a relational approach to job design and the coordination ... - CiteSeerX

IS THE DOCTOR IN? A RELATIONAL APPROACH TO JOB DESIGN AND THE COORDINATION OF WORK JODY HOFFER GITTELL, DANA BETH WEINBERG, A D R I E N N E L . B E N N E T T, A N D J O S E P H A . M I L L E R When designing jobs, the degree of specialization is a key consideration. Though functional specialization allows workers to develop deep areas of expertise, it also increases the challenge of coordinating their work. In this article, we propose the concepts of stage- and site-based specialization and posit that together they can counteract the divisive effects of functional specialization. Taking advantage of a natural experiment in physician job design at a Massachusetts hospital, we explore the impact of stage- and site-based specialization on coordination and performance outcomes. Building on recent interest in relational approaches to job design, this study is the first to link relational job design to relational outcomes such as coordination. Our findings have practical implications for job design in professional service settings such as education, consulting, and health care. © 2008 Wiley Periodicals, Inc.

hen designing jobs, the degree of specialization is a key consideration. Some jobs are designed to be broad, encompassing a wide range of tasks that span an entire work process from beginning to end. Other jobs are more specialized, focusing on a narrower set of tasks. Two distinct approaches to job design—the technical (or mechanistic) approach and the psychological (or motivational) approach—offer competing arguments regarding the benefits of broad versus specialized jobs (Morgeson & Campion, 2002). The technical approach to job design argues that specialization and the

W

simplification of work helps organizations to achieve maximum efficiency (Smith, 1991; Taylor, 1911). The psychological approach to job design argues that broad jobs are more intrinsically motivating, satisfying, and conducive to achieving desired outcomes (Ambrose & Kulik, 1999; Likert, 1961; McGregor, 1960) because they provide higher levels of autonomy, significance, and feedback (Hackman & Oldham, 1980). Neither of these approaches has focused on how job design affects the coordination of work, however. Even seemingly minor changes to job design may affect the work process—not only within particular jobs, but

Correspondence to: Jody Hoffer Gittell, Brandeis University, Heller School for Social Policy and Management, 415 South Street, Waltham, MA 02454, Phone: 781-736-3680, E-mail: [email protected]. H u m a n R e s o u r c e M a n a g e m e n t , Winter 2008, Vol. 47, No. 4, Pp. 729–755 © 2008 Wiley Periodicals, Inc. Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/hrm.XXXXX

730

HUMAN RESOURCE MANAGEMENT, Winter 2008

also among them. By influencing the nature, frequency, and quality of interactions among workers, job design may therefore have implications for coordination. The potential for job design to affect the coordination of work is particularly clear from the perspective of relational coordination. Traditionally, coordination has been seen as an informationprocessing problem to be resolved by designing appropriate coordinating mechanisms to ensure the According to the necessary flow of information between people who play different theory of relational roles in the division of labor (Galcoordination, braith, 1977; Tushman & Nadler, 1978). Now, however, coordinacoordination that tion is understood to be a relational process that occurs occurs through through a network of relationfrequent, highships among people who perform interdependent tasks. Because coquality ordination is the management of task interdependence, it is fundacommunication, mentally a relational process supported by (Crowston & Kammerer, 1998; Faraj & Sproull, 2000; Faraj & relationships of Xiao, 2006; Gittell, 2002; Weick & Roberts, 1994). shared goals, One relational perspective on shared knowledge, coordination—relational coordination—identifies specific dimenand mutual respect, sions of relationships that are integral to the coordination of enables interdependent work. Defined as “a mutually reinforcing process of organizations to interaction between communicabetter achieve their tion and relationships, carried out for the purpose of task integradesired outcomes. tion” (Gittell, 2002), relational coordination is expected to provide the information-processing capacity needed to coordinate highly interdependent work. According to the theory of relational coordination, coordination that occurs through frequent, high-quality communication, supported by relationships of shared goals, shared knowledge, and mutual respect, enables organizations to better achieve their desired outcomes (Gittell, 2003). But specialization may weaken relational coordination by breaking down communica-

tion and relationships among participants who work in different areas of specialization. In this article, we consider three types of specialization—functional, stage-based, and sitebased—and their potential impact on relational coordination. Although previous research has explored the impact of functional specialization on relational coordination and the ability of alternative job designs—such as flexible job boundaries (Gittell, 2000) and boundary-spanning roles (Gittell, 2002)—to mitigate its negative effects, the impact of stage- and site-based specialization on relational coordination has not yet been explored. Building on recent interest in relational approaches to job design, this study is the first to link relational job design to relational outcomes such as coordination. In the following sections, we develop hypotheses regarding the impact of these three different forms of specialization on relational coordination and performance. We then test these hypotheses in a patient care setting where coordination is a tremendous challenge, taking advantage of a natural experiment in physician job design at NewtonWellesley Hospital in Newton, Massachusetts. Our findings have practical implications for job design in professional service settings such as education, consulting, and health care.

Functional Specialization Job designs that increase functional specialization are expected to increase quality and efficiency outcomes because of repetition, focus, and the resulting ability to gain higher levels of expertise and skill (e.g., Smith, 1991; Taylor, 1911). Functional specialization also creates distinct occupational communities (Van Maanen & Barley, 1984), distinct thought worlds (Dougherty, 1992), and distinct communities of practice (Brown & Duguid, 1991), thus facilitating relational forms of coordination among people who work in the same function. The shared experience of carrying out the same job function has the potential to create stronger relational ties, such as a greater sense of shared goals, higher levels of shared knowledge, and greater respect for each other’s work, Human Resource Management DOI: 10.1002/hrm

Is the Doctor In? A Relational Approach to Job Design and the Coordination of Work

thereby facilitating frequent, high-quality communication and together resulting in higher levels of relational coordination. However, the same thought worlds, occupational communities, and communities of practice that increase relational coordination within functions are also expected to weaken relational coordination with people who work in different functions. H1: Functional specialization is positively associated with relational coordination within functions, but negatively associated with relational coordination between functions.

Stage-Based Specialization Jobs also can be specialized based on the stage of a work process (such as specialization by time, as conceptualized by Miller, 1959). For example, health care jobs are specialized not only by functional expertise— such as physician, nurse, physical therapist, occupational therapist, social worker, and so on—but often are further specialized by stage of care—primary care, rehabilitative care, acute care, emergency care, and so on. Specialization in a particular stage of work is expected to create a shared experience with others who work at the same stage. For example, nurses and physicians who work in acute hospital care have different areas of functional specialization, but they have in common that they work at the same stage of care and thus face many similar issues and challenges, compared to their colleagues who specialize in, say, primary care. The shared experience that emerges from working at the same stage of a work process has the potential to create stronger relational ties, such as a greater sense of shared goals, shared knowledge, and mutual respect, thereby facilitating higher-quality communication, and giving rise to higher levels of relational coordination.

Site-Based Specialization In addition to function-based and stagebased specialization, jobs can also be specialized according to the site at which a job is to Human Resource Management DOI: 10.1002/hrm

731

be performed (for example, specialization by territory, as conceptualized by Miller, 1959). In health care, for example, a job can be assigned to a particular hospital or clinic, rather than floating across multiple sites. Site-based specialization is expected to create stronger ties among those who work at the same site, because of increased opportunity for contact. For example, nurses or physicians who work together in the same location have greater opportunity for contact with each other than with their colleagues who work in different locations. Proximity, or colocation, has been found to create stronger ties in the form of more frequent, higher-quality Specialization in a communication (Okhuysen & Eisenhardt, 2002; Van Den Bulte particular stage of & Moenaert, 1998). Other scholars have shown that over time, work is expected to workgroup members become encreate a shared trained to external pacers and to one another (Ancona & Chong, experience with 1996; Karau & Kelly, 1992). Group stability—that is, the others who work at frequency and duration of contact the same stage. among group members (Hackman, 1982)—has been shown to increase mood convergence among group members (Bartel & Saavedra, 2000), as well as creativity (Amabile & Conti, 1999) and learning (Edmondson, Bohmer, & Pisano, 2001). Similarly, recent work suggests that group stability fosters group learning about the expertise of each group member (Moreland, Argote, & Krishnan, 1998). Following a similar logic, we expect that frequency and duration of contact may also influence the ability to achieve relational forms of coordination. The contact that occurs through working together at the same site is expected to foster more frequent, higher-quality communication among participants over time, leading to stronger relational ties, and giving rise to higher levels of relational coordination.

Combining Functional, Stage-, and Site-Based Specialization In sum, functional specialization is the division of work into clusters of tasks that re-

732


quire similar skills. Stage-based specialization is the further division of work according to the stage of the work process at which it is carried out. Site-based specialization is the division of work according to the site at which it is carried out. Jobs may be relatively specialized on some dimensions and relatively unspecialized on other dimensions, or they may be specialized on all three dimensions.

