developing a workplace resilience instrument
TRANSCRIPT
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/303798137
Developing a workplace resilience instrument
Article in Work · May 2016
DOI: 10.3233/WOR-162297
CITATIONS
23READS
4,545
2 authors:
Some of the authors of this publication are also working on these related projects:
"The Four Factors of Workplace Resilience" View project
Essentials of Gathtering Insight View project
Larry Mallak
Western Michigan University
35 PUBLICATIONS 1,467 CITATIONS
SEE PROFILE
Mustafa Yıldız
Amasya University
3 PUBLICATIONS 25 CITATIONS
SEE PROFILE
All content following this page was uploaded by Larry Mallak on 21 July 2016.
The user has requested enhancement of the downloaded file.
Developing a Workplace Resilience Instrument
Larry A. Mallak, Ph.D.
Professor, Department of Industrial and Entrepreneurial Engineering & Engineering Management
Mustafa Yildiz
Doctoral student, Department of Educational Leadership, Research, and Technology Western Michigan University Kalamazoo, Michigan, USA 49008
Mallak, L.A. & Yildiz, M. (2016). The Development of a Workplace Resilience Instrument. WORK: A Journal of Prevention, Assessment and Rehabilitation, 54(2), 241-253.
1
Developing a Workplace Resilience Instrument
Abstract
BACKGROUND: Resilience benefits from the use of protective factors, as opposed to
risk factors, which are associated with vulnerability. Considerable research and
instrument development has been conducted in clinical settings for patients. The need
existed for an instrument to be developed in a workplace setting to measure resilience of
employees.
OBJECTIVE: This study developed and tested a resilience instrument for employees in
the workplace.
PARTICIPANTS AND METHODS: The research instrument was distributed to
executives and nurses working in the United States in hospital settings. Five-hundred-
forty completed and usable responses were obtained. The instrument contained an
inventory of workplace resilience, a job stress questionnaire, and relevant
demographics. The resilience items were written based on previous work by the lead
author and inspired by Weick’s [1] sense-making theory.
RESULTS: A four-factor model yielded an instrument having psychometric properties
showing good model fit. Twenty items were retained for the resulting Workplace
Resilience Instrument (WRI). Parallel analysis was conducted with successive iterations
of exploratory and confirmatory factor analyses. Respondents were classified based on
their employment with either a rural or an urban hospital. Executives had significantly
higher WRI scores than nurses, controlling for gender. WRI scores were positively and
significantly correlated with years of experience and the Brief Job Stress Questionnaire.
2
CONCLUSIONS: An instrument to measure individual resilience in the workplace
(WRI) was developed. The WRI’s four factors identify dimensions of workplace
resilience for use in subsequent investigations: Active Problem-Solving, Team Efficacy,
Confident Sense-Making, and Bricolage.
Keywords:, , nursing, , sense-making, hospital, bricolage, coping
1. Introduction
The primary objective of this study was to build and test a resilience instrument
for use in the workplace. Most studies have focused on patients in clinical settings rather
than employees in the workplace. Our study focused on employees and how we can
characterize the level of resilience in how they approach work. This study used the
resilience scales developed by Mallak [2] as the basis for such a tool, using the relevant
research and application of resilience tools developed to date.
With resilient individuals able to withstand stress better than others, the ability
to reduce stress through the ability to measure and improve resilience has enormous
consequences. Stress costs U.S. businesses an estimated $300 billion annually [3].
With resilient individuals able to withstand stress better than others [4], the
ability to reduce stress through the ability to measure and improve resilience has
enormous consequences. Estimates are that 67-90% of all office visits to a physician can
be traced to stress-related symptoms [5] [6]. Stress creates adverse effects on 43% of all
adults [6]. Stress is a major contributor to heart disease, cancer, stomach problems, lung
problems, accidents, cirrhosis of the liver, and suicide; the common cold and skin rashes
3
can often be traced back to stress conditions [5] [6]. There is much to be gained from an
instrument that can effectively measure individual resilience in the workplace and lead
to interventions to increase resilience.
Resilience is a key construct in the performance of targeted behaviors for solving
problems and taking action in the face of adversity. The increasing need for quicker
decision making in complex systems having severe consequences requires individuals
and organizations to have the capacity to make high quality decisions and take effective
actions. In the U.S., the recent increase in the frequency of costly natural disasters and
continued vigilant action to thwart terrorist actions represent high-profile situations
benefiting from resilient behavior.
Resilience research, especially the measurement of resilience dimensions, is
found predominantly in the psychological, medical, and nursing professions and their
associated journals. Resilience has been identified as having “enormous utility for
nursing” [7]. Rew and Horner [8] found that resilient individuals have positive
outcomes in the face of adversity. The classic study in resilience is a 32-year effort led by
Werner [9], following 698 children born on Hawaii’s island of Kauai. Her study
participants were exposed to perinatal stress, poverty, and a “family environment
troubled by chronic discord and parental psychopathology” [9] (p. 503). Her work
surfaced the use of adaptive processes to adverse conditions by children facing these
situations and fueled subsequent work by researchers attempting to build valid and
reliable measurement instruments to characterize an individual’s level of resilience.
The resilience instrument literature shows 1) these instruments were developed
in clinical/counseling settings, not workplace settings; 2) the instruments have some
4
common dimensions, but are differentiated; and 3) the need to develop a resilience
instrument for the workplace.
2. Theory & Models
Many models and theories have been advanced in the psychological and clinical
literature. Several of these models and theories are shared here for their relevance to a
study of workplace resilience.
The resilience literature appears to have three developmental phases over the
past 50 years. Although this is not meant to be a comprehensive literature review, the
literature relevant to this study is representative of the material published during each
timeframe. These are the three phases of resilience construct development:
1) Foundation: the 32-year longitudinal study (1955-1987) by Werner [9] set the
stage by discovering protective factors among a group of 698 children born in
Kauai.
