Summary This dissertation presents a cross-sectional survey exploring burnout in care staff working in dementia-registered residential homes in Cardiff. The aging population and anticipated prevalence of dementia makes this exploration of particular relevance. Stress and burnout among care staff working with people with dementia can result in greater sickness and worse care outcomes. This survey analyses the responses of 163 staff to questions on dementia knowledge, attitudes and psychosocial stressors. The data are explored to reveal the underlying themes and concepts that are particular to this population. These concepts are used to produce a model of burnout that has 3 facets, ‘Physical and Emotional Burnout’, ‘Work Burnout’ and ‘Resident Burnout’ and all 3 have ‘Stress’ as their core component. In addition, ‘Physical and Emotional Burnout’ is mediated through ‘Hopeful’ attitudes, while ‘Work Burnout’ and Resident Burnout’ are mediated through ‘Professional’ values. Further variables associated with ‘Physical and Emotional Burnout’, include exposure to ‘Physical Violence’ in work and less ‘Time in Profession’. ‘Work Burnout’ was also associated with less ‘Time in Profession’ and ‘Resident Burnout’ was negatively associated with having ‘British Ethnicity/Nationality’. The results of this survey indicate that stress is a central component to burnout as found in previous research. The factors found to mediate burnout also reflect current research on organisational engagement and positive psychological states. The associations of ‘Time in Profession’, exposure to ‘Physical Violence’ and ‘British Ethnicity/Nationality’ are also relevant to this population and would benefit from further study to explore potential confounders. Further research into burnout in this population, would benefit from assessment of the direction of causality for the above associations and this could be of use in evaluating interventions to improve these working environments. Burnout remains an important concept to understand to improve the lives of both care staff and people living with dementia.
223
Embed
Summary - Amazon S3 to Dementia ... including some of the difficulties experienced through caring for people ... number of different diseases that affect the ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Summary
This dissertation presents a cross-sectional survey exploring burnout in care staff working in dementia-registered residential homes in Cardiff. The aging population and anticipated prevalence of dementia makes this exploration of particular relevance. Stress and burnout among care staff working with people with dementia can result in greater sickness and worse care outcomes.
This survey analyses the responses of 163 staff to questions on dementia knowledge, attitudes and psychosocial stressors. The data are explored to reveal the underlying themes and concepts that are particular to this population.
These concepts are used to produce a model of burnout that has 3 facets, ‘Physical and Emotional Burnout’, ‘Work Burnout’ and ‘Resident Burnout’ and all 3 have ‘Stress’ as their core component. In addition, ‘Physical and Emotional Burnout’ is mediated through ‘Hopeful’ attitudes, while ‘Work Burnout’ and Resident Burnout’ are mediated through ‘Professional’ values.
Further variables associated with ‘Physical and Emotional Burnout’, include exposure to ‘Physical Violence’ in work and less ‘Time in Profession’. ‘Work Burnout’ was also associated with less ‘Time in Profession’ and ‘Resident Burnout’ was negatively associated with having ‘British Ethnicity/Nationality’.
The results of this survey indicate that stress is a central component to burnout as found in previous research. The factors found to mediate burnout also reflect current research on organisational engagement and positive psychological states.
The associations of ‘Time in Profession’, exposure to ‘Physical Violence’ and ‘British Ethnicity/Nationality’ are also relevant to this population and would benefit from further study to explore potential confounders.
Further research into burnout in this population, would benefit from assessment of the direction of causality for the above associations and this could be of use in evaluating interventions to improve these working environments.
Burnout remains an important concept to understand to improve the lives of both care staff and people living with dementia.
Dissertation
An Exploration into Burnout in Care Staff Working in
Dementia-Registered Residential Homes in Cardiff
Submitted by
David Mark Howells
2013
for
MSc Ageing, Health and Disease
Cardiff University
Cardiff, Wales
United Kingdom
Acknowledgement This dissertation has been the culmination of work that has involved a great number of people who have unconditionally offered their time, support and expertise. The origins of this work developed from the ‘Enhanced Dementia Care Project’, a Cardiff County Council-run scheme, funded through a ‘Promoting Independence and Well-being’ grant from the Welsh Assembly Government. Thanks should be given here to Jude Viney and her team for having the faith in my ability to evaluate the project and for providing support and guidance at every stage of the process of the survey development. Thanks should also be made to the project manager, Becky Vangasse for responding to my numerous e-mails on the minutiae of the questionnaire and care home engagement. The entire project team worked tirelessly to make the scheme a success and should be congratulated for maintaining their enthusiasm throughout. Thanks should also go out to the numerous key staff from the care homes that were enthusiastic for improving care for their residents and helped to design the questionnaire and distribute to their colleagues. Thanks also to those care staff responding to the survey, without whom, this evaluation would not have been possible. I hope this work helps you to understand your important role and the need to support each other in your very stressful occupation. Thanks to Dr Marion Gray and Dr Rhiannon Callaghan for their support and the opportunities to develop this project over more years than was originally envisaged! Thanks to Dr Stanley Zammit and Dr John Gallacher for their assistance in illuminating the dark arts of statistical analysis when this was most needed! A great deal of thanks go to my dissertation supervisor, Professor Antony Bayer who has given advice and clarity of thought on the project and provided direction for making the most of this experience. I would also like to thank Dr Win Tadd for suggesting this MSc course 5 years ago, while acting as celebrant for my daughter, Ariana’s naming ceremony. A big thank you to Ariana, who has ensured that my reading has not been too narrow over the course of the last few years. Also a special thank you to my wife, Samantha who has been a far greater support than either of us anticipated was needed and deserves special recognition for her wisdom and humour through a seemingly endless process. I do not know how I will repay you!
