Professional Documents
Culture Documents
A Relational Model of How High-Performance Work Sy
A Relational Model of How High-Performance Work Sy
net/publication/220520994
CITATIONS READS
307 2,485
3 authors, including:
Some of the authors of this publication are also working on these related projects:
Testing Innovative Interventions to Improve Relational Care for People with Chronic Severe Brain Injury View project
All content following this page was uploaded by Rob Seidner on 21 May 2014.
Rob Seidner
College of Urban Planning and Public Affairs, University of Illinois at Chicago, Chicago, Illinois 60607,
seidner@alumni.brandeis.edu
Julian Wimbush
Health Services and Policy Analysis Program, University of California, Berkeley, Berkeley, California 94270,
wimbush@berkeley.edu
n this paper we explore a causal mechanism through which high-performance work systems contribute to performance
I outcomes. We propose that high-performance work systems can improve organizational performance by strengthening
relationships among employees who perform distinct functions, a pathway that is expected to be particularly important in
settings characterized by highly interdependent work. In a nine-hospital study of patient care, we identify high-performance
work practices that positively predict the strength of relational coordination among doctors, nurses, physical therapists,
social workers, and case managers, in turn predicting quality and efficiency outcomes for their patients. Relational coor-
dination mediates the association between these high-performance work practices and outcomes, suggesting a relational
pathway through which high-performance work systems work.
Key words: high-performance work systems; relational coordination; patient care
History: Published online in Articles in Advance.
and skills of employees or on the commitment of employ- Others have argued that in addition to building the
ees to their organization, this third view focuses on knowledge and skills of employees, high-performance
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
relationships between employees as the primary causal work systems also enhance the motivation and commit-
mechanism that connects high-performance work sys- ment of employees. Commitment-based human resource
tems and performance outcomes (e.g., Collins and Clark practices create an organizational climate that motivates
2003, Collins and Smith 2006). employees to act in the best interest of the organi-
In this paper we adopt this third view and propose zation, thus enhancing performance (Appelbaum et al.
a model of high-performance work systems in which 2000, Arthur 1992, Osterman 1988, Rousseau 1995).
each component practice reaches across multiple func- A key argument in this literature is that human resource
tions to engage employees in a coordinated effort. All practices build a psychological contract by signaling
the high-performance work practices identified in this an employer’s commitment to a long-term relation-
study are focused on building employee-employee rela- ship, in turn yielding a long-term commitment from
tionships. We argue that these high-performance work the employee (Tsui et al. 1997). Consistent with this
practices contribute to performance by supporting the argument, studies have found that particular work prac-
development of relational coordination, a mutually rein- tices are associated with higher levels of commitment
forcing web of communication and relationships carried (e.g., Tsui et al. 1997, Whitener 2001) and that commit-
out for the purpose of task integration (Gittell 2002b). ment in turn is positively associated with performance.
We test our model with multilevel data from a nine- In particular, Bowen and Ostroff (2004) provide argu-
hospital study of patient care that includes administrator ments suggesting that motivation and discretionary effort
interviews to measure work practices, care provider sur- underlie the association between human resource prac-
veys to measure relational coordination, and patient sur- tices and performance and are triggered by a strong
veys to measure patient outcomes. We explore the effects human resource system. Note that the human capital
of these high-performance work practices on quality and and commitment pathways are not mutually exclusive.
efficiency outcomes for patients and the mediation of Although research and theory often focus on one or the
these effects through relational coordination among care other, some theorists have argued that high-performance
providers. Hospitals are notorious for operating with work systems can contribute to performance through
well-defined silos that engender turf battles between both pathways (e.g., Appelbaum et al. 2000).
them. We expect that relational coordination will be crit-
ical for achieving desired performance outcomes in this Relational Models of How High-Performance
setting due to the high levels of task interdependence, Work Systems Work
uncertainty, and time constraints found there (Gittell Relationships among employees have also been theo-
2000). We expect these high-performance work prac- rized to play a role in achieving high levels of orga-
tices will foster relational coordination, thus bridging nizational performance. Some scholars have made this
the boundaries between the distinct professions that are argument by drawing on the concept of organizational
responsible for carrying out the work. social capital, a type of social capital that exists in and
can be developed by organizations as a distinctive orga-
Human Capital and Commitment Models of nizational capability and source of competitive advan-
How High-Performance Work Systems Work tage (Leana and Van Buren 1999, Nahapiet and Ghoshal
Models of high-performance work systems often draw 1998). Organizational social capital has been shown to
on human capital theory, whose central implication is improve performance by enabling employees to access
that human resource practices can improve organiza- the resources that are embedded within a given network
tional performance by increasing the knowledge and and by facilitating the transfer and sharing of knowledge
skills of employees (Becker 1975). To be successful, (Levin and Cross 2004, Tsai and Ghoshal 1998).
firms must invest in and maintain the workforce just Other theorists have argued that employee-employee
as they invest in and maintain the capital infrastruc- relationships are important for coordinating work (Adler
ture. Researchers have found that companies can achieve et al. 2008, Faraj and Sproull 2000, Gittell 2000), based
sustained performance advantages by leveraging the on the argument that coordination is the management of
knowledge of their employees. High-performance work task interdependence (Malone and Crowston 1994) and
systems can foster the development of human capital therefore fundamentally a relational process (Bechky
in the form of firm-specific idiosyncratic skills (Gibbert 2006, Faraj and Sproull 2000, Gittell 2002b, Weick and
2006), creating a performance advantage for organiza- Roberts 1993). One of these relational perspectives—
tions (Fried and Hisrich 1994, MacMillan et al. 1987, relational coordination—identifies specific dimensions
Tyebjee and Bruno 1984) through processes such as of relationships that are integral to the coordination of
increased employee problem solving (Snell and Dean work. According to the theory of relational coordina-
1992) and improved customization by employees in ser- tion, coordination that occurs through frequent, high-
vice industries (Batt 2002). quality communication supported by relationships of
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS 3
shared goals, shared knowledge, and mutual respect networks. They then demonstrate that the impact of these
enables organizations to better achieve their desired out- high-performance work practices on firm performance is
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
comes (Gittell 2006). Defined as “a mutually reinforcing mediated by the strength of firms’ top management team
process of interaction between communication and rela- social networks. More recently, Vogus (2006) has argued
tionships carried out for the purpose of task integration” that high-performance work practices such as selec-
(Gittell 2002a, p. 301), relational coordination is a type tion, training, performance appraisal, performance-based
of employee-employee relationship that is particularly rewards, and job security contribute to high-quality inter-
relevant for coordinating work that is highly interdepen- actions and mindfulness by signaling to employees the
dent, uncertain, and time-constrained. importance of relationships. Vogus continues by postu-
Substantial progress has been made toward identifying lating that these high-quality interactions contribute to
the work practices through which organizations influence higher-quality outcomes for hospital patients. Empirical
the development of employee-employee relationships. tests of this model demonstrated mediation.
