Content uploaded by Bengt Arnetz
Author content
All content in this area was uploaded by Bengt Arnetz on Nov 13, 2018
Content may be subject to copyright.
Social Science and Medicine 52 (2001) 417–427
Violence towards health care staff and possible effects on the
quality of patient care
Judith E. Arnetz*, Bengt B. Arnetz
Department of Public Health and CaringSciences, Section for Social Medicine, Uppsala University, Uppsala Science Park,
751 85 Uppsala, Sweden
Abstract
Much of the research on violence in the health care sector has focused on the immediate and long-term effects of
patient violence on staff victims. There is a lack of studies, however, examining whether individual reactions to violent
episodes, such as anger and increased fear in one’s work, have any measurable effect on staff behaviour toward their
patients, and ultimately on the quality of patient care. The aim of the present study was to investigate whether an
association exists between staff experiences with violence and patient-rated quality of patient care. A theoretical model
was presented, suggesting that violence or threats experienced by health care staff have a negative effect on the quality
of health care services offered, as measured by patients. In addition, it was theorised that there would be an association
between staff work environment and staff reports of violence. Six questionnaire studies, three concerning hospital staff’s
views of their work environment and three dealing with patients’ perceptions of the quality of care, provided the data
for evaluating the model. Work environment and quality of care studies were carried out simultaneously at a single
hospital in 1994, 1995, and again in 1997. Regression analysis was used to see which combination of work environment
and quality of care variables would best predict a positive overall grade for quality of care from the patient perspective.
Violence entered consistently as an important predictor into each of the three best regression equations for 1994, 1995,
and 1997, respectively. The results of this analysis suggest that the violence experienced by health care staff is associated
with lower patient ratings of the quality of care. The study indicates that violence is not merely an occupational health
issue, but may have significant implications for the quality of care provided. #2001 Elsevier Science Ltd. All rights
reserved.
Keywords: Violence; Health care workers; Quality of care; Patient satisfaction
Introduction
Violent behaviour towards health care personnel has
been shown to often have long-term psychological
effects on its victims, including post-traumatic stress
disorder, even when physical injury is not present
(Caldwell, 1992; Wykes & Whittington, 1991, 1994).
Aggressive or violent behaviour from patients can cause
staff to put their role as caregivers into question,
resulting in feelings of guilt and self-doubt (Holden,
1985; Graydon, Kasta & Khan, 1994; Mezey &
Shepherd 1994; Wykes & Whittington, 1994), or what
Holden termed ‘‘cognitive dissonance’’ (Holden, 1985).
Several studies have examined the immediate and long-
term effects of violence on staff (Lanza, 1983; Rix, 1987;
Ryan & Poster, 1989; Wykes & Whittington, 1991;
Croker & Cummings, 1995). However, few studies have
examined whether individual reactions, such as in-
creased fear, anger, irritability, sleeplessness and cogni-
tive disturbances, have any affect on staff behaviour
toward patients, and ultimately on patient ratings of the
quality of care provided.
How serious an occupational stressor is violence, and
what are its possible consequences for health care staff,
their work environment, and their patients? The present
study investigated the association between violence
*Corresponding author. Tel.: +46-18-66-35-92; fax: +46-18-
51-16-57.
E-mail address: judy.arnetz@socmed.uu.se (J.E. Arnetz).
0277-9536/01/$ - see front matter #2001 Elsevier Science Ltd. All rights reserved.
PII: S 0277-9536(00)00146-5
towards staff and the quality of patient care in a single
general hospital in Sweden. Violence in this study was
broadly defined, encompassing threatening behaviour,
verbal aggression as well as physical assault, and was
based on staff self-reporting in written questionnaires.
While questionnaire respondents were asked to describe
the nature of the violence they had experienced, the
analyses presented here do not distinguish between
verbal and physical violence. This lack of distinction is
based on previous research that has indicated that risk
factors for threats of violence and actual physical
violence are virtually identical (Flannery, Hanson &
Penk, 1995; Arnetz, Arnetz & Petterson, 1996; Arnetz,
Arnetz & So
¨derman, 1998). In addition, threats of
violence and verbal aggression can have as negative an
effect on health care staff as physical assault (Arnetz et
al., 1996; Flannery et al., 1995; Graydon et al., 1994;
Smith & Hart, 1994). Quality of care was measured from
the patient’s perspective, also by means of written
questionnaires. Work environment and quality of care
studies were carried out simultaneously at the hospital,
first in 1994, again in 1995, and a third time in 1997.
Theoretical model: violence and its association to the
quality of patient care
Previous research has focused on risk factors for
violence and has aimed at understanding the aetiology
of aggressive and violent behaviour towards health care
workers (Lanza, 1988; Lanza, Kayne, Hicks & Milner,
1991, 1994; Whittington & Wykes, 1994; Arnetz et al.,
1996; Arnetz, 1998). Building further on this research,
the present article presents a model of the possible
consequences of violence towards health care staff on
the quality of health care that they provide (Fig. 1).
According to this model, violence experienced by health
care staff has a negative association to patient ratings of
the quality of health care services offered.
In a general essay on work satisfaction, stress and
performance in the caring professions, Wallis (1987)
proposed a schematic model linking the negative effects
of work stress on job satisfaction, and thereby on
performance. Performance in the caring professions was,
in this working model, the quality of patient care.
Although the model lacked methods for measuring
performance, it cited patient satisfaction, psychological
health of patients, and interpersonal relations between
staff and patients as possible quality outcome measures.
Wallis suggested that ‘‘patient-avoiding behaviours’’
were a possible staff strategy for coping with stress that
might have negative consequences for the quality of
patient care.