2) Conceptualization: the 1990s saw the publication of management books on
resilience (cf. [10] [11], a now-classic analysis of a 1949 disaster [1], and resilience
scales [12] [2]).
3) Measurement: the 2000s produced resilience scales for use primarily in clinical
settings for patients [13] [14] [15] [16].
Although resilience articles have been published since 2004, the most-cited have
focused on the scales shown in Table 1 and discussed later in the literature review for
5
this study. This study extends the resilience work by developing resilience scales for use
by employees in a workplace setting.
Insert Table 1 about here
Resilience is often studied in contrast with vulnerability, with the associated
elements of protective factors (those that promote resilient behaviors) and risk factors
(those that promote vulnerable behaviors) [17]. Risk factors often emanate from one’s
personal life. How one proceeds from the point of being confronted with adverse events
and the associated risk factors defines the extent of the individual’s resilience. The types
of protective factors deployed vary whether the individual is studied in a clinical setting,
a work setting, or a disaster setting.
In general, many of the empirical findings in each of these three settings support
a more inclusive theory of resilience where positive, adaptive behaviors that directly
address the needs of the situation are viewed as protective factors and where negative or
neutral behaviors that are relatively “fixed” or do not directly address the needs of the
situation are viewed as risk factors. Several models embody the resilience-vulnerability
phenomena along with protective factors.
Several research streams have led to the current theoretical foundation for the
study of workplace resilience. These research streams are focused on the processes
deployed in response (or even in advance) of situations requiring resilience. In physical
systems, resilience refers to a material’s ability to store and return elastic energy [18].
Similarly, in the workplace, we seek the ability for an employee to absorb energy from a
stressful situation and to return to their original (or improved) condition once the
stressor is removed. Unlike an inanimate material, a person needs to perform one or
6
more processes to be able to return to their original state—these processes typically take
the form of protective factors [17]. Protective factors exist in contrast with risk factors
[19] which are associated with vulnerability. In Werner’s classic study [17] [9], the risk
factors facing children born in Kauai (Hawaii) included absent or alcoholic parents,
abuse, and teen motherhood, to name a few. In the workplace, protective factors
emanate from the theories of coping [20] and how job stress is handled [21]. Within the
context of resilience, coping and responses to job stress move the person’s psychological
state to a different “place” than that before the adverse situation was encountered. It is
akin to the quote often attributed to Friedrich Nietzsche, “That which does not kill me
makes me stronger.”
Similarly, the construct of stress originates in engineering—stress is defined as
the force per unit area, but can be conceived as “internal forces that neighboring
particles of a continuous material exert on each other” [22]. Transferring this
engineering definition to the individual, stress is indeed an internal phenomenon and,
like engineering materials, it is manifested physically. Stress is often contrasted with
anxiety; anxiety is a cognitive phenomenon of uncertain origin while stress involves
physical symptoms having a known origin (adapted from definitions in [23] [24]).
Several models from the literature illustrate the relationships between risk
factors and protective factors with the construct of resilience. The Adolescent Resilience
Model [25] contains both individual and family components for risk and protective
factors aimed toward the outcomes of increased resilience and quality of life. The Youth
Resilience Model [8] portrays the interaction between risk factors (vulnerability) and
protective resources (protection), while treating family and community as part of the
7
sociocultural context for resilience. Hunter and Chandler’s [26] continuum of resilience
in adolescents has risk factors at one extreme and adaptive behaviors/self-efficacy at its
other extreme. The resilience scales developed by Wagnild and Young [12] were based
on Block and Block’s [27] ego-resilience (a high level of resilience) and ego-brittleness
(vulnerability) and on Rutter’s [28] “buffering effect.”
Resilience models and theories also recognized the interaction among body,
mind, and spirit in producing effective behaviors and outcomes. These appear as
“biopsychosocial factors” [29], “biopsychospiritual homeostasis” [13] and
“psychoneuroimmunology” [30]. Tusaie and Dyer’s [30] model has its roots in the
psychological aspects of coping and in the physiological aspects of stress. Their model
shows the 1990s emergence of the resilience construct as a direct successor to the
concept of psychoneuroimmunology. Their model shows protective factors on the
psychological side and homeostasis on the physiological side. Carver et al.’s [20] study
resulted in an instrument to measure coping; its items range from risk factors (e.g.,
turning to substance abuse) to protective factors (e.g., active coping).
A study of military personnel’s response to adverse conditions found that
resilience is based on complex biopsychosocial factors in short- and long-term reactions
to traumatic stress [29] and is not restricted to events such as disaster response or post-
traumatic stress disorder. Using the Mann Gulch smokejumping disaster of 1949 as a
case study, Weick [1] modeled resilience as a function of sense-making, attitude of
wisdom, virtual role system, and Levi-Strauss’ concept of bricolage [31] [32]—or the
ability to use available materials and methods to solve a problem. Weick’s work
provided a basis for designing the original items in Mallak’s [2] instrument. Similarly,
8
Kobasa [33] found that hardy (or resilient) executives exhibited a stronger commitment
to self, an attitude of vigorousness toward their environment, a sense of meaningfulness,
and an internal locus of control. Whereas Kobasa [33] measured commitment, control,
and challenge as the larger factors from which she drew her resilience conclusions,
Bartone et al. [29] studied commitment, control, and change among military personnel
dealing with trauma from a military plane crash involving fatalities.
The dominant resilience scales found in the literature have been developed
primarily with clinical populations, not workplace populations. As such, the validity of
these instruments for use in the workplace is questionable until psychometric properties
can be established with a workplace population. Resilience scales have been developed
based on work with the military [29], elderly women [12], adolescents [15], general
population/psychiatric patients [13], mental health outpatients in Norway/control
group [14], adult rheumatoid patients [16], and nursing staff [2]. Only the instruments
by Mallak [2] and Bartone et al. [29] were developed solely with data from working
adults. With Bartone’s instrument anchored in the military, the need exists for a
resilience instrument tailored to civilian workplaces.