Dedication
To the centre of my world, Sam and Ari.
Contents
Chapter 1: Introduction 1
Chapter 2: Background 2
Introduction 2
Dementia 2
Dementia Demographics 3
Care Homes 4
Care Home Population of Older People 4
People Living with Dementia in Care Homes 4
Dementia Registered Care Homes 5
Difficulties in Care 6
Care Staff 7
Care Staff Characteristics 7
Care Staff Characteristics in Cardiff 9
Burnout 10
Consequences of Burnout 11
Factors Associated with Burnout 12
Demographics 12
Personality 14
Mental Health 15
Offensive Behaviour 16
Attitudes 17
Knowledge 18
Psychosocial/Organisational Factors 20
Preventing Burnout 23
Background Summary 25
Chapter 3: Aims and Research Hypothesis 26
Aims of Study 26
Research Hypothesis 28
Aims and Research Hypothesis Summary 28
Chapter 4: Methodology 29
Introduction 29
Design 29
Participants 30
Measures 30
Demographic Information 31
Copenhagen Burnout Inventory 31
Dementia Knowledge Questionnaire 32
Approaches to Dementia Questionnaire 33
Copenhagen Psychosocial Questionnaire 34
Data Analysis 35
Missing Data 35
Demographic Information 36
Descriptive Statistics 36
Exploratory Factor Analysis 36
CBI Associations 38
Logistic Regression 39
Ethical Approval 41
Methodology Summary 42
Chapter 5: Results 43
Introduction 43
Questionnaire Response 44
Demographic Variables 45
Psychometric Properties of Variables 47
Copenhagen Burnout Inventory (CBI) 47
CBI Descriptives 47
CBI Exploratory Factor Analysis 48
CBI Checking Assumptions 48
CBI Factor Extraction 49
CBI 3 Factor Model 50
CBI Summary 55
Dementia Knowledge Questionnaire (DKQ) 56
DKQ Descriptives 56
DKQ Exploratory Factor Analysis 57
DKQ Checking Assumptions 57
DKQ Factor Extraction 57
DKQ 2 Factor Model 58
DKQ Summary 62
Approaches to Dementia Questionnaire (ADQ) 63
ADQ Descriptives 63
ADQ Exploratory Factor Analysis 64
ADQ Checking Assumptions 64
ADQ Factor Extraction 65
ADQ 2 Factor Model 66
ADQ Summary 70
Copenhagen Psychosocial Questionnaire II (COPSOQ) 71
COPSOQ Exploratory Factor Analysis 71
COPSOQ Checking Assumptions 71
COPSOQ Factor Extraction 72
COPSOQ 3 Factor Model 73
COPSOQ: ‘Offensive Behaviour’ 77
COPSOQ Summary 78
Statistical Associations with CBI Factors 79
Demographic Associations 79
Covariate Associations 80
‘Offensive Behaviour’ Associations 81
Multivariate Analysis 82
Checking Assumptions 82
Logistic Regression Associations 83
Logistic Regression of CBI ‘Physical and Emotional Burnout’ 83
Logistic Regression of CBI ‘Work Burnout’ 87
Logistic Regression of CBI ‘Resident Burnout’ 90
Comparing Logistic Regression Burnout Models 93
Graphical Representations of the Burnout Models 94
CBI ‘Physical and Emotional’ Burnout Model 94
CBI ‘Work Burnout’ Model 96
CBI ‘Resident Burnout’ Model 98
Results Summary 101
Chapter 6: Discussion 102
Introduction 102
Critique of Background 102
Critique of Methodology 104
Copenhagen Burnout Inventory (CBI) 106
Dementia Knowledge Questionnaire (DKQ) 107
Approaches to Dementia Questionnaire (ADQ) 107
Copenhagen Psychosocial Questionnaire II (COPSOQ) 108
Demographic Information 109
Data Analysis 110
Critique of Results 111
Burnout as Determined by CBI 113
CBI Descriptives and Factor Analysis 114
Covariate Descriptives and Factor Analysis 117
Demographic Associations 119
Covariate Associations 120
Offensive Behaviour Associations 122
Explorations Using Logistic Regression 123
Logistic Regression of ‘Physical and Emotional Burnout’ 124
Logistic Regression of ‘Work Burnout’ 126
Logistic Regression of ‘Resident Burnout’ 127
Predicting Burnout 129
Discussion 130
Limitations of the Survey 130
Strengths of the Survey 132
Opportunities for Further Research 133
Implications for this Data 133
Implications for this Population 134
Implications for Research Theme 135
Discussion Summary 136
Chapter 7: Conclusion 137
Background 137
Methodology 137
Results: Descriptives 138
Results: Exploration 138
Results: Regression 140
Discussion of Burnout 142
Implications for Further Research 144
References 147
Appendix I 163
Questionnaire Information Leaflet 164
Questionnaire Booklet 167
Appendix II 187
Recoding Demographic Variables 188
Care Homes 189
Sex 189
Age 190
Marital Status 191
Children 192
Education 193
NVQ Level 194
Dementia Training 196
Job Status 197
Time in Current Job 198
Time in Profession 199
Shift Pattern 200
Hours Worked 202
Ethnicity/Nationality 204
Appendix III 205
COPSOQ: ‘Offensive Behaviours’ Frequency 206
Behaviour 207
Bullying: Frequency/Protagonist 207
Sexually Inappropriate: Frequency/Protagonist 208
Threats of Violence: Frequency/Protagonist 209
Physical Violence: Frequency/Protagonist 210
1
Chapter 1: Introduction
This dissertation will explore the concept of burnout of care staff working in dementia
registered residential homes in Cardiff, through a study involving a postal survey of
staff.