Building further on Wallis’ model, we focus on
violence towards staff as a specific source of stress in
health care work. We propose that, even in non-
psychiatric settings, violence becomes the mode of
communication between patient and caregiver when
normal communication is lacking. In Fig. 1, violence is
illustrated in the form of concentric circles, or ellipses,
the common centre being the interaction between staff
and patient. The patient–staff interaction is defined as
the central aspect in the development of violence. This
interaction is affected by the immediate environment,
which we call the ward environment. This is the work
environment for the health care worker, at the same time
being the care environment for the patient. We suggest
that violence has a negative effect not only on caregivers,
but on patients as well. Violence from patients has a
negative effect on health care staff, causing more
negative attitudes toward work tasks and the patients
themselves. This negative climate affects the patient–
staff relationship, with the caregiver on his/her guard,
Fig. 1. Proposed model of the association between violence towards health care staff and patient-rated quality of care.
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427418
spending less time with patients, and less responsive to
patients’ needs. Similar reactions are experienced by the
patient, who thus feels less satisfied with the quality of
health care services being offered in a more negative
caring environment. Thus, violence has an indirect
negative effect on the ward environment, and ultimately
on the quality of care, as perceived by patients.
The course of events described above may have some
association to staff burnout, which has been defined as
‘‘a syndrome of physical and emotional exhaustion,
involving the development of negative self-concept,
negative job attitudes, and loss of concern and feeling
for clients’’ (Pines & Maslach, 1978). Several studies
have suggested that burnout in staff can lead to
deterioration in the individual’s involvement in his or
her work and in the quality of care or services provided
(Maslach & Jackson, 1981; Weisman & Nathanson,
1985; Firth & Britton, 1989; Schaufeli & Enzmann,
1998). ‘‘Stress’’’’ from patients, including violent beha-
viour, has been associated with the depersonalisation
component of burnout in nurses (Leiter & Harvie, 1996,
Prosser, Johnson, Kuipers, Szmukler, Bebbington &
Thornicroft, 1997). Each of these studies measured only
staff members’ own perceptions of staff–patient relation-
ships, with no measurement of patient views. Caplan
(1993) compared nursing staff and patient perceptions of
the ward atmosphere in a maximum security forensic
hospital. She found that both staff and patients were
affected by the potential for violence, but were none-
theless positive about the therapeutic ward atmosphere.
The feedback loop in Fig. 1 suggests that negative
aspects of the ward environment may also be conducive
to the development of violent behaviour in patients. In
the present study, our schematic model of violence as a
negative aspect of the ward environment was proposed
and evaluated. The primary aim of the study was to
investigate whether an association exists between staff
experiences with violence and patient-rated quality of
care. Specifically, would staff reports of patient violence
towards staff predict lower quality of care ratings from
patients? In addition, we theorised that changes in staff
ratings of their work environment over time would be
associated with changes in reported violence towards
staff.
Methods
Evaluation of the described model was based on six
studies carried out at the O
¨rebro Regional Hospital
(RSO
¨), a large general (non-psychiatric) hospital in
central Sweden. At the time the studies were first
initiated, RSO
¨was an 850-bed hospital employing
approximately 4500 staff. In 1994, as part of its overall
quality improvement efforts, the hospital conducted
simultaneous questionnaire studies of the work environ-
ment of hospital staff, as well as of patients’ views of the
quality of care. Results from both investigations were
studied in detail, and served as catalysts for quality
improvement efforts hospital-wide during the ensuing
year. Both measurements were then repeated in 1995,
and again in 1997. The work environment and quality of
care instruments developed for this project have
previously been described in detail (Arnetz, 1999, 1997;
Petterson & Arnetz, 1997; Arnetz & Arnetz, 1996).
These instruments provided the basis for evaluation of
the proposed model.
Work environment studies
All categories of hospital staff were included in the
work environment studies. Questionnaires were anon-
ymous, with response rates of 76% (n=3500) in 1994,
61% (n=2617) in 1995, and 71% (n=2414) in 1997. The
survey instrument was analysed by means of nine key
indices, termed ‘‘improvement areas’’ (Table 1). These
indices have been tested on large groups of health care
personnel and described previously (Arnetz, 1999, 1997).
In 1995 and 1997 the instrument included an additional
index, comprised of questions that measured the hospital
staff’s perception of the quality of care that they provided.
Violence at work
All three work environment questionnaires included a
question concerning the individual staff member’s
experience with violence. The question used in the
1994 version of the questionnaire was without any time
limitation: ‘‘Have you ever been a victim of violence or
threat of violence at your workplace?’’ The question
used in the follow-up questionnaires in 1995 and 1997
referred to the previous year: ‘‘Have you been a victim
of violence or threat of violence at work during the past
year?’’ Response alternatives (identical all three years)
were ‘‘No,’’ ‘‘Yes, once or twice,’’ or ‘‘Yes, several
times.’’ Violence and threat of violence were combined
in a single question since earlier research has shown
substantial overlap between health care staff who report
being victims of violence and of threats of violence
(Flannery, et al., 1995; Arnetz et al., 1996; Arnetz et al.,
1998; Arnetz & Arnetz, 2000). The 1995 questionnaire
included 17 additional questions concerning the perpe-
trator of violence, nature of the violent incident, and
reactions of staff victims.