However, the resilience scales developed by these authors provide input into the
redesign of the Mallak [2] scales, called the Workplace Resilience Instrument (WRI).
See Table 1 for an analysis of the resilience instruments’ content.
3. Resilience in the Workplace
The resilience instrument literature shows 1) these instruments were developed
in clinical/counseling settings, not workplace settings; 2) the instruments have some
9
common dimensions, but are differentiated; and 3) the need to develop a resilience
instrument for the workplace.
The most-used resilience instruments were developed in clinical/counseling
settings, not more general work settings. The resilience instruments in use today and
shown in Table 1 were developed primarily in clinical and counseling settings. While
these are helpful for instrument design and to establish psychometric properties, the
generalization to work settings is not clear. A second major difference is that the
resilience scales in Table 1 were designed and used primarily with patients. This study
builds an instrument to measure the resilience of employees in the workplace.
The main resilience scales in use are CD-RISC [13], Resilience Scales [12], and
Dispositional Resilience Scales [29]. There are many articles and books on the concept
of resilience in work settings (e.g., [10] [4] [11]) and an emerging research stream on
workplace resilience [34] [35]. Resilience instruments share common dimensions, but
are differentiated. A review of the psychology, management, medical, and nursing
literature produced research on a small set of resilience scales. As mentioned earlier,
these scales have been used primarily in clinical and counseling settings (Table 1) as
shown in the column labeled “Primary Population(s).” Of the 294 studies listed in the
CD-RISC user manual [36], only 18 (about 6%) were conducted with employees in the
workplace.
Analysis of the resilience instruments shows some common or related
dimensions. Personal competence appears in three of them as the highest-loading
factor, with a related factor of commitment in the Bartone et al. [29] instrument
showing the highest loading.
10
A workplace resilience instrument needs relevant dimensions that have solid
psychometric properties. The instruments shown in Table 1 may be used in workplace
settings, but the generalizability of the instrument items for other than the Bartone et al.
[29] and Mallak [2] scales is an open question. Beyond perception of one’s own self, the
items in these instruments primarily concern interactions and relationships with family
members and friends [13] [14] [12].
4 Methods
4.1 Instrumentation
The instrument package contained the revised resilience scales (25 items), the 16-
item Brief Job Stress Questionnaire [21], demographic items—gender, location of
hospital (urban or rural), respondent age, respondent years of healthcare experience,
and US state where employed. The resilience scales were based on the Mallak [2] scales
and modified to accommodate work settings in various sectors, not just healthcare.
Resilience items were rewritten to focus on the individual’s response concerning
resilience. The original scales had some items focused on the individual and others on
the team in which the individual worked.
During the design phase of this study, a concern regarding the negatively-worded
(reverse-coded) items was raised. Several authors [37] [38] [39] have discovered that
negatively-worded items, when mixed with positively-worded items, create their own
variance due to their negative nature. In order to avoid possible method effects, those
that were negatively-worded were converted to positive statements so that a positive
endorsement corresponded to a high score on that specific item as in the other items.
11
Response scales were changed from an agree-disagree format to an extent-of-
truth format (e.g., “not true at all” to “true all the time”). The Brief Job Stress
Questionnaire (BJSQ) [21] was modified from statements starting with “you” to starting
with “I.” Some wording was modified to work better with the English-speaking
respondent population, as the BJSQ was translated from Japanese into English for
publication. Urban or rural hospital location was included to discover any differences
between those locations. Years of healthcare experience was included to discover the
degree of correlation between experience and the subscales of WRI.
4.2 Samples
The resilience scales were distributed to 3,291 employees in the healthcare sector
in two campaigns. The first campaign targeted hospital officers in the Great Lakes
region of the United States and produced 177 usable responses out of 2601 sent, for a
6.8% response rate. A second campaign targeted hospital-based nursing staff in the
United States and produced 363 usable responses out of 690 sent, for a 52.6% response
rate. In aggregate, 540 usable responses were received for an overall response rate of
16.4%.
Respondents ranged in age from 16-75, with two-thirds of respondents between
the ages of 45-64. Analyses were conducted on all age ranges. Informed consent was
obtained from each of the study participants, and the study protocol was approved by
the university’s Institutional Review Board. Respondents were 84% female. Three-
fourths of the respondents worked in urban hospitals, 99% had some college education,
with a mean 25 years of healthcare experience. See Table 2 for demographics on the
study samples.
12
Insert Table 2 about here
4.3 Data Preparation
Instrument responses were reviewed for their level of completeness and missing
data. Fifty-four observations had some missing responses. Three of those 54
observations had more than 50% or more of responses missing. Those three
observations were deleted entirely. The remaining 51 observations had 3 or fewer
responses missing on the resilience scales which corresponded to less than 1% of the
whole dataset. The demographic information was used to understand the pattern of
missing data. Descriptive statistics and some basic analysis at the individual item level
provided the evidence to decide that the data were missing at random (MAR) [40].
Because the data were missing at random, the missing values were imputed by using
Markov Chain Monte Carlo (MCMC) multiple imputations implemented in Mplus [41].
Five datasets were generated. The one with largest amount of variance was chosen for
the investigation of the psychometric qualities of the WRI, although the differences
among the five datasets were rather small. The other four datasets are available upon
request.
4.4 Analytical Strategy, Estimation, and Fit
The 1998 [2] resilience scale was factor analyzed in the framework of exploratory
factor analysis via a varimax rotation that set the inter-factor correlations to be zero.
With this condition, it is impossible to estimate the model-data fit in the framework of
confirmatory factor analysis due to identification reasons. For example, the factors
Source Reliance (SR), and Role Dependence (RD) were measured by two items for each.
As a result, another condition of the initial model was violated in addition to revising the
13
wording of some items from negative to positive statements, and changing the scale
property of the original scale. These rearrangements actually changed the nature of this
study from a strictly confirmatory approach to an alternative models or model-
generating approach as described in Jöreskog and Sörbom [42].