When workers coordinate their work through frequent, highquality communication, connected by shared goals, shared knowledge, and mutual respect, they reduce delays, duplicated efforts, and rework, thus

Impact of Stage and SiteBased Specialization on Relational Coordination The previous arguments suggest that stage and site-based specialization improve relational coordination primarily because they help to counteract the divisive effects of functional specialization. We therefore predict that workers in different areas of functional specialization who work together consistently at the same stage and site of a work process will experience higher levels of relational coordination with each other. Although these other forms of specialization may create divisions of their own (between stages or sites of work), they will also serve to increase relational coordination at a given stage and site of work.

improving efficiency.

H2: Stage- and site-based specialization are positively associated with relational coordination between functions.

Stage- and site-based specialization are therefore expected to improve efficiency performance by improving relational coordination between workers in different functions. H3a: Stage- and site-based specialization are positively associated with efficiency performance. H3b: The association between stage- and sitebased specialization and efficiency performance is mediated by relational coordination between functions. Furthermore, when workers coordinate their work through frequent, high-quality communication, connected by shared goals, shared knowledge, and mutual respect, they reduce the risk of errors, thus improving quality performance as well. This argument is consistent with Edmondson’s (1996, 1999) work linking communication and respect to error reduction. Stage- and sitebased specialization are therefore expected to improve quality performance by improving relational coordination between workers in different functions. H4a: Stage- and site-based specialization are positively associated with quality performance. H4b: The association between stage- and sitebased specialization and quality performance is mediated by relational coordination between functions. In sum, we expect stage- and site-based specialization to strengthen relational coordination between functions, thereby resulting in higher levels of efficiency and quality performance. These hypotheses are illustrated in Figure 1.

Impact of Stage- and Site-Based Specialization on Performance

The Setting for This Study

Because of its impact on relational coordination, stage- and site-based specialization are expected to improve both efficiency and quality performance. When workers coordinate their work through frequent, high-quality communication, connected by shared goals, shared knowledge, and mutual respect, they reduce delays, duplicated efforts, and rework, thus improving efficiency.

Coordination problems plague the current health care system (e.g., Audet, Doty, Shamasdin, & Schoenbaum, 2005; Institute of Medicine, 2003). Patients are often required to sort their way through the system, receiving diagnoses and treatments from a fragmented, loosely connected set of providers (e.g., Kenagy, Berwick, & Shore, 1999). Even within the hospital setting, where resources are preHuman Resource Management DOI: 10.1002/hrm


Relational Coordination Between Functions Job Design Stage- and Site-Based Specialization

aThis

Shared Goals Shared Knowledge Mutual Respect Frequent Communication Timely Communication Accurate Communication Problem-Solving Communication

Efficiency Performance Excess Length of Stay Total Costs

Quality Performance Mortality Readmit Within 7 Days Readmit Within 30 Days

figure illustrates how the impact of job design on quality and efficiency performance is mediated through relational coordination.

FIGURE 1. Mediated Model of Job Design, Relational Coordination and Performancea

sumably brought together within a single organization to improve the coordination of their deployment, coordination often falls to the patient and his or her family members (Cleary, 2003). Physicians are expected to be at the center of coordination and decision making, but they tend to be present in the hospital intermittently because of their external responsibilities. While other members of the care team, including nurses, residents, therapists, social workers, and case managers, are largely hospital-based and therefore work together on an ongoing basis, physicians have traditionally maintained private practices outside the hospital, coming to the hospital primarily when their patients are hospitalized. Moreover, as reimbursement levels have been reduced, physicians have been required to see more and more patients in their private practices to maintain their incomes. At the same time, managed care pressures have resulted in fewer hospital admissions, resulting in fewer but sicker patients in the hospital and increasing the challenge of coordinating their care. One response to these pressures has been the creation of a new job design for physicians called “hospitalist” or “hospital specialist” (Wachter & Goldman, 1996). Rather than following their patients from outpaHuman Resource Management DOI: 10.1002/hrm

tient care to inpatient care, physicians now have the option of handing them off to a hospitalist physician who becomes responsible for that patient’s care during the hospital stay. The result of this new job design is that some physicians are dedicated to hospitalbased care, working exclusively in the same hospital with the same staff, so that patients can be admitted to the care of a dedicated hospitalist physician rather than remaining under the care of their own private practice physician while in the hospital. Our study explores this new job design in one hospital, using the natural experiment provided by the fact that some patients there were assigned to hospitalist physicians while others remained under the care of their own private practice physicians. Although this new physician job design was not developed specifically to address problems with the coordination of care, the hypotheses developed above suggest that this job design will in fact improve outcomes by improving coordination between physicians and other members of the care team.

Methods To test our hypotheses, we chose a hospital that had implemented a hospitalist job de-

733

734


sign five years before the study but that continued to have physicians with the traditional job design treating patients in the same unit. At the time of the study, hospitalists treated approximately one-third of the hospital’s medical patients, and private practice physicians treated the other two-thirds. To test our hypotheses regarding the impact of job design on relational coordination, we used a sample of 335 patients, about half of all patients who were cared for in the hospital’s medical unit from April 2003 to June 2003. For each patient in this sample, we surveyed each of the nurses, case managers, and therapists responsible for their care, asking detailed Patients are often questions about the coordination of their care with other members required to sort their of the care team, including the physician responsible for that paway through the tient (the attending physician). The attending physician for each system, receiving patient was either a private pracdiagnoses and tice physician (traditional physician job design) or a hospitalist treatments from a (new physician job design). All patients attended by a hospitalist fragmented, loosely during the three-month study peconnected set of riod were included in our sample. We initially selected a random providers. sample of half of the patients attended by a private practice physician, but later in the survey period we began to sample every patient to ensure an adequate sample size. To guard against recall problems, we surveyed the patient teams immediately after patient discharge. Both the patient and the care providers about whom respondents were surveyed were named on the survey forms, and their names appeared in the patient’s records, particularly in their medical notes. Patient records were available electronically to aid recall related to events during the patients’ stay. To test our hypothesis regarding the impact of job design on performance outcomes, we used data on all 6,686 patients with complete data who were cared for in the medical unit from July 2001 to June 2003. A large sample of patients is useful for predicting pa-

tient outcomes—particularly when the patient population is as varied as it is in a medical unit—to distinguish the effects of organizational factors from the effects of individual patient characteristics. To test our final hypothesis regarding mediation of the impact of job design on performance by relational coordination, we returned to the smaller patient sample described above because of the availability of relational coordination measures for that sample.

Job Design To capture the difference in physician job design, we created a dummy variable called “hospitalist job design” with a value equal to 1 if the physician assigned to the patient in question had the new hospitalist job design, and equal to 0 otherwise.

Relational Coordination This is a multilevel network construct developed and tested in previous research, operating at the individual, group, and interorganizational levels (e.g., Gittell, 2002; Gittell & Weiss, 2004; Gittell, Weinberg, Pfefferle, & Bishop, 2008; Weinberg, Lusenhop, Gittell, & Kautz, 2007). Relational coordination was measured for this study using a previously developed survey (Gittell et al., 2000) that was administered to nurses, therapists, and case managers in the hospital unit, asking respondents seven questions about their interactions with providers who shared responsibilities for a particular patient, including the attending physician. Because of our focus on variation with the unit, these questions were asked with respect to the providers assigned to a specific patient, rather than at the more aggregated level of the organizational unit. They included: How frequently did you communicate with each of these people about the status of Patient X? Did these people communicate with you in a timely way about the status of Patient X? Did these people communicate with you accurately about the status of Patient X? When problems occurred regarding the care of Patient X, did these people blame others or work to solve the problem? To what exHuman Resource Management DOI: 10.1002/hrm


tent did these people share your goals for the care of Patient X? How much did these people know about the work you did with Patient X? How much did these people respect the work you did with Patient X? Items were measured on a five-point Likert-type scale. Please see the Appendix for the original survey document. Of the 2,727 surveys sent, 1,800 responses were received, for an overall provider response rate of 66%. Responses were dropped if they referred to the care of a patient who was “observation only” and therefore not technically admitted to the hospital (a fact that often was not known until after the patient had already been discharged, because of delays in payer decision making). We also dropped responses that did not complete the questions regarding coordination with the physician on the team. We were left with 893 complete provider responses regarding the care of 335 patients. Relational coordination with the physician was constructed for each respondent based on responses regarding the physician on a particular patient care team, including frequency of communication, timeliness of communication, accuracy of communication, problem-solving communication, shared goals, shared knowledge, and mutual respect with the physician on the team. Exploratory factor analysis suggested that relational coordination is best characterized as a single factor with the following loadings: frequency, 0.6759; timeliness, 0.8725; accuracy, 0.8566; problem solving, 0.8943; shared goals, 0.9053; shared knowledge, 0.8596; and mutual respect, 0.9057, with an eigenvalue of 5.13 and no cross-loadings greater than 0.20. The seven measures were then averaged into a single index, relational coordination with the physician (alpha = 0.93), and aggregated across all providers who worked with the same patient. Similar indices were constructed for relational coordination with residents (alpha = 0.94), nurses (alpha = 0.87), therapists (alpha = 0.87), and case managers (alpha = 0.87).