Leana and Van Buren (1999) argue that stable employ- Though the types of employee-employee relationships
ment relationships and reciprocity norms facilitate the explored in these studies are varied, including relational
formation of social capital among employees. Evans and coordination (Gittell 2000), social networks (Collins and
Davis (2005) argue that work practices such as selec- Clark 2003, Gant et al. 2002), social capital (Evans
tive staffing, self-managed teams, decentralized deci- and Davis 2005, Leana and Van Buren 1999), and
sion making, extensive training, flexible job assignments, mindful interacting (Vogus 2006), these studies suggest
open communication, and performance-contingent com- that high-performance work practices can enhance per-
pensation influence multiple dimensions of an organi- formance through the pathway of employee-employee
zation’s social structure, including the development of relationships.
bridging ties, norms of reciprocity, shared mental mod-
els, role making, and organizational citizenship behavior. Theory Building
Gittell (2000) argues that work practices such as cross- The work practices found in the studies described above
functional selection, cross-functional conflict resolution, resemble in many ways the work practices found in
cross-functional performance measurement, flexible job the earlier high-performance work systems literature
design, and cross-functional boundary spanner roles can —they include selection, training, performance mea-
foster the development of relational coordination. These surement, rewards, knowledge-sharing mechanisms, and
work practices were shown to predict significantly higher so on. But they differ in an important way. Although
levels of relational coordination among airline employees the work practices found in these studies have the
from 12 distinct functions who were engaged in the flight potential to influence employee skills and commit-
departure process, though their impact on performance ment, they are focused primarily on strengthening
was not explored. relationships between employees. This understanding
Similarly, Gant et al. (2002) show that on steel- of high-performance work practices therefore responds
finishing lines with high-performance work systems, implicitly to an argument by post-bureaucracy theo-
defined as selection, training, incentive pay, job design, rists that traditional work practices often create divi-
problem-solving teams, and extensive labor/management sions between employees even when relationships are
communication, production employees have denser com- critically important due to the need for coordination
munication networks with each other and that these steel- (Heckscher 1994, Piore 1993). According to Piore
finishing lines also exhibit higher performance, measured (1993, p. 15), the bureaucratic organizational practices
in terms of fewer delays and higher yields. They argue that have become widespread through the rise of Tay-
that these human resource practices influence perfor- lorism “have pushed us to restrict communication among
mance outcomes because they influence the social net- the people responsible for the way in which the differ-
works of production employees. Their results suggest ent parts are performed.” Heckscher (1994, p. 24) envi-
that social networks may mediate the link between sions a postbureaucratic, interactive organizational form
work practices and outcomes, though mediation was not in which “everyone takes responsibility for the success
demonstrated. of the whole” and in which “workers need to understand
Collins and Clark (2003) have provided one of the key objectives in depth in order to coordinate their
the best empirical tests to date of the argument that actions intelligently ‘on the fly’.”
human resource practices influence outcomes through High-Performance Work Practices as Predictors of
their impact on relationships among employees. They Relational Coordination. Rather than rejecting the role
argue that the social networks of top management of formal work practices as the postbureaucratic litera-
teams enhance a firm’s information-processing capabil- ture has tended to do, we argue that formal work prac-
ity and that human resource practices, including men- tices can be redesigned to foster the employee-employee
toring, incentives, and performance appraisals, can be relationships through which work is effectively coordi-
designed to encourage the development of these social nated “on the fly.” High-performance work practices can
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
4 Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS
serve to overcome the silos of bureaucratic organizations solving have been found to strengthen working relation-
by connecting employees directly with each other to ships, while the reactive assignment of blame has been
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
enable them to coordinate their work. We focus here on found to undermine those relationships (Edmondson
six high-performance work practices—cross-functional 1996). In particular, these cross-functional performance
selection, cross-functional conflict resolution, cross- measurement practices strengthen the shared goals and
functional performance measurement, cross-functional problem-solving communication dimensions of relational
rewards, cross-functional meetings, and cross-functional coordination.
boundary spanners—and their impact on relational coor- Likewise, rewards have traditionally been tied to
dination, reflected in the frequency, timeliness, accuracy, individual or functional outcomes, thereby encouraging
and problem-solving nature of communication among subgoal optimization at the expense of organizational
employees and the degree to which their relationships outcomes. Employees engaged in interdependent tasks
are characterized by shared goals, shared knowledge, are most likely to coordinate their tasks effectively if
and mutual respect. their rewards are also interdependent (Wageman and
Selection has long been recognized as a powerful way Baker 1997). But because rewards are often put in place
to achieve fit between a prospective employee and a job without a clear understanding of how they are expected
(Lawrence and Lorsch 1968). Though selection tradi- to work, misalignment is common. Research indicates
tionally focused on skills, other attributes such as per- that more individualized rewards are associated with
sonality traits (Day and Silverman 1989), organizational lower levels of integration across units, whereas shared
fit (Kwiatkowski 2003), and teamwork ability (Cappelli rewards have been found to support coordination and
and Rogovsky 1994) are also relevant to job perfor- goal commitment among parties involved in the same
mance. In the context of highly interdependent work, work process (Guthrie and Hollensbe 2004, Zenger and
selection for cross-functional teamwork is expected to be Hesterly 1997). In particular, cross-functional rewards
particularly relevant and has been found to affect coor- strengthen the shared goals dimension of relational
dination across functional boundaries (Gittell 2000), in coordination.
particular strengthening the mutual respect dimension of Meetings are a coordinating mechanism that fos-
ters real time coordination, incorporating information
relational coordination.
as it becomes available (Argote 1982). Meetings give
Conflicts are likely to occur in the presence of
participants the opportunity to coordinate their tasks
high levels of task interdependence and/or diversity—
interactively, on the spot. Face-to-face interactions are
including functional diversity—among participants
expected to have particular relevance for assuring effec-
(Pelled et al. 1999, Walton and Dutton 1967). Conflicts
tive communication because of their high bandwidth,
have been found to improve performance when they
their immediacy, and their ability to build connections
take place in a group that values task-related conflict; among participants through the use of nonverbal cues
however, unresolved conflicts undermine relationships (Goffman 1961, Nohria and Eccles 1992). While infor-
and hinder performance over time (Jehn 1995). Cross- mal meetings are sometimes argued to be more effec-
functional conflict resolution can play a constructive role tive than formal meetings (Mangrum et al. 2001), formal
by providing a way to articulate and accommodate mul- meetings may be needed to connect participants who
tiple points of view, each with the potential to add value work in distinct functions. In particular, cross-functional
to the work process. Consistent with this argument, con- meetings strengthen the accuracy of communication as
flict resolution has been found to provide opportuni- well as the shared goals and shared knowledge dimen-
ties for building a shared understanding of the work sions of relational coordination.
process among participants, thereby strengthening the Boundary spanners are staff members whose primary
relationships through which coordination is carried out task is to integrate the work of other people around a
(Gittell 2000, Mareschal 2003), in particular strengthen- project, process, or customer (Galbraith 1995, Lawrence
ing the shared knowledge and mutual respect dimensions and Lorsch 1968). Because boundary spanners enable
of relational coordination. new information to be incorporated on an ongoing basis,
Traditional performance measurement practices assign they are typically used when tasks cannot be fully pro-
accountability for outcomes to individuals or functions, grammed in advance. Because they build understanding
despite the task interdependencies that often make out- between areas of functional expertise, they are expected
comes the responsibility of a larger group. This focus on to add value when existing boundaries are highly divi-
individual or functional accountability encourages sub- sive (Galbraith 1995, Mohrman 1993). In particular,
goal optimization (March and Simon 1958), whereas cross-functional boundary spanners strengthen the fre-
cross-functional accountability encourages workers to quency and timeliness of communication as well as the
adopt a broader perspective and to focus on problem shared knowledge dimensions of relational coordination.