Quality of care studies
Hospital staff distributed quality of care question-
naires to both inpatients and outpatients in 23 hospital
departments during a two-week period in 1994, 1995 and
again in 1997. These three studies coincided with the
work environment studies described above, i.e., both
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427 419
staff and patients responded to questionnaires simulta-
neously on three separate occasions. Patient question-
naires were anonymous, with response rates of 50%
(n=1834) in 1994, 57% in 1995 (n=2466) and 71%
(n=3593) in 1997. Six indices, also termed improvement
areas, provided the basis for analysis of the patient
questionnaire (Table 1). The work environment index
represents the patient’s perception of the staff work
environment. This aspect of the instrument was created
on the theory that patients will be more satisfied with the
care they receive in an environment that they perceive as
positive (Weisman & Nathanson, 1985; Shortell et al.,
1994; Arnetz & Arnetz, 1996). The questionnaire also
included a global satisfaction question, asking patients
to rate their overall view of the quality of care on a scale
of one (very negative) to ten (very positive). This overall
grade was treated as a scale (index) and used as a
seventh improvement area in the analysis of the quality
instrument. This single item of patient-perceived quality
of care had been previously validated against eight more
specific domains of patient-perceived quality (Arnetz &
Arnetz, 1996).
All work environment and quality of care indices were
tested for internal consistency using factor analysis with
oblique rotation. All but two indices were found to have
Cronbach’s alphas of 0.70 or higher (Arnetz & Arnetz,
1996; Arnetz, 1997, 1999) and were as follows: Mental
energy 0.87; Goal quality 0.81; Efficiency 0.76; Work
climate 0.86; Performance feedback 0.62; Leadership
0.91; Participation 0.79; Development 0.70; Stress 0.75;
Quality of care (staff perception) 0.70; Information-
illness 0.86; Information-routines 0.62; Accessibility
0.81; Medical care 0.79; Care processes 0.80; Work
environment (patient perception) 0.79. Index values
were calculated by totalling an individual respondent’s
scores on the component index items, and then
converting that sum to a percentage of the maximal
achievable index score. All indices were presented as a
percentage, with 0 as the lowest possible value and 100
as the maximum. Percentages represented mean values
for all hospital departments combined, or for each
respective department.
Statistical analysis
Violence towards hospital staff: data based on individual
questionnaire responses
All data concerned with staff experience with violence
at work was based on individual questionnaire
Table 1
Data sources for work environment (WE) and quality of care (QOC) studies
WE 1994: individual responses n=3500 WE 1995: individual responses n=2617 WE 1997: individual responses n=2414
Indices: Mental energy Indices: Mental energy Indices: Mental energy
Goal quality Goal quality Goal quality
Efficiency Efficiency Efficiency
Work climate Work climate Work climate
Performance feedback Performance feedback Performance feedback
Leadership Leadership Leadership
Participation Participation Participation
Development Development Development
Stress Stress Stress
Quality of care Quality of care
Single item: Violence Single item: Violence Single item: Violence
Single item: Hospital department Single item: Hospital department Single item: Hospital department
###
WE aggregated data: mean values for 10
variables, 23 hospital depts
WE aggregated data: mean values for 11
variables, 23 hospital depts
WE aggregated data: mean values for 11
variables, 23 hospital depts
QOC aggregated data: mean values for 7
variables, 23 hospital depts
QOC aggregated data: mean values for 7
variables, 23 hospital depts
QOC aggregated data: mean values for 7
variables, 23 hospital depts
"""
QOC 1994: individual responses n=1834 QOC 1995: individual responses n=2466 QOC 1997: individual responses n=3593
Indices: Information-illness Indices: Information-illness Indices: Information-illness
Information-routines Information-routines Information-routines
Accessibility Accessibility Accessibility
Medical care Medical care Medical care
Care processes Care processes Care processes
Work environment Work environment Work environment
Overall quality grade Overall quality grade Overall quality grade
Single item: Hospital department Single item: Hospital department Single item: Hospital department
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427420
responses. Descriptive statistics were used to analyse the
items related to violence for all responding hospital staff
in 1994, 1995 and 1997, respectively. In earlier studies,
younger staff (Whittington & Wykes, 1994; Arnetz et al.,
1996) and male health care staff (Aiken, 1984; Carmel &
Hunter, 1989; Arnetz et al., 1996) have reported
significantly more experience with workplace violence.
Reported prevalence of violence by hospital department
was examined using Chi square statistics, stratifying for
employee age and sex. Responses to the violence
question were dichotomised (yes/no), since no more
than 3% of respondents had experienced several
incidents. One-way analysis of variance (ANOVA) was
used to examine the possible effects of exposure to
violence on the work environment indices for each year
respectively.
Combined sample: aggregated data representingmean
values for hospital departments
In order to study the possible association between
violence reported by staff and perceived quality of care
by patients, it was necessary to join the data files from
the work environment and quality of care questionnaires
for each respective year. Hospital department was the
one variable that was the same in both questionnaires,
and thus provided the link between the two data sets.
Joined files for 1994, 1995 and 1997, respectively, were
created by aggregating the data, which implied that the
three final data sets were comprised of mean values for
each hospital department for each variable or index
(Dixon, 1992). Each of these three files was comprised of
the 10 work environment indices (only 9 in 1994, which
lacked the index measuring staff’s perception of the
quality of care), the question on violence, and the seven
quality of care indices. Finally, the three joined files were
combined into a single file, including data from all six
studies. Aggregating the data in this way made it
possible to evaluate the proposed model concerning
the ward environment, which was best reflected in mean
values for all departments. Graphic illustration of the
aggregation of data sets is presented in Table 1.
Predictors of quality of care
All possible subsets multiple regression analysis was
used to see which combination of aggregated work
environment and quality of care variables would best
predict a positive overall grade for quality of care from
the patient perspective. More specifically, would mean
values for exposure to violence enter in to any of the
regression equations? The overall patient grade for the
quality of care (aggregate department values) was the
dependent variable in all of the multiple regression
analyses. Patients graded the overall quality of care on a
visual analog scale (VAS) from 1 to 10. The overall
grade was thus treated as a continuous variable. As with
all indices, points for the overall grade were converted to
a percentage. Independent work environment variables
were mean department values for violence, and for
mental energy, efficiency, participation, and stress. The
ANOVA had shown these indices to be consistently
associated with reported violence in each of the three
measurement periods (Table 2). Independent variables
from the patient quality of care studies were care
processes and staff work environment (patients’ per-
spective). These variables were viewed as offering the
best reflection of the ward environment from the
patient’s perspective. The care processes index concerns
to what degree patients felt the staff had time for them,
were supportive, and were responsive to the patient’s
needs and requests. The work environment index reflects
the patients’ perception of the work climate, such as
whether staff seem stressed, co-operate well amongst
themselves, assume responsibility, and seem positive
about their work. The same combination of independent
variables was used in all multivariate regression ana-
lyses, in an effort to evaluate the strength of the
proposed model. Best subsets were identified using
Mallows’ C
p
. The Mallows’ C
p
statistic compares the
relative magnitudes of the standard error in a particular
subset with the standard error in the complete model
with all predictors present (Howell, 1992). According to
this method, the best subsets are thus defined as having
the lowest C
p
values (Dixon, 1992), i.e., the ratio of error
terms is as low as possible.
Predictors of changes in reported violence
In the final analysis using data aggregated by hospital
department, regression analysis was used to examine
whether changes over time in mean values for staff work
environment would predict changes over time in mean
aggregate values for reported violence. Only data from
1995 and 1997 were used in this analysis, since both
periods examined violence during the past year. The
independent variables were aggregate delta values (1997
minus 1995) for each of the ten work environment
indices.
The biostatistics software program BMDP (1993
version 7.0 for personal computers) was used in
analysing the data. Significance was set at p50.05
(two-tailed) for all analyses.
Results
Staff reports of violence at work
Staff experience with violence or threats of violence at
work as reported in the 1994, 1995, and 1997
questionnaires, respectively, is summarised in Table 3.
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427 421
Responses to the question regarding exposure to
violence during the past year, i.e., responses in 1995
and 1997, were similar, with just under 12% reporting
experiences with violence. However, there were a greater
number of non-respondents to the violence question in
1995 (6%, n=141), compared to 1997 (3%, n=60). A
greater percentage of hospital staff (26%) in 1994
reported having ever experienced violence or threats at
work (2% non-response, n=73). In all three studies, the
majority of staff responding affirmatively to the question
on violence had experienced only occasional incidents.
The additional questions regarding violence in the
1995 study revealed that patients were aggressive toward
staff in 76% of violent incidents, patient relatives in 20%
of incidents, other staff members in 9% of incidents, and
‘‘other’’ accounted for 5%. Of the staff who had
experienced violence in the preceding year (12% of all
staff, n=284), 51% reported experiences with threats or
verbal aggression. Eighteen percent of staff victims had
experienced some form of physical violence, such as
slapping or punching. Twelve per cent of those
victimised had been physically injured, and 1% of
violence victims had taken sick leave as a result of an
incident. Help or support after a violent incident was
most commonly from co-workers (49%), thereafter from
someone outside the workplace (18%), and was less
often from one’s supervisor (14%). Most common
reactions among staff victims were anger (59%), sadness
(47%), disappointment (43%) and fear (40%). As a
result of a violent incident, 47% of staff victims reported
having become more cautious, 13% felt fearful, and
15% felt less enjoyment in working with patients.
Multiple responses were possible on all of the above
questions.
Prevalence of violence by hospital department
The Chi square analysis established that there were
significant differences between hospital departments
with regard to staff-reported violence, even when
stratifying by staff age and sex (p50.001 for all levels).
Results of the one-way ANOVAS, used to study
possible associations between reported violence and
the work environment indices, all based on individual
responses, are presented in Table 2.
Table 2
ANOVA for work environment indices and reported violence (yes/no) 1994, 1995, 1997
1994 1995 1997
Index Mean yes/no FSignificance level Mean yes/no FSignificance level Mean yes/no FSignificance level
Mental energy 79.3/82.4 22.26 ****
a
72.2/76.3 13.14 *** 71.1/75.8 11.82 ***
Goal quality 63.1/64.3 2.24 ns 58.2/61.9 7.32 ** 60.8/64.3 5.47 *
Efficiency 58.8/61.1 13.98 *** 54.6/60.0 28.17 **** 52.0/54.7 6.62 *
Work climate 70.2/70.9 1.29 ns 64.8/69.0 15.60 *** 66.5/68.7 3.32 ns
Feedback 60.2/61.0 0.94 ns 57.3/60.4 5.32 * 49.7/49.3 0.08 ns
Leadership 72.5/73.3 0.78 ns 68.5/73.4 11.60 *** 63.9/66.0 1.44 ns
Participation 60.0/63.3 15.65 *** 58.1/65.8 41.71 **** 60.6/64.8 11.19 ***
Development 50.8/51.4 0.75 ns 49.2/53.6 19.12 **** 50.0/50.7 0.39 ns
Stress 63.4/57.7 71.88 **** 65.1/59.8 25.30 **** 73.0/70.0 6.49 *
Quality of care(staff rating) 64.3/71.2 43.88 **** 69.5/73.3 10.40 **
a
*p50.05, **p50.01, ***p50.001, ****p50.0001.
Table 3
Reported violence toward hospital staff 1994, 1995, 1997.
Victim of violence or threat at work (%)
1994
(n=3433)
1995
(n=2476)
1997
(n=2354)
No 73.9 88.5 88.3
Yes, once or twice 23.0 10.4 10.4
Yes, several times 3.1 1.1 1.3
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427422
For each measurement period, experience with
violence at work was significantly associated with lower
ratings for mental energy, work efficiency, and partici-
pation in work processes and decisions, and higher
ratings for stress. Violence was also associated with
significantly lower staff ratings of the quality of care
provided by the hospital in both 1995 and 1997. In the
1995 study, all ten of the work environment indices were
significantly associated with reported violence. Four of
the indices }work climate, feedback, leadership, and
development }showed no significant association to
reported violence in either 1994 or 1997.
Predictors of quality of care
Data from all three work environment and quality of
care studies were aggregated by hospital department and
the data sets were joined. All possible subsets regression
analysis was then used to find the best predictors of a
positive overall patient quality grade for each measure-
ment period, respectively. Statistics for the best subset
for each respective year are compiled in Table 4.
Although the best predictors of a positive overall
patient quality grade differed for each measurement
period, staff experience with violence entered into each
of the three best regression equations. For 1994, the best
subset explained 29% of the variance for the overall
quality grade. However, none of the patient question-
naire items entered into the equation. For 1995, the best
subset was comprised of only two items, violence (i.e.,
negative slope) and patients’ ratings of the staff work
environment (positive), explaining 78% of the total
variance. The best predictors of a more positive grade in
1997, aside from violence (negative), included the work
environment indices mental energy (positive) and
efficiency (negative), and the patient survey indices
concerning caring processes and staff work environment
(patient’s perspective). This combination of factors
accounted for 90% of the variance.
Staff ratings for efficiency were inversely correlated
with patient ratings of the quality of care in both 1994
and 1997, based on data aggregated by hospital
department (Table 4). As mentioned above, work
efficiency ratings were lower amongst staff who had
experienced patient violence, based on individual
responses (Table 2).
Predictors of changes in reported violence
Stepwise regression analysis was performed to in-
vestigate the possible association between changes over
time in reported violence and changes in staff ratings of
their work environment. This analysis also used
aggregate data, i.e., mean values by hospital depart-
ment, and studied change between 1995 and 1997. The
ten work environment indices (delta values, i.e., 1997
Table 4
Multiple regression analysis using all possible subsets regression (aggregate data). Best subset 1994, 1995, 1997 dependent variable: hospital department mean for overall quality of
care grade from patients
Mallows’ C
p
Contribution
to R
2a
T-statistic Adjusted R
2
(%) Residual mean
square
F-statistic d.f. (numerator,
denominator)
Significance
(tail probability)
1994 1.62 29 62.3 4.74 3, 25 0.0094
Violence 0.15 2.43
Stress 0.24 ÿ3.09
Efficiency 0.19 ÿ2.71
1995 4.42 78 3.2 42.70 2, 21 0.0000
Violence 0.15 ÿ3.94
Staff work environment (patients’ view) 0.64 8.27
1997 5.60 90 2.34 46.63 5, 20 0.0000
Violence 0.03 ÿ2.58
Mental energy 0.02 2.51
Efficiency 0.04 ÿ3.03
Caring processes 0.34 9.32
Staff work environment (patients’ view) 0.12 5.42
a
The contribution to R
2
for each variable is the amount by which R
2
would be reduced if that variable were removed from the regression equation (Dixon, 1992).
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427 423
values minus 1995 values) were entered as independent
variables in the regression equation. Delta values for
four work environment indices }mental energy, stress,
feedback and quality of care (staff perspective) }
explained 81% of the variance in reported violence over
time (Table 5).
Regression coefficients were positive for mental
energy, stress and feedback from supervisors, but
negative for staff ratings of quality of care.
Discussion
The results presented here indicate that there is an
association between violence experienced by health care
staff and patient-rated quality of care. This study was
possible due to the unique quality improvement efforts
at the Regional Hospital (RSO
¨)inO
¨rebro, Sweden,
where three parallel evaluations of the hospital staff’s
views of their work environment and patients’ percep-
tions of the quality of care have been conducted. Other
studies have linked violence with work environment
factors (Lanza et al., 1994; Snyder, 1994; Flannery,
Hanson, Penk, Pastva, Navon & Flannery, 1997).
However, none of those attempted to measure any
possible effects of violence on the quality of care. Moos
(1972) found that patients’ and staff views of the
psychiatric treatment environment were more similar
on wards with lower levels of patient anger and
aggression. Beech and Norman (1995) reported higher
patient ratings of psychiatric nurses who were able to
control violent behaviour. To our knowledge, no other
studies have examined the association between violence
towards health care workers and patient-rated quality of
care. This study has focused on what we call the ward
environment, encompassing both the work environment
(staff perspective) and the care environment (patient
perspective). The setting for the studies presented here
was a non-psychiatric health care environment, and it
could be argued that the prevalence rates of violence
among staff were too low to have any measurable effect
on the ward environment. However, we were able to
establish a clear association between staff-reported
violence and more negative staff ratings of their work
environment, and these associations were consistent
over time. This is in keeping with reports by Dougherty,
Bolger, Preston, Jones and Payne (1992), who found
that staff exposure to aggressive behaviour in long-term
geriatric patients was the best predictor of diminished
job satisfaction. The additional questions in the 1995
staff questionnaire concerning the individual’s specific
experience with violence give further support to the
negative effects of violence, with possible consequences
for the quality of care. Although the majority of violent
incidents did not result in serious physical injury or time
away from work, reactions of anger, sadness, disap-
pointment, and fear were not uncommon. After a
violent incident, nearly half the staff victims reported
being more cautious or on their guard in their work with
patients, 13% were continuously fearful, and 15% felt
less satisfaction in direct patient work. These findings
are in line with results of previous studies (Lanza, 1983;
Holden, 1985; Ryan & Poster, 1989; Graydon et al.,
1994; Smith & Hart, 1994; Arnetz, 1998) and should be
seen in light of research that pinpoints fear of violence as
having a mediating effect on psychological well-being,
job satisfaction, and organisational commitment (Ro-
gers & Kelloway, 1997; Leather, Beale & Lawrence,
1997). Moreover, all of these reactions may affect the
way in which the individual health care professional
behaves toward and treats his or her patients. Increased
fear and caution in staff supports the ‘‘patient-avoiding’’
coping strategy (Wallis, 1987) and depersonalisation
aspect of burnout (Leiter & Harvie, 1996, Prosser et al.,
1997) mentioned earlier. Ultimately, these staff beha-
viours may be reflected in more negative patient
perceptions of the quality of health care provided.
Paradoxically, we found that after controlling for
changes in ward stress, increased mental energy over
time was associated with increases in reported violence
(Table 5). In order to further understand the association
between burnout and violence towards staff, we believe
that future studies need to incorporate measures of work
stress as well.
The associations between reported violence and the
work environment indices (Table 2) were not entirely
consistent over time. This may in part be a reflection of
the variation in response rates. Experience with violence
Table 5
Multiple regression analysis predicting changes in reported violence (1995–1997). Independent variables represent changes (Dvalues)
for the period 1995 to 1997
Independent variables Beta Adjusted R
2
Fto enter DR
2
Mental energy 0.03 0.67 58.20
Stress 0.03 0.72 5.71 0.06
Feedback 0.02 0.77 5.89 0.05
Quality of care (staff rating) ÿ0.02 0.81 5.03 0.04
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427424
at work was significantly associated with lower ratings
on all indices in 1995, when the response rate was
substantially lower (61%), as compared to 1994
(76%) and 1997 (71%). The 1995 study coincided
with a general announcement that the hospital would
be entering a period of substantial downsizing, and
this may have attributed to generally lower morale
among staff, as well as to the lower response rate
(Petterson & Arnetz, 1998). The significant associations
between all work environment indices and reported
violence may reflect more negative attitudes from
hospital staff generally. The best subset regression for
1995 included only violence and the patients’ view of
staff work environment as predictors of the quality
grade from patients, indicating that these variables
outweighed all others in predictive value. However, the
variation in associations between violence and the work
environment indices over time may also in part be
explained by the different time aspects in the violence
questions. Those responding in 1994 may have been
reporting violent events that occurred long ago, while
staff responding in 1995 and 1997 were referring to more
recent events.
The best multivariate regression equations, based on
aggregated mean values for hospital departments for all
study variables, included violence in each of the best
subsets. In view of the fact that both staff ratings of their
work environment (Weisman & Nathanson, 1985;
Arnetz, 1997) and patient ratings of the quality of care
(Linder-Pelz, 1982; Weisman & Nathanson, 1985;
Arnetz & Arnetz, 1996) clearly change over time, it is
not surprising that predictors of a positive quality grade
are different at different points in time. It therefore
becomes all the more interesting that a factor such as
violence towards staff, although not as prevalent in a
general hospital environment as in psychiatric disci-
plines, emerged as a predictor of quality each year.
These findings support our proposed model, that there is
a negative association between violence experienced by
health care staff and the quality of health care services,
measured from the patients’ point of view. However,
there are several factors that need to be considered
regarding these results. First, in the best subset regres-
sion for 1994, the regression coefficient for violence was
positive in relation to the overall quality grade in 1994,
while stress and efficiency were inversely related. In the
regression subsets for both 1995 and 1997, the regression
coefficients for violence were negative. The best subset in
1994 also explained only 29% of the variance, and the
residual mean square was substantially higher, com-
pared to the subsequent measurements. This might in
part be explained by the fact that the staff question
regarding experience with violence and threats was
different in 1994, with no reference to any specific time
period of exposure, while the patient ratings concerned a
specific two-week period in 1994. In 1995 and 1997, the
violence question referred specifically to the previous
12-month period. While there was a higher percentage
of positive responses to the violence question in
1994, it is also likely that those who responded
affirmatively referred to incidents that occurred some
time in the past. Generally in this study, staff
responses to the work environment questionnaires
reflected perceptions of a longer time period, compared
to patient responses to quality of care questionnaires,
which reflected a more immediate period in time.
Nevertheless, tolerance levels for violence exceeded
0.90 in all three best subsets, indicating good stability
and relatively little overlap with other predicting
variables (Howell, 1992).
It is interesting to note that staff ratings for efficiency
were inversely correlated with patient ratings of the
quality of care, in both 1994 and 1997 (Table 2). This
particular index concerns work planning, working
toward a common goal, finding time for continuing
education, and decision-making processes. Positive staff
views of an efficient workplace may not coincide with
patient views of an optimal care environment. Those
ward environments considered ‘‘efficient’’ by the staff
themselves may be less sensitive to patient needs, and
perhaps conducive to aggressive patient behaviour. In
future quality studies, it might therefore be informative
to compare patient and staff views of efficiency in health
care services.
The results in Table 5 offer support for the feedback
loop in our proposed model. Using this model, changes
over time in staff’s ratings for mental energy, stress,
feedback from supervisors, and quality of care at their
workplace predicted changes over time in the outcome
variable, reported violence. In light of these findings we
believe that further research using longitudinal data is
warranted.
This study has several limitations that should be
addressed. First, while it was necessary to use aggre-
gated (departmental) data in order to combine work
environment (staff) with quality of care (patient) data,
this may have resulted in decreased sensitivity, causing a
weakening of some associations that in reality were
stronger. In addition, these analyses are at the organisa-
tional level, and do not focus on circumstances within
specific hospital departments where there may have been
considerable variation in staff and patient ratings over
time. This is an issue discussed by Weisman and
Nathanson (1985), who used aggregate data in a study
of the relationship between professional satisfaction and
client satisfaction with family planning. However, the
use of aggregate data has also enabled us to treat
reported violence, essentially a dichotomous variable, as
a continuous variable, representing mean values for all
hospital departments.