Performance Outcomes Performance measures for this study included two efficiency outcomes and three Human Resource Management DOI: 10.1002/hrm

735

quality outcomes. Efficiency outcomes were excess length of stay in the hospital and total costs of stay. Excess length of stay was measured as the difference between the patient’s actual length of stay and the Massachusetts state average for a comparable patient. We created this measure using three steps. First, using 3M’s APR-DRG grouper software, we classified each patient into a diagnostic group and within that diagnostic group, assigned a severity of illness measure between 1 and 4, based on the patient’s secondary diagnosis. Using a reference database of all acute-care hospitalizations (836,126) in Massachusetts in 2003, we created a normative case-mix measure equal to the mean length of One response to stay for patients in the larger population who were in both the these pressures has same severity of illness and diagnostic group. We used this nor- been the creation of mative “severity of illness” measure as a case-mix adjuster in all a new job design for our models. To create the excess physicians called length of stay variable, we rounded the average length of “hospitalist” or stay for each patient profile to the nearest integer (since the length “hospital specialist.” of stay data supplied by the hospital was also recorded as an integer) and subtracted it from the patient’s actual length of stay to create a comparison of the efficiency of their care to the state average for a comparable patient. Thus, the dependent variable is the difference between actual and expected length of stay. The total cost measure was available through the hospital’s activity-based cost accounting system. We also estimated models with a log-transformed value for total costs. Quality outcomes were patient mortality, patient readmission to the hospital within seven days, and patient readmission within 30 days. Mortality is an important quality measure, though incidence tends to be so low that large samples are needed to enable statistically significant results to be found. Patient readmissions are another common measure of quality performance in medical units, indicating the possibility that an error occurred during hospitalization, or that a pa-

736


tient was discharged prematurely. Each of these measures is dichotomous (for example, the patient was readmitted in seven days, yes or no), and each was available from the hospital information system. To serve as reliable measures, all quality and efficiency measures in our models were risk-adjusted.

Control Variables Several control variables were used in this study to risk-adjust performance outcomes. Control variables included patient severity of illness, age and gender, number of days the patient stayed in the intensive care unit (ICU), and a To control for propensity score for being treated potential differences by a hospitalist rather than a traditional physician. Severity of illin the selection of ness was expected to increase length of stay, total costs of the patients treated by stay, and likelihood of readmission and mortality. To create a hospitalists, we continuous measure of severity created a that could be used across diagnoses, we used a reference datapropensity score base of all acute-care hospitalizations (836,126) in Massachusetts that measures the in 2002 to create a measure equal probability of being to the mean length of stay for patients in the larger population treated by a who were in both the same diagnostic group (based on APR-DRG hospitalist. classification derived from primary diagnosis) and severity of illness group (between 1 and 4, based on secondary diagnosis). For example, the severity of illness measure for a level-3 pneumonia patient is 5.2, the statewide average length of stay for patients in that class. Though our models included only the outcomes of patients in the target hospital, the statewide sample was used as a basis for risk adjustment due to the benefits of a larger sample. Similar to severity of illness, patient age and number of days in the ICU were expected to affect costs, length of stay, and likelihood of readmission and mortality. The expected effects of gender were less certain, but in keeping with common practice we in-

cluded gender in our models. Patient age, ICU admission, and gender were extracted from hospital records. In the models of readmission after seven and 30 days, we included an additional control for whether the patients’ private practice physician had an affiliation with the hospital (94% did). Since our data included readmission only to this particular hospital, we do not know whether patients were admitted to other facilities during this time period. Physicians with ties to this hospital, however, are more likely to admit their patients here than elsewhere for subsequent hospitalizations. To control for potential differences in the selection of patients treated by hospitalists, we created a propensity score that measures the probability of being treated by a hospitalist. The propensity score was calculated using logistic regression of treatment by a hospitalist on a dummy variable for each of the APR-DRG classifications of patients in the sample, the patient’s severity of illness (1 to 4) and risk of mortality measures within that category (1 to 4), patient age and gender, and a dummy variable for the day of the week the patient was admitted. Using these results, we calculated the predicted probability of treatment by a hospitalist for each patient.

Analyses We conduct one-tailed t-tests to examine the differences in relational coordination within and between different types of specialization. These analyses used reports of relational coordination by individual providers. The unit of analysis for all other statistical analyses in this article was the patient or the patient’s team of providers. In both relational coordination and performance outcome models, we used random-effects regression analysis with patient or patient team as the unit of analysis and physician as the random effect, since patients treated by the same physician may experience similar outcomes, and their care teams may experience similar levels of coordination with the physician. The nested nature of our data, in which the same physicians oversaw the care of mulHuman Resource Management DOI: 10.1002/hrm


tiple patients in the sample, called for an approach that addressed the physician-specific effects in the data. Thus, we chose randomeffects modeling because of its ability to adjust standard errors for the lack of independence between observations (Raudenbush & Bryk, 1992). Although structural-equation modeling would provide an opportunity to model relational coordination as a latent variable and to explore the full theoretical model at once, we do not currently know of a widely used method that incorporates both the flexibility of structural equation modeling and the correction for nested data with so many clusters of varying sizes. We estimated the effects of job design on relational coordination and performance outcomes controlling for physician volume of admissions and patient severity of illness, age, and gender. In the models using our larger sample, we also controlled for number of days in the intensive care unit, though we excluded this measure from the models using the coordination sample, since none of the patients in that smaller sample were admitted to the ICU. Logistic random-effects regressions were used for estimating the likelihood of readmission and mortality, given that these dependent variables are dichotomous. In the readmission models, patients who died in the hospital were removed from the sample, since they could not be readmitted. In the readmission models, we used a dummy variable to denote patients who had no preexisting ties to the hospital, given the possibility that they were readmitted to a different hospital and, therefore, not captured in our readmission data. In models that tested for the mediation of job design and performance by relational coordination, patients were again the unit of analysis. To demonstrate that relational coordination mediates the relationship between job design and performance requires that three conditions be met: (1) variations in job design are significantly related to variations in relational coordination, (2) variations in relational coordination significantly relate to variations in performance, and (3) when both job design and relational coordination are controlled, the relationship beHuman Resource Management DOI: 10.1002/hrm

737

tween job design and performance is no longer significant or is substantially decreased (Baron & Kenney, 1986). We conducted a Sobel test to determine whether those conditions for mediation were met.

Threats to Validity in a QuasiExperimental Design Because our research design involved a static posttest comparison with no pretest data, it was particularly vulnerable to four of the 12 validity threats discussed by Campbell and Stanley (1966).

Two Threats to Internal Validity

Given the possibility

that factors other Given the possibility that factors other than the change in job dethan the change in sign are responsible for the patterns we observed, history was a job design are serious threat to the internal varesponsible for the lidity of our findings. The results might be explained, for example, patterns we by the Hawthorne effect, in which increased attention from observed, history managers or knowledge of participating in an experiment is re- was a serious threat sponsible for the benefits obto the internal served. Given the fact that this particular experiment in job devalidity of our sign had been in progress for five findings. years before we observed it, however, we expected that the novelty of participating in the experiment would have worn off considerably and, therefore, that the Hawthorne effect would play, at most, a minor role in the observed results. Selection was a second threat to the internal validity of our findings, given that physicians were not randomly assigned to the two job designs, and given that those who selected hospital specialization were likely to have differed in meaningful ways from those who did not. Moreover, the physicians’ ability to determine whether to treat their own patients or refer them to a hospital specialist introduced another selection issue into this study, particularly if there

738


was a type of patient for whom a hospital specialist would be well suited and another type of patient for whom a hospital specialist would not be well suited. Without specifically addressing physician decision making, we handled this issue by controlling for the probability that each patient would be assigned to a hospitalist given factors like the patient’s age, primary diagnosis, and illness.

Two Threats to External Validity Selection-treatment interaction was a potential threat to the external validity of our findings, given that physicians who were interested in higher levels of coordination with hospital staff were more likely to select or be selected for this job. This concern is consistent with human resource management theories that suggest employee selection should be carried out deliberately to match the type of person to the type of job. A second potential threat to the external validity of our study was the reactive effect of the experimental arrangement, given the possibility that physicians who were not hospital specialists perceived unequal treatment, result-

TABLE

I

ing in resentful demoralization (Cook & Campbell, 1979; Cook & Shadish, 1994). This potential threat was reduced, however, by the fact that all referring physicians in our study had the option of treating the patient themselves rather than handing off the patient to a hospitalist. The fact that their referral was voluntary can be expected to reduce potential feelings of resentment. If they did not find handing off to be more convenient or favorable for themselves or their patient, they were not forced to do so.

Job Design and Relational Coordination Models Table I shows mean levels of relational coordination reported by nurses, therapists, and case managers with each member of the care team. Each cell shows relational coordination as reported by the function in the lefthand column with respect to the function in the top row. Physicians and residents were not surveyed and therefore do not appear in the left-hand column. Using one-tailed t-tests with unequal variance, we tested the significance of the

Relational Coordination Within and Between Functionsa

Nurses

RC With Physician 3.77

RC With Residents 3.93

RC With Nurses 4.35

RC With Therapists 3.86

RC With Case Managers 4.05

Therapists

(p = .0000) (t = –8.97) (df = 703) 2.36

(p = .0000) (t = –7.92) (df = 717) 2.46

3.97

(p = .0000) (t = –26.92) (df = 452) 4.28

(p = .0000) (t = –5.20) (df = 693) 3.74

(p = .0000) (t = –16.46) (df = 418) 3.65

(p = .0000) (t = –15.12) (df = 375) 3.25

(p = .0001) (t = –3.79) (df = 404) 4.23

3.17

(p = .0000) (t = –5.89) (df = 462) 4.52

Case Managers

(p = .0000) (t = –8.74) (df = 625)

(p = .0000) (t = –12.26) (df = 547)

(p = .0002) (t = –3.57) (df = 450)

(p = .0000) (t = –10.97) (df = 412)

a

Unit of observation is the provider response (n = 893). Each cell shows relational coordination as reported by the provider type in the lefthand column with respect to the provider type in the top row. Physicians and residents were not surveyed and therefore do not appear in the left-hand column. Within-function ties are in bold, and between-function ties are in regular typeface. A one-tailed t-test with unequal variance was used to test for significance of difference between the within-function ties and between-function ties in the same row. Human Resource Management DOI: 10.1002/hrm


differences between within-function ties (nurse/nurse, therapist/therapist, case manager/case manager) and between-function ties (nurse/physician, nurse/resident, nurse/therapist, nurse/case manager, and so on) in each row. In every case, the cells representing within-function ties report significantly higher levels of relational coordination than the cells representing between-function ties. For example, nurses’ relational coordination with other nurses (4.35) was significantly stronger than their coordination with physicians (3.77, p < 0.001), residents (3.93, p < 0.001), therapists (3.86, p < 0.001), and case managers (4.05, p < 0.001). Therapists’ relational coordination with other therapists (4.28) was significantly stronger than their coordination with physicians (2.36, p < 0.001), residents (2.46, p < 0.001), nurses (3.97, p < 0.001), and case managers (3.74, p < 0.001). Case managers’ relational coordination with other case managers (4.52) was significantly stronger than their coordination with physicians (3.65, p < 0.001), residents (3.25, p < 0.001), nurses (4.23, p < 0.001), and

TABLE

II

Nurses

Therapists

Case Managers

therapists (3.17, p < 0.001). These results are significant even if we use a higher standard of significance to correct for global type-1 error (0.05 divided by 12 tests = 0.004). These findings therefore provide strong support for Hypothesis 1, which stated that specialization by function is associated with higher levels of relational coordination within functions than between functions. Table II shows mean levels of relational coordination reported by nurses, therapists, and case managers with each member of the care team under the traditional job design (first number in each cell) versus the hospitalist job design (second number in each cell). Using one-tailed t-tests with unequal variance where needed, we tested the significance of the differences between relational coordination under the traditional job design versus the hospitalist job design. We see that relational coordination with the physician was significantly stronger under the hospitalist job design than under the traditional job design. For nurses, relational coordination with the physician was 3.34 under

Relational Coordination Under Traditional Job Design Versus Hospitalist Job Designa RC With Physician 3.34 4.30

RC With Residents 3.90 3.97

RC With Nurses 4.31 4.40

RC With Therapists 3.80 3.94

RC With Case Managers 3.96 4.16

(p = .0000) (t = –9.74) (df = 440) 2.10 2.69

(p = .1939) (t = –0.86) (df = 430) 2.19 2.79

(p = .0958) (t = –1.31) (df = 285) 3.89 4.07

(p = .1368) (t = –1.10) (df = 292) 4.34 4.21

(p = .0151) (t = –2.17) (df = 410) 3.70 3.81

(p = .0008) (t = –3.18) (df = 225) 3.22 4.17

(p = .0011) (t = –3.10) (df = 204) 3.29 3.20

(p = .0256) (t = –1.96) (df = 276) 4.26 4.18

(p = .8270) (t = 0.94) (df = 227) 3.16 3.20

(p = .1828) (t = –0.91) (df = 258) 4.38 4.62

(p = .0000) (t = –6.86) (df = 378)

(p = .7013) (t = 0.53) (df = 314)

(p = .8303) (t = 0.96) (df = 344)

(p = .4250) (t = –0.19) (df = 254)

(p = .0199) (t = –2.07) (df = 225)

a

Unit of observation is the provider response (n = 893). Each cell shows relational coordination as reported by the provider type in the left-hand column with respect to the provider type in the top row, first under traditional job design (first number in each cell) then under hospitalist job design (second number in each cell). Physicians and residents were not surveyed and therefore do not appear in the left-hand column. One-tailed t-tests, with unequal variance where needed, were used to test for significance of difference between the two means.

Human Resource Management DOI: 10.1002/hrm

739

740


the traditional job design and 4.30 under the hospitalist job design (p < 0.001); for therapists, relational coordination with the physician was 2.10 under the traditional job design and 2.69 under the hospitalist job design (p < 0.001); and for case managers, relational coordination with the physician was 3.22 under the traditional job design and 4.17 under the hospitalist job design (p < 0.001). There were some differences in relational coordination among other members of the team as well. Under the hospitalist job design, nurses reported stronger ties with case managers (p < 0.05), therapists reported Readmissions are stronger ties with residents (p < 0.01), and case managers resignificantly ported stronger ties with each other (p < 0.05). But these differcorrelated with ences were less significant than whether the patient’s the differences in relational coordination with the physician. physician had Indeed, if we correct for global type-1 error by requiring a preexisting ties to higher standard of significance (0.05 divided by 15 tests = the hospital, 0.003), then the only significant confirming the differences in relational coordination associated with the new importance of job design were between the physician and other members of including those ties the care team. This is not surprisin our readmission ing, given that the change in job design studied here is specifically models. a change in physician job design. We therefore included only relational coordination between the physician and other members of the care team in our subsequent models. Table III presents descriptive data for the job design and relational coordination models. Table IV gives the results from the random-effects regression models. Controlling for other factors, the hospitalist job design, distinguished from the traditional job design by its stage- and site-based specialization, results in higher levels of relational coordination between other members of the team and the physician (r = 0.66, p < 0.01). These findings are consistent with Hypothesis 2.

Job Design and Performance Models Table V gives descriptive data for the job design and performance models. Patients assigned to hospitalist physicians tended to be younger and to have less severe illnesses. Additionally, patients who were younger or had less severe illnesses tended to have significantly better outcomes. These correlations together confirm the importance of including patient age and severity of illness in our outcome models and also the importance of including a propensity score for treatment by a hospitalist to control for potential selection of healthier patients into the hospitalist service. Readmissions are significantly correlated with whether the patient’s physician had preexisting ties to the hospital, confirming the importance of including those ties in our readmission models. Table VI presents the results from the random-effects regression models. Physician job design was negatively associated with excess length of stay (r = –0.46, p < 0.001), total costs per stay (r = –655, p < 0.001) and the log-transformed value of costs per stay (r = –0.07, p < 0.001). Physician job design was not significantly associated with patient mortality, though the sign is in the expected direction. But physician job design marginally predicted reduced likelihood of readmission after seven days (r = –0.24, p < 0.10) and significantly predicted reduced likelihood of readmission after 30 days (r = –0.33, p < 0.05). In sum, the physician job design that incorporates stage- and site-based specialization appears to have improved efficiency outcomes and some quality outcomes, providing strong support for Hypothesis 3a and mixed support for Hypothesis 4a, with results differing depending on the nature of the outcome variable.

Mediated Model of Job Design, Relational Coordination, and Performance Table III shows descriptive data for the job design, relational coordination, and performance models. The sample included all patients for whom measures of both performance and Human Resource Management DOI: 10.1002/hrm


Patient severity of illness

Patient age

Patient gender

Probability of assignment to hospitalist

Excess length of stay

Total costs

Log total costs

3.

4.

5.

6.

7.

8.

9.

+p < .10 *p < .05 **p < .01 ***p < .001 a Unit of observation is the patient (n = 335).

11. Readmission after seven days 12. Readmission after 30 days

10. Mortality

Relational coordination between other team members and the physician

2.

4.72 (1.80) 73.4 (17.00) .56 (.50) .28 (.11) –1.15 (2.09) 5281.87 (3655.75) 8.39 (.59) .02 (.15) .04 (.20) .12 (.33)

3.99 (.84)

.42 (.49)

Mean (SD)

.04 .05

–.11

.09+

–.12*

–.09

–.21***

.00

–.01

–.09

.03

1.00

2.

–.04

.06+

–.09+

–.06***

–.12*

.23***

–.04

–.15**

–.01

.39***

1.00

1.

.15**

.06

.19***

.45***

.40***

–.35***

–.33***

–.05

.19***

1.00

3.

Descriptive Data for Relational Coordination Modelsa

Physician job design

III

1.

TABLE

.09

.02

.05

.05

.01

–.11+

–.42***

.15*

1.00

4.

–.04

.02

.02

.03

.04

.09+

–.32***

1.00

5.

–.07

–.02

–.09+

–.26***

–.23***

.02

1.00

6.

–.09+

–.13*

–.09+

.52***

.51***

1.00

7.

.02

–.04

.05

.92***

1.00

8.