The importance of this subject area will be detailed in the background to this
exploration and will help to place the study within current understanding in this field.
The academic literature will be examined to understand the concepts likely to be
associated with burnout in this population and relevant key factors selected for
inclusion in the study. Difficulties in the measurement of these factors will also be
discussed along with the challenges faced in choosing appropriate psychometric
instruments to provide reliable estimates of the selected concepts.
The methodology for conducting the survey will be detailed and will additionally
describe the statistical methods used to assess the suitability of the instruments for
this purpose. The results obtained through following the above steps will enable a
greater understanding of the reliability and applicability of the instruments for each of
the desired concepts in this population.
The aim of this exploration is to produce a simple model to explain burnout, in terms
of the most significant concepts derived from a multivariate regression analysis. The
model of burnout resulting from this exploration will be assessed and the implications
for the participants and wider populations discussed.
2
Chapter 2: Background
Introduction
The background to this dissertation will detail the growing need for residential
dementia care due to ongoing demographic changes in the U.K. population. The
need for a highly skilled workforce in providing good quality dementia care will be
highlighted, including some of the difficulties experienced through caring for people
that may have behavioural difficulties.
The impact of these problems on care staff, including burnout, will be recognised as
well as the role that individual characteristics can play in modifying these. The
academic literature on burnout will be examined with key concepts identified that
influence burnout in comparable work environments.
Difficulties in implementing burnout interventions will be also be described, with
limited research having explored this area in general and this population in particular.
Dementia
Dementia is a term that is used to describe a pattern of neurological impairment,
typically involving deficits of cognition, not least memory. The term captures a
number of different diseases that affect the brain, with Alzheimer’s disease being the
most common, at an estimated 62% of cases (Knapp and Prince 2007). Many of
these illnesses become more prevalent with age, with 1 in 14 people over 65 and 1
in 6 people over 80 having some form of dementia (Knapp and Prince 2007).
3
Dementia Demographics
The ageing population in the U.K. has given rise to increased estimates of dementia,
with associated morbidity, mortality, carer stress and societal costs (Lakey et al.
2012). As the prevalence of dementia in Wales is anticipated to rise, changes in the
provision of services are needed to ensure optimal care for this vulnerable
population (See Figure 1. Knapp and Prince 2007). See also Figure 2. for U.K.
estimates of dementia prevalence from 2006 to 2051 (WAG 2009).
4
Care Homes
Care Home Population of Older People
In the U.K., around 400,000 people have their home as part of a residential complex,
(27,700 people in 1,164 care homes in Wales) equating to 2.8% of all people aged
over 65 years (BGS 2011; Knapp and Prince 2007). The reasons for living in a care
home are typically as a result of the individual needing increased support, a more
common occurrence with advancing age and co-morbid medical conditions. The
changing demographics of the U.K. population suggest that by 2031, 22% will be
aged over 65 and in the next 50 years, demand for care homes may increase by up
to 150% (BGS 2011).
People Living with Dementia in Care Homes
Cognitive impairment leading to difficulty with independent living has been viewed
as, “one of the biggest issues” in Wales (WAG 2011). It has been estimated that a
third to half of all people with dementia live in a care home and approximately 40% of
all care home residents have care needs as a result of dementia (BGS 2011; Knapp
and Prince 2007). Figure 3. illustrates the proportions of people in the UK with late
onset dementia living in residential care and in the community (Knapp and Prince
2007).
5
Dementia Registered Care Homes
The anticipated demands for residential accommodation suggest that increasing
numbers of people will have or develop dementia whilst in a care home. The ability
of these homes to meet the challenges of this population vary, however and 70% of
British geriatricians surveyed believed that management of dementia is sub-optimal
in care homes (BGS 2011).
In Wales, care homes that have a ‘dementia-registered’ status are expected to care
for residents with high levels of dementia-related morbidity (Cardiff County Council
2010). Many private care homes have been criticised as being unable to meet the
care needs of people with dementia, however, not least during episodes of
behavioural disturbance (Ballard et al. 2001).
6
Difficulties in Care
Complications of dementia can include both physical and psychological symptoms
that place significant demands on those carers looking after them. Some of the most
distressing and difficult to manage symptoms involve behavioural disturbance and
can include, “aggression… and psychosis… [with a] risk of 90% across the course of
the illness” (Banerjee 2009, p. 16-17).
The impact of these behaviours is considerable, with 65% of family carers reporting
being exposed to aggression and 16% of this occurring daily (O’Callaghan et al.
2010). It is significant cause of caregiver burden, distress and depression and can
directly result in institutionalisation (Black and Almeida 2004; De Vugt et al. 2005;
Miller et al. 2010).