solving rather than the assignment of blame (Chenhall In summary, the high-performance work practices
2005, Deming 1986, Locke and Latham 1990). Per- described above are expected to foster relational coor-
formance measurement practices that focus on problem dination, which is reflected in the frequency, timeliness,
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS 5
accuracy, and problem-solving nature of communica- staff, facilities, and equipment—to be utilized more pro-
tion among employees and the degree to which their ductively, leading to more efficient outcomes such as
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
relationships are characterized by shared goals, shared faster turnaround times and shorter throughput times.
knowledge, and mutual respect. We know from previous Because high-performance work practices strengthen
research that firms “can improve performance either by relational coordination among employees, we expect the
increasing the number of practices they employ within association between high-performance work practices
the system or by using the practices in an AR system in and efficiency outcomes to be mediated by relational
a more comprehensive and widespread manner” (Youndt coordination.
et al. 1996, p. 849) Relational coordination therefore
Hypothesis 3. The association between high-perfor-
depends not only on the adoption of high-performance
mance work practices and efficiency outcomes is medi-
work practices, but also on the intensity of their adoption
ated by relational coordination among employees.
and the degree to which they reach across all relevant
employee functions. Our concept of high-performance Together, these hypotheses serve as the basis for a
work practices therefore reflects (1) the number of work relational model of how high-performance work systems
practices that are adopted, (2) the intensity of their adop- work.
tion, and (3) the degree to which they reach across all
relevant employee functions, together summarized as the
strength of high-performance work practices. Methods
Hypothesis 1. The strength of high-performance Setting
work practices positively predicts relational coordination To test these hypotheses, a study of patient care was
among employees. conducted using a convenience sample of nine major
urban hospitals. Previous studies have shown that coor-
High-Performance Work Practices and Relational dination between care providers is positively related to
Coordination as Predictors of Quality and Efficiency both quality and efficiency. Specifically, coordination is
Performance. Through their impact on relational coor- associated with provider-perceived (Argote 1982) and
dination, high-performance work practices are expected patient-perceived quality of care (Gittell 2000), and with
to improve quality and efficiency performance for orga- reduced lengths of hospital stay (Gittell et al. 2000,
nizations. Previous research has shown how relation- Shortell et al. 1994). However, the contribution of high-
ships of shared goals, shared knowledge, and mutual performance work practices to these outcomes and to the
respect enabled employees from different functions coordination of care has not been explored. We chose a
to coordinate work by supporting frequent, timely, work process for which outcomes were well understood
problem-solving communication among them (Gittell and readily measured—surgical care for joint replace-
2006). Relational coordination is expected to result in ment patients. We selected nine orthopedics units, each
fewer missed signals between employees with different located in a different hospital that performed relatively
areas of functional expertise, due to the information- large numbers of joint replacements. In each orthope-
processing capacity that is created through shared dics unit, there was a group of care providers—including
goals, shared knowledge, and mutual respect. Rela- physicians, nurses, physical therapists, case managers,
tional coordination enables more consistent communica- and social workers—who were responsible for provid-
tion and a reduction in the probability of errors, leading ing care to joint replacement patients over a six-month
to higher-quality outcomes. Because high-performance study period.
work practices strengthen relational coordination among
employees, we expect the association between high- Data Sources
performance work practices and quality outcomes to be Data from the participating orthopedics units included
mediated by relational coordination. administrator interviews, a care provider survey, a patient
survey, and patient hospitalization records. Administra-
Hypothesis 2. The association between high-perfor-
tor interviews were used to measure high-performance
mance work practices and quality outcomes is mediated
work practices at the unit level. The care provider sur-
by relational coordination among employees.
vey was used to measure relational coordination at the
In the same way, high-performance work practices level of individual care providers. Patient surveys and
are also expected to improve efficiency outcomes for hospitalization records were used to measure outcomes
organizations. Because relational coordination results in at the level of individual patients. To measure high-
fewer missed signals between employees, it is expected performance work practices, front-line administrators
to reduce the time that is wasted carrying out redun- were interviewed in each orthopedics department, includ-
dant communication, searching for missing information, ing at least one physician, nurse, physical therapist, social
and waiting to hear from coworkers. Relational coordi- worker, and case manager. For each unit, unstructured
nation thus enables organizational resources—including interviews and observations were conducted in person at
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
6 Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS
the time of the initial site visits, followed up by more resolution, cross-functional performance measurement,
systematic structured telephone interviews after the site cross-functional rewards, cross-functional team meet-
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
visits. The interview protocol that we developed based on ings, and cross-functional boundary spanners. Descrip-
our first stage of interviews and observations was used as tive data for these work practices are shown in Table 1.
a guide for our second stage of interviews. Selection was measured by asking administrators in
To measure relational coordination, surveys were each orthopedics unit about selection criteria for physi-
mailed to all eligible care providers in the nine ortho- cians, nurses, and physical therapists, probing as to
pedics units who had responsibilities for joint replace- whether cross-functional teamwork ability was consid-
ment patients during the study period in five core ered an important selection criterion. This variable was
functions: physicians, nurses, physical therapists, social coded from 0 to 2 for each of these three workgroups,
workers, and case managers. A unit administrator desig- 0 indicating that cross-functional teamwork ability was
nated by the chief of orthopedics identified all eligible not considered, 1 indicating that it was considered to
care providers in each unit. The administrator received some extent, and 2 indicating that it was a consistent
written guidelines as to whom should be included (all criterion for selection.
providers from the above five functions who were Conflict resolution was measured by asking about
directly or indirectly involved with providing care for conflict-resolution processes. Questions probed as to
joint replacement patients). Surveys were mailed to whether any formal cross-functional conflict resolution
all eligible care providers initially during the second process was in place for physicians, nurses, or physical
month of the study period, with one repeat mailing therapists. This variable was coded 0 or 1 for physi-
during the study period for nonrespondents. Responses cians, nurses, and physical therapists; 0 indicated that the
were received from 338 of 666 providers for an overall workgroup had no access to a formal cross-functional
provider response rate of 51%. conflict resolution process and 1 indicated that the work-
To measure patient outcomes, the patient survey was group did have access.
adapted from a validated instrument that is widely used Performance measurement was measured by asking
to assess the quality of care in health care settings (Cleary about the quality-assurance process and the utilization
et al. 1991). We received responses to 878 of 1,367 review process in each hospital, probing as to whether
questionnaires sent to patients in the target population— each of these processes were largely focused on iden-
patients with a diagnosis of osteoporosis who received tifying the single function that was responsible for a
primary unilateral hip or knee replacement in one of the quality or utilization problem or whether the approach
nine hospitals during the six-month study period—for a was more cross-functional. Responses were coded on a
response rate of 64%. In addition, hospital administrators 5-point scale, where 1 = highly functional, 2 = fairly
provided hospitalization records for each patient. These functional, 3 = equally functional/cross-functional, 4 =
records were used to determine length of stay for each fairly cross-functional, and 5 = highly cross-functional.
patient and to extract information regarding patient char- Questions also probed interviewees as to whether these
acteristics to use as control variables in models of quality two performance-measurement processes were reactive
and efficiency outcomes. Of the 878 survey respondents, (focused on affixing blame) or proactive (focused on
69 were dropped because they failed to meet one or more problem solving). Responses were coded on a 5-point
conditions of the study (primary unilateral hip or knee scale: 1 = highly reactive, 2 = fairly reactive, 3 =
replacement with a diagnosis of osteoporosis), leaving equally reactive/proactive, 4 = fairly proactive, and 5 =
809 viable respondents. Of these, all 809 respondents had highly proactive.