Maximum likelihood was the first estimation method that was considered
because of the desirable test statistics it provides. However, maximum likelihood
requires that the items are both univariate and multivariate normal. Tests of univariate
and multivariate normality indicated that the data were not normally distributed
(Shapiro-Wilk test for univariate normality ranged from 0.65 to 0.90, p<0.0001;
multivariate normality of Mardia Skewness=5897, p<0.0001; multivariate normality of
Mardia Kurtosis= 29.66, p<0.0001). Therefore, other estimation methods that rely on
normal theory were not considered for further analysis. Considering that the data were
not normal, and the five-point Likert scale as ordinal in nature, weighted least squares
mean and variance (WLSMV) implemented in Mplus [43] was selected for estimating
model data fit.
To evaluate the model-data fit, root mean square error of approximation
(RMSEA) [44] , comparative fit index (CFI) [45], and Tucker-Lewis fit index (TLI) [46]
were used. According to Hu and Bentler [47], model-data fit should be evaluated based
on multiple fit indexes, and acceptable levels of model-data fit are RMSEA≤0.06,
CFI≥0.95, TLI≥0.95.
5 Results
The 1998 [2] resilience model was developed based on the assumption that the
inter-factor correlations were zero. The model-data fit of this simple structure derived
14
from this model was under-identified (three inputs in the variance-covariance matrix, 3
or 4 parameters to be estimated) for some factors such as SR and RD. Therefore, the
goodness-of-fit statistics for this model were not available. Therefore, to get an
approximate estimate of the goodness-of-fit tests in the framework of confirmatory
factor analysis, the assumption of orthogonal solutions was violated and all of the
factors were allowed to correlate, so that the model could be estimated. The test of the
1998 model [2] containing the original 25-item instrument with additional correlated
factors was analyzed in the framework of confirmatory factor analysis and provided the
following fit statistics: RMSEA = 0.092, CFI = 0.918, TLI = 0.905. In addition to poor
model data fit statistics, the two factors, SR and RD, showed Heywood cases [48],
meaning that the standardized item-to-factor correlations were greater than one.
Further investigation of the modification indices and the possible sources of misfit led to
the decision that the factor structure of the resilience scales had to be reinvestigated via
a model building approach by employing both the tools of exploratory and confirmatory
factor analysis.
An exploratory factor analysis conducted on the polychoric correlation matrix
revealed two items with too low communality estimates (square multiple correlation less
than 0.40): I21 and I24 (Table 3). I21 is essentially a reverse-coded item, where a higher
score indicates lower resilience levels. I24 concerns the use of resources, even if
unauthorized to use them, and therefore the responses indicating higher resilience
levels can be confounded with responses indicating conformance to rules. These two
items do not relate well with the other scale items. These two items were flagged, and
further analyses were conducted without them.
15
Insert Table 3 about here
Eigenvalues, an eigenvalue plot, and parallel analysis were used to determine the
number of possible factors. (See Figure 1 for parallel analysis.) These analyses indicated
there could be three or four factors. As a result, an exploratory factor analysis starting
from a single factor model to a six-factor model were investigated. All of these
investigations were conducted in Mplus 6.1 [43]. The estimation method was WLSMV,
with geomin rotation for multi-factor solutions (Table 4). Items I21 and I24 were not
included in these analyses.
Insert Figure 1 about here
Insert Table 4 about here
The purpose of the exploratory factor analysis was to set up a simple structure
where one item is allowed to load on only one factor and error variances of the items are
uncorrelated. A combination of two paths was followed to achieve a simple structure.
One path was to use statistical reasoning which was to allow an item to load only on a
factor on which it had a higher loading. The second path was the theoretical necessity
(content validity) which was always a priority. As methodological decisions were
required, the information considered was statistical and theoretical, with respect to the
design of a valid instrument to measure workplace resilience.
Table 4 clearly shows that the single-factor EFA model did not have desirable fit
statistics. This single-factor EFA was identical to its counterpart CFA solution. The two
factor model’s factor structure was used to achieve a simple structure by using the
16
salient loadings (loading≥0.40) from the EFA two-factor model. The test of this model
in the framework of CFA via WLSMV revealed the following fit statistics: RMSEA=0.13,
CFI=0.83, TLI=0.81. Modification indices were examined. It was found out that the
residuals of item 8 and 9, and items 11 and 12 would improve model-data fit if these two
pairs were allowed to correlate. The addition of these correlations one-at-a-time
improved model-data fit to the following fit statistics: RMSEA=0.09, CFI=0.92,
TLI=0.91. The fit statistics were still below the acceptable range. Then, the examination
of a three-factor model that is based on the EFA (loadings≥ 0.40) yielded the following
fit statistics: RMSEA=0.093, CFI=0.92, TLI=0.91. Modification indices were examined.
It was found out that the addition of correlated error variances of items 8 and 9 would
improve model-data fit. The addition of the correlated residual variances of item 8 and 9
improved the fit statistics to RMSEA=0.085, CFI=0.935, TLI=0.927. Then, a four-factor
model was examined that was also developed based on the EFA solution. This model
was a complex model which means that there were some items that were allowed to load
on multiple factors. This model had the following fit statistics: RMSEA = 0.081, CFI =
0.944, TLI = 0.934.
As the steps were taken in order to arrive to a simple structure, a total of four
loadings were deleted by the use of the previously described paths. As a result, when the
simple structure was achieved, the model had the following fit statistics: RMSEA =
0.084, CFI = 0.938, TLI = 0.929. The examination of the modification indices indicated
that the residuals of item 8 and 9, then items 19 and 20 would improve the fit if they
were allowed to correlate. The model with the correlated errors had the following fit
statistics: RMSEA = 0.077, CFI = 0.948, TLI = 0.941. Because items 8 and 9 were quite
similar in terms of content, item 9 was removed from the instrument due to having a
17
smaller loading than item 8. The same procedure was applied with the items 19 and 20,
and then item 20 was deleted. With this condition, the model-data fit became: RMSEA
= 0.078, CFI = 0.948, TLI = 0.940.