In care homes, behavioural disturbance is estimated to be present in up to two thirds
of residents with dementia and in hospital, 73% of nursing staff on dementia wards
report having been assaulted (Boustani et al. 2005; O’Callaghan et al. 2010). It has
also been suggested that these behaviours adversely affect the health of care staff
and can increase the risk of burnout, however other studies have failed to find these
effects (Brodaty et al. 2003; Nagatomo et al. 2001; Schmidt et al. 2012)
7
Care Staff
Care Staff Characteristics
Caring for people with dementia is challenging and requires staff that have the
necessary skills to meet these complex demands. Care homes are usually privately-
run companies and vary in their environmental and organisational structure. They do
have common elements, however with care staff of varying grades and experience.
A care home will typically have a greater number of less qualified ‘junior’ staff (who
nonetheless may have extensive ‘hands-on’ experience), supervised by a more
qualified ‘senior’, not infrequently with a nursing background.
The nature and degree of dementia training of care staff is highly variable both within
and between care homes, however there are minimum standards and the care home
has a duty to prove to inspectorate services that they are capable of meeting their
residents’ needs (WAG 2004).
One description of the working conditions of ‘junior staff’ suggests that they, “work
long hours, are poorly paid, receive minimal benefits, and are prone to injury and
depression” and have insufficient training or support (Zimmerman et al. 2005, p. 96).
Other descriptions of assistants in nursing homes, have identified them as a
population particularly vulnerable to burnout, with high rates of turnover, low pay,
limited involvement in decisions and minimal autonomy (Gruss et al. 2004).
8
Greater stress has also been suggested for staff working with more cognitively
impaired residents as well as those on day shifts, where workload is high and with
part-time status (Brodaty et al. 2003). Increased care staff age and more experience
in nursing homes have also been associated with greater strain (relating to
behaviours associated with dementia) and greater age was also associated with
reduced job satisfaction in this sample (Brodaty et al. 2003).
Characteristics of the care homes associated with greater stress in workers include
larger size (greater than 16 beds) and specialised dementia care status, i.e. homes
likely to include people with greater needs relating to dementia (Zimmerman et al.
2005).
9
Care Staff Characteristics in Cardiff
The profiles of care staff and the care homes they work in are changing, not least
due to demographic, economic and service-driven needs. These profiles are also
changing considerably across Cardiff.
In 2008, there were 9 care homes registered for dementia care in Cardiff but with
only 151 of the 485 beds (31%) having the dementia registration (Cardiff County
Council 2008). A survey by Cardiff Local Authority noted wide variations of approach
in the delivery of dementia care, however all homes expressed the desire to develop
‘a person centred care’ model (Cardiff County Council 2008). Care homes in the
survey varied in their design and environment, with some being purpose-built and
others planning improvements and all had a unique mix of resident characteristics
and dementia care needs (Cardiff County Council 2008).
The survey recorded that 41% of care staff had NVQs (National Vocational
Qualifications) and 63% of these had NVQ level 2 (Cardiff County Council 2008).
30% of staff reported having had some form of dementia care training, however the
content of this varied from e-learning or in-house training to training from external
agencies (Cardiff County Council 2008). Staff turnover, consistency of approach
and openness to change were described as being heterogeneous between the
homes and 7 of the 9 Cardiff homes in the survey were specifically documented as
needing improvements in staff supervision (Cardiff County Council 2008).
10
Burnout
The care of people with dementia is recognised as being difficult and stressful,
particularly where behavioural problems are prevalent (Donaldson et al. 1996). For
informal or family carers, this is frequently termed ‘burden’, however in residential
settings, these stressors can contribute to a pattern of exhaustion, known as
‘burnout’ (Sorensen et al. 2006).
‘Burnout’ is considered to be a psychological response of a worker to chronic strain
in their job resulting in negative consequences for both employee and employer. It is
typically thought of as a state that affects people in the human service sector
(possibly as burnout was first described here) and inter-personal strain is placed
centrally to the concept (Borgogni et al. 2012).
Burnout, as described by Maslach (2003), is characterised by the worker
experiencing feelings of exhaustion, cynicism and inefficacy. These three
components are thought to arise through workplace stress, such as excessive
demands, interpersonal conflict and inadequate support (Maslach 2003). The result
is that workers reduce their efficiency to expend the minimum amount of physical
and psychological resources on their day to day tasks (Maslach 2003). The burnout
concept also allows for factors that reduce burnout, with ‘engagement’ being viewed
as key to initiatives that reduce work stress (Maslach 2011; Schaufeli and Salanova
2011).
11
The exact concept(s) that describe the phenomenon of ‘burnout’ are, however
variable, dependent on the model being used and continue to be in flux in the
academic literature on the subject (Cox et al. 2005). A central issue involves the
need to form a consensus on establishing burnout as a distinct concept specific to
employees and independent of exhaustion, stress or affective disorders (Cox et al.
2005). The need to establish burnout as a definitive state (i.e. present or absent) or
as a trait (present to varying degrees) is also outstanding (Cox et al. 2005).
Consequences of Burnout
Burnout has been suggested as predicting employee turnover, ill health and work
efficacy (Maslach et al. 2001). Severe burnout is estimated to be present in over 7%
of the working population in western countries and has major implications in terms of
social, psychological and economic costs (Shirom 2005).
