing professionals in multidisciplinary teams: A case study from psychiatry. Leadership & Organizational Development Journal, 14(6), 12–16. Snelling, J. M. (1974). Training of social workers. Proceedings of the Royal Society of Medicine, 67(9), 943–945. Stanton, A. H., & Schwartz, M. S. (1954). The mental hospital: A study of institutional participation in psychiatric illness and treatment. New York: Basic Books. Stogdill, R. M. (1974). Handbook of leadership: A survey of theory and research. New York: Free Press. Thomas, E. J., & Fink, C. F. (1963). Effects of group size. Psychological Bulletin, 60(4), 371–384. Tjosvold, D., & Fottler, M. D. (1996). Joint hospital management by physicians and nursing administrators. Health Care Management Review, 21(3), 43–54. Toseland, R. W., Palmer-Ganeles, J., & Chapman, D. (1986). Teamwork in psychiatric settings. Social Work(January–February), 46–52. Tuckman, B. W., & Jensen, M. A. C. (1977). Stages of smallgroup development revisited. Group & Organization Studies, December, 419–427. VA (2005). Veterans Health Administration National Center for Patient Safety. Retrieved April 17, 2005, from http:// www.patientsafety.gov/. Vinokur-Kaplan, D. (1995). Treatment teams that work (and those that don’t): An application of Hackman’s group effectiveness model to interdisciplinary teams in psychiatric hospitals. Journal of Applied Behavioral Science, 31(3), 303–327. Webster’s II. (1995). Webster’s II new college dictionary. New York: Houghton Mifflin Company. Weick, K. E., & Roberts, K. H. (1993). Collective mind in organizations: Heedful interrelating on flight decks. Administrative Science Quarterly, 38(3), 357–381. Willshire, L. (1999). Psychiatric services: Organizing impossibility. Human Relations, 52(6), 775–804.