.03

–.04

.07

1.00

9.

--

–-

1.00

10.

.06***

1.00

11. Is the Doctor In? A Relational Approach to Job Design and the Coordination of Work 741

742


TABLE

IV

Job Design and Relational Coordination Modela

Physician job design Probability of assignment to hospitalist Patient severity of illness Patient age Patient gender Constant R squared

Relational Coordination Between OtherTeam Members and the Physician .66** (.004) –.86+ (.062) –.00 (.967) –.00 (.125) –.08 (.342) 4.21*** (.000) .16

+p < .10 *p < .05 **p < .01 ***p < .001 a

Unit of observation is the patient (n = 312). Observations were dropped if coordination with the physician was not reported. All models are random effects regressions with physician as the random effect (n = 49).

relational coordination were available (n = 312). Because of the smaller subset of patients for whom relational coordination outcomes were available, this subsample was smaller than the full sample used in our original job design and outcomes models (n = 6,866). Among our outcome measures, only the two efficiency outcomes had sufficient variation to enable us to test our models with sufficient power to detect significant effects of job design. Not only were there very few cases of readmissions and mortalities, but the random-effects models we used also decreased statistical power. We encountered a different issue related to statistical power when we examined costs per stay. The variation in the variable was too large to detect patterns in our small sample. This problem was tempered when we used the log-transformed measure. Thus, we present the results for excess length of stay and the log of total costs. Table VII presents the results from the mediated model of job design, relational co-

ordination, and efficiency outcomes. Physician job design was negatively associated with excess length of stay in this smaller sample (r = –0.59, p < 0.05), consistent with our findings in the larger sample. When relational coordination was included in the model, relational coordination was negatively associated with excess length of stay (r = –0.46, p < 0.01) and the coefficient on physician job design became smaller and insignificant. We found similar results for log total costs. Physician job design was negatively associated with log total costs in this smaller sample (r = –0.13, p < 0.05), also consistent with our findings in the larger sample. When relational coordination was included in the model, relational coordination was negatively associated with log total costs (r = –0.08, p < 0.05), and the coefficient on physician job design became smaller and insignificant. We used the Sobel test to assess whether the association between physician job design and outcomes was reduced significantly when controlling for the mediator of relational coordination. Drawing on the critical values recommended by MacKinnon, Lockwood, Hoffman, West, and Sheets (2002), we found that the results for both outcomes supported mediation (excess length of stay: z’ = 2.12, p < 0.01; total costs: z’ = 1.73, p < 0.01). Together, these results suggest that the physician job design explored in this study is associated with higher levels of both relational coordination and efficiency outcomes, with neutral to positive associations with quality outcomes. For the excess length of stay and total cost outcomes, we found that job design is associated with higher levels of performance through relational coordination.

The Significance of What We Have Found We took advantage of a natural experiment in physician job design to explore the impact of job design on relational forms of coordination. In doing so, we uncovered two types of specialization that are often ignored. In Human Resource Management DOI: 10.1002/hrm


Excess length of stay

+p < .10 *p < .05 **p < .01 ***p < .001 a Unit of observation is the patient (n = 6,866).

12. Mortality

11. Readmit within 30 days

10. Readmit within 7 days

9. Log total costs

8. Total costs

7.

6. Probability of assignment to hospitalist

5. Patient ICU days

4. Patient gender

3. Patient age

5.2 (2.90) 72.6 (17.30) 0.59 (.49) .4 (1.80) .27 (.11) –.91 (2.53) 7,122 (9,501) 8.52 (.56) .06 (.24) .16 (.36) .03 (.17)

.26 (.44)

Mean (SD)

–.11***

–.05***

–.03**

–.08***

–.03**

–.07***

.23***

.04***

–.06***

–.08***

–.03**

1.00

1.

.11***

.07***

.23***

.57***

.64***

–.20***

–.25***

.55***

–.03**

.13***

1.00

2.

.05***

.04**

.08***

.06

–.01

–.06***

–.37***

.00

.10***

1.00

3.

–.01

.00

.00

–.01

–.03*

.03**

–.23***

–.04***

1.00

4.

Descriptive Data for Job-Design and Performance Modelsa


V

2. Patient severity of illness

1.

TABLE

.00

.00

.10***

.44***

.69***

.11***

.06***

1.00

5.

–.12***

–.07**

–.10***

–.23

–.11***

.04**

1.00

6.

.02+

.00

–.11***

.51***

.39***

1.00

7.

.07***

.03**

.13***

.76***

1.00

8.

.11***

.05***

.10***

1.00

9.

.08***

.04**

1.00

10.

.59***

1.00

11. Is the Doctor In? A Relational Approach to Job Design and the Coordination of Work 743

744


TABLE

VI

Job-Design and Performance Modelsa

Physician job design Probability of assignment to hospitalist Patient severity of illness Patient days in ICU Patient age Patient gender Patient ties to hospital Constant R squared Chi squared Observations

Efficiency Excess Length of Log Total Stay Costs –.46*** –.07** (.000) (.004) –1.70*** –1.03*** (.000) (.000) –.36*** .11*** (.000) (.000) .52*** .09*** (.000) (.000) –.01*** .002*** (.000) (.000) .08 –.03+ (.192) (.090) -1.74*** (.000) .11 NA 6,699

-8.37*** (.000) .37 NA 6,866

Mortality –.26 (.184) –4.85*** (.000) .20*** (.000) –.05 (.124) .03*** (.000) –.23 (.127) -–5.61*** (.000) NA .0000 6,866

Quality Readmit Within 7 Days –.24+ (.089) –1.45* (.015) .09*** (.000) –.08* (.031) .00 (.170) –.07 (.513) 3.02** (.003) –6.07*** (.000) NA .0000 6,866

Readmit Within 30 Days –.33* (.033) –1.89*** (.000) .10*** (.000) –.08** (.001) .00 (.526) –.22** (.002) 2.33*** (.000) –3.94*** (.000) NA .0000 6,866

+p < .10 *p < .05 **p < .01 ***p < .001 a Unit of observation is the patient. All models are random effects regressions with physician (n = 116) as the random effect. Quality models are logistic random effects regressions.

addition to functional specialization, there is also stage-based specialization, in which a job is dedicated to a particular stage of work, and site-based specialization, in which a job is dedicated to a particular location. We hypothesized that a job that is specialized by function will be more conducive to cross-functional coordination if it is also specialized by stage and site of work. The shared experience that emerges from working on the same stage of work has the potential to create stronger relationships with other functions, including a greater sense of shared goals, shared knowledge, and mutual respect, thereby facilitating frequent, highquality communication and resulting in higher levels of relational coordination across functions. Specialization by site increases frequency and duration of contact, which has the potential to create more frequent, higher-quality communication over

time, leading to stronger relationships and together resulting in higher levels of relational coordination across functions. We found that a job design that was specialized by stage and site in addition to function had the effect of strengthening weak ties between functions. We hypothesized that this job design would generate better performance than a job design based on functional specialization alone. The results supported these expectations, with the new job design associated with improved performance. We conclude that these lesser-known forms of specialization (stage- and site-based specialization) can help to overcome the fragmentation that arises from functional specialization. Our findings have several important implications for theory and practice. Critics of bureaucracy have long recognized that functional specialization undermines coordination between functions, increasing fragmenHuman Resource Management DOI: 10.1002/hrm


TABLE

V I I Mediated Model of Job Design, Relational Coordination, and Performanceaa


Excess Length of Stay –.59* (.011)

–.53*** (.000)

Excess Length of Stay –.29 (.250) –.46*** (.000)

–2.59* (.048) –.47*** (.000) –.02* (.012) .37 (.118) 3.31*** (.000) .18 304

–3.27* (.011) –.46*** (.000) –.02** (.009) .34 (.143) 5.39*** (.000) .21 304

–2.99* (.021) –.46*** (.000) –.02** (.009) .35 (.126) 5.20*** (.000) .21 304

Log Total Costs –.13* (.037)

Log Total Costs

Log Total Costs –.07 (.273)

–.10** (.005)

–.08* (.030)

–1.09** (.001) .13*** (.000) –.01** (.005) .04 (.564) 8.90*** (.000) .23 312

–1.01** (.004) .13*** (.000) –.01** (.004) .04 (.520) 8.85*** (.000) .23 312

Relational coordination between other team members and the physician Probability of assignment to hospitalist Patient severity of illness Patient age Patient gender Constant R squared Observations

Physician job design Relational coordination between other team members and the physician Probability of assignment to hospitalist Patient severity of illness Patient age Patient gender Constant R squared Observations

–.94** (.007) .13*** (.000) –.01** (.050) .04 (.493) 8.51*** (.000) .22 312

Excess Length of Stay

+p < .10 *p < .05 **p < .01 ***p < .001 a Unit of observation is the patient. All models are random effects regressions with the physician (n = 49) as the random effect.