One meta-analysis of studies showed correlations between employee burnout and
negative work performance, in particular relatiing to their role, the organisation and
customer satisfaction (Taris 2006). It has also been suggested that elements of
burnout can also be transferred between workers, through processes of ‘priming’ and
‘empathic identification’ (Bakker et al. 2007)
12
Factors Associated with Burnout
Burnout is believed arise out of an imbalance between the individual and their work
environment. Various models that have been used to understand burnout in the
work environment have examined factors relating to demands, autonomy, support,
justice and effort-reward imbalance (Borritz et al. 2010; Kristensen 2010). The
following sections summarise some of the factors thought to be relevant when
exploring burnout;
Demographics
Individual factors suggested as significant in burnout have included younger age,
male gender, single relationship status and working for less than 2 years (Maslach
2003; Milfont et al. 2008; Zimmerman et al. 2005). Other studies have considered
age to be a minor or inconsequential component in predicting burnout and female
gender to be associated with ‘personal’ burnout and male gender associated with
‘client-related’ burnout (on the Copenhagen Burnout Inventory or CBI scales,
comprising of ‘Personal’, ‘Work-Related’ and ‘Client-Related’ Burnout) (Borritz et al.
2005; Nagatomo et al. 2001; Shirom 2005).
Greater job satisfaction (negatively associated with burnout) has been reported for
those with greater training and in non-white care staff, although turnover is greater
for non-white workers (Rosen et al. 2011; Zimmerman et al. 2005).
Burnout has also shown some familial clustering, however twin studies have
favoured a shared environmental explanation, rather than genetic (Shirom 2005).
13
Other factors potentially related to burnout include ‘socio-economic status’
(supervisor status or advanced education), family status (cohabiting and children at
home), health related lifestyle (smoking, alcohol, exercise, weight) and illness
(Borritz et al. 2010).
Temporary workers have higher psychological distress and worse health outcomes
than permanent workers, as have shift workers compared with regular daytime
workers (Llorens et al. 2010). Working more than 40 hours per week and working
long (greater than 10 hour) shifts has been associated with affective disorders and
burnout (Albert et al. 2013; Llorens et al. 2010). Recovery time, both out of hours
and days off, has also been suggested as a protective factor for burnout (Sonnentag
2005).
Given the above research, including information on demographic factors and working
conditions in this survey may provide valuable information on burnout.
14
Personality
Personality factors have also been associated with burnout, particularly ‘neurotic’
personality traits, as well as those with less social and highly individualistic traits
(Gandoy-Crego et al. 2009; Shimizutani et al. 2008). Other associations with
burnout have included, “openness to changes and anxiety”, with non-burnt out staff
showing traits of, “emotional stability, liveliness, privateness and tension”
(Gustafsson et al. 2009).
The presence of personality traits acting as a confounder should be considered
however, as personality may influence the individual’s self-evaluation (of their
psychosocial work environment) rather than affecting ‘burnout’, such that people with
a negative outlook may hold more negative evaluations of their own coping and
health.
Neuroticism has also been considered a minor component in predicting burnout in
some studies, although this association was increased where the leadership was
based on an ‘autocratic’ style (De Hoogh and Den Hartog 2009; Shirom 2005). In
addition, those with a low ‘internal locus of control’ showed lower burnout where the
leadership style was ‘charismatic’ (De Hoogh and Den Hartog 2009).
15
Mental Health
Depression has been strongly associated with burnout and has also been suggested
as a potentially significant confounder in explaining many of its consequences
(Borritz et al. 2010; Shirom 2005). Taking account of this correlation when designing
burnout research studies has been recommended (Shirom 2005).
Of note, some psychometric instruments used to explore workplace psychosocial
factors (e.g. Copenhagen Psychosocial Questionnaire or COPSOQ (Long Version))
have recognised this confounder and include a section on ‘depressive symptoms’
and this has been significantly associated with sickness absence (Pejtersen et al.
2010).
Both the COPSOQ and the Copenhagen Burnout Inventory (or CBI) contain a
question based on ‘emotional exhaustion’ and this has also been associated with
depressive symptoms, while COPSOQ scales on ‘emotional demands’ were
positively associated and ‘meaning of work’ negatively associated with mental health
problems (Burr et al. 2010; Pejtersen et al. 2010). In other studies, organisational
injustice has been associated with depression (Andersen et al. 2010).
The direction of causality in these models suggest that workplace factors (notably
emotional exhaustion and job satisfaction) have stronger effects on mental health
than mental health does on workplace factors (De Lange et al. 2004).
A potential confounder for mental health problems and burnout, has been suggested
as workplace violence, with this predicting fatigue and emotional demands in
addition to depression (Burr et al. 2010).
16
Offensive Behaviour
Violence in the workplace is classed under a broad category of ‘Offensive
Behaviour’, which can be perpetrated by supervisors, colleagues, subordinates,
service users or other persons. These behaviours can be verbally or physically
aggressive, sexually inappropriate and/or bullying and are associated with greater
rates of turnover, sickness and reduced health and psychological well-being
(Clausen et al. 2012).
Offensive behaviour is reported as being more common in human service
occupations and particularly where the workforce is dominated by a majority of a
single gender, such as nursing (Clausen et al. 2012).
Nurses are an employee group that is frequently subject to verbal and physical
aggression and the frequency of these incidents are associated with burnout in
general, and with ‘depersonalisation’ (emotional distancing) in particular (Winstanley
and Whittington 2002).