full data available for the variables taken from the hospi- Rewards were measured by asking about the criteria
talization records (age, race, gender, surgical type, length for rewards for physicians, nurses, and physical thera-
of stay), but some had missing data on the patient sur- pists, probing as to whether rewards were based purely
vey variables. The number of respondents with complete on individual performance or whether they were based
responses for the quality outcome models was n = 588 on some cross-functional performance criteria as well.
and for the efficiency outcome models was n = 599. Test- This variable was coded from 0 to 2. For nurses and
ing for missing data bias, we found that respondents who physical therapists, 0 indicated no performance-based
were excluded from our final models were older, more rewards, 1 indicated individual rewards only, and 2 indi-
likely to be female, and had lower levels of preoperative cated some cross-functional team rewards. For physi-
functioning, but that they did not differ significantly on cians, 0 indicated individual rewards only, 1 indicated
any of the other variables in our models, including qual- surplus sharing with the hospital (potential for sharing
ity of care or length of stay. positive financial outcomes), and 2 indicated risk shar-
ing with the hospital (potential for sharing both positive
High-Performance Work Practices and negative financial outcomes).
High-performance work practices included in this study Meetings were measured by asking key informants
were cross-functional selection, cross-functional conflict about participation in physician rounds and nursing
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS 7
Cross-functional selection
Physicians selected for cross-functional teamwork 0701 0–2 044 088 9
Nurses selected for cross-functional teamwork 0760 0–2 144 073 9
Physical therapists selected for cross-functional teamwork 0570 0–2 167 088 9
Cross-functional conflict resolution
Physician access to cross-functional process 0916 0–1 044 053 9
Nurse access to cross-functional process 0700 0–1 022 044 9
Physical therapists access to cross-functional process 0438 0–1 033 050 9
Cross-functional performance measurement
Cross-functional approach to quality measurement 0544 1–5 333 141 9
Problem-solving approach to quality measurement 0729 1–5 278 139 9
Cross-functional approach to efficiency measurementc 1–5 256 188 9
Problem-solving approach to efficiency measurement 0834 1–5 300 158 9
Cross-functional rewards
Physicians rewarded for cross-functional teamwork 0438 0–3 022 067 9
Nurses rewarded for cross-functional teamwork 0560 0–2 056 088 9
Physical therapists rewarded for cross-functional teamwork 0803 0–2 111 105 9
Cross-functional meetings
Nurses included in physician rounds 0548 0–2 133 087 9
Physical therapists included in physician rounds 0691 0–2 056 088 9
Case managers included in physician rounds 0677 0–2 067 087 9
Physicians included in nursing roundsc −0210 0–2 078 044 9
Physical therapists included in nursing roundsc −0112 0–2 144 073 9
Case managers included in nursing rounds 0642 0–2 133 1 9
Cross-functional boundary spanners
Case manager caseload −0740 6.7– 40 2630 1080 9
Case manager discharge planning role 0515 0–1 089 033 9
Case manager coordination role c 0368 0–1 044 053 9
Primary nursing model 0746 0–1 056 053 9
High-performance work practices index ( = 093)
Relational coordination
Shared goals 0629 1–5 421 061 331
Shared knowledge 0629 1–5 393 059 333
Mutual respect 0659 1–5 381 059 327
Frequency of communication 0566 1–5 384 073 334
Timeliness of communication 0782 1–5 408 062 334
Accuracy of communication 0796 1–5 423 062 333
Problem-solving focus of communication 0784 1–5 405 046 320
Relational coordination index ( = 086)
a
N = 9 hospital units. Variables coded from interviews with administrators.
b
N = 336 care providers. Variables coded from survey of care providers.
c
These four items were dropped from the high-performance work practices index because of weak factor loadings.
rounds, probing to find out which functional groups par- for a patient throughout his or her stay and to serve
ticipated in those rounds and the consistency of their as a point person for coordinating that patient’s care).
participation. Rounds are the primary form of meeting Caseload, the number of patients for whom case man-
used for coordinating patient care. These variables were agers were typically responsible at a time, was measured
coded on a 0 – 2 scale, with 0 indicating that the func- as a continuous variable, ranging across hospitals from
tional group did not participate in the rounds, 1 indicat- 6.7 to 40. Each of the case manager roles—leadership of
ing that they participated sometimes, and 2 indicating rounds and planning for patient discharge—was coded
that they participated usually or always. as 0 or 1, with 0 indicating that the role was not expected
Boundary spanner was measured by asking about the and 1 indicating that it was expected of case managers.
caseload and roles of the case managers who worked Primary nursing was coded as 1 if the model was in
with joint replacement patients and whether the primary place and 0 if not.
nursing model was in place on that unit (i.e., the practice Because these six work practices were correlated with
of assigning one nurse to assume primary responsibility one another, forming a “bundle” of work practices, we
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
8 Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS
combined the above measures into an index of high- comprehensive and widespread manner, for example, by
performance work practices. Exploratory factor analysis extending their reach to cover a wider array of employee
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
An additive scaling method was used in which each item measure of the efficiency of care for each individual
was standardized with a mean of 0 and a standard devi- patient, controlling for the patient characteristics that are
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
ation of 1 so that each item was equally weighted. Cron- believed to necessitate longer lengths of stay (described
bach’s alpha was 0.86, suggesting that this construct has below). Length of stay was calculated from hospital
a high level of reliability. No items were dropped due to records for each patient as the number of whole days
weak factor loadings, and no cross-loadings greater than between the date of admission and the date of discharge.
0.40 were found. Furthermore, all items had item-to-
total correlation scores of 0.40 or greater. We conclude Control Variables
that the relational coordination index meets standards for
For Predicting Relational Coordination. Control vari-
reliability and convergent validity.
ables for predicting relational coordination include
Using one-way analysis of variance, significant cross-
dummy variables that indicate the functional identity
unit differences in relational coordination were found—
of the care provider respondent, given that different
F 8327 = 532, p < 0001—as well as significant
functions are expected to engage differently in rela-
cross-functional differences in relational coordination—
tional coordination due to the differences in their pro-
F 5330 = 289, p < 005. When unit-level and
fessional identities and their differential coverage by
function-level differences were considered jointly, unit-
high-performance work practices. Ideally, control vari-
level differences remained significant—F 8322 =
ables would also include demographic characteristics
451, p < 0001—and function-level differences became
such as tenure and gender that might influence care
Insignificant—F 5322 = 175, p = 012. To further
provider engagement in relational coordination; how-
assess treating relational coordination as a unit-level
ever, these variables were not included on the survey.
construct, we computed intraclass correlations ICC(1)
We included the total number of joint replacements con-
and ICC(2). ICC(1) is the proportion of total variance
ducted by each unit in the six-month period prior to
explained by unit membership with values ranging from
the study period to control for possible learning effects
−1 to +1 and values between 0.05 and 0.30 being most
(Luft 1990) that could improve relational coordination
typical. This number provides an estimate of the reliabil-
between the functions involved in patient care.
ity of a single respondent’s assessment of the unit mean.