Further analysis showed the deletion of item 18 would improve model fit
statistics. A theoretical review of item 18’s fit with its corresponding factor showed that,
although the item measures an underlying concept of value to resilience, it did not have
a close fit with the other items in the factor. Item 18 was therefore deleted and the
resulting model-data fit became: RMSEA = 0.078, 90% CI = 0.071-0.083, CFI=.951,
TLI=.943 (Figure 2). This final model produced a 20-item instrument called the
Workplace Resilience Instrument (WRI).
Insert Figure 2 about here
The five-factor model was examined in the framework of EFA. In that case, the
fifth factor had an insufficient number of items loading on it. The six-factor solution had
the same condition. Therefore, the further examination of these candidate models was
terminated and the four-factor model was adopted: active problem-solving, team
efficacy, confident sense-making, and bricolage. Table 5 shows the four factors and an
example item from each of those factors.
Insert Table 5 about here
18
6. Discussion
6.1 The Meaning Behind the Analyses
Through confirmatory factor analysis, four factors related to workplace resilience
emerged. These results show that the four factors of workplace resilience, namely, active
problem-solving, team efficacy, confident sense-making, and bricolage, as assessed by
the WRI were applicable to this target population in the workplace. Each of the four
factors showed evidence of internal consistency (alpha: 0.77-0.83; omega: 0.77-0.83).
Essentially, the four factors of the WRI are protective factors, in terms of the resilience
literature. Protective factors work to increase the resilience capacity of an individual and
exist in contrast to risk factors, which work to increase one’s vulnerability.
The inter-factor correlations of the WRI subscales are mostly moderate and
significant at p<.05 (Table 6). This indicates that the subscales are related, but have
sufficient statistical evidence that they are measuring distinct dimensions of workplace
resilience. The correlations between the WRI subscales and the BJSQ subscales provide
evidence of convergent validity, which is expected because of the theoretical
relationships between job stress and resilience.
A counterintuitive finding in the convergent validity testing with job stress was
the positive correlations between WRI factors and the BJSQ factor of job demand. We
expected measures of job stress to be negatively correlated with resilience on the
assumption that a more resilient individual is likely to score lower on job stress. Put
another way, the resilience person’s use of protective factors should indeed protect
him/her from the forces of job stress. However, the significant positive correlation of the
job demand factor with all four WRI factors may provide some insight into how
19
protective factors are deployed effectively. Perhaps a more resilient individual
experiences stress differently than the person with lower resilience. The job demand
items in the BJSQ may represent behaviors that are important to performance at higher
levels of resilience in the workplace. These job demand items concern the focus of
attention, the difficulty of the job, and the inability to complete all of one’s tasks in the
time given. Based on the underlying dimensions being tested, better performance
against these job demands bears a logical relationship with the expectation of higher
workplace resilience.
Insert Table 6 about here
6.2 The Four Factors of WRI
The resulting model had four factors: Active Problem-Solving, Team Efficacy,
Confident Sense-Making, and Bricolage. These four factors track well with existing
resilience and coping research.
Active Problem-Solving. An active approach to problem-solving demonstrates a
need to do something positive, rather than merely talking about the problem or hoping
it will go away. In the workplace, this requires employees to have a bias for action and
the ability to focus on the problem instead of worrying about why things aren’t going
well. This factor corresponds with Carver et al.’s [20] scale on coping. Their highest
scale assignment in the coping instrument was “active coping,” which consisted of items
concerning the taking of action to solve a problem.
20
Team Efficacy. A resilient individual not only “works well in teams,” but has a
systemic understanding of how the team operates and achieves its objectives. Rather
than assume that a fellow team member knows what he or she is supposed to do, the
resilient individual discusses team member roles with other team members. Goals are
made known and shared with everyone on the team and, in turn, guide each team
member’s actions.
Confident Sense-Making. The ability to extract order out of chaos is a mark of the
resilient individual. Making sense of one’s reality requires accessing the right resources
quickly; to do so confidently is a key factor in workplace resilience. Classically, the types
of behavior exhibiting this factor have saved lives and led to long-lasting innovations [1].
More notably, in today’s workplaces, confident sense-making requires the individual to
quickly filter out unnecessary signal and information and to focus on the relevant
stimuli for decision making and action.
Bricolage. This French term, from Levi-Strauss’ The Savage Mind [32], captures
another unique factor of the resilient individual. The bricoleur practices a highly applied
engineering approach, much like the 1980s U.S. television character MacGyver.
Resilience benefits from fashioning solutions creatively to address the situation. When
confronted with chaotic, extreme, and dangerous situations, the resilient individual
takes intelligent risks and realizes there is time to “STOP”—stop, think, observe, and
plan1.
1 As used by the U.S. Army [53] and outdoor survival agencies to communicate the need for being mindful in the face of danger.
21
Tables 7-9 display some of the demographic variables and their relationship to
the WRI subscales. In terms of gender, there are statistically significant differences
between males (range: 0.15 to 0.20) and females (range: -0.03 to -0.04) with respect to
each of the WRI’s four factors (Table 7). This finding shows males scoring higher as a
group than females on the resilience factors. In comparison, Connor and Davidson’s [13]
use of the CD-RISC resilience instrument showed no significant differences by gender,
race, or age. When looking at factor scores across the two samples, executives scored
higher on all four resilience factors compared with nurses (Table 8). Although the
executive sample had a higher percentage of males than the nursing sample (39% vs.
5%), the executive sample was still predominantly female, suggesting that the executive
position was the dominant factor in this comparison. When males were removed from
the two samples, the same relationships held—the executive sample scored significantly
higher than the nursing sample on all four factors. Finally, we tested for differences
between those in urban versus rural hospital locations and found no statistical
differences (Table 9).
Insert Tables 7, 8, & 9 about here
Analyses conducted by age range showed only minor differences on Factor 2:
Team Efficacy and Factor 3: Confident Sense-Making. For both factors, respondents
aged 65-74 scored significantly higher than those aged 25-34. Years of healthcare
experience was positively correlated with each of the four WRI factors.