One study of care home workers notes that the associations with turnover are
strongest for ‘bullying’ (commonly from colleagues and supervisors) but that these
effects are mediated to an extent by employee ‘well-being’ (Clausen et al. 2012).
17
Attitudes
Employee attitudes to their workplace is thought to have a number of aspects, not
least job satisfaction and organisational commitment, both of which are associated
with burnout (Clausen 2009; Judge and Kammeyer-Mueller 2012). Attitudes are also
viewed as having a good predictive value for intentions and subsequent actions,
proving a useful model for research into specific behaviours (Judge and Kammeyer-
Mueller 2012).
A concern for staff with burnout caring for people with dementia is that the elements
of depersonalization and cynicism could predispose to harmful attitudes that may
lead them to regard patients as objects (Lee et al. 2012). This psychological
distancing may be used as a coping strategy by the individual to protect themselves
against further stress, resulting in them performing tasks mechanically, rather than in
a person-centred manner, thereby avoiding therapeutic interactions, to the detriment
of both parties (Sonnentag 2005).
One study involving residential home care staff did not find associations between
depersonalisation or emotional exhaustion and the quantity or quality of staff-resident
interactions but did find improved interactions with greater ‘personal efficacy’ and
‘involvement in decisions’ (Jenkins and Allen 1998). Hopefully this suggests that the
potential outcome as described above, of cynicism and objectification of residents, is
an extreme and infrequent reaction.
18
Person-centred attitudes have been associated with job satisfaction, particularly
amongst, “staff working in newer facilities and those who feel better trained”
(Zimmerman et al. 2005, p. 102-3). It has also been noted, amongst informal carers
that hopeful attitudes have been associated with less burden and distress and
greater resiliency and social support (Cumming 2011).
Given the potentially important contribution to influencing both burnout and quality of
care, assessment of attitudes towards people with dementia should also be a key
component of this survey.
Knowledge
Dementia knowledge has been linked with attitudes to dementia and through this,
the behaviour of care staff towards people with dementia (Lintern 2001). Knowledge
about dementia has been shown to be highly variable between different grades and
occupations of healthcare employees as well as between specialties (Barrett et al.
1997).
Studies of informal carers have also shown that ‘irrational beliefs’ about dementia
predict depression in the carer, potentially through uncertainty about future
expectations of the illness (Graham et al. 1997). Carers with greater knowledge
were more likely to have reduced expectations of people with dementia and to make
positive comparisons, however they were also more likely to have increased anxiety
(Graham et al. 1997b).
19
The Dementia Knowledge Questionnaire (DKQ), used in the above studies has
received conflicting reports, with some studies failing to show either positive or
negative associations with carer stress (Goncalves-Pereira et al. 2010). Ethnic
differences in DKQ scores have also been seen, with older people of Indian origin
scoring lower on this test than a caucasian sample, in the general population
(Purandare et al. 2007).
Care staff working with people with dementia in care homes and day centres have
demonstrated better care and improved quality of life of their service users with
greater professional knowledge (Kazui et al. 2008). Educational support is
considered important in attaining this goal and educational interventions have been
shown to reduce burden, again in informal carers (Graham et al. 1997).
Knowledge of dementia is frequently assumed to be a vital component for any
professional training course into dementia care, however the evidence from
academic literature is somewhat weak. Dementia knowledge was therefore viewed
as an ‘intuitively’ important factor to include in exploring stress/burnout in care home
staff, however lacks an evidence base with which to anticipate outcomes.
20
Psychosocial/Organisational Factors
The strain of conflicts in employee ‘work-life balance’ has also been suggested as a
predictor of burnout, although the direction of causality may be bidirectional (Brauchli
et al. 2011). Stronger associations with burnout have been observed, however,
between the impact of demands of work on life, rather than life on work (Brauchli et
al. 2011; Fuz et al. 2008). Suggested interventions to address burnout may involve
changing work-related factors, such as reduced hours and increased flexibility and
autonomy over working patterns (Brauchli et al. 2011; Llorens et al. 2010).
Overall, workplace organisational factors, rather than individual characteristics have
been shown to be more significant in predicting burnout, and are thought to include,
“chronically difficult job demands, an imbalance between high demands and low
resources, and the presence of conflict” (Maslach 2003, p. 191).
This was also suggested through associations between stress and work intensity in
staff working in nursing homes, with protective effects described from ‘effective
coping strategies’ (Schmidt et al. 2012; Schmidt and Diestel 2013). Interviews with
nursing staff have also suggested that a source of stress may be related to
discrepancies between the work that staff felt was necessary and the resources that
they had been allocated to complete that work (Edberg et al. 2008).
21
High levels of burnout on all 3 of the CBI’s scales (‘Personal’, ‘Work-Related’ and
‘Client-Related’ Burnout) have been linked with ‘emotional’ and ‘quantitative
demands’, and ‘role conflicts’ and negatively associated with ‘meaning of work’
(Borritz et al. 2005). Elsewhere, ‘role conflicts’ were associated with increased
turnover, with ‘influence’ at work and ‘leadership quality’ reducing this risk (Clausen
et al. 2012).
Poor work ‘predictability’ has been correlated with high ‘Personal’ and ‘Work-Related’
burnout, while ‘emotional demands’ and reduced ‘role clarity’ were associated with
‘Work-Related’ and ‘Client-Related’ burnout (Borritz et al. 2005). Factors that have
been shown to be significant for only ‘Work-Related’ burnout include ‘work pace’,
poor ‘potential for development’, and poor ‘leadership’ (Ibid.).