ICC(2) provides an overall estimate of the reliability of For Predicting Quality and Efficiency Outcomes. Con-
unit means, with values equal to or above 0.70 being trol variables for predicting quality and efficiency out-
acceptable. For relational coordination, ICC(1) = 0.25 comes were selected to adjust for factors that have been
and ICC(2) = 0.81. We concluded that relational coordi- shown in the health care literature to affect quality of
nation performs well on both forms of intraclass corre- care and length of stay for joint replacement patients.
lation. Taken together, these results are consistent with Control variables included the following patient char-
treating relational coordination as a unit-level construct. acteristics: age, comorbidities, psychological well-being,
preoperative status, surgical procedure (hip versus knee),
Performance Outcomes marital status, race, and gender.
Performance outcomes for this study included both the Patient age was determined from hospital records.
quality and efficiency of patient care. Hospitals have Older patients were expected to require longer hospital
been striving to improve the quality of care as perceived lengths of stay. Preoperative clinical status was assessed
by patients (Cleary et al. 1991). All hospitals in this in the patient survey using the pain and functioning ele-
study had been conducting patient surveys for years, but ments of the WOMAC. Patients with lower preopera-
differences between their surveys required them to adopt tive status were expected to require longer lengths of
a new survey for this study. We used a single item mea- stay. Comorbidities were assessed in the patient survey
sure of patient-perceived quality of care (“Overall, how with a series of questions asking patients whether they
would you rate the care you received at the hospital?”), suffered from heart disease, high blood pressure, dia-
measured on a five-point Likert-type scale, consistent betes, ulcer or stomach disease, kidney disease, anemia
with other studies that have used single-item measures or other blood disease, cancer, depression, or back pain
of patient-perceived quality of care (Ware and Hays (Katz et al. 1996). Individual patients with a greater
1988, Young et al. 2000, Rohrer and Wilkinson 2008). number of comorbid conditions were expected to require
Responses were coded as 1 = poor, 2 = fair, 3 = good, longer hospital lengths of stay.
4 = very good, and 5 = excellent. Surgical procedure was measured through procedure
Hospitals have also been striving to improve the effi- code in the hospital record and was either a hip or a
ciency of care by reducing the lengths of patient stays. knee replacement. Knee replacements were expected to
Length of stay is the number of inpatient days of care require longer lengths of stay than hips, due to greater
used by a given patient. Days of inpatient care are a difficulty of achieving postoperative mobility.
resource that external payers are intent on reducing. This Psychological well-being was assessed in the patient
study therefore uses the length of hospital stay as a survey using the mental health component of the SF-36
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
10 Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS
(Stewart et al. 1988). Patients with higher levels of psy- outcomes is reduced significantly when controlling for
chological well-being were expected to report receiving relational coordination, then drew on the critical values
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
higher quality of care: psychological theory suggests that recommended by MacKinnon et al. (2002) to determine
people with high levels of positive affect tend to per- whether the results for quality and efficiency outcomes
ceive experiences in a more favorable light. Patient gen- supported mediation.
der, race, and marital status were determined through For all the above analyses, random effects modeling
the patient survey and were included because some stud- was used to adjust standard errors for the multilevel
ies have found demographic influences on health care nature of the data, accounting for non-independence of
outcomes. the error terms (Raudenbush and Bryk 1992). To deter-
mine the days of hospital stay associated with a change
Data Analyses in relational coordination, we conducted a Poisson ran-
To test Hypothesis 1, we regressed relational coor- dom effects regression on the length of stay model.
dination (n = 336 care providers in 9 units) on the
high-performance work practices index (n = 9 units), Findings
controlling for the functional identity of the care Descriptive data and intercorrelations between key vari-
provider respondents (with “nurse” as the omitted ables are reported in Table 2 in aggregate form, and
respondent category) and for the volume of joint replace- broken out by individual hospital unit. Due to the high
ments conducted on the unit in the six-month period correlation found between high-performance work prac-
prior to the study. tices and relational coordination, we tested for multi-
To test Hypothesis 2, we regressed quality outcomes collinearity in the outcomes models that include both
(n = 588 patients in 9 units) on the high-performance high-performance work practices and relational coordi-
work practices index (n = 9 units), controlling for the nation as independent variables by examining variance
patient characteristics expected to affect these outcomes, inflation factors. Evidence of multicollinearity exists
as well as the volume of joint replacements conducted if (1) the largest variance inflation factor is greater
on the unit in the six-month period prior to the study; than 10, or (2) the mean value of all the variance infla-
we then aggregated relational coordination to the unit tion factors is considerably larger than 1, where 4.02
level (n = 9 units) and entered it along with the high- is given as an example of a variance inflation factor
performance work practices index (n = 9 units) into the that is not considerably larger than 1 (Chatterjee and
above equation for quality outcomes (n = 588 patients Price 1991). We found that the variance inflation fac-
in 9 units). tors for the quality of care model ranged from 1.06 to
To test Hypothesis 3, we regressed efficiency out- 8.45, with an average variance inflation factor of 2.46;
comes (n = 599 patients in 9 units) on the high-perfor- for the length of stay model they ranged from 1.06 to
mance work practices index (n = 9 units), controlling 8.50, with an average variance inflation factor of 2.47.
for the patient characteristics expected to affect these Comparing our results to the Chatterjee and Price crite-
outcomes, as well as the volume of joint replacements ria, we conclude that multicollinearity is not likely to be
conducted by the unit in the six-month period prior to a substantial problem in our models.
the study. We then aggregated relational coordination to
the unit level (n = 9 units) and entered it along with High-Performance Work Practices and
the high-performance work practices index (n = 9 units) Relational Coordination
into the above equation for efficiency outcomes (n = 599 Hypothesis 1 test results are reported in Table 3.
patients in 9 units). The results show that high-performance work practices
We therefore test mediation in this study across multi- are positively associated with relational coordination
ple levels of analysis, consistent with previous studies of (r = 031, p < 0001). The physician respondent dummy
relational coordination (e.g., Gittell 2001, 2002b). Given variable was negative and significant (r = −016,
its status as a multilevel theory that operates across mul- p < 0001), suggesting that physicians were significantly
tiple levels of analysis, relational coordination is partic- less engaged in relational coordination than nurses. The
ularly amenable to testing mediation across levels. If the other dummy variables were not significant, suggesting
coefficient on high-performance work practices becomes that residents, physical therapists, case managers, and
insignificant when relational coordination is added to social workers did not differ significantly from nurses in
the outcomes equations, this result can be taken to sug- their engagement in relational coordination.
gest that relational coordination mediates between high- The estimated effect of high-performance work prac-
performance work practices and outcomes, or in other tices on relational coordination is statistically significant
words that high-performance work practices influence and moderately large. The standardized coefficient of
outcomes through their effect on relational coordination. 0.31 on high-performance work practices suggests that
We used the Sobel test to determine whether the asso- for a hospital at the mean level of relational coordi-
ciation between high-performance work practices and nation, a one-point change in high-performance work
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS 11
n = 39 n = 46
n = 45
n = 48
061 058
119
151
n=1
−042
014 −002
362
498
9
Relational Coordination
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
Relational
n = 68
n = 70
085
100
n=1
−032
424
430
coordination
8
High-performance work practices 031∗∗∗
n = 27 n = 33
n = 93
n = 97
066 086
094
192
0000
n=1
012
056 −004
398
560
7
n = 63
n = 67
−002
104
394
n=1
Resident respondent
080
408
437
6
0613
Physical therapist respondent 007+
Hospitals
n = 15
n = 64
n = 65
052
096
133
n=1
0091
083
012
419
417
5
n = 135
n = 40
−007+
064
070
035
443
444
4
0095
Surgical volume −003
n = 123
n = 125
0386
n = 33
054
099
174
n=1
−028
−017
408
590
3
Constant 003
0450
Within-unit R2 007
n = 51
n = 90
n = 93
084
108
239
n=1
−054
−027
368
580
Between-unit R2 090
2
n = 109
n = 52
111
206
n=1
−095
−019
362
557
1
−017∗∗∗
000
+
p < 010; ∗ p < 005; ∗∗ p < 001; ∗∗∗ p < 0001.