6.3 Generalizability and Limitations of the WRI
The 1998 model [2] was a good model for its time, but resilience theory and
psychometric methods have evolved since then. This has allowed for the design for an
22
instrument that has validity for use in the workplace with employees. The improvement
of resilience in the workplace can be aided by the use of a relevant, valid, and reliable
instrument. The WRI has shown promising psychometric qualities and is grounded in
the resilience research.
Improving individual resilience is one area where managers can play a role in
their employees’ development. With knowledge of the primary resilience factors and the
results from the use of the WRI, those efforts can be focused on specific actions.
Expected outcomes of improved workplace resilience include: more effective actions
taken in a crisis, reduced stress, higher quality decision making, decreased use of sick
days, and higher job satisfaction. Future research on resilience and outcomes can verify
the extent and conditions in which these outcomes exist.
With the study sample being solely in the healthcare sector, care must be taken
when attempting to generalize these findings to other sectors. Additionally, the majority
of the study sample was nurses, a distinct job class in a specific sector. Future work will
study the behavior of the WRI in sectors such as manufacturing, service, and education
to assess the validity of using the WRI in sectors other than healthcare. Although no
significant findings emerged from the analysis of resilience by U.S. regions, future work
could focus on resilience differences across countries known to differ on other
workplace variables. Finally, future work could investigate whether the measurement
precision of the WRI is the same across sample groups such as gender, age, and type of
work.
23
7. Conclusion
We have shown the psychometric qualities for an instrument to measure
resilience in the workplace. The WRI was shown to have four factors and convergent
validity with a job stress instrument.
The self-administered instrument was completed by 540 participants across two
samples—healthcare executives and hospital-based nursing staff. The instrument
contained items measuring resilience and job stress, and captured demographics of
gender, age interval, employment location, and years of healthcare experience. The WRI
used a five-point “extent-of-truth” scale.
The WRI’s psychometric properties provide evidence for a four-factor model.
This model has an acceptable RMSEA and has good fit indices as indicated by CFI and
TLI. Analyses by hospital location (urban vs. rural) showed no significant differences in
resilience levels, but males scored significantly higher as a group than females among all
four WRI factors and hospital executives scored significantly higher than nurses on all
four factors, even when analyzed on female-only subsets. Years of healthcare experience
was positively correlated with each of the four WRI factors. These findings suggest
future research may wish to focus on investigating if and why protective factors
(inducing higher resilience) are more likely to be deployed by males than females and
how protective factors are developed through the course of one’s career and how
development of protective factors could be accelerated among those showing lower
levels of resilience.
This instrument development study produced a resilience instrument designed
with employees in the workplace, not patients or clients in a clinical setting.
24
Psychometrics provided validation and support for the quality of the tool for use in
workplace settings. The WRI has the potential to provide organizations and managers a
useful tool for improving workplace resilience and helping employees achieve their
potential.
Acknowledgments
This work was supported by the Faculty Research and Creative Activities Award
program at Western Michigan University, Kalamazoo, Michigan, USA.
References
1. Weick KE. The collapse of sensemaking in organizations: the Mann Gulch disaster. Administrative Science Quarterly. 1993; 38: p. 628-652.
2. Mallak LA. Measuring Resilience in Healthcare Organizations. Health Manpower Management. 1998; 24(4): p. 148-152.
3. Rosch PJ. The quandary of job stress compensation. Health and Stress. 2001; 3: p. 1-4.
4. Coutu D. How Resilience Works. Harvard Business Review. 2002; May: p. 2-8.
5. Mosley Jr. DC, Mosley DC, Pietri PH. Supervisory Management: The Art of Empowering and Developing People (9e) Cincinnati: Southwestern Publishing; 2015.
6. WebMD. The Effects of Stress on Your Body. [Online].; 2014 [cited 2012 12 09. Available from: http://www.webmd.com/balance/stress-management/effects-of-stress-on-your-body.
7. Ahern N, Keihl E, Sole M, Byers J. A Review of Instruments Measuring Resilience. Issues in Comprehensive Pediatric Nursing. 2006; 29: p. 103-125.
25
8. Rew L, Horner SD. Youth Resilience Framework for Reducing Health-Risk Behaviors in Adolescents. Journal of Pediatric Nursing. 2003; 18(6): p. 379-388.
9. Werner EE. Risk, Resilience, and Recovery: Perspectives from the Kauai Longitudinal Study. Development and Psychopathology. 1993; 5: p. 503-515.
10. Conner DR. Managing at the Speed of Change New York: Villard Books; 1993.
11. Deevy E. Creating the Resilient Organization: A Rapid Response Management Program Engelwood Cliffs, NJ: Prentice-Hall; 1995.
12. Wagnild GM, Young HM. Development and psychometric evaluation of the resilience scale. Journal of Nursing Measurement. 1993; 1: p. 165-178.
13. Connor KM, Davidson JR. Development of a New Resilience Scale: The Connor-Davidson Resilience Scale (CD-RISC). Depression and Anxiety. 2003; 18: p. 76-82.
14. Friborg O, Hjemdal O, Rosenvinge JH, Martinussen M. A New Rating Scale for Adult Resilience: What are the Central Protective Resources Behind Healthy Adjustment? International Journal of Methods in Psychiatric Research. 2003; 12(2): p. 65-76.
15. Oshio A, Kaneko H, Nagamine S, Nakaya M. Construct Validity of the Adolescent Resilience Scale. Psychological Reports. 2003; 93: p. 1217-1222.
16. Sinclair VG, Wallston KA. The Development and Psychometric Evaluation of the Brief Resilient Coping Scale. Assessment. 2004; 11(1): p. 94-101.
17. Werner EE. Protective Factors and Individual Resilience. In Meisels SJ, Shonkoff JP, editors. Handbook of Early Childhood Intervention. Cambridge: Cambridge University Press; 1990. p. 97-116.
18. Ashby MF, Greer AL. Metallic glasses as structural materials. Scripta Materialia. 2006; 54: p. 321-326.
19. Cohen S, Komarck T, Mermelstein R. A global measure of perceived stress. Journal of Health and Social Behavior. 1983; 24: p. 386-396.
20. Carver CS, Scheier MF, Weintraub JK. Assessing Coping Strategies: A Theoretically Based Approach. Journal of Personality and Social Psychology. 1989; 56: p. 267-283.
26
21. Kawada T, Otsuka T. Relationship Between Job Stress, Occupational Position, and Job Satisfaction Using a Brief Job Stress Questionnaire. Work. 2011; 40: p. 393-399.
22. GoEngineer. Stress-GoEngineer. [Online].; 2014 [cited 2014 November 14. Available from: http://www.goengineer.com/glossary/stress/.
23. National Institutes of Health (NIH). MedLine Plus. [Online].; 2014 [cited 2014 November 14. Available from: http://www.nlm.nih.gov/medlineplus/ency/article/003211.htm.
24. Huffington Post. Huffington Post. [Online].; n.d. [cited 2014 November 14. Available from: http://www.huffingtonpost.com/2014/02/25/stress-anxiety-difference_n_4833172.html.
25. Haase J. The Adolescent Resilience Model as a Guide to Interventions. Journal of Pediatric Oncology Nursing. 2004; 21(5): p. 289-299.
26. Hunter AJ, Chandler GE. Adolescent Resilience. The Journal of Nursing Scholarship. 1999; 31(3): p. 243-247.
27. Block JH, Block J. The Role of Ego-Control and Ego in the Organization of Behavior. In Collins W, editor. Development of Cognition, Affect, and Social Relations. Hillsdale: Lawrence Erlbaum Associates; 1980. p. 48.
28. Rutter M. Resilience in the face of adversity: protective factors and resistance to psychiatric disorders. British Journal of Psychology. 1985; 147: p. 598-611.
29. Bartone PT, Ursano RJ, Wright KM, Ingraham LH. The impact of a military air disaster on the health of assistance workers: a prospective study. Journal of Nervous and Mental Disorders. 1989; 177: p. 317-328.
30. Tusaie K, Dyer J. Resilience: A Historical Review of the Construct. Holistic Nursing Practice. 2004;(Jan/Feb): p. 3-8.
31. Hatton E. Levi-Strauss’s Bricolage and Theorizing Teachers’ Work. Anthropology & Education Quarterly. 1989; 20: p. 74-96.
32. Levi-Strauss C. The Savage Mind (2nd Ed.) London: Weidenfeld and Nicholson; 1974.
27
33. Kobasa SC. Stressful life events, personality, and health: an inquiry into hardiness. Journal of Personality and Social Psychology. 1979; 37: p. 1-11.
34. McLarnon M, Rothstein MG. Development and Initial Validation of the Workplace Resilience Inventory. Journal of Personnel Psychology. 2013; 12(2): p. 63-73.
35. Winwood P, Colon R, McEwen K. A Practical Measure of Workplace Resilience. Journal of Environmental Medicine. 2013 October; 55(10): p. 1205-1212.
36. Connor KM, Davidson JR. Overview: Connor-Davidson Resilience Scale (CD-RISC). ; 2013.
37. Barnette JJ. Effects of Stem and Likert Response Option Reversals on Survey Internal Consistency: If You Feel the Need, There is a Better Alternative to Using those Negatively Worded Stems. Educational and Psychological Measurement. 2000; 60: p. 361-370.
38. Merritt SM. The Two-Factor Solution to Allen and Meyer’s (1990) Affective Commitment Scale: Effects of Negatively Worded Items. Journal of Business Psychology. 2012; 27: p. 421-436.
39. Roszkowski MJ, Soven M. Shifting gears: consequences of including two negatively worded items in the middle of a positively worded questionnaire. Assessment & Evaluation in Higher Education. 2009; 35(1): p. 113-130.
40. McKnight PE, McKnight KM, Sidani S, Figueredo AJ. Missing data: A gentle introduction New York: Guilford; 2007.
41. Asparouhov T, Muthen B. Multiple Imputation with Mplus. [Online].; 2010 [cited 2014 July 1. Available from: http://www.statmodel.com/download/Imputations7.pdf.
42. Jöreskog KG, Sörbom D. LISREL 8: User’s guide Chicago: Scientific Software; 1993.
43. Muthen LK, Muthen BO. Mplus User’s Guide, Seventh Edition. [Online].; 2012 [cited 2014 September 30. Available from: http://www.statmodel.com/download/usersguide/Mplus%20user%20guide%20Ver_7_r6_web.pdf.
44. Steiger JH. Structural Model Evaluation and Modification: An Interval Estimation Approach. Multivariate Behavioral Research. 1990; 25(2): p. 173-180.
28
45. Bentler PM. Comparative fit indexes in structural models. Psychological Bulletin. 1990; 107: p. 238-246.
46. Tucker LR, Lewis C. A Reliability Coefficient for Maximum Likelihood Factor Analysis. Psychometrika. 1973; 38(1): p. 1-10.
47. Hu LH, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal. 1999; 6(1): p. 1-55.
48. Hoyle RH. Handbook of structural equation modeling New York: Guilford Press; 2012.
49. Sood A, Prasad K, Schroeder D, Varkey P. Stress Management and Resilience Training Among Department of Medicine Faculty: A Pilot Randomized Clinical Trial. Journal of General Internal Medicine. 2011; 26(8): p. 858-861.
50. McDonald RP. Test Theory: New York: Taylor & Francis; 1999.
51. Goldberg DP, Hillier VF. A Scaled Version of the General Health Questionnaire. Psychological Medicine. 1979; 9: p. 139-145.
52. Somers S. Measuring Resilience Potential: An Adaptive Strategy for Organizational Crisis Planning. Journal of Contingencies and Crisis Management. 2009; 17(1): p. 12-23.