Many of these factors represent a shift in understanding of organizational stressors
away from task and intensity related to that of inter-personal relationship based
understanding or ‘Social Capital’ (Kristensen 2010). Meaning at work, justice (or
equity) and job satisfaction are considered a key components of this concept and the
concept of ‘affective organisational commitment’, which has been negatively
associated with exhaustion and cynicism amongst nurses (Clausen 2009; Taris et al.
2002).
‘Affective organisational commitment’ has also been associated with employee well-
being, job performance and ability to cope with work stress and inversely associated
with turnover (Clausen 2009; Rosen et al. 2011). Interestingly, greater ‘meaning of
work’ and ‘quality of leadership’ have also been associated with greater ‘Personal’
burnout, potentially as a consequence of it being a protective factor for continuing
work despite higher levels of burnout (Borritz et al. 2005).
22
Leadership has been demonstrated as a vital component in the
psychosocial/organisational environment, with benefits suggested for styles that are,
‘participatory, supporting and/or fair’ and potential harms for styles that are, ‘laissez
faire, autocratic and/or abusive’ and this holds true for nursing homes (Castle and
Decker 2011; Wild et al. 2010; Llorens et al. 2010).
Nursing assistants in care homes that had ‘nonempowered environments’ described
more job-focused stressors than in ‘empowered environments’ and senior
supervision has been suggested as a key factor in retaining these staff (Bishop et al.
2008; Gruss et al. 2004). The benefits here also extended to the care home
residents with greater work commitment corresponding to improved quality of life and
greater satisfaction in their relationships with nursing staff (Bishop et al. 2008).
Further associations have been noted between nursing assistant job satisfaction and
having enough time to do their job, having a challenging role and having satisfactory
working hours (Bishop et al. 2009).
The psychosocial/organisational environment is therefore central to concepts of
workplace burnout and is to undergo further analysis in this project, although the
topic is substantially broad so as to require a clear focus.
23
Preventing Burnout
Some studies have noted that ‘burnout’ is changeable over time, suggesting that an
individual’s susceptibility to burnout is modifiable with the potential to improve well-
being and reduce sickness absence from appropriate interventions (Borritz et al.
2006).
Although research on interventions to reduce burnout is limited, it is suggested that a
combination of strategies to improve both personal and organisational characteristics
would be of greatest benefit (Maslach 2003). Other research has focused on
identifying individuals or even ‘clusters’ of workers at risk of burnout and of
developing targeted interventions based on their individual need (Maslach and Leiter
2008).
Maslach and Leiter (2008) distinguish between those with early warning signs
(significant exhaustion or cynicism), those at a ‘tipping point’ (significantly low
fairness scores) and those already in burnout (significant exhaustion and cynicism)
and makes suggestions for addressing these 3 states.
24
Many interventions have been shown to have no effect on burnout or to actually
have negative effects, particularly in workplace reorganisation where staff have no
active involvement (Anderson et al. 2010; Visser et al. 2008). Another study,
exploring the effects of an intervention in a large hospital, noted more negative
evaluations from staff, including of ‘leadership quality’, ‘supervisor support’ and
‘possibilities for development’ (Aust et al. 2010).
Another significant factor in this intervention, was of staff reporting reduced
‘emotional demands’, although this may be due to disengagement (a feature of
burnout) (Ibid.). The mechanism of these negative observations was considered as
a result of, ‘disappointing expectations’ (Ibid.).
Research into training and/or interventions for both burnout and dementia care is
sparse and very much needed, with some ‘positive psychology’ approaches showing
promise (Elliott et al. 2012; Meyers et al. 2012).
25
Background Summary
Dementia is becoming more prevalent in the U.K. due to the ageing population and
with it comes challenges for those caring for them, not least in care homes. The
amount of research literature on burnout in care staff is poorly representative of the
importance of this field given the morbidity and economic implications. There is
enough of a correlation, however between the available research in dementia care
homes and more generalised research on burnout, to be able to draw some
conclusions.
The academic research suggests that a combination of factors is involved in burnout,
with influence from both individual and organisational elements. This interaction of
factors can be difficult to disentangle and interventions to improve the workplace may
result in unintended negative consequences if inexpertly managed.
26
Chapter 3: Aims and Research Hypothesis
Aims of Study
Burnout of staff working in dementia-registered care homes has substantial
implications for maintaining a healthy and committed workforce, as well as for high
quality resident care. The aim of this study is to explore the burnout concept as it
relates to care staff working in dementia registered residential homes in Cardiff.
The academic literature, as described, identifies a number of individual and
organisational factors that have been linked to burnout. Some of these key factors
are used to examine burnout in this population, with selection based on their
importance to the burnout model and the burden on participants. The risk of
introducing ‘Type I’ statistical errors into the analysis due to excessive data collection
from limited participants is also acknowledged.
Demographic information of care staff has been linked to varying rates of burnout.
The aim for collecting this information is to explore these associations based on a
pragmatic ‘best-fit’ for responses to these questions, rather than theoretical
considerations.
The exploration of both burnout and the factors associated with burnout involves
identifying suitable, validated psychometric instruments to act as proxies for the
underlying concepts. As research in dementia-registered residential homes is
limited, the instruments selected may not have been validated for use in these
populations.