tation (Gouldner, 1954; Merton, 1940; Selznick, 1949). Still, human resource management theorists and practitioners recognize the benefits of functional specialization for achieving high levels of expertise and skill, particularly as knowledge proliferates. Our findings suggest that one powerful way to maintain these benefits of functional specialization while counteracting its negative effects is to Human Resource Management DOI: 10.1002/hrm

specialize by stage and site of work, in addition to function. More broadly, our findings suggest that jobs should be designed with explicit attention to how well they coordinate with other jobs. Although job-design research has had much to say about the design of individual jobs, it has been lacking a good theory to explain combinations of jobs and mixtures of multiple

745

746


job designs. Relational coordination helps to address this gap because it accounts for the communication and relationship ties among different members of a team, each with his or her own area of expertise. We can therefore see how job design influences not only an individual’s performance, but, through its effect on coordination, also influences organizational performance. This article offers a relational approach to job design, responding to the recognition that relationships can play an important role in job design (Grant, 2007; Grant et al., 2007; Humphrey, Nahrgang, & Morgeson, 2007; Morgeson & Humphrey, 2006; Wrzesniewski & …jobs should be Dutton, 2001). Relational approaches were popular in early designed with job-design research (Hackman & Lawler, 1971; Karasek, 1979; Sims, explicit attention to Szilagyi, & Keller, 1976; Turner & Lawrence, 1965) but have received how well they little attention in the past three coordinate with decades. Now researchers are revisiting relational approaches to job other jobs. design, focusing on the opportunities they create for social interaction (Grant et al., 2007; Morgeson & Humphrey, 2006), social support, and friendship (Humphrey et al., 2007), as well as their ability to respond to task interdependence (Humphrey et al., 2007; Kiggundu, 1983; Wageman, 1995). Thus far, researchers have linked relational approaches to job design to individual outcomes such as satisfaction, motivation, and individual performance. The contribution of our research is to link relational job design to a relational outcome—coordination—and to improvements in organizational performance. This study also advances our understanding of the process of organizing, given that the central task of organizing is to coordinate the actions of individuals to accomplish collective goals (Weick, 1979) and given that we need more research on coordination to better understand the process of organizing (Heath & Sitkin, 2001). We have learned in this study that design factors can positively or negatively influence the process of organizing, thus taking a step beyond the “emer-

gent” view found in the work of Karl Weick and colleagues (e.g., Weick, 1979, 1995; Weick & Roberts, 1993) to recognize the importance of structure and design. Though we have acknowledged it as a threat to the validity of our research design, the selection-treatment interaction also has important theoretical and practical implications. Selection-treatment interactions are recognized in human resource management theories that discuss the importance of adopting bundles of related human resource practices because of the limited ability of a single practice to bring about desired changes. In particular, high-performance work systems theorists argue that human resource practices work better when they have a similar underlying logic (e.g., Batt, 1999; Ichniowski, Shaw, & Prennushi, 1997; MacDuffie, 1995). That is, if the hospitalist job design is intended to increase relational coordination between physicians and other hospital staff, physicians who are selected for this job design should be selected for characteristics that are consistent with increased relational coordination, such as high levels of relational competence. Although this study contributes primarily to theories of relational coordination and job design, it also has value for health services research. According to recent research on hospitalists, “most evaluations found that patients managed by hospitalists had lower costs or charges than patients in comparison groups and that these savings were achieved primarily by reducing length of stay” (Coffman & Rundall, 2005). Furthermore, most evaluations found no statistically significant differences in quality of care. It is therefore well established that hospitalists offer efficiency advantages with no apparent disadvantages in quality and, according to our results, also had fewer readmissions, a key indicator of quality. But health services researchers have not yet determined the reasons for these better outcomes, according to Coffman and Rundall (2005), who have called for research to better identify the underlying reasons. Our article answers this call by identifying stage- and site-based specialization as a defining characteristic of the hospitalist job design we observed, and by demonstrating that higher levHuman Resource Management DOI: 10.1002/hrm


els of relational coordination account for the performance advantages of this job design. The hospitalist literature has argued that physician availability is the mechanism that accounts for the superior performance outcomes observed for hospitalists. But we would argue instead that physician availability is a key characteristic of stage- and sitebased specialization that contributes to higher levels of relational coordination between the physician and other hospital-based staff. Our argument therefore encompasses some of the arguments that others have made regarding the advantages of hospital specialists, but goes a step further to make a broader contribution to job-design theory. It is important to note that our hypotheses are specifically about jobs that are specialized by stage and site of care. The hospitalist job design has taken different forms in different hospitals, with some hospitalists assigned to float across hospitals, specialized by stage but not by site of care. It follows from our hypotheses that hospitalists who are not site-specific will have lower levels of relational coordination with their hospitalbased colleagues and therefore will produce less efficient, lower-quality outcomes.

Limitations of This Study This study was limited in several ways. First, it was based on a quasi-experimental research design that was vulnerable to four validity threats. We therefore suggest additional research with alternative research designs to further test the effects of job-design innovations on relational coordination and performance. In particular, we recommend a more rigorous quasi-experimental design, a true experimental design with random assignment to job design, or a time-series design that reduces concerns about selection. Alternatively, we recommend the collection of data from multiple sites, which would distance participants in one condition from those in other conditions, minimizing the salience of differential treatment (Wall, Kemp, Jackson, & Clegg, 1986). Of course, a multisite design would present its own challenges, given the large number of factors that can vary across sites. Human Resource Management DOI: 10.1002/hrm

747

A major strength of our study was its ability to measure relational forms of coordination through participant surveys. A limitation, however, was that we did not survey physicians and residents because of the anticipated difficulty of getting them to complete the large number of patient-specific surveys that were required for this study. As a result, we have relational coordination measures from the perspective of other team members, but no reports from physicians or residents. Another strength of this study was its ability to identify two Stage- and siteforms of specialization in addition to functional specialization. A limitation, however, was that we were based specialization not able to distinguish between can help to repair the effects of stage- and site-based specialization in our analyses, these relationships since the new job design observed by building stronger in this study included both. connections among

Implications for Managing Professionals

health care

providers with Our findings have important practical implications for managdifferent areas of ing professionals. Work relationships between physicians, nurses, functional expertise. therapists, social workers, and case managers are often dysfunctional due to differences in professional identity and conflicts over professional autonomy (Abbott, 1988; Barley, 1986; Weinberg, 2003). But our findings suggest that physician job design is another culprit in creating these dysfunctional work relationships. Stage- and site-based specialization can help to repair these relationships by building stronger connections among health care providers with different areas of functional expertise. The same challenges can be found in other professional service settings. Functional specialization leads to fragmentation in the delivery of other professional services, such as education and consulting. For consultant job design, the implication of the current study is for consultants with different areas of functional expertise to be assigned to work together with the same client, focused on a particular phase of the project in which task

748


interdependencies are particularly intense, to develop better-integrated business solutions for that client. Further developing the practical implications for professional job design will require further exploration of other features of job design that affect relational coordination, such as flexible job boundaries and boundary spanners, as well as other human resources practices that can help support relational coordination, such as hiring and performance measurement practices (see Gittell, Seidner, & Wimbush, in press). With these additional steps, we can gain greater insight into how high-performance work systems can be designed to build relational coordination among professional service providers to deliver more efficient, higher-quality outcomes for their clients.

Acknowledgments We thank participants at the MIT Organization Studies Group, the Harvard School of Public Health Quality of Care Seminar, the Organization Studies Workshop at Brandeis University, the Research Seminar at the University of Kentucky’s Martin School, and the Labor Lunch at the Wharton School, as well as Deborah Ancona, Forrest Briscoe, Samer Faraj, Adam Grant, Seok-Woo Kwon, and Anita Tucker. Thanks also to the clinical staff and patients of NewtonWellesley Hospital for their participation in this study. We thank Leslie G. Selbovitz, MD, chief medical officer and senior vice president for medical affairs, and Lawrence S. Friedman, MD, chair, Department of Medicine, for their support of this research.

JODY HOFFER GITTELL is an associate professor of management and MBA program director at Brandeis University’s Heller School for Social Policy and Management. She received her PhD from the Massachusetts Institute of Technology. Her research explores how relational coordination among frontline workers contributes to quality and efficiency outcomes in service settings, with a particular focus on the airline and health care industries. Her forthcoming book, High Performance Healthcare (McGraw-Hill, 2009), provides a comprehensive look at how health care organizations can foster relational coordination among frontline care providers, following up on her earlier book The Southwest Airlines Way: Using the Power of Relationships to Achieve High Performance (McGraw-Hill, 2003). DANA BETH WEINBERG, PhD, is an assistant professor of sociology at Queens CollegeCity University of New York (CUNY). She received her doctorate in sociology from Harvard University. She is the author of Code Green: Money-Driven Hospitals and the Dismantling of Nursing. Her research has focused on issues related to professions and occupations; the interaction of organizational factors that shape experiences and performance of frontline workers; patients’ experiences of care; and coordination of care both within and across settings. As a faculty fellow at CUNY for 2008–2009, she serves as director of interdisciplinary research in the Office of the University Dean for Health and Human Services at CUNY. ADRIENNE L. BENNETT, MD, PhD, is the director of the Division of Hospital Medicine and associate professor of clinical medicine in the Department of Internal Medicine at the Ohio State University Medical Center (OSUMC). She is co-editor of the textbook Just The Facts Hospital Medicine. A charter member of the National Association of Inpatient Physicians (now Society of Hospital Medicine [SHM]), she is a member of the SHM Career Satisfaction Task Force and a coauthor of the SHM Career Satisfaction White Paper. Her current clinical and research interests include quality improvement/patient safety (she is the physician leader for the OSUMC goal to improve deep venous thrombosis risk assessment and prevention) and teaching systems-based medicine.


JOSEPH A. MILLER is senior vice president for the Society of Hospital Medicine. In this role, he oversees membership/marketing, data management/Web services, finance/accounting, and business operations. In addition, he coordinates SHM’s biannual survey, staffs the Benchmarks Committee and Career Satisfaction Task Force, and is course director for SHM’s Practice Management Pre-Course. Miller has more than 30 years of management, consulting, and market research experience with provider organizations, health plans, academia, and vendor organizations. In 2007, he coauthored Hospitalists: A Guide to Building and Sustaining a Successful Program, published by SHM and the American College of Healthcare Executives.

REFERENCES Abbott, A. (1988). The system of professions: An essay on the expert division of labor. Chicago: University of Chicago Press. Amabile, T. M., & Conti, R. (1999). Changes in the work environment for creativity during downsizing. Academy of Management Journal, 42, 630–640. Ambrose, M. L., & Kulik, C. T. (1999). Old friends, new faces: Motivation in the 1990s. Journal of Management, 25, 231–292. Ancona, D. G., & Chong, C. (1996). Entrainment: Pace, cycle and rhythm in organizational behavior. In B. Staw & L. L. Cummings (Eds.), Research in organizational behavior (Vol. 18) (pp. 251–284). New York: JAI Press. Audet, A., Doty, M., Shamasdin, J., & Schoenbaum, S. (2005). Physicians’ views on quality of care: Findings from the Commonwealth Fund National Survey of Physicians and Quality of Care. New York: Commonwealth Fund. Barley, S. (1986). Technology as occasion for structuring: Evidence from observations of CT scanners and the social order of radiology departments. Administrative Science Quarterly, 31, 78–108. Baron, R., & Kenney, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Bartel, C. A., & Saavedra, R. (2000). The collective construction of work moods. Administrative Science Quarterly, 45, 197–231. Batt, R. (1999). Work design, technology and performance in customer service and sales. Industrial and Labor Relations Review, 52, 539–564. Brown, J., & Duguid, P. (1991). Organizational learning and communities-of-practice: Toward a unified Human Resource Management DOI: 10.1002/hrm

view of working, learning, and innovation. Organization Science, 2, 40–57. Campbell, D. T., & Stanley, J. L. (1966). Experimental and quasi-experimental designs for research. Chicago: Rand McNally. Cleary, P. (2003). A hospitalization from hell: A patient’s perspective on quality. Annals of Internal Medicine, 138(1), 33–39. Coffman, J., & Rundall, T. G. (2005). The impact of hospitalists on the cost and quality of inpatient care in the United States: A research synthesis. Medical Care Research and Review, 62, 379–406. Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design and analysis issues for field settings. Chicago: Rand McNally. Cook, T. D., & Shadish, W. R. (1994). Social experiments: Some developments over the past fifteen years. Annual Review of Psychology, 45, 545–580. Crowston, K., & Kammerer, E. E. (1998). Coordination and collective mind in software requirements development. IBM Systems Journal, 37(2), 227–245. Dougherty, D. (1992). Interpretive barriers to successful product innovation in large firms. Organization Science, 3, 179–202. Edmondson, A. (1996). Learning from mistakes is easier said than done: Group and organizational influences on the detection and correction of human error. Journal of Applied Behavioral Science, 32(4), 5–28. Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44, 350–383. Edmondson, A., Bohmer, R., & Pisano, G. (2001). Disrupted routines: Effects of team learning on new technology adaptation. Administrative Science Quarterly, 46, 685–716. Faraj, S., & Sproull, L. (2000). Coordinating expertise in software development teams. Management Science, 46, 1554–1568.

749

750

HUMAN RESOURCE MANAGEMENT, Winter 2008 Faraj, S., & Xiao, Y. (2006). Coordination in fast response organizations. Management Science, 52, 1155–1169. Galbraith, J. R. (1977). Organization design. Reading, MA: Addison-Wesley. Gittell, J. H. (2000). Organizing work to support relational coordination. International Journal of Human Resource Management, 11, 517–539. Gittell, J. H. (2002). Coordinating mechanisms in care provider groups: Relational coordination as a mediator and input uncertainty as a moderator of performance effects. Management Science, 48, 1408–1426. Gittell, J. H. (2003). A theory of relational coordination. In K. S. Cameron, J. E. Dutton, & R. E. Quinn (Eds.), Positive organizational scholarship: Foundations of a new discipline (pp. 279–295). San Francisco, CA: Berrett-Koehler. Gittell, J. H., Fairfield, K., Bierbaum, B., Head, W., Jackson, R., Kelly, M., et al. (2000). Impact of relational coordination on quality of care, post-operative pain and functioning, and the length of stay: A nine-hospital study of surgical patients. Medical Care, 38, 807–819. Gittell, J. H., Seidner, R., & Wimbush, J. (in press). A relational model of how high performance work systems work. Organization Science. Gittell, J. H., Weinberg, D .B., Pfefferle, S., & Bishop, C. (2008). Impact of relational coordination on job satisfaction and quality outcomes: A study of nursing homes. Human Resource Management Journal, 18, 154–170. Gittell, J. H., & Weiss, L. (2004). Coordination networks within and across organizations: A multilevel framework. Journal of Management Studies, 41, 127–153. Gouldner, A. W. (1954). Patterns of industrial bureaucracy. Glencoe, IL: Free Press. Grant, A. M. (2007). Relational job design and the motivation to make a prosocial difference. Academy of Management Review, 32, 393–417. Grant, A. M., Campbell, E. M., Chen, G., Cottone, K., Lapedis, D., & Lee, K. (2007). Impact and the art of motivation maintenance: The effects of contact with beneficiaries on persistence behavior. Organizational Behavior & Human Decision Processes, 103(1), 53–67. Hackman, J. R. (1982). A set of methods for research on work teams (Technical Report #1). New Haven, CT: School of Organization and Management, Yale University. Hackman, J. R., & Lawler, E. E. (1971). Employee reac-

tions to job characteristics. Journal of Applied Psychology, 55, 259–286. Hackman, J. R., & Oldham, G. (1980). Work redesign. New York: Addison-Wesley. Heath, C., & Sitkin, S. (2001). Big-B versus Big-O: What is organizational about organizational behavior? Journal of Organizational Behavior, 22(1), 43–58. Humphrey, S. E., Nahrgang, J. D., & Morgeson, F. P. (2007). Integrating motivational, social, and contextual work design features: A meta-analytic summary and theoretical extension of the work design literature. Journal of Applied Psychology, 92, 1332–1356. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The effects of human resource practices on manufacturing performance: A study of steel finishing lines. American Economic Review, 87, 291–313. Institute of Medicine. (2003). Crossing the quality chasm: A new health system for the 21st century. Washington, DC: National Academies Press. Karasek, R. A. (1979). Job demands, job decision latitude, and mental strain: Implications for job redesign. Administrative Science Quarterly, 24, 285–307. Karau, S. J., & Kelly, K. (1992). The effect of time scarcity and time abundance on group performance quality and interaction process. Journal of Experimental and Social Psychology, 28, 542–571. Kenagy, J., Berwick, D., & Shore, M. (1999). Service quality in health care. Journal of the American Medical Association, 28(1), 661–665. Kiggundu, M. N. (1983). Task interdependence and job design: Test of a theory. Organizational Behavior and Human Performance, 31(2), 145–172. Likert, R. (1961). New patterns of management. New York: McGraw-Hill. MacDuffie, J. P. (1995). Human resource bundles and manufacturing performance: Organizational logic and flexible production systems in the world auto industry. Industrial and Labor Relations Review, 48, 197–221. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83–104. McGregor, D. (1960). The human side of enterprise. New York: McGraw-Hill. Merton, R. K. (1940). Bureaucratic structure and personality. Social Forces, 18, 560–568. Miller, E. J. (1959). Technology, territory and time: The Human Resource Management DOI: 10.1002/hrm

Is the Doctor In? A Relational Approach to Job Design and the Coordination of Work internal differentiation of complex production systems. Human Relations, 12, 243–272. Moreland, R. L., Argote, L., & Krishnan, R. (1998). Training people to work in groups. In R. S. Tindale & L. Heath (Eds.), Theory and research on small groups: Social psychological applications to social issues (pp. 37–60). New York: Plenum Press. Morgeson, F. P., & Humphrey, S. E. (2006). The work design questionnaire (WDQ): Developing and validating a comprehensive measure for assessing job design and the nature of work. Journal of Applied Psychology, 91, 1321–1339. Morgeson, F. P., & Campion, M. A. (2002). Minimizing tradeoffs when redesigning work: Evidence from a longitudinal quasi-experiment. Personnel Psychology, 55, 589–612. Okhuysen, G. & Eisenhardt, K. (2002). Integrating knowledge in groups: How formal interventions enable flexibility. Organization Science, 13, 370–386. Raudenbush, S. W., & Bryk, A. S. (1992). Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage. Selznick, P. (1949). TVA and the grassroots: A study in the sociology of formal organization. Berkeley, CA: University of California Press. Sims, H. P., Szilagyi, A. D., & Keller, R. T. (1976). The measurement of job characteristics. Academy of Management Journal, 19, 195–212. Smith, A. (1991). The wealth of nations. New York: Knopf. Original work published 1776. Taylor, F. W. (1911). The principles of scientific management. New York: Harper and Row.

Van Den Bulte, C., & Moenaert, R. (1998). The effects of R&D team co-location on communication patterns among R&D, marketing and manufacturing. Management Science, 44(11), 1–18. Van Maanen, J., & Barley, S. R. (1984). Occupational communities: Culture and control in organizations. In B. M. Staw & L. L. Cummings (Eds.), Research in organizational behavior (Vol. 6, pp. 287–365). Greenwich, CT: JAI Press. Wachter, R. M., & Goldman, L. (1996). The emerging role of “hospitalists” in the American health care system. New England Journal of Medicine, 33, 514–517. Wageman, R. (1995). Interdependence and group effectiveness. Administrative Science Quarterly, 40, 145–181. Wall, T. D., Kemp, N. J., Jackson, P. R., & Clegg, C. W. (1986). Outcomes of autonomous workgroups: A long-term field experiment. Academy of Management Journal, 29, 280–304. Weick, K. E. (1979). The social psychology of organizing. Reading, MA: Addison-Wesley. Weick, K. E. (1995). Sensemaking in organizations. Thousand Oaks, CA: Sage. Weick, K. E., & Roberts, K. (1993). Collective mind in organizations: Heedful interrelating on flight decks. Administrative Science Quarterly, 38, 357–381. Weinberg, D. B. (2003). Code green: Money driven hospitals and the dismantling of nursing. Ithaca, NY: Cornell University Press.

Turner, A. N., & Lawrence, P. R. (1965). Industrial jobs and the worker. Boston: Harvard University Press.

Weinberg, D. B., Lusenhop, W., Gittell, J. H., & Kautz, C. (2007). Coordination between formal providers and informal caregivers. Health Care Management Review, 32(2), 140–150.

Tushman, M., & Nadler, D. (1978). Information processing as an integrating concept in organizational design. Academy of Management Review, 3, 613–624.

Wrzesniewski, A., & Dutton, J. E. (2001). Crafting a job: Revisioning employees as active crafters of their work. Academy of Management Review, 26, 179–200.


751

752


APPENDIX

Relational Coordination Survey

Newton-Wellesley Hospital An Evaluation of Patient Care Coordination DEAR ______________________________________________________________ This survey asks about your experience with the other care providers involved in the care of PATIENT NAME ______________________ MEDICAL RECORD# ___________. Our records indicate this patient was admitted for ________________________________________________ on _____/_____/_____, and discharged on ______/_______/______. Please answer all questions as they relate to your experience with the patient’s care. Care providers included: ATTENDING PHYSICIAN: [MD____] ________________________________________________________ NURSE [RN____] _________________________________________________________________________ NURSE [RN____] _________________________________________________________________________ CASE MANAGER [CM____] _______________________________________________________________ CASE MANAGER [CM____] _______________________________________________________________ SOCIAL WORKER [SW____] _______________________________________________________________ SOCIAL WORKER [SW____] _______________________________________________________________ RESIDENT [RE____] _____________________________________________________________________ INTERN [IN____] _________________________________________________________________________ PHYSICAL THERAPIST [PT____] ___________________________________________________________ OCCUPATIONAL THERAPIST [OT____] _____________________________________________________ SPEECH LANGUAGE PATHOLOGIST [ST____] _______________________________________________ RESPIRATORY THERAPIST [RT____] _______________________________________________________

Please select only one response for each question in the survey. If you have any questions or concerns, please contact …. THANK YOU FOR TAKING THE TIME TO COMPLETE THIS SURVEY! PLEASE DROP COMPLETED SURVEY INTO THE BOX LABELLED “CARE COORDINATION STUDY.” AS A TOKEN OF OUR APPRECIATION FOR TAKING THE TIME TO COMPLETE THIS SURVEY, WE WILL SEND YOU $5 FOR EACH SURVEY YOU COMPLETE AT THE CONCLUSION OF THE SURVEY PERIOD.



APPENDIX

753

Relational Coordination Survey (contd.)

PROVIDER ID ____________

PATIENT NAME ____________________________

MR# ____________

Consent: I understand that I am being asked to participate in a research study that is evaluating how a team of providers—physicians, nurses, case managers, and therapists—works together to deliver inpatient care. I understand that I will receive $5 for each survey I complete, and that the information gained will help NWH understand how the quality of care can be improved for patients, and how the experience of providing care can be improved for providers. I understand that the information gained in the survey is anonymous and confidential—staff members at NWH and the researchers will not see my individual survey responses. The responses will be analyzed as part of the research and may be published, but at no time will individual respondents’ names be identified. I understand that my participation in the research survey is voluntary. I may refuse to participate or to answer specific questions on the survey. My refusal will not affect my relationship with NWH in any way. By completing and returning this survey I am consenting to participate in this research study. 1. How frequently did you communicate with each of these care providers about this patient? Attending Physician

Never

Rarely

Occasionally

Often

Constantly

Case Manager(s)/Social Worker(s)

Never

Rarely

Occasionally

Often

Constantly

Floor Nurse(s)

Never

Rarely

Occasionally

Often

Constantly

House Staff (Resident and/or Intern)

Never

Rarely

Occasionally

Often

Constantly

Therapist(s) (PT, OT, SLP, RT)

Never

Rarely

Occasionally

Often

Constantly

2. Did these care providers communicate with you in a timely way about this patient? Attending Physician

Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always

Floor Nurse(s)

Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always


754


APPENDIX



PATIENT NAME ____________________________

MR# ____________

3. Did these care providers communicate with you accurately about this patient? Attending Physician

Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always

Floor Nurse(s)

Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always

4. When issues arose regarding the care of this patient did these care providers work with you to solve the problem? Attending Physician

Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always

Floor Nurse(s)

Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always


Never

Rarely

Occasionally

Often

Always

5. How much did these care providers know about your role in caring for this patient? Attending Physician

Nothing

Little

Some

A lot

Everything


Nothing

Little

Some

A lot

Everything

Floor Nurse(s)

Nothing

Little

Some

A lot

Everything


Nothing

Little

Some

A lot

Everything


Nothing

Little

Some

A lot

Everything



APPENDIX

755



PATIENT NAME ____________________________

MR# ____________

6. How much did these care providers respect your role in caring for this patient? Attending Physician

Not at all

A little

Somewhat

A lot

Completely


Not at all

A little

Somewhat

A lot

Completely

Floor Nurse(s)

Not at all

A little

Somewhat

A lot

Completely


Not at all

A little

Somewhat

A lot

Completely


Not at all

A little

Somewhat

A lot

Completely

7. How much did these care providers share your goals for the care of this patient? Attending Physician

Not at all

A little

Somewhat

A lot

Completely


Not at all

A little

Somewhat

A lot

Completely

Floor Nurse(s)

Not at all

A little

Somewhat

A lot

Completely


Not at all

A little

Somewhat

A lot

Completely


Not at all

A little

Somewhat

A lot

Completely


a relational approach to job design and the coordination ... - CiteSeerX

a relational approach to job design and the coordination ... - CiteSeerX

Suggest Documents

Relational Coordination - CiteSeerX

A Relational Approach to Defining Transformations in a ... - CiteSeerX

A Relational Approach to Defining Transformations in a ... - CiteSeerX

Impact of relational coordination on job satisfaction ... - Semantic Scholar

85 A Bottom-Up Approach to Design Coordination Michael Cumming ...

A Magic Approach to Optimizing Incremental Relational ... - CiteSeerX

A Relational Approach to Monitoring Complex Systems

A relational approach to direct mail consumption

A Multistrategy Approach to Relational Knowledge ...

A Fuzzy Relational Approach to Event

A Relational Approach to Bidirectional Transformation

Coordination and Tailorability Issues in the design of a ... - CiteSeerX

Relational coordination among nurses and other providers - CiteSeerX

Relational Modes between Industrial Design and ... - CiteSeerX

A Sea of Trouble: A Relational Approach to the ...

Heterogeneous component coordination: the CLF approach - CiteSeerX

Situational awareness, relational coordination and integrated care ...

The importance of relational coordination for ...

A Relational Approach to the Design of an Arabic Lexical ...www.researchgate.net › publication › fulltext › A-Relatio

The Transactional and the Relational Approach to Contracting Out for ...

Beyond the Bake Sale: A Community- Based Relational Approach to ...

A relational approach to supervision: Addressing ruptures in the alliance

A Systemic and Relational Approach to Pricing Services

A Relational Approach to Materiality and Organizing - Audencia - Hal