53. United_States_Army_Combat_Readiness_Center. Leader Engagement Kit-U.S. Army. [Online].; 2008 [cited 2014 November 7. Available from: http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=4&ved=0CDIQFjAD&url=http%3A%2F%2Fwww.armyg1.army.mil%2Fdcs%2Fdocs%2Fleader%2520engagement%2520kit_complete.pdf&ei=rgpdVN7TO4ugyAT4p4K4Bw&usg=AFQjCNGUdPh6dvGon-LMUPLE8WxmlhQ6UA&bvm=bv.79184187,d.
29
Table captions
Table 1. Summary of the major resilience scales.
Table 2. Demographics of the study respondents.
Table 3. Initial communality estimates
Table 4. EFA solutions with Geomin rotation
Table 5. WRI factors and example items
Table 6. Correlations among WRI subscales, BJSQ subscales, experience, and subscale reliability
Table 7. Factor scores and gender
Table 8. Factor scores and sample
Table 9. Factor scores and hospital location
30
Figure captions
Figure 1. Eigenvalues and parallel analysis plot
Figure 2. Factor structure of the Workplace Resilience Instrument.
31
Table 1. Summary of the major resilience scales.
Instrument Source Dimensions Primary Population(s)
CD-RISC 25 items
Connor & Davidson [13]
Sood et al. [49]
Personal competence Trust in one’s instincts, strengthening by stress Positive acceptance of change Control Spiritual influences
Psychiatric/ PTSD population
Medical staff General population
DRS 45 items (32-short form)
Bartone et al. [29]
Commitment Control Change
Military survivor assistance officers
Resilience Scales (RS)
Wagnild & Young [12]
Personal competence Acceptance of self and life
Elderly women
Resilience Scale for Adults (RSA)
Friborg et al. [14]
Personal competence Social competence Family coherence Social support Personal structure
Mental health outpatients & control group
32
Table 2. Demographics of the study respondents.
Sample Frequency % U.S. hospital-based nurses 362 67 Midwest U.S. hospital executives 175 33 Gender Female 447 84 Male 85 16 Hospital Location Urban 337 66 Rural 174 34 Age 16-24 19 4 25-34 38 7 35-44 59 11 45-54 150 28 55-64 225 42 65-74 32 6 75+ 2 0 Prefer not to answer 9 2 Regions of U.S. West 67 13 Midwest 269 50 Northeast 90 17 South 108 20
33
Table 3. Initial Communality Estimates
I1 I2 I3 I4 I5 I6 I7 I8 I9 I10 0.60 0.65 0.70 0.56 0.73 0.57 0.65 0.66 0.64 0.70 I11 I12 I13 I14 I15 I16 I17 I18 I19 I20 0.81 0.74 0.48 0.64 0.68 0.70 0.61 0.70 0.80 0.77 I21 I22 I23 I24 I25 0.17 0.40 0.46 0.19 0.51
34
Table 4. EFA solutions with Geomin rotation*
Solution RMSEA CFI TLI Inter-factor correlation range 1 Factor 0.16 0.74 0.72 2 Factor 0.12 0.88 0.85 0.27 3 Factor 0.10 0.92 0.89 0.03-0.45 4 Factor 0.08 0.95 0.92 0.17-0.59 5 Factor 0.07 0.97 0.94 0.17-0.64 6 Factor 0.06 0.98 0.96 0.14-0.58
*EFA solutions were based on WLSMV
35
Table 5. WRI factors and example items
Factor Example Item Active Problem-Solving I take delight in solving difficult problems. Team Efficacy I understand my team’s overall goals. Confident Sense-Making I make sense of the situation when it becomes chaotic. Bricolage When under extreme pressure, I still take time to try
new methods.
36
Table 6. Correlations among WRI subscales, BJSQ subscales, experience, and subscale reliability
WRI Factors
BJSQ Factors
Active Problem-Solving
Team Efficacy
Confident Sense-Making
Bricolage
Job Control -0.31 -0.30 -0.30 -0.28 Support -0.27 -0.40 -0.36 -0.24 Job Demand 0.15 0.17 0.13 0.10 WRI Factors Active
Problem-Solving
0.68
0.68
0.75
Team Efficacy 0.77 0.69 Confident
Sense-Making 0.73
Job Satisfaction
0.30 0.35 0.37 0.29
Correlation Experience 0.15 0.21 0.22 0.11 Internal Alpha* 0.80 0.79 0.77 0.83 Consistency/ Reliability
Omega** 0.80 0.80 0.77 0.83
*Reliability coefficient alpha, **Reliability coefficient omega [50]
37
Table 7. Factor scores and gender
Gender Male Female t df p
Active Problem Solving
0.20 -0.04 -2.99 1 0.002
Team Efficacy
0.15 -0.03 -2.61 1 0.009
Confident Sense-Making
0.16 -0.04 -2.41 1 0.01
Bricolage 0.17 -0.03 -2.83 1 0.004
38
Table 8. Factor scores and sample
Sample Executives Nurses t df p Active Problem-Solving
0.34 -0.18 -8.49 1 <0.001
Team Efficacy
0.30 -0.16 -8.92 1 <0.001
Confident Sense-Making
0.32 -0.17 -8.21 1 <0.001
Bricolage 0.28 -0.14 -7.57 1 <0.001
39
Table 9. Factor scores and hospital location
Location Rural Urban t df p Active Problem-Solving 0.06 -0.03 -1.44 1 0.15 Team Efficacy 0.009 -0.006 -0.24 1 0.81 Confident Sense-Making 0.02 -0.02 -0.69 1 0.48 Bricolage 0.03 -0.01 -0.88 1 0.37
40
Figure 1. Eigenvalues and parallel analysis plot2
2 The parallel analysis was based on maximum likelihood.
41
Figure 2. Factor structure of the Workplace Resilience Instrument.
RMSEA=0.077, CFI=0.953, TLI=0.945
42
View publication statsView publication stats