27
The instruments therefore need to be examined using exploratory statistical methods
to establish their factor (or subscale) construction, as determined by the participant
response patterns. These subscales are compared to those described in the
academic literature to assess the applicability of those concepts to this population.
The individual items within each of the subscales are also examined in order to
describe patterns that could reflect underlying concepts within the subscale. This
may be of particular relevance where subscale items for the population are divergent
from those expected to be found from the literature.
An examination of the instruments and demographic information relating to burnout
in this manner aims to provide an overview of the applicability of these factors to this
specific population. The main aim is to produce a burnout model (or models) that
best describe the burnout concept, using the demographic associations and the
derived factor constructs.
The aim of producing the burnout model (or models) is to explore the concept of
burnout through examining those factors found to reflect the underlying concepts for
burnout in this population. It is anticipated that this knowledge will be useful in
further understanding the burnout of care staff in dementia registered residential
homes in Cardiff. The strengths and limitations of the study will be acknowledged
and the information used to suggest further directions for burnout research in these
environments and for a more generalised population.
28
Research Hypothesis
The hypothesis of this research is that burnout, in care staff working in dementia
registered residential homes in Cardiff, varies with certain individual and
organisational factors. The academic literature suggests that dementia knowledge,
attitudes, job satisfaction, leadership, emotional demands and exhaustion are
associated with burnout, along with various demographic variables. This study will
explore the hypothesis that these factors are associated with burnout as described in
the literature and will produce a model (or models) that provide the most
parsimonious explanation of the variability of burnout with these associated factors.
Aims and Research Hypothesis Summary
The aim of this dissertation is to explore the concept of burnout as applicable to care
staff in dementia-registered residential homes in Cardiff. Selected demographic and
conceptual factors that have been suggested by the academic literature to be
associated with burnout will be examined to establish their applicability for use in this
population. These factors include dementia knowledge, attitudes, job satisfaction,
leadership, emotional demands and exhaustion, amongst others.
The hypothesis of this exploration is that through this process, a model will be
produced that explains the concept of burnout in terms of the most significantly
associated factors. The implications of the burnout model for this and other
populations will be discussed.
29
Chapter 4: Methodology
Introduction
This section will detail the methodology followed to explore the concept of burnout in
this population, as influenced by the academic literature on the subject. This will
include information on the design of the study and the process of recruitment, as well
as details of the measures felt to best reflect the variables under examination.
The construction of these measures will be explored to assess their validity and
reliability when used in previous research with comparable populations. The
statistical methods used in data analysis will also be described as this process is
central to understanding the data and its implications.
Design
The design of this observational study was a cross-sectional survey. The method
was a pseudoanonymised questionnaire to be returned by post.
30
Participants
The participants for this survey were care staff working in 18 dementia-registered
residential homes in Cardiff in 2010. Identification of the homes was through the
information resources available to Cardiff Local Authority (Cardiff County Council
2010). Engagement with the care home managers and/or senior staff was at events
organised as part of the ‘Enhanced Dementia Care’ project.
The recruitment of the participants took place in 2 phases, with ‘phase 1’ involving
recruitment from an initial 9 care homes and 6 months later, ‘phase 2’ involving a
further 9 care homes. This division was an artefact of the registration of dementia-
care status in the ‘phase 2’ residential homes only after the study had commenced.
Measures
The questionnaire used in the survey comprised of a number of validated
instruments that had shown good reliability in testing from previous research, as well
as selected demographic information. The outcome variable measuring ‘burnout’
was the ‘Copenhagen Burnout Inventory’. Co-variates used to explore the burnout
concept in this sample were the ‘Dementia Knowledge Questionnaire’, the
‘Approaches to Dementia Questionnaire’ and the ‘Copenhagen Psychosocial
Questionnaire II’.
31
Demographic Information
The demographic information requested from the participants included items that had
been suggested in the academic literature as being significant in assessing burnout
in care staff. The aim was to assess these variables further and to assist in adjusting
for confounding factors in later analysis.
Copenhagen Burnout Inventory
The Copenhagen Burnout Inventory or ‘CBI’ is a psychometric instrument designed
to explore the concept of ‘burnout’ in populations, with components that relate to
people in general (the ‘Personal’ burnout subscale), people in work (the ‘Work-
Related’ burnout subscale) and people in human service work (the ‘Client-Related’
burnout subscale) (Kristensen et al. 2005).
The CBI consists of 19 questions relating to ‘burnout’, answered on a 5-point likert
scale, scoring between 100 (‘Always’) and 0 (‘Never’) in 25-point increments, with 1
inversely scored question. ‘Personal’, ‘Work’ and ‘Resident’ burnout sub-scales
consist of 6, 7 and 6 questions, respectively (with Cronbach’s alpha scores of 0.87,
0.87 and 0.85), although answers have typically been positively skewed towards
lower burnout scores (Ibid.).
The core concept of ‘burnout’, as described by the CBI, relates to, ‘fatigue and
exhaustion’ (Ibid.). Using the CBI in a population of approximately 1900 workers
(with numerous professions in 7 different organisations and locations), both the
‘Personal’ and ‘Work’ burnout subscales show high correlations with a ‘Vitality’ scale
and good correlations with a ‘Mental health’ scale (Ibid.). .
32
All burnout sub-scales (but particularly ‘work’ related) were associated with job
satisfaction and also predicted frequency and duration of sickness absence, sleep
problems, