This cross-sectional study was conducted in one 700- bed general hospital and one 600-bed university

hospital in the Dutch speaking part of Belgium (Flanders), as well as in one hospital group comprised of
six hospitals (number of beds ranged from 125 to 320) in the French speaking part of Belgium
(Wallonia). The studied participants were registered nurses working in a direct-care nursing unit:
medical and surgical units, intensive care and medium care units, emergency room (ER), operation
theatre (OR) and post anaesthetic care units (PACU). Nurses working in both adult and paediatric care
settings were included. The total study sample was 1201 nurses (response rate 56.5%) of 116 units. We
selected 96 units for further analyses with 30% unit response rate (range from 30% to 100%; 437
respondents per unit). This resulted in a study sample of 1108 nurses. Twenty nursing units were
dropped from analysis because of a low unit response rate (<30%). Previous study results including units
with low response rates showed larger effect sizes and wider condence bounds (Van Bogaert et al.,
2010, 2013c). Members of nursing units were invited by a coordinator/contact person at each institution
to voluntarily complete the questionnaires between June 2011 and June 2012. The survey was offered in
one hospital and the hospital group on paper, and was offered electronically in one hospital.
Respondents could complete the questionnaire at home and/or in their hospital .

Conventional regression analyses ignore the correlated structure of the observations on clustered data
because they underestimate standard errors and increase the likelihood of a false rejection of the null
hypothesis or acceptance of a relationship when in fact it does not exist (Type I error). Meanwhile, a
two-level model incorporating a nested structure of staff members with nursing units corrects for the
dependency of observations. Therefore, the effects of the independent variables on the dependent
variables were tested with two-level linear mixed effects models with a random intercept. Level One
involved variables related to the staff members on a given nursing unit, and Level Two involved variables
related to the nursing unit (Fitzmaurice et al., 2004; Van Bogaert et al., 2010, 2013a). Generalized linear
mixed effects models were tted analysing discrete dependent variables (simple multilevel models). To
determine optimal predictive models, the nal models were assessed with backward procedures
dropping variables that did not improve goodness of t (multiple multilevel models). Coefcients for all
the independent measures were estimated in both unadjusted models as well as models adjusted for
several nurse characteristics that had signicant associations with at least one of the dependent
variables. This was done in an attempt to adjust for potential confounders at the individual level such as
age, years in nursing, years on the present unit, gender, education, and work schedule as previously
applied in prior studies (Van Bogaert et al., 2010, 2013b). In addition, adjustments were made for
response rates at the unit level as well as four types of units ((1) medicalsurgical units, n = 51; (2) ICU
medium care units, n = 19; (3) OR and PACU, n = 9; and (4) ER, n = 9) and. The Statistical Package for the
Social Sciences (SPSS Inc, Chicago) version 20.0 software was used for descriptive analysis. PROC MIXED
and PROC NLMIXED under SAS 9.2 (SAS Institute Inc, Cary, NC) were used to t the multilevel models



Overview
This prospective observational study tested the hypotheses that selected characteristics of
rehabilitation tean lunctioning predict stroke patient functional impmvenwnt. discharge desl iinatio.?
and LOS. Rehabilitation team members at 50 veterans Administration (VA) hospitals were surveyed on
10 measures or tean functioning and on 2 neasures or hospital influences. Results on the relationship of
hospital culture to tean functiioinn? are reported elsewhere. Data on primary outcomes of change in
motor functimn, discharge disposition, and LOs of stmoke parients treated by the sane teams were
abstracted from a national dutabase. Team level scores were computed and used as independemt
variables in regression equations for each of the patient outcomes, controlling for patient case-ins
seventy and demographic characteristics.

Team Functkrning and Hospital Influences lndependent Variables) nhe independent sanahks
were created in reference to a concepuual model of rehabilitation treatnent effeiveness"I as a function
of team relations (4 scales) and team actions (6 scales ). Team relations exist on a spectrum anchored m
the domains of social clinate (subscales of tean order and organ nizaiion. task oentation. and
innovation) and interpmfessonal rebrions suhscak of interprofessional relationsy. Liuewise. team actions
exist on a spectrum anchored in the donains of managerial practices (subscales of conimunication.
teamness, effectneness, and utility or quality information) and leadership (physician involvenent and
physician support). nhe context of tenm functioning was assessed on 2 scales or hospital influe
eniessaamminitrratvee support and supervisor expectations. The specifics or 1I or the 12 individual
scales have been jd21111 The utility of quality infonnation was developed concurmently but as not
reported with tle preseously des ccribe? scales. nhis scale measures staff percepuion or the degree to
which quality nanagement activities within the team contribute to quality of care (coefficient cze.69).
For example. team members were asked the degree to which impmved access to patienr functional care
ontconc data-. . suld imp rroee quality of care. Copies of the survey are available from the primury
author.

Our regression results showed that several process and outcome indicators such as access to
after-hours care, quality of care, continuity of care, confidence in the syst eem, utilization of physician
and nurse services, patient centeredness, comprehensiveness of care. and disease prevention and
promotional initiatives were positively associated with team-based practice. However, no signifi icant
associations were found for access difficulry. two dimensions of follow-up coordination and overall
coordin nation. Since these regression results cannot be interpreted in a causal way, we have employed
the propensity score matching approach and conducted sensitivity analyses. The estimated average
treatment effect of team-based primary care were positively significant for access to atter-hours care,
quality of care, confidence in the syst eem? overall coordination of care and patient centeredness. and
these results are reliable based on sensitivity analy ysss. Although the estimated average treatment
effect of team-based care for access difficulry, rwo dimensions of follow-up coordination, continuity of
care. health prom moioon and disease prevention initiatives, utilization of physician and nurse services
were statistically significant. these results are sensitive to the presence of unobserved confounders and
hence are unreliable. These findings demonstrate that neither the regression technique nor the
propensity score matching technique can provide robust results for policy without analyzing the
sensitivity of these results with respect rn the presence of unobserved selection.

With widespread changes affecting all business sectors, organizations were focusing greater
attention and resources on improving the quality of their products and services as part of a strategy to
improve their competitive position (Kueng, 2000). Among the most widely adopted approaches for
improving quality were total quality management (TQM) and continuous quality improvement (CQI),
but research suggested that quality improvement (QI) implementation is a gradual process (Handfield &
Ghosh, 1994; Perrott, 2002). Organizations therefore will differ in the extent of their QI
implementation. Researchers exploring QI implementation, the successes and failures, had identified
the important role organizational culture plays (Detert et al., 2000; Huq & Martin, 2000, Krishnan et al.,
1993; Mohanty & Yadov, 1996). Their findings indicated that the culture of an organization can either
support or inhibit change (Ferlie & Shortell, 2001; Putz, 1991) and cultures that support QI
implementation were those that were innovative, risk-taking, empowering and participative
(Hildebrandt, 1991; Huq & Martin; Shortell et al, 1995). These findings suggested that organizations
with a QI-supportive culture would have greater QI activity.

Researchers (McCamus, 1990; Reger et al., 1994) had found a profound connection between
employee satisfaction and improved organizational performance. One organizational performance
measure in health care is patient outcomes. Therefore, organizational implementation of QI may have
been associated with better patient outcomes. Quality improvement had been the subject of study by
other researchers so that a body of literature existed with known variables and existing theories on
which the current study could build. To advance that existing knowledge base, the researcher employed
quantitative research methods following a methodology comparable to that used by Berlowitz et al.
(2003) in their examination of QI implementation in Veterans Administration (VA) nursing homes.
Berlowitz and his colleagues surveyed a sample of employees from each of a sample of 35 VA nursing
homes to obtain information on organizational culture, QI implementation and employee satisfaction.
Quality of care was calculated from an administrative database and measured by the risk-adjusted rate
of development of pressure ulcers on the skin of bed-bound nursing home patients.



As noted, research has shown the link between job satisfaction and job performance to be
weak. However, researchers have demonstrated a much stronger relationship between job satisfaction
and firm performance. In a conceptual discussion of why loyal employees and customers improve the
bottom line (financial performance), Brooks (2000) noted that conditions that support high internal
customer service predict high customer satisfaction. Internal customer service refers to how employees
are treated in the organization and how they treat each other in work-related situations. Those
employees who feel they are treated well and receive high levels of customer service from other
personnel and departments are less likely to leave the organization. Consequently, companies with low
employee turnover and high customer loyalty have larger profits. The conclusion suggests, then, that
there are direct, quantifiable links between employee (and customer) variables and financial results.
The empirical literature supporting the relationship between employee satisfaction and firm
performance is extensive. For example, in a study of 298 schools encompassing in excess of 13,000
teachers, Ostroff (1992) found that employee satisfaction and attitudes were correlated with
organizational performance (measured by variables specific to schools such as academic achievement
and student behaviors). For instance, commitment to the organization showed a significant correlation
with satisfaction

In contrast to other studies that indicate a correlation between employee attitudes (used
interchangeably with job satisfaction) and organizational performance, Schneider, Hanges, Smith and
Salvaggio (2003) found an inverse relationship. In a study of the relationship between employee job
satisfaction and performance, they found that overall job satisfaction and satisfaction with job security
were predicted by Return on Assets and Earnings per Share. In other words, higher return on assets and
higher earnings per share were positively correlated with higher job satisfaction. This included three
specific attitude characteristics, Satisfaction with Pay, Satisfaction with Security, and Overall Job
Satisfaction. Porter (1985) identified employee productivity as the most important component in
measuring a companys efficiency. He suggested that if employee productivity is high, the company will
have low production costs. However, in Silvestros evaluation of this relationship between productivity
and job satisfaction, results supported Schneider et als. (2003) findings. There was an inverse
correlation between employee satisfaction and productivity, but as Porter (1985) noted, a direct
relationship between productivity and profits. Higher productivity levels resulted in higher profits but
lower job satisfaction.