0013
0009
−080∗∗
078∗
—
0000
0023
0045
−092∗∗
074∗
−068∗
073
101
213
401
511
0
−330–161
Range
2–35
1–5
Table 2 Descriptive Data and Correlations
336
788
809
3. Quality of care
4. Length of stay
coordination
p < 005;
Table 4 Impact of High Performance Work Practices and Relational Coordination on Quality and Efficiency Outcomes
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
1a 1b 1c 2a 2b 2c
Notes. Unit of analysis is the joint replacement patient (n = 588 for quality of care, n = 599 for length of stay). Random effects regression
is used to account for clustering of patients by hospital unit (n = 9). High-performance work practices, relational coordination, and surgical
volume are entered at the hospital unit level (n = 9).
+
p < 010; ∗ p < 005; ∗∗ p < 001; ∗∗∗ p < 0001.
suggests that for a hospital at the mean level of quality, ing relational coordination among employees in different
a one-point change in high-performance work practices functions (Hypothesis 2).
would correspond to a 0.38-point change in patient-
perceived quality of care. The coefficient of 1.93 on High-Performance Work Practices, Relational
relational coordination suggests that for a hospital at Coordination, and Efficiency Outcomes
the mean level of quality, a one-point change in rela- Results from testing Hypothesis 3 are shown in columns
tional coordination would correspond to a 1.93-point 4–6 of Table 4. Findings indicate that high-performance
change in patient-perceived quality of care. Because the work practices are associated with shorter lengths of stay
primary explanatory variables—high-performance work (r = −016, p = 0001). When relational coordination is
aggregated to the unit level and included in the equa-
practices and relational coordination—are measured at
tion (column 6), relational coordination is associated
the unit level, the equations explain little within-unit
with shorter lengths of stay (r = −119, p = 0005); the
variation in quality of care (R2 = 005), but they explain coefficient on high-performance work practices becomes
a large percentage of between-unit variation in qual- insignificant, again suggesting mediation.
ity of care (R2 = 073 for the most complete model). The coefficient of −016 on high-performance work
Results of the Sobel test suggest that the association practices suggests that a one-point increase in high-
between high-performance work practices and quality performance work practices is associated with a 0.16-
of care is significantly mediated by relational coordi- day reduction in patient length of stay. The coefficient
nation (z = 187, p < 001). Together, these results of −119 on relational coordination suggests that a one-
suggest that high-performance work practices predict point increase in relational coordination is associated
quality outcomes and that they do so by strengthen- with a 1.2-day reduction in patient length of stay. Among
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS 13
High-performance
work practices
Quality
Selection for outcomes
1.93*
cross-functional teamwork Relational coordination Patient-perceived
Cross-functional quality of care
Shared goals
conflict resolution Shared knowledge
0.31*** Mutual respect
Cross-functional
performance measurement
Frequent communication Efficiency
Cross-functional rewards Timely communication outcomes
Accurate communication
Cross-functional meetings Problem-solving communication –1.19** Patient
length of stay
Cross-functional
boundary spanners
∗ ∗∗ ∗∗∗
p < 005; p < 001, p < 0001.
U.S. hospitals, the median for the average charge per potentially positive role of formal work practices. Post-
day was $4,357 in 2006 (Medicare Cost Reports 2006), bureaucracy theorists (e.g., Heckscher 1994, Piore 1993)
or $4,803 in 2008, adjusted for annual health care cost argue that the networks needed for getting work done
increases of 4.5% in 2007 and 5.5% in 2008. So a one- tend to emerge informally and that formal organiza-
point increase in relational coordination is associated tional practices, traditionally designed to segment and
with $5,764 ($4,803/day ∗ 1.2 days) in average cost sav- divide employees from their counterparts in different
ings for each joint replacement patient served. Results of functions, serve more as obstacles than facilitators of
the Sobel test suggest that the association between high- these networks. Along with other recent work (Collins
performance work practices and length of stay is signif- and Clark 2003, Gant et al. 2002, Leana and Van Buren
icantly mediated by relational coordination (z = 240, 1999, Vogus 2006), our study provides a countervail-
p < 001). Together, these results suggest that high- ing argument that formal practices are not necessar-
performance work practices predict efficiency outcomes ily obstacles and can indeed be designed to encourage
and that they do so by strengthening relational coordi- the development of relationships between employees in
nation among employees in different functions (Hypoth- different functions, producing significant performance
esis 3). See Figure 1 for a summary of all results. advantages for organizations. Of the existing research
on high-performance work systems, Collins and Clark’s
Discussion (2003) study is the most similar to ours. Like us,
In this paper we explored a relational pathway through Collins and Clark conceptualize relationships as media-
which high-performance work systems were predicted to tors between human resource practices and performance.
contribute to performance outcomes. We proposed that Despite that similarity, our work differs from that of
high-performance work practices can enhance organiza- Collins and Clark both theoretically and methodologi-
tional performance by encouraging the development of cally. Theoretically, we focus on relationships between
relational coordination between employees who perform employees in different functions rather than on busi-
distinct functions—in contrast to traditional bureau- ness relationships more generally. Furthermore, our the-
cratic work practices that divide and separate employ- ory building is focused at the level of production or
ees and in contrast to other types of high-performance service delivery rather than of relationships among top
work practices that focus primarily on the development managers, because of our interest in the coordination
of employee skills and commitment. High-performance of work. Methodologically, we differ from Collins and
work practices that focus only on employee skills and Clark by using multiple data sets to test our hypothe-
commitment are not expected, on their own, to yield ses. Collins and Clark interviewed CEOs to determine
the coordinated, synergistic behaviors that are needed to human resource practices, social networks, and perfor-
achieve the highest levels of performance in interdepen- mance, thus introducing the potential for biases that can
dent work settings. arise when relying on a single source of data; in contrast,
we interviewed front-line managers to determine human
Contributions resource practices, surveyed direct service workers to
This study has made two key contributions. First, we measure relational coordination, and examined customer
have linked the high-performance work systems lit- data to determine performance outcomes. Because of
erature to postbureaucracy theory by articulating the these theoretical and methodological differences from
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
14 Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS
previous work, our study uniquely demonstrates how providers who responded to our survey. The omission
formal work practices play an important role in produc- of care provider demographics from the model that pre-
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
ing high performance in multidisciplinary work groups dicts relational coordination among care providers may
that are engaged in interdependent work. result in omitted variable bias. Finally, our focus on
A second contribution of this study is that we artic- cross-functional relationships pays less attention to the
ulate a novel relational pathway through which high- within-function relationships that have also been shown
performance work practices contribute to performance. to be important (Vogus 2006).