Second, all six studies were carried out at a single
hospital. While it was this fact that enabled us to link the
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427 425
work environment with the quality of care data, and
thereby to test and evaluate the proposed model, the
results presented here may not be representative of
all health care environments. However, subsequent
national studies of ours, of both health care staff
and patients, indicate that RSO
¨is representative for
Swedish health care in general (unpublished data).
Third, response rates varied in these studies over the
three years. In the patient studies, especially, response
rates were substantially lower in 1994 and 1995 than
in the third year. While response rates of 50–60%
have been considered adequate in patient studies of
this kind (Carey & Seibert, 1993), it remains unknown
whether non-respondents differed significantly from
respondents in their views of the quality of care.
Fourth, reports of exposure to violence or threats of
violence are from staff only. In order to better
understand the theorised association between violence
and the quality of care, future studies should also
include patient reports on violent events. Fifth,
quality of care is not clearly defined. Quality of care
from the patient point of view is undoubtedly subject to
change over time, and is determined by a multitude of
factors (Linder-Pelz, 1982). The overall grade used here
as the dependent variable provided a convenient
summary of patient ratings, but may have been too
broad. This overall rating has been previously validated
using more specific domains of patient-perceived quality
of care (Arnetz & Arnetz, 1996). While more specific
aspects of the patients’ views are embodied in the other
six quality of care indices, our main focus in evaluation
of the proposed model was on the indices concerning
care processes and staff work environment from the
patient perspective. However, it would be of interest in
future studies to assess whether violence also has an
effect on other measures of quality of care, such as
professional ratings of quality and medical treatment
outcomes.
Finally, the study presented here represents cross-
sectional data. While repeated measurements were
conducted of both the staff work environment and
patient satisfaction, it is not possible to establish
causality. However the data presented do indicate an
association between violence towards health care staff
and patient ratings of the quality of health care services
offered. In addition, certain aspects of the health care
work environment may be associated with the occur-
rence of violent behaviour towards staff. These results
have implications for future violence prevention pro-
grams at the health care workplace. Limiting work stress
may be as important as learning to manage violent
incidents. Management intervention strategies should
focus on work environment as well as patient care
factors in order to prevent the possible negative effects
of violent behaviour on staff well-being and on the
quality of care.
Acknowledgements
This work was supported by grants from The Swedish
Working Life Fund. We would like to thank Mr. Ove
Petersson, and Ms. Carine Norstro
¨m, senior project
management at the Regional Hospital, and all the staff
and patients of the Regional Hospital in O
¨rebro,
Sweden. The helpful comments of two anonymous
referees are gratefully acknowledged.
References
Aiken, G. J. M. (1984). Assaults on staff in a locked ward:
prediction and consequences. Medicine, Science and the
Law,24, 199–207.
Arnetz, B. B. (1999). Staff perception of the impact of health
care transformation on quality of care. International Journal
for Quality in Health Care,11, 345–351.
Arnetz, B. B. (1997). Physicians’ view of their work environ-
ment and organisation. Psychotherapy and Psychosomatics,
66, 155–162.
Arnetz, J. E. (1998). The Violent Incident Form (VIF): a
practical instrument for the registration of violent incidents
in the health care workplace. Work & Stress,12, 17–28.
Arnetz, J. E., & Arnetz, B. B. (1996). The development and
application of a patient satisfaction measurement system for
hospital-wide quality improvement. International Journal
for Quality in Health Care,8, 555–566.
Arnetz, J. E., & Arnetz, B. B. (2000). Implementation and
evaluation of a practical intervention programme for
dealing with violence towards health care workers. Journal
of Advanced Nursing,31, 668–680.
Arnetz, J. E., Arnetz, B. B., & Petterson, I. -L. (1996). Violence
in the nursing profession: occupational and lifestyle risk
factors in Swedish nurses. Work & Stress,2, 119–127.
Arnetz, J. E., Arnetz, B. B., & So
¨derman, E. (1998). Violence
toward health care workers. Prevalence and incidence at a
large regional hospital in Sweden. American Association of
Occupational Health NursingJournal,46, 107–114.
Beech, P., & Norman, I. J. (1995). Patients’ perceptions of the
quality of psychiatric nursing care: findings from a small-
scale descriptive study. Journal of Clinical Nursing,4,
117–123.
Caldwell, M. F. (1992). Incidence of PTSD among staff victims
of patient violence. Hospital and Community Psychiatry,43,
838–839.
Caplan, C. A. (1993). Nursing staff and patient perceptions of
the ward atmosphere in a maximum security forensic
hospital. Archives of Psychiatric Nursing,7, 23–29.
Carey, R. G., & Seibert, J. H. (1993). A patient survey system to
measure quality improvement: questionnaire reliability and
validity. Medical Care,31, 834–845.
Carmel, H., & Hunter, M. (1989). Staff injuries from inpatient
violence. Hospital and Community Psychiatry,40, 41–46.
Croker, K., & Cummings, A. L. (1995). Nurses’ reactions to
physical assault by their patients. Canadian Journal of
NursingResearch,27, 81–93.
Dixon, W. J. (1992). BMDP Statistical Software Manual, Vol.
2, University of California Press, Berkeley, pp. 1085–1104.
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427426
Dougherty, L. M., Bolger, J. P., Preston, D. G., Jones, S. S., &
Payne, H. C. (1992). Effects of exposure to aggressive
behavior on job satisfaction of health care staff. The Journal
of Applied Gerontology,11, 160–172.
Firth, H., & Britton, P. (1989). ‘Burnout’, absence and turnover
amongst British nursing staff. Journal of Occupational
Psychology,62, 55–59.
Flannery, R. B., Hanson, M. A., & Penk, W. (1995). Patients
threats. Expanded definition of assault. General Hospital
Psychiatry,17, 451–453.
Flannery, R. B., Hanson, M. A., Penk, W. E., Pastva, G. J.,
Navon,M.A.,&Flannery,G.J.(1997).Hospitaldownsizing
and patients’ assaults on staff. Psychiatric Quarterly,68, 67–
76.
Graydon, J., Kasta, W., & Khan, P. (1994). Verbal and physical
abuse of nurses. Canadian Journal of NursingAdministra-
tion,7, 70–89.
Holden, R. J. (1985). Aggression against nurses. The Australian
Nurses Journal,15, 44–48.
Howell, D. C. (1992). In Statistical methods for psychology(pp.
512–519). Belmont, CA: Duxbury Press.
Lanza, M. L. (1983). The reactions of nursing staff to physical
assault by a patient. Hospital and Community Psychiatry,34,
44–47.
Lanza, M. L. (1988). Factors relevant to patient assault. Issues
in Mental Health Nursing,9, 239–257.
Lanza, M. L., Kayne, H. L., Hicks, C., & Milner, J. (1991).
Nursing staff characteristics related to patient assault. Issues
in Mental Health Nursing,12, 253–265.
Lanza, M. L., Kayne, H. L., Hicks, C., & Milner, J. (1994).
Environmental characteristics related to patient assault.
Issues in Mental Health Nursing,15, 319–335.
Leather, P., Beale, D., & Lawrence, C. (1997). Effects of
exposure to occupational violence and the mediating impact
of fear. Work & Stress,11, 329–340.
Leiter, M. P., & Harvie, P. L. (1996). Burnout among mental
health workers: a review and a research agenda. Interna-
tional Journal of Social Psychiatry,42, 90–101.
Linder-Pelz, S. (1982). Toward a theory of patient satisfaction.
Social Science & Medicine,16, 577–582.
Maslach, C., & Jackson, S. E. (1981). The measurement of
experienced burnout. Journal of Occupational Behaviour,2,
99–113.
Mezey, G., & Shepherd, J. (1994). Effects of assault on health-
care professionals.. In J. Shepherd (Ed.), Violence in health
care. A practical guide to copingwith violence and caringfor
victims (pp. 1–11). Oxford: Oxford Medical Publications.
Moos, R. (1972). Size, staffing, and psychiatric ward treatment
environments. Archives of General Psychiatry,26, 414–418.
Petterson, I. -L., & Arnetz, B. B. (1997). Measuring psychoso-
cial work quality and health. Development of health care
measures of measurement. Journal of Occupational Health
Psychology,2, 229–241.
Petterson, I. -L., & Arnetz, B. B. (1998). Psychosocial stressors
and well-being in health care workers. The impact of an
intervention program. Social Science and Medicine,47,
1763–1772.
Pines, A., & Maslach, C. (1978). Characteristics of staff burnout
in mental health settings. Hospital and Community Psychia-
try,29, 233–237.
Prosser, D., Johnson, S., Kuipers, E., Szmukler, G., Bebbing-
ton, P., & Thornicroft, G. (1997). Perceived sources of work
stress and satisfaction among hospital and community
mental health staff, and their relation to mental health,
burnout and job satisfaction. Journal of Psychosomatic
Research,43, 51–59.
Rix, G. (1987). Staff sickness and its relationships to violent
incidents on a regional secure psychiatric unit. Journal of
Advanced Nursing,12, 223–228.
Rogers, K. -A., & Kelloway, E. K. (1997). Violence at work:
personal and organizational outcomes. Journal of Occupa-
tional Health Psychology,2, 63–71.
Ryan, J. A., & Poster, E. C. (1989). The assaulted nurse: short-
term and long-term responses. Archives of Psychiatric
Nursing,3, 323–331.
Schaufeli, W., & Enzmann, D. (1998). The burnout companion
to study and practice: A critical analysis. London: Taylor &
Francis.
Shortell, S. M., Zimmerman, J. E., Rousseau, D. M., Gillies, R.
R., Wagner, D. P., Draper, E. A., Knaus, W. A., & Duffy, J.
(1994). The performance of intensive care units: does
good management make a difference?. Medical Care,32,
508–525.
Smith, M. E., & Hart, G. (1994). Nurses’ responses to patient
anger: from disconnecting to connecting. Journal of
Advanced Nursing,20, 643–651.
Snyder, W. (1994). Hospital downsizing and increased fre-
quency of assaults on staff. Hospital and Community
Psychiatry,45, 378–380.
Wallis, D. (1987). Satisfaction, stress, and performance: issues
for occupational psychology in the ‘caring’ professions.
Work & Stress,1, 113–128.
Weisman, C. S., & Nathanson, C. A. (1985). Professional
satisfaction and client outcomes. Medical Care,23,
1179–1192.
Whittington, R., & Wykes, T. (1994). Violence in psychiatric
hospitals: are certain staff prone to being assaulted?. Journal
of Advanced Nursing,19, 219–225.
Wykes, T., & Whittington, R. (1991). Coping strategies used by
staff following assault by a patient: an exploratory study.
Work & Stress,5, 37–48.
Wykes, T., & Whittington, R. (1994). Reactions to assault. In
T. Wykes (Ed.), Violence and health care professionals (pp.
105–126). London: Chapman & Hall.
J.E. Arnetz, B.B. Arnetz / Social Science and Medicine 52 (2001) 417–427 427
