Content uploaded by Judith Johnson
Author content
All content in this area was uploaded by Judith Johnson on Jan 12, 2017
Content may be subject to copyright.
1
Burnout mediates the association between symptoms of depression and patient safety
perceptions
Judith JOHNSON, PhD, ClinPsyD* ,Gemma LOUCH, PhD, Alice DUNNING, Olivia
JOHNSON, Angela GRANGE, PhD, RSCN, RGN, Caroline REYNOLDS, BSc, Louise
HALL, MSc and Jane O’HARA, PhD
*Corresponding author. School of Psychology, University of Leeds, Leeds, LS29JT, UK; Tel:
+44 (0)113 3435724; Fax: +44 (0)113 3435749 AND Bradford Institute for Health Research,
Bradford Royal Infirmary, Bradford, BD9 6RJ, UK; Tel: +44 (0)1274 38 34 18;
j.johnson@leeds.ac.uk @DrJTJohnson
Article accepted by Journal of Advanced Nursing:
http://onlinelibrary.wiley.com/wol1/doi/10.1111/jan.13251/abstract
DOI: 10.1111/jan.13251
2
ABSTRACT
Background
There is growing interest in the relationships between depressive symptoms and
burnout in healthcare staff and the safety of patient care. Depressive symptoms are higher in
healthcare staff than the general population and overlap conceptually with burnout. However,
minimal research has investigated these variables in nurses. Given the conceptual overlap
between depressive symptoms and burnout, there is also a need for an explanatory model
outlining the relative contributions of these factors to patient safety.
Aims
To investigate the relationships between depressive symptoms, burnout and
perceptions of patient safety. A mediation model was proposed whereby the association
between symptoms of depression and patient safety perceptions was mediated by burnout.
Design
A cross-sectional questionnaire was distributed at three acute NHS Trusts.
Method
Three-hundred and twenty-three hospital nursing staff completed measures of
depressive symptoms, burnout and patient safety perceptions (including measures at the level
of the individual and the work area/unit) between December 2015 - February 2016.
Results
When tested in separate analyses, depressive symptoms and burnout facets were each
associated with both patient safety measures. Furthermore, the proposed mediation model
3
was supported, with associations between depressive symptoms and patient safety
perceptions fully mediated by burnout.
Conclusion
These results suggest that symptoms of depression and burnout in hospital nurses may
have implications for patient safety. However, interventions to improve patient safety may be
best targeted at improving burnout in particular, with burnout interventions known to be most
effective when focused at both the individual and the organisational level.
Keywords: burnout, depression, healthcare, health services, nursing, patient safety,
workforce issues
4
SUMMARY STATEMENT
Why is this research needed?
•A small number of studies have found that burnout and depressive symptoms are
linked with patient safety outcomes in healthcare staff, but further research is needed
in nursing populations.
•There is a need to better understand how depressive symptoms and burnout relate to
each other and to patient safety in nursing staff.
What are the key findings?
•When tested in separate analyses, burnout and depressive symptoms were each
directly associated with patient safety perceptions in nursing staff.
•A mediation model was tested and results indicated that burnout fully mediated the
association between depressive symptoms and patient safety perceptions.
How should the findings be used to influence policy/practice/research/education?
•These results highlight the importance of supporting nurse wellbeing to enhance the
quality and safety of patient care.
•In particular, these results suggest that burnout may be the more important staff
wellbeing factor for patient safety interventions to focus on.
5
•Burnout interventions should focus on both work-restructuring interventions in
addition to individual-level interventions such as mindfulness courses.
6
INTRODUCTION
Background
With around 10% of hospital inpatient episodes affected by clinical errors (Baines et
al., 2013, Vincent et al., 2001) improving patient safety remains a priority in acute healthcare
settings. To develop interventions to reduce errors, a growing body of research has sought to
understand underlying causes and contributing factors. One area has focused on the role that
human factors such as staff burnout and depressive symptoms may play (Hall et al., 2016).
The burnout concept originated in health services and describes a chronic stress
response comprising three dimensions of emotional exhaustion, depersonalisation and low
accomplishment (Maslach and Jackson, 1981). Due concerns about the validity and relevance
of the personal accomplishment scale (Demerouti and Bakker, 2008, Demerouti et al., 2001),
much research has focused on the remaining two dimensions. Emotional exhaustion captures
feelings of emotional and physical resource depletion and depersonalisation captures
excessively detached attitudes towards patients (Maslach et al., 2001). A range of factors can
lead to burnout, including excessive workload, work conflicts and personality factors
(Garrosa et al., 2008). When it occurs, burnout may have wide-ranging consequences such as
lower staff empathy and poorer patient experience (Vahey et al., 2004, Leiter et al., 1998,
Passalacqua and Segrin, 2012). Several studies indicate that burnout may also be linked with
clinical errors (Hall et al., 2016), although the majority of this research has been amongst
doctors.
Depression is a broader concept which is statistically and conceptually distinct from
burnout (Iacovides et al., 2003, Thuynsma and de Beer, 2016). Rather than focusing on work-
related attitudes, the concept of depression encompasses a more holistic sense of low mood
and psychological distress (Henry and Crawford, 2005). Research suggests that depression is
7
not a discrete category, but one that exists on a continuum, ranging from low depression (or
fewer depressive symptoms), to high depression (or more depressive symptoms) (Wood et al.,
2010, Hankin et al., 2005). Similarly to burnout, depressive symptoms in healthcare staff may
have consequences for patient care such as a greater likelihood of cognitive failures and
medical errors (Allan et al., 2014, West et al., 2006). However, there has been less research
into the consequences of depressive symptoms than burnout and only two studies have
investigated depressive symptoms and patient safety amongst nursing staff (Tanaka et al.,
2012, Saleh et al., 2014).
Symptoms of mental health problems such as depression have long been known to be
elevated amongst healthcare staff, particularly in nurses (Su et al., 2009, Letvak et al., 2012,
Wall et al., 1997) and recent reports suggest that burnout rates are increasing (Shanafelt et al.,
2015, Ives et al., 2015). As such, identifying associations between burnout, depressive
symptoms and patient safety could both highlight an area for patient safety interventions to
target and suggest that such interventions are of escalating importance.
Contribution of the research
There are two main areas that need addressing. First, most studies investigating
burnout or depressive symptoms and patient safety have focused on medical staff, so there is
a need for further research in nursing staff, as a considerable amount of patient care is centred
around the work of nurses (Hughes, 2008). To date, no research has investigated either
depressive symptoms or burnout and patient safety in UK hospital nurses. Of the studies
which have been conducted in nursing, results suggest that higher levels of burnout and
depressive symptoms are associated with lower perceptions of patient safety (Halbesleben et
al., 2008, Laschinger and Leiter, 2006) and higher rates of objectively measured errors (Saleh
8
et al., 2014), although these findings have not been replicated consistently (Holden et al.,
2011).
Second, there is a need to understand how burnout and depressive symptoms relate to
each other in their association with patient safety. The two concepts are overlapping but
distinct and interventions to reduce depressive symptoms differ from those aiming to reduce
burnout. In particular, while interventions addressing depressive symptoms may focus at the
level of the individual (e.g., mindfulness), interventions for burnout should also involve
organisational-level change, considering work-related changes such as work process
restructuring, shift readjustments and performance appraisals (Awa et al., 2010, West and
Dawson, 2012). A small number of studies have investigated both burnout and depressive
symptoms (de Oliveira Jr et al., 2013, Dyrbye et al., 2013, Fahrenkopf et al., 2008, Garrouste-
Orgeas et al., 2015), but none have proposed an explanatory model of how these variables
contribute together to patient safety outcomes. Furthermore, each of these studies has been
amongst doctors and no research has investigated these variables together in nursing staff.
Based on evidence that i) depressive symptoms increase risk of burnout (Ahola et al., 2005),
ii) burnout is the more consistent predictor of subsequently reported errors (West et al., 2006,
West et al., 2009) and ii) healthcare staff attribute errors to work-related stressors, fatigue and
burnout (Shanafelt et al., 2010), we propose that the two variables relate to patient safety
outcomes via a mediating model (see Figure 1a). This model does not suggest that depressive
symptoms result in burnout, but rather that it is the proportion of depressive symptoms that
overlap with burnout that is associated with patient safety perceptions (see Figure 1b). If this
is the case, it may suggest that while staff-level interventions for depressive symptoms, such
as mindfulness, could have patient safety benefits, interventions to enhance patient safety
may be better targeted at addressing burnout.
9
Figure 1
Theoretical framework
This work was grounded in Hobfoll’s Conversation of Resources model (COR)
(Hobfoll, 2001). This model suggests that a person experiences psychological stress: (1)
when their resources are threatened with loss; (2) when their resources are actually lost; and
(3) when they fail to gain anticipated resources following significant resource investment.
The term ‘resource’ loosely refers to commodities related to individual values and can allude
to personal, social and economic assets. Once an individual is experiencing psychological
distress, the model proposes that they will be more cautious with investing future resources.
10
In the healthcare literature, research using the COR has focused on the concept of
burnout as psychological distress and has investigated the relationship between burnout and
patient safety (Halbesleben et al., 2008, Welp et al., 2014). It has been suggested that once a
healthcare professional is experiencing burnout – a form of psychological stress – they may
be likely to ‘pull away’ from their patients and focus their efforts on specific aspects of their
job which they find enjoyable (Halbesleben et al., 2008). Halbesleben and colleagues (2008)
suggest burnout contributes to patient safety risks by diverting attention towards preferred
aspects of work and reducing extra behaviours that would benefit the organisation. The
current study tested the COR model in a sample of UK hospital nurses, investigating whether
the model’s predictions regarding an association between burnout and patient safety was
confirmed. Furthermore, we sought to extend this model by testing whether these findings
held when depressive symptoms were included as the form of psychological distress, rather
than burnout.
THE STUDY
Aims
The present study investigated the relationships between depressive symptoms,
burnout and patient safety in hospital nursing, taking a cross-sectional approach. In line with
previous research, the two burnout facets of Emotional Exhaustion and Depersonalisation
were included (Halbesleben et al., 2008). Safety perceptions were used to measure patient
safety, as previous research has suggested these are associated with objectively measured
patient safety outcomes (Hansen et al., 2011, Hofmann and Mark, 2006, Mardon et al., 2010)
and are useful for detecting variations in perceived safety between individual practitioners
(Halbesleben et al., 2008, Louch et al., 2016). There were two hypotheses:
11
Hypothesis 1: There will be a direct association between depressive symptoms and
patient safety perceptions.
Hypothesis 2: The two facets of burnout (Emotional Exhaustion and
Depersonalisation) will mediate the association between depressive symptoms and
patient safety perceptions.
Design
A cross-sectional questionnaire design was used.
Participants
A convenience sample of UK hospital nurses, midwives and healthcare assistants
were recruited from three acute NHS Trusts from December 2015 - February 2016.
Data collection
The study was advertised through posters displayed in hospital staff rooms.
Questionnaires were distributed on wards and participants completed these in a location and
at a time of their choice. Participants were informed of the date when questionnaires would
be collected and could return these to the researchers directly, or leave them in sealed
envelopes in drop-boxes in secure, locked rooms. Participants could choose to be entered into
a prize draw for shopping vouchers (20 prizes with a total value of £900).
Measures
Demographic information
Information regarding age, length of time in post, gender, qualifications and typical
hours worked per week were recorded in a questionnaire proforma.
12
Depressive symptoms
Depressive symptoms were measured using the Depression subscale of the 21-item
version of the Depression, Anxiety and Stress Scale (DASS-21) (Henry and Crawford, 2005).
This contains seven items including ‘I found it difficult to work up the initiative to do things’.
Participants rated the extent to which they experienced each item in the past week on a 4-
point scale from 0 (‘Did not apply to me at all’) - 3 (‘Applied to me very much, or most of the
time’). Possible scores ranged from 0-21.
Burnout
Burnout was measured using the Emotional Exhaustion and Depersonalisation
subscales of the Maslach Burnout Inventory (MBI) (Maslach and Jackson, 1981). The
Emotional Exhaustion subscale comprises nine items including ‘I feel emotionally drained
from my work’. The Depersonalisation subscale contains five items including ‘I feel I treat
some patients as if they were impersonal objects’. Items were rated on a 7-point scale from 0
(‘Never’) - 6 (‘Every day’). Possible scores ranged from 0-63 on the Emotional Exhaustion
subscale and 0 to 35 on the Depersonalisation subscale.
Patient safety perceptions
Perceptions were reported at both the individual level and the work area/unit level,
based on evidence that these measures provide useful and complementary information that
vary between nurses in response to stress and individual differences (Louch et al., 2016).
Individual safety perceptions. To assess perceptions of safety at the level of the
individual, participants completed the Safe Practitioner Measure (Louch et al., 2016), a one-
item statement (‘My practice is not as safe as it could be because of work related
factors/conditions’). The was rated from 1 (‘Strongly disagree’) - 5 (‘Strongly agree’) (Louch
13
et al., 2016). Responses were recoded such that higher scores indicated a more favourable
perception.
Work area/unit safety perceptions. Perceptions of patient safety at the work area/unit
level were measured using a four-item measure from the Agency for Healthcare Research and
Quality (AHRQ) Hospital Survey on Patient Safety Culture (HSOPC) (Sorra and Nieva,
2004). Items included, ‘It is just by chance that mistakes don't happen around here’ and were
scored from 1 (‘Strongly disagree’) 5 (‘Strongly agree’). Possible scores ranged from 0-20.
Higher scores indicated more favourable perceptions.
Ethical considerations
Participants were given an information sheet about the study and asked to provide
written informed consent. On collection of the questionnaires, identifiable information was
separated from non-identifiable questionnaire data and stored securely. The research was
approved by the University of Leeds, School of Psychology Ethics Committee (15-0183).
Validity, reliability and rigour
Depressive symptoms
The DASS-21 depression subscale (Henry and Crawford, 2005) demonstrated good
internal consistency in the current study (α = 0.87). Previous research indicates that the
subscale has convergent validity with longer depression measures (Norton, 2005).
Burnout
The MBI subscales (Maslach and Jackson, 1981) demonstrated acceptable internal
consistency in the current study (α = 0.83 for Emotional Exhaustion, α = 0.70 for
Depersonalisation).
14
Patient safety perceptions
Individual safety perceptions. Previous research has found convergent reliability
between this item and nurse-patient ratios and longer patient safety measures (Louch et al.,
2016).
Work area/unit safety perceptions. The AHRQ subscale (Sorra and Nieva, 2004)
demonstrated acceptable internal consistency in the current study (α = 0.74) and has
previously been established as an effective measure for gathering patient safety perceptions in
nursing staff (Halbesleben et al., 2008, Louch et al., 2016).
Data analysis
Preliminary data analysis was undertaken using SPSS version 22. Descriptive
statistics (means and standard deviations) and bivariate associations were conducted for study
variables and the demographic variables ‘length of time in current position’ and ‘typical hours
worked per week’. Spearman’s Rho correlations were performed, as the distributions of most
variables in the dataset did not conform to the assumptions of normality.
Missing data analyses were then undertaken for study variables. Rates of missing data
on individual measures ranged from 0% (safe practitioner measure) to 10.2% (Emotional
Exhaustion subscale). In total, 15.48% of cases were missing data for one or more measures.
Little’s chi-square statistic for testing whether values are missing completely at random
(MCAR) (Little, 1988) was not significant (x=30.06, df=31, p=0.51), indicating there was no
systematic pattern to the missing data. When data is not missing systematically and overall
rates of missing data are <20%, data imputation is preferred over case deletion, as this
preserves statistical power (Garson, 2015). Data imputation was conducted using regression
imputation in AMOS 22. This method first fits the proposed statistical model using maximum
15
likelihood and model parameters are set as equal to their maximum likelihood estimates.
Linear regression is then used to predict each missing value as a linear combination of the
observed values for that same case. Predicted values are imputed for missing values
(Arbuckle, 2013).
To examine whether the associations between depressive symptoms and patient safety
perceptions were mediated by the two burnout subscales (Emotional Exhaustion and
Depersonalisation), Structural Equal Modelling (SEM) was undertaken using AMOS 22. This
approach has three main benefits over the ‘causal steps’ hierarchical regression approach
outlined by Baron and Kenny (1986). First, it has greater flexibility and enables the testing of
multiple mediators in one model (Preacher and Hayes, 2008). Second, measurement errors
can bias the parameter estimates for mediation resulting from the hierarchical regression
approach, but these can be controlled for in the SEM approach (Cheung and Lau, 2007).
Third, the AMOS 22 programme enables hypothesis tests to be conducted using the
bootstrapping method. Bootstrapping is a nonparametric resampling procedure which does
not require data distributions to conform to the assumptions of parametric tests (e.g.,
normality). Bootstrapping provides the most powerful method for obtaining confidence limits
for mediation effects (Preacher and Hayes, 2008) and Cheung and Lau (2007) suggest that a
sample size of 100 provides adequate power to detect a medium to large effect size.
We used the bootstrapping approach (5000 bootstrap samples; 95% confidence
interval) to test six models. All models controlled for age, gender and length of time in
current position. The first model tested the direct association between depressive symptoms
and safety perceptions at the individual level and the second model repeated this but instead
used patient safety perceptions at the work area/unit level. The second two models (one for
each outcome measure) then tested whether any direct association was mediated by the two
burnout facets (Emotional Exhaustion and Depersonalisation; Figure 1). For completeness,
16
the final two models then tested for the opposite mediation pattern, whereby depressive
symptoms mediated the association between the two burnout facets (Emotional Exhaustion
and Depersonalisation) and each patient safety outcome variable. In line with
recommendations by Cheung and Lau (2007), bias-corrected bootstrap confidence intervals
were calculated and used to test hypotheses.
We also calculated fit statistics for each model. The use of model fit indices in studies
of mediation using SEM is much debated, as proposed cut-off values have often been based
on research using simulated models with a large number of degrees of freedom (df) (Kenny et
al., 2014). The commonly used RMSEA statistic is particularly unreliable in the context of
small df models, such as those used to test for mediation. In line with recommendations by
Kenny and colleagues (Kenny et al., 2014, Kenny, 2016), we reported the chi-squared
statistic (X2), the Comparative Fit Index (CFI), where values above 0.9 were considered
acceptable and the Akaike Information Criterion (AIC). There are no recommended cut-offs
for the AIC, but lower values represent better fit.
RESULTS
Participant characteristics
Three-hundred and twenty-three eligible participants completed the
questionnaires (M age=39.79, SD= 12.31, 92.3% female). Ninety-one percent of participants
were white, 1.9% were Afro-Carribean, 2.2% were Asian, 3.1% were of another ethnicity,
1.5% preferred not to state their ethnicity and data were missing for 0.3% of participants. For
education, 4% of participants reported a Masters degree as their highest qualification, 48.9%
reported a bachelors degree, 18.6% reported A-Levels or equivalent, 16.1% reported GCSEs
or equivalent, 7.4% had completed some secondary school and 4.6% reported ‘other’. Data
were missing for 0.3% participants. In terms of staff position in the hospital, 42.1% of
17
participants were staff nurses, 31.9% were healthcare assistants, 13.6% were ward
sisters/charge nurses, 4% were ward managers, 3.7% were midwives, 2.5% were agency staff
and 1.9% were student nurses.
Descriptive statistics
Descriptive statistics for study variables are presented in Table 1. Spearman’s Rho
correlations indicated significant positive associations between both the Safe Practitioner
measure and the AHRQ with depressive symptoms (rs=-0.28, p<0.001 and rs=-0.34, p<0.001,
respectively), Emotional Exhaustion (rs=-0.41, p<0.001 and rs=-0.51, p<0.001, respectively)
and Depersonalisation (rs= -0.43, p<0.001 and rs=-0.45, p<0.001, respectively). There were
also significant positive associations between hours worked with Emotional Exhaustion
(rs=0.15*, p=0.01) and Depersonalisation (rs=0.16, p=0.005), but not depressive symptoms
(rs=0.05, p=0.42).
Table : Means, Standard deviations a and correlations for variables
Mean
2 3 4 5 6 7
1.Length of time in
current position
(months)
96.05
95.67
-.26*** -.02 -.14* .08 .02 -.06
2.Typical hours per
week
35.39
20.07
.02 -.05 .05 .15* .16**
3. Individual-level
safety (Safe
3.11 .48*** -.28*** -.41*** -.43***
18
practitioner measure) 1.19
4. Work area/unit
level safety (AHRQ
subscale)
12.49
3.73
-.34*** -.51*** -.45***
5.Depression (DASS
subscale)
4.27
4.30
.67*** .52***
6.MBI Emotional
Exhaustion
26.08
13.00
.60***
7. MBI
Depersonalisation
6.50
6.00
Note. *p<0.05, **p<.01, ***p<0.001. AHRQ = Agency for Healthcare Research and Quality; DASS =
Depression, Anxiety and Stress Scale; MBI = Maslach Burnout Inventory. aStandard deviations appear in
italics below the means.
Structural Equation Models
We then constructed a series of six structural equation models to test our mediation
hypotheses. We used bootstrapping, a nonparametric resampling procedure which provides
the most powerful method of obtaining confidence limits for mediation effects.
Depressive symptoms and patient safety perceptions
When depressive symptoms were the only predictor variable in the models, there was
a direct association between these and individual level safety perceptions (c = -0.277, CI=-
0.364,-0.181, p =0.001; X2(3)=2.537, p=0.469, CFI=1, AIC=36.537; Figure 2) and work
area/unit level safety perceptions (c=-0.303, CI=-0.404,-0.193, p<0.001; X2(3)=9.304,
p=0.026, CFI=0.951, AIC=43.304; Figure 3).
19
Figure 2
20
Figure 3
Burnout as a mediator of the association between depressive symptoms and patient
safety perceptions
When the burnout facets (Emotional Exhaustion and Depersonalisation) were
included in the model with individual safety perceptions as the outcome variable (Figure 2B),
the direct association between depressive symptoms and safety ceased to be significant
(c’=0.027, CI=-0.103, 0.163, p=0.730). The direct associations between depressive symptoms
and Emotional Exhaustion (a1=0.602, CI=0.495, 0.685, p=0.001) and Depersonalisation (a2=
0.492, CI=0.402, 0.576, p <0.001) were significant, as were the direct associations between
Emotional Exhaustion and Depersonalisation and individual safety (b1=-0.228, CI=-0.403,
-0.075, p=0.002 and b2=-0.356, CI=-0.465,-0.246, p <0.001, respectively). These results
indicated that the two facets of burnout fully mediated the association between depressive
symptoms and perceived patient safety. Model fit indices were X2(10)=58.643, p<0.001,
CFI=0.902, AIC=108.643.
Similar results were found when the model was tested with work area/unit level safety
perceptions as the outcome variable (Figure 3B). With Emotional Exhaustion and
Depersonalisation included as mediators, the direct association between depressive symptoms
and safety ceased to be significant (c’=0.065, CI=-0.074, 0.219, p=0.411). The direct
associations between depressive symptoms and Emotional Exhaustion (a1=0.598, CI=0.486,
0.679, p=0.001) and Depersonalisation (a2= 0.485, CI=0.393, 0.569, p<0.001) were
significant, as were the direct associations between Emotional Exhaustion and
Depersonalisation and work area/unit level safety perceptions (b1=-0.411, CI=-0.554, -0.294,
p=0.001 and b2=-0.267, CI=-0.387, -0.145, p<0.001, respectively). Again, these results
21
indicated that the two facets of burnout fully mediated the association between depressive
symptoms and perceived patient safety. Model fit indices were X2(10)=66.146, p<0.001,
CFI=0.893, AIC=116.146.
Testing depressive symptoms as a mediator of the association between burnout and
patient safety perceptions
For completeness, we then repeated these two models but reversed the mediation
pattern, in effect testing whether the association between the burnout facets (Emotional
Exhaustion and Depersonalisation) and the patient safety outcome variables ceased to be
significant once depressive symptoms were included as a mediator in the models. Neither of
these found evidence for the reverse mediation pattern.
In the model testing individual patient safety perceptions as the outcome variable
(Figure 4A), Emotional Exhaustion and Depersonalisation were both significantly associated
with depressive symptoms (a1=0.482, CI=0.324, 0.623, p=0.001 and a2=0.206, CI=0.059,
0.355, p=0.006). Emotional Exhaustion and Depersonalisation were also both significantly
associated with patient safety perceptions (c’1=-0.224, CI=-0.390, -0.075, p=0.002 and c’2=-
0.348, CI=-0.455, -0.240, p<0.001). However, depressive symptoms were not directly
associated with patient safety perceptions, suggesting they were not a mediator (b=0.025, CI=
-0.105, 0.156, p=0.767).
22
Figure 4
In the model testing patient safety perceptions at the area/unit level safety as the
outcome (Figure 4B), Emotional Exhaustion and Depersonalisation were both significantly
associated with depressive symptoms (a1=0.480, CI=0.322, 0.620, p=0.001 and a2=0.209,
CI=0.062, 0.356, p=0.004). Emotional Exhaustion and Depersonalisation were also both
significantly associated with patient safety perceptions (c’1=-0.392, CI=-0.532, -0.281,
p=0.001 and c’2=-0.255, CI=-0.371, -0.135, p<0.001). However, depressive symptoms were
not directly associated with patient safety perceptions, suggesting they were not a mediating
variable (b=0.058, CI=-0.076, 0.205, p=0.449).
23
Interestingly, the fit indices for these models were better than the initial models
(X2(6)=5.605, p<0.469, CFI=1, AIC=63.605 for individual patient safety perceptions and
X2(6)=18.679, p=0.005, CFI=0.975, AIC=76.679 for ward level perceptions).
DISCUSSION
This paper presents findings from a cross-sectional questionnaire survey in UK
hospital nurses from multiple clinical areas, across three acute NHS Trusts. The results
suggested that there was a direct association between depressive symptoms and safety
perceptions at both the individual and area level. However, this association was fully
mediated by two facets of burnout, Emotional Exhaustion and Depersonalisation.
Interestingly, the best overall model fit was found when burnout variables were modelled as
predictors in the models, with depressive symptoms modelled as an outcome of burnout.
However, no evidence was found for the reverse mediation pattern (where depressive
symptoms mediated the association between burnout and patient safety perceptions).These
findings add to the literature in three main ways.
First, these results support previous research indicating an association between
burnout and patient safety in medical staff (Fahrenkopf et al., 2008, West et al., 2006) and
extend this by finding the first evidence for this link amongst UK hospital nursing staff. Only
a small number of studies have investigated whether burnout is associated with patient safety
perceptions amongst nurses in any nation (Halbesleben et al., 2008, Laschinger and Leiter,
2006) and one previous study by Holden and colleagues reported no evidence for this link
(Holden et al., 2011). Holden et al.’s study measured patient safety perceptions using a single
item asking about perceived risk of medication error (Holden et al., 2011), whereas the
present study measured patient safety using the Safe Practitioner Scale (SPS) (Louch et al.,
2016) and the AHRQ Perceptions of Safety Subscale (AHRQ PSS) (Sorra and Nieva, 2004)
24
and found evidence of this association according to both measures. It is possible that the
divergence in results between the present study and the one by Holden and colleagues (2011)
is owing to the measure of patient safety perceptions used, suggesting that the SPS and
AHRQ PSS may be more sensitive measures suitable for future research in this area.
Second, these results provide the first evidence of an association between depressive
symptoms and perceptions of patient safety in UK hospital nurses. Only a limited number of
studies have previously investigated this and have focused on number of perceived errors
(Tanaka et al., 2012, Saleh et al., 2014). The current study finds further support for a link
between depressive symptoms and safety perceptions in nurses, suggesting this is also present
in UK nurses. Furthermore, by measuring safety perceptions using the SPS and AHRQ PSS,
the current study suggests that this link generalises to both individual safety perceptions and
perceptions at the area level.
Third, this study proposed and tested the first explanatory model of the relationships
between burnout, depressive symptoms and patient safety perceptions and found evidence
that the association between depressive symptoms and safety is explained by the overlap
between depressive symptoms and burnout. Few studies have investigated both burnout and
symptoms of depression together (Fahrenkopf et al., 2008, Dyrbye et al., 2013, de Oliveira Jr
et al., 2013, Garrouste-Orgeas et al., 2015) and none have proposed mechanisms for how
these variables contribute together to patient safety perceptions. Furthermore, each of these
studies was amongst medical staff. Results from the present study support previous findings
that both variables are separately associated with patient safety perceptions and extend this to
suggest that this association is also present in nursing staff groups. Importantly, the current
study suggests that burnout may be the more important variable in relation to patient safety.
Not only was the association between depressive symptoms and patient safety perceptions
entirely mediated by burnout, but the best statistical model fit was found when depressive
25
symptoms were considered an outcome of burnout, rather than a predictor of burnout and
patient safety perceptions. However, it should be noted that as a cross-sectional study, it is not
possible to draw conclusions regarding cause and effect and the associations described could
be bi-directional.
Theoretical framework
The study was grounded in the Conversation of Resources model (COR) (Hobfoll,
2001), which suggests that once a person is experiencing psychological distress, they will
become cautious regarding resource investment. Theorists have proposed that burnout is a
form of psychological stress (Halbesleben et al., 2008, Welp et al., 2014) and have used the
model to explain the links between burnout and patient safety perceptions. In particular, it has
been suggested that once a healthcare staff member is experiencing burnout, they may stop
investing their personal resources by ‘pulling away’ from their patients, instead focusing their
efforts only on aspects of their job which they enjoy (Halbesleben et al., 2008). The current
study tested the predictions of this model in a UK nursing sample and found evidence that as
the model would suggest, both forms of psychological distress measured (depressive
symptoms and burnout) were linked with patient safety perceptions. The current study also
extends the model by finding evidence that work-related psychological distress (i.e., burnout)
may be the more important variable leading to inhibition of resource investment amongst
nurses, explaining the association between more general psychological distress (i.e.,
depressive symptoms) and safety perceptions.
Limitations
The study was limited by its cross-sectional design, which prevents the extent to
which any causal interpretations can be made regarding the data. Previous longitudinal
research has indicated that the relationship between depressive symptoms, burnout and errors
26
may be cyclical, with depressive symptoms or burnout increasing risk for subsequent errors
and involvement in errors leading to increased risk of depressive symptoms or burnout (West
et al., 2006, West et al., 2009). In the current study we were interested in understanding
patient safety as the outcome variable, but it is possible that the relationships between each of
the variables studied may be bi-directional.
The study was also limited by reliance on self-report measures, which may have led to
bias in responses. However, each of the measures was carefully selected and had been
previously validated as suitable self-report tools. Furthermore, there are no objective rating
tools for burnout and collecting accurate objective patient safety outcome data for nurses at
the individual level is rarely possible, owing to the team-based nature of nursing work.
Previous research has suggested that safety perceptions are associated with objectively
measured patient safety outcomes (Mardon et al., 2010, Hofmann and Mark, 2006, Hansen et
al., 2011) and are useful for capturing individual-level variations in patient safety
(Halbesleben et al., 2008, Louch et al., 2016).
Finally, as questionnaires were not addressed to individual nurses, the study was
limited by a lack of response rate information. However, as questionnaires were distributed to
three acute NHS Trusts, our sample will only reflect a minority of eligible participants, which
may have led to a higher rate of extreme responders (e.g., participants particularly high or
low on measures of depressive symptoms/burnout). Whilst conclusions from inferential
statistics are likely to be robust to this bias, this limits the interpretation of overall
burnout/depressive symptoms reported.
Implications
There is an increasing recognition of the importance of healthcare staff wellbeing in
contributing to patient outcomes (NHS, 2013, DoH, 2013). However, the concept of
27
wellbeing has been poorly defined and as such, it has been unclear where interventions to
support and promote staff wellbeing should be focused. Due to the limited sample size, the
current results must be interpreted with caution. However, they offer the tentative suggestion
staff burnout, rather than depressive symptoms, may be particularly important in relation to
patient safety and interventions focused on reducing burnout may be beneficial for patient
safety outcomes. Burnout interventions are thought to be most effective when they not only
address staff stress via individual-level interventions, for example staff counselling, but when
they also address aspects of the workplace (Awa et al., 2010). Such interventions might
consider restructuring work processes, adjusting shifts or implementing or changing
performance appraisals (Awa et al., 2010, West and Dawson, 2012). These types of
interventions require organisational involvement and can be challenging to deliver, however
the current research suggests they may be more beneficial for patient safety than those
targeted only at addressing personal staff stress levels.
Future research
Longitudinal research is necessary to understand the relationships between depressive
symptoms and burnout on patient safety outcomes over time. Previous research in doctors
suggests that the relationship may be cyclical (West et al., 2006, West et al., 2009), but further
research is needed to explore this in a nursing population. Research is also needed to test
whether interventions targeting staff burnout are effective for burnout reduction and lead to
concomitant improvements in staff safety perceptions. No studies have yet tested this in
either medical or nursing populations.
CONCLUSION
When tested in separate analyses, both depressive symptoms and burnout were
associated with patient safety perceptions amongst UK hospital nurses. Furthermore, the
28
association between depressive symptoms and perceptions of patient safety was fully
mediated by burnout. These findings suggest that nurse wellbeing may have implications for
patient safety and that interventions focused on addressing burnout, rather than symptoms of
depression, may be most effective for patient safety improvements.
29
CONFLICT OF INTEREST STATEMENT
No conflict of interest has been declared by the authors.
30
REFERENCES
AHOLA, K., HONKONEN, T., ISOMETSÄ, E., KALIMO, R., NYKYRI, E., AROMAA, A.
& LÖNNQVIST, J. 2005. The relationship between job-related burnout and
depressive disorders—results from the Finnish Health 2000 Study. Journal of
Affective Disorders, 88, 55-62.
ALLAN, J. L., FARQUHARSON, B., JOHNSTON, D. W., JONES, M. C., CHOUDHARY,
C. J. & JOHNSTON, M. 2014. Stress in telephone helpline nurses is associated with
failures of concentration, attention and memory and with more conservative referral
decisions. British Journal of Psychology, 105, 200-213.
ARBUCKLE, J. L. 2013. IBM® SPSS® Amos™ 22 User’s Guide, IBM Corp.
AWA, W. L., PLAUMANN, M. & WALTER, U. 2010. Burnout prevention: A review of
intervention programs. Patient Education and Counseling, 78, 184-190.
BAINES, R. J., LANGELAAN, M., DE BRUIJNE, M. C., ASSCHEMAN, H.,
SPREEUWENBERG, P., VAN DE STEEG, L., SIEMERINK, K. M., VAN ROSSE,
F., BROEKENS, M. & WAGNER, C. 2013. Changes in adverse event rates in
hospitals over time: a longitudinal retrospective patient record review study. BMJ
Quality & Safety, 22, 290-298.
BARON, R. M. & KENNY, D. A. 1986. The moderator–mediator variable distinction in
social psychological research: Conceptual, strategic and statistical considerations.
Journal of Personality and Social Psychology, 51, 1173.
CHEUNG, G. W. & LAU, R. S. 2007. Testing mediation and suppression effects of latent
variables: Bootstrapping with structural equation models. Organizational Research
Methods.
DE OLIVEIRA JR, G. S., CHANG, R., FITZGERALD, P. C., ALMEIDA, M. D., CASTRO-
ALVES, L. S., AHMAD, S. & MCCARTHY, R. J. 2013. The prevalence of burnout
31
and depression and their association with adherence to safety and practice standards: a
survey of United States anesthesiology trainees. Anesthesia & Analgesia, 117, 182-
193.
DEMEROUTI, E. & BAKKER, A. B. 2008. The Oldenburg Burnout Inventory: A good
alternative to measure burnout and engagement. In: HALBESLEBEN, J. R. B. (ed.)
Handbook of Stress and Burnout in Health Care. 0. Hauppauge, NY: Nova Science.
DEMEROUTI, E., BAKKER, A. B., NACHREINER, F. & SCHAUFELI, W. B. 2001. The
job demands-resources model of burnout. Journal of Applied Psychology, 86, 499.
DOH 2013. A promise to learn– a commitment to act: Improving the Safety of Patients in
England. In: HEALTH, D. O. (ed.).
DYRBYE, L. N., SATELE, D., SLOAN, J. & SHANAFELT, T. D. 2013. Utility of a brief
screening tool to identify physicians in distress. Journal of General Internal
Medicine, 28, 421-427.
FAHRENKOPF, A. M., SECTISH, T. C., BARGER, L. K., SHAREK, P. J., LEWIN, D.,
CHIANG, V. W., EDWARDS, S., WIEDERMANN, B. L. & LANDRIGAN, C. P.
2008. Rates of medication errors among depressed and burnt out residents:
prospective cohort study. British Medical Journal, 336, 488-491.
GARROSA, E., MORENO-JIMENEZ, B., LIANG, Y. & GONZÁLEZ, J. L. 2008. The
relationship between socio-demographic variables, job stressors, burnout and hardy
personality in nurses: An exploratory study. International journal of nursing studies,
45, 418-427.
GARROUSTE-ORGEAS, M., PERRIN, M., SOUFIR, L., VESIN, A., BLOT, F., MAXIME,
V., BEURET, P., TROCHÉ, G., KLOUCHE, K. & ARGAUD, L. 2015. The Iatroref
study: medical errors are associated with symptoms of depression in ICU staff but not
burnout or safety culture. Intensive Care Medicine, 41, 273-284.
32
GARSON, D. G. 2015. Missing Values Analysis and Data Imputation, Asheboro, USA,
Statistical Associates Publishing.
HALBESLEBEN, J. R., WAKEFIELD, B. J., WAKEFIELD, D. S. & COOPER, L. B. 2008.
Nurse burnout and patient safety outcomes: nurse safety perception versus reporting
behavior. Western Journal of Nursing Research, 30, 560-577.
HALL, L. H., JOHNSON, J., WATT, I., TSIPA, A. & O’CONNOR, D. B. 2016. Healthcare
Staff Wellbeing, Burnout and Patient Safety: A Systematic Review. PloS One, 11,
e0159015.
HANKIN, B. L., FRALEY, R. C., LAHEY, B. B. & WALDMAN, I. D. 2005. Is depression
best viewed as a continuum or discrete category? A taxometric analysis of childhood
and adolescent depression in a population-based sample. Journal of abnormal
psychology, 114, 96.
HANSEN, L. O., WILLIAMS, M. V. & SINGER, S. J. 2011. Perceptions of hospital safety
climate and incidence of readmission. Health Services Research, 46, 596-616.
HENRY, J. D. & CRAWFORD, J. R. 2005. The short-form version of the Depression Anxiety
Stress Scales (DASS-21): Construct validity and normative data in a large non-
clinical sample. British Journal of Clinical Psychology, 44, 227-239.
HOBFOLL, S. E. 2001. The influence of culture, community and the nested‐self in the stress
process: advancing conservation of resources theory. Applied Psychology, 50, 337-
421.
HOFMANN, D. A. & MARK, B. 2006. An investigation of the relationship between safety
climate and medication errors as well as other nurse and patient outcomes. Personnel
Psychology, 59, 847-869.
HOLDEN, R. J., SCANLON, M. C., PATEL, N. R., KAUSHAL, R., ESCOTO, K. H.,
BROWN, R. L., ALPER, S. J., ARNOLD, J. M., SHALABY, T. M. &
33
MURKOWSKI, K. 2011. A human factors framework and study of the effect of
nursing workload on patient safety and employee quality of working life. BMJ
Quality & Safety, 20, 15-24.
HUGHES, G. 2008. Nurses at the ‘Sharp End’ of Patient Care. In: HUGHES, G. (ed.) Patient
Safety and Quality: An Evidence-Based Handbook for Nurses. Rockville (MD):
Agency for Healthcare Research and Quality (US).
IACOVIDES, A., FOUNTOULAKIS, K., KAPRINIS, S. & KAPRINIS, G. 2003. The
relationship between job stress, burnout and clinical depression. Journal of Affective
Disorders, 75, 209-221.
IVES, J., SHAER, D., SHERRING, S. & MARKS-MARAN, D. 2015. Increasing burnout
among nurses: How can nurse managers respond? British Journal of Mental Health
Nursing, 4, 288-293.
KENNY, D. A. 2016. Measuring model fit [Online]. Available:
http://davidakenny.net/cm/fit.htm [Accessed 20th September 2016.
KENNY, D. A., KANISKAN, B. & MCCOACH, D. B. 2014. The performance of RMSEA in
models with small degrees of freedom. Sociological Methods & Research,
0049124114543236.
LASCHINGER, H. K. S. & LEITER, M. P. 2006. The impact of nursing work environments
on patient safety outcomes: The mediating role of burnout engagement. Journal of
Nursing Administration, 36, 259-267.
LEITER, M. P., HARVIE, P. & FRIZZELL, C. 1998. The correspondence of patient
satisfaction and nurse burnout. Social Science & Medicine, 47, 1611-1617.
LETVAK, S., RUHM, C. J. & MCCOY, T. 2012. Depression in hospital-employed nurses.
Clinical Nurse Specialist, 26, 177-182.
34
LITTLE, R. J. 1988. A test of missing completely at random for multivariate data with
missing values. Journal of the American Statistical Association, 83, 1198-1202.
LOUCH, G., O’HARA, J., GARDNER, P. & O’CONNOR, D. B. 2016. The daily
relationships between staffing, safety perceptions and personality in hospital nursing:
A longitudinal on-line diary study. International Journal of Nursing Studies, 59, 27-
37.
MARDON, R. E., KHANNA, K., SORRA, J., DYER, N. & FAMOLARO, T. 2010.
Exploring relationships between hospital patient safety culture and adverse events.
Journal of Patient Safety, 6, 226-232.
MASLACH, C. & JACKSON, S. E. 1981. The measurement of experienced burnout. Journal
of Occupational Behavior, 2, 99-113.
MASLACH, C., SCHAUFELI, W. B. & LEITER, M. P. 2001. Job burnout. Annual review of
psychology, 52, 397-422.
NHS 2013. Report of the Mid Staffordshire NHS Foundation Trust Public Inquiry. London:
The Stationery Office.
NORTON, P. J. 2005. A psychometric analysis of the Intolerance of Uncertainty Scale among
four racial groups. Journal of Anxiety Disorders, 19, 699-707.
PASSALACQUA, S. A. & SEGRIN, C. 2012. The effect of resident physician stress, burnout
and empathy on patient-centered communication during the long-call shift. Health
Communication, 27, 449-456.
PREACHER, K. J. & HAYES, A. F. 2008. Asymptotic and resampling strategies for
assessing and comparing indirect effects in multiple mediator models. Behavior
Research Methods, 40, 879-891.
SALEH, A. M., AWADALLA, N. J., EL-MASRI, Y. M. & SLEEM, W. F. 2014. Impacts of
nurses’ circadian rhythm sleep disorders, fatigue and depression on medication
35
administration errors. Egyptian Journal of Chest Diseases and Tuberculosis, 63, 145-
153.
SHANAFELT, T. D., BALCH, C. M., BECHAMPS, G., RUSSELL, T., DYRBYE, L.,
SATELE, D., COLLICOTT, P., NOVOTNY, P. J., SLOAN, J. & FREISCHLAG, J.
2010. Burnout and medical errors among American surgeons. Annals of Surgery, 251,
995-1000.
SHANAFELT, T. D., HASAN, O., DYRBYE, L. N., SINSKY, C., SATELE, D., SLOAN, J.
& WEST, C. P. Changes in burnout and satisfaction with work-life balance in
physicians and the general US working population between 2011 and 2014. Mayo
Clinic Proceedings, 2015. Elsevier, 1600-1613.
SORRA, J. S. & NIEVA, V. F. 2004. Hospital survey on patient safety culture (prepared by
Westat, under Contract No. 290-96-0004; AHRQ Publication No. 04-0041),
Rockville, MD, Agency for Healthcare Research and Quality.
SU, J. A., WENG, H. H., TSANG, H. Y. & WU, J. L. 2009. Mental health and quality of life
among doctors, nurses and other hospital staff. Stress and Health, 25, 423-430.
TANAKA, M., TANAKA, K., TAKANO, T., KATO, N., WATANABE, M. & MIYAOKA, H.
2012. Analysis of risk of medical errors using structural-equation modelling: a 6-
month prospective cohort study. BMJ Quality & Safety, 21, 784-790.
THUYNSMA, C. & DE BEER, L. T. 2016. Burnout, depressive symptoms, job demands and
satisfaction with life: discriminant validity and explained variance. South African
Journal of Psychology, 0081246316638564.
VAHEY, D. C., AIKEN, L. H., SLOANE, D. M., CLARKE, S. P. & VARGAS, D. 2004.
Nurse Burnout and Patient Satisfaction. Medical Care, 42, II57-II66.
36
VINCENT, C., NEALE, G. & WOLOSHYNOWYCH, M. 2001. Adverse events in British
hospitals: preliminary retrospective record review. British Medical Journal, 322, 517-
519.
WALL, T. D., BOLDEN, R., BORRILL, C., CARTER, A., GOLYA, D., HARDY, G.,
HAYNES, C., RICK, J., SHAPIRO, D. & WEST, M. 1997. Minor psychiatric
disorder in NHS trust staff: occupational and gender differences. The British Journal
of Psychiatry, 171, 519-523.
WELP, A., MEIER, L. L. & MANSER, T. 2014. Emotional exhaustion and workload predict
clinician-rated and objective patient safety. Frontiers in Psychology, 5.
WEST, C. P., HUSCHKA, M. M., NOVOTNY, P. J., SLOAN, J. A., KOLARS, J. C.,
HABERMANN, T. M. & SHANAFELT, T. D. 2006. Association of perceived medical
errors with resident distress and empathy: a prospective longitudinal study. JAMA,
296, 1071-1078.
WEST, C. P., TAN, A. D., HABERMANN, T. M., SLOAN, J. A. & SHANAFELT, T. D.
2009. Association of resident fatigue and distress with perceived medical errors.
JAMA, 302, 1294-1300.
WEST, M. & DAWSON, J. 2012. Employee engagement and NHS performance. The King's
Fund.
WOOD, A. M., TAYLOR, P. J. & JOSEPH, S. 2010. Does the CES-D measure a continuum
from depression to happiness? Comparing substantive and artifactual models.
Psychiatry Research, 177, 120-123.