We have argued that high-performance work practices
can contribute to performance by supporting the devel- Future Research
opment of relational coordination, “a mutually reinforc- Next steps in theory development should include the
ing process of interaction between communication and development of high-performance work system models
relationships carried out for the purpose of task inte- that include all three causal mechanisms—skills, com-
gration” (Gittell 2002a, p. 301). Like the role-based mitment, and employee-employee relationships. Oth-
coordination found in Thompson (1967) and Bechky ers have begun to take steps in this direction. Smith
(2006), and like role-based relationships more gener- et al. (2005) developed a model that explores the
ally as explored by Meyerson et al. (1996) in their contribution of human capital and social capital to
work on swift trust and by Klein and coauthors (2006) perceived knowledge-creation capability in technology
in their work on de-individualization, relational coor- firms, based on the argument that one without the other
dination focuses on relationships between roles rather is not very useful. Leana and Pil (2006) developed and
than individual role inhabitants. Although there are other tested a similar model in the context of public schools.
relational perspectives on coordination (Bechky 2006, But to our knowledge, no one has yet explored the high-
Faraj and Sproull 2000, Weick and Roberts 1993), the performance work systems that would support the devel-
theory of relational coordination is unique in iden- opment of skills, commitment, and employee-employee
tifying specific dimensions of relationships that are relationships. We believe that the next frontier for theo-
integral to the coordination of interdependent work, ries of high-performance work systems is the design of
while focusing on the development of these relation- work systems that explicitly support the development of
ships between roles rather than individuals. By show- all three pathways, and we hope that this paper, by fur-
ing that relational coordination constitutes a pathway ther explicating the relational path through which high-
between high-performance work practices and outcomes, performance work systems work, will serve as a building
our study provides a new relational model of how high- block in that direction.
performance work systems work.
Conclusion
Limitations In conclusion, this study suggests that adopting high-per-
Despite its contributions and strengths, this study is formance work practices to foster relational coordination
limited in several ways. First, this study is limited by constitutes one viable path for improved organiza-
the use of interviews rather than survey instruments tional performance. But organizations have other options
to measure work practices, rendering the results less when choosing paths for improving performance. What
amenable to replication because of the time-consuming are the relative advantages of the relational approach
process of conducting interviews and constructing vari- explored here? Relational coordination enables employ-
ables. Second, our data are limited by the lack of mea- ees to more effectively coordinate their work with each
sures for individual skill and commitment. We have other, thus pushing out the production possibilities fron-
counterbalanced this omission by arguing that individ- tier to achieve higher-quality outcomes while using
ual skill and commitment are not sufficient to yield the resources more efficiently—for example, as we found
coordinated, synergistic behaviors needed to achieve the here, enabling hospital workers to achieve better patient-
highest levels of performance in interdependent work perceived quality of care along with shorter patient
settings. Although this study has identified a unique lengths of stay. Relational coordination and the high-
and important mechanism—relational coordination— performance work practices that support its development
through which high-performance work systems work in are therefore particularly relevant in industries that must
such settings, we recommend that future studies explore maintain or improve quality outcomes while responding
the interplay among skill, commitment, and relational to cost pressures. In an increasingly competitive econ-
coordination. omy, nearly all industries are likely to face this dual
A third limitation of our data is the lack of employee- challenge.
level control variables other than the functional identity Second, unlike relationships that are based on per-
of the respondent. We have fairly extensive demographic sonal ties, the relationships found in relational coordina-
information for our patient respondents but no demo- tion are based instead on ties between roles. The high-
graphics other than functional identity for the care performance work practices explored in this paper are
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS 15
expected to foster relationships of shared goals, shared Batt, R. 1999. Work design, technology and performance in customer
knowledge, and mutual respect among employees whose service and sales. Indust. Labor Rel. Rev. 52 539–564.
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
work is interdependent, with or without the presence of Batt, R. 2002. Managing customer services: Human resource prac-
personal ties. This feature allows for the interchange- tices, quit rates, and sales growth. Acad. Management J. 45
ability of employees, allowing employees to come and 587–598.
go without missing a beat, an important consideration Bechky, B. A. 2006. Gaffer, gofers, and grips: Role-based coordina-
for organizations that strive to achieve high levels of tion in temporary organizations. Organ. Sci. 17(1) 3–21.
performance while allowing employees the scheduling Becker, B., B. Gerhart. 1996. The impact of human resource man-
flexibility to meet their outside commitments. Role- agement on organizational performance: Progress and prospects.
Acad. Management J. 39(4) 779–801.
based relationships may require greater organizational
investments to foster than personal ties—for example, Becker, G. S. 1975. Human Capital. Columbia University Press,
New York.
designing cross-functional boundary spanner roles and
cross-functional performance measurement systems ver- Bowen, D. E., C. Ostroff. 2004. Understanding HRM-firm perfor-
sus hosting after-work parties—but they are also more mance linkages: The role of the “strength” of the HRM system.
Acad. Management Rev. 29 203–221.
robust to staffing changes that occur over time. High-
performance work systems that foster these role-based Cappelli, P., D. Neumark. 2001. Do high performance work practices
improve establishment level outcomes? Indust. Labor Rel. Rev.
relationships may therefore provide organizations with a 54 737–775.
relatively sustainable source of competitive advantage.
Cappelli, P., N. Rogovsky. 1994. New work systems and skill require-
ments. Internat. Labour Rev. 133(2) 205–220.
Acknowledgments
Chatterjee, S., B. Price. 1991. Regression Analysis by Example. John
The authors thank participants in the MIT Institute for
Wiley and Sons, New York.
Work and Employment Research seminar, particularly Robert
McKersie and Paul Osterman; the Conference on HRM and Chenhall, R. H. 2005. Integrative strategic performance measure-
Knowledge-Related Performance at the Copenhagen Business ment systems, strategic alignment of manufacturing, learning
and strategic outcomes. Account. Organ. Soc. 30(5) 395–422.
School, particularly Scott Snell and Dana Minbaeva; and the
Health Care Industry Council symposium at the Labor and Cleary, P. D., S. Edgman-Levitan, M. Roberts, T. W. Moloney,
Employment Relations Annual Meetings, particularly Dana W. Mullen, J. D. Walker. 1991. Patients evaluate their hospital
Weinberg and Tim Vogus, for their comments on this paper. care: A national survey. Health Affairs 10 254–267.
The authors are grateful for the efforts of the editor Kyle Lewis Collins, C. J., K. Clark. 2003. Strategic human resource practices, top
and three anonymous reviewers, who provided extensive feed- management team social networks, and firm performance. Acad.
back that served to improve the contribution and clarity of this Management J. 46 740–751.
paper. Most of all, the authors thank patients, providers, and Collins, C., K. Smith. 2006. Knowledge exchange and combination:
administrators in the nine hospitals that participated in data The role of human resource practices in the performance of
collection. high-technology firms. Acad. Management J. 49(3) 544–560.
Datta, D. K., J. P. Guthrie, P. M. Wright. 2005. HRM and labor
References productivity: Does industry matter? Acad. Management J. 48(1)
Adler, P., S. Kwon, C. Heckscher. 2008. Professional work: The emer- 135–145.
gence of colloborative community. Organ. Sci. 19(2) 359–376. Day, D., S. Silverman. 1989. Personality and job performance: Evi-
Appelbaum, E., T. Bailey, P. Berg, A. L. Kalleberg. 2000. Manu- dence of incremental validity. Personnel Psych. 42(1) 25–36.
facturing Advantage: Why High-Performance Work Systems Pay Delery, J. E. 1998. Issues of fit in strategic human resource manage-
Off. ILR Press, Ithaca, NY. ment: Implications for research. Human Resource Management
Argote, L. 1982. Input uncertainty and organization coordination in Rev. 8(3) 289–310.
hospital emergency units. Admin. Sci. Quart. 27 420–434. Delery, J. E., D. H. Doty. 1996. Modes of theorizing in strategic
Arthur, J. B. 1992. The link between business strategy and industrial human resource management: Tests of universalistic, contin-
relations systems in American steel minimills. Indust. Labor Rel. gency, and configurational performance predictions. Acad. Man-
Rev. 45 488–506. agement J. 39(4) 802–835.
Arthur, J. B. 1994. Effects of human resource systems on manu- Delery, J. E., J. D. Shaw. 2001. The strategic management of people
facturing performance and turnover. Acad. Management J. 37 in work organizations: Review, synthesis and extension. G. R.
670–687. Ferris, ed. Research in Personnel and Human Resource Man-
agement, Vol. 20. JAI Press, Greenwich, CT, 165–197.
Bailey, T., P. Berg, C. Sandy. 2001. The effect of high-performance
work practices on employee earnings in the steel, apparel, and Deming, J. E. 1986. Out of the Crisis. MIT University Press,
medical electronics and imaging industries. Indust. Labor Rel. Cambridge, MA.
Rev. 54 525–44.
Dunlop, J. T., D. Weil. 1996. Diffusion and performance of modu-
Bartel, A. P. 1994. Productivity gains from the implementation of lar production in the U.S. apparel industry. Indust. Rel. 35(3)
employee training programs. Indust. Rel. 33 411–425. 334–355.
Bartel, A. P. 2004. Human resource management and performance Edmondson, A. 1996. Learning from mistakes is easier said than
outcomes: Evidence from retail banking. Indust. Labor Rel. Rev. done: Group and organizational influences on the detection and
57 181–203. correction of human error. J. Appl. Behav. Sci. 32(1) 5–32.
Gittell, Seidner, and Wimbush: Relational Model of How High-Performance Work Systems Work
16 Organization Science, Articles in Advance, pp. 1–17, © 2009 INFORMS
Evans, W. R., W. D. Davis. 2005. High-performance work systems Katz, J. N., L. C. Chang, O. Sangha, A. H. Fossel, D. W. Bates. 1996.
and organizational performance: The mediating role of internal Can comorbidity be measured by survey rather than medical
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
Nahapiet, J., S. Ghoshal. 1998. Social capital, intellectual capital Thompson, J. D. 1967. Organizations in Action. McGraw-Hill,
and the organizational advantage. Acad. Management Rev. 232 New York.
not be posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to institutional subscribers. The file may
242–266.
Tsai, W., S. Ghoshal. 1998. Social capital and value creation: The
Nohria, N., R. G. Eccles. 1992. Face-to-face: Making network orga- role of intrafirm networks. Acad. Management J. 41 464–476.
nizations work. N. Nohria, R. G. Eccles, eds. Networks and
Tsui, A. S., J. L. Pearce, L. V. Porter, Tripoli. 1997. Alternative
Organizations. Harvard Business School Press, Boston.
approaches to the employee-organization relationship: Does
Osterman, P. 1988. Employment Futures. Oxford University Press, investment in employees pay off? Acad. Management J. 40(5)
New York. 1089–1121.
Pelled, L. H., K. M. Eisenhardt, K. R. Xin. 1999. Exploring the black Tyebjee, T. T., A. V. Bruno. 1984. A model of venture capitalist
box: An analysis of work group diversity, conflict and perfor- investment activity. Management Sci. 30 1051–1066.
mance. Admin. Sci. Quart. 44 1–28.
Vogus, T. 2006. What is it about relationships? A behavioral theory of
Piore, M. 1993. The social embeddedness of labor markets and cog- social capital and performance. Labor Emp. Rel. Proc. 58 164–173.
nitive processes. Labour 7(3) 3–18.
Wageman, R., G. Baker. 1997. Incentives and cooperation: The joint
Ramsey, H., D. Scholarios, B. Harley. 2000. Employees and high- effects of task and reward interdependence on group perfor-
performance work systems: Testing inside the black box. British mance. J. Organ. Behav. 8(2) 139–158.
J. Indust. Rel. 38 501–532.
Walton, R., J. Dutton. 1967. The management of interdepartmental
Raudenbush, S. W., A. S. Bryk. 1992. Hierarchical Linear Mod- conflict: A model and review. Admin. Sci. Quart. 14(1) 73–83.
els: Applications and Data Analysis Methods. Sage Publications,
Newbury Park, CA. Ware, J. E., R. D. Hays. 1988. Methods for measuring patient satis-
faction. Med. Care 26(4) 393–402.
Richard, O. C., N. B. Johnson. 2004. High performance work prac-
tices and human resource management effectiveness: Substitutes Weick, K. E., K. Roberts. 1993. Collective mind in organizations:
or complements? J. Bus. Strat. 21(2) 133–148. Heedful interrelating on flight decks. Admin. Sci. Quart. 38
357–381.
Rohrer, J. E., J. Wilkinson. 2008. Physical limitations and perceived
quality of care among family medicine patients. Clinical Rehab. West, M., C. Borrill, J. Dawson, J. Scully, M. Carter et al. 2002. The
22 283–287. link between the management of employees and patient mortal-
ity in acute hospitals. Internat. J. Human Resource Management
Rousseau, D. M. 1995. Psychological Contracts in Organizations: 13 1299–1311.
Understanding Written and Unwritten Agreements. Sage Publi-
cations, Thousand Oaks, CA. Whitener, E. M. 2001. Do “high commitment” human resource prac-
tices affect employee commitment? A cross-level analysis using
Shortell, S. M., J. E. Zimmerman, D. M. Rousseau et al. 1994. The hierarchical linear modeling. J. Management 27(5) 515–535.
performance of intensive care units: Does good management
make a difference? Med. Care 38 508–525. Wright, P. M., T. M. Gardner, L. Moynihan. 2006. Impact of HR
practices on the performance of business units. Human Resource
Smith, K. G., C. J. Collins, K. D. Clark. 2005. Existing knowledge,
Management J. 13(3) 21–36.
knowledge creation capability, and the rate of new product intro-
duction in high-technology firms. Acad. Management J. 48(2) Youndt, M. A., S. Snell, J. W. Dean, Jr., D. P. Lepak. 1996. Human
346–357. resource management, manufacturing strategy, and firm perfor-
mance. Acad. Management J. 39(4) 836–866.
Snell, S. A., J. W. Dean. 1992. Integrated manufacturing and human
resource management: A human capital perspective. Acad. Man- Young, G., M. Meterko, K. Desai. 2000. Patient satisfaction with hos-
agement J. 35 467–504. pital care. Med. Care 38(3) 325–334.
Stewart, A. L., R. D. Hays, J. E. Ware. 1988. The MOS 36-item Zenger, T. R., W. S. Hesterly. 1997. The disaggregation of cor-
short form health survey (SF-36): Conceptual framework and porations: Selective intervention, high-powered incentives, and
item selection. Med. Care 26 724–735. molecular units. Organ. Sci. 8(3) 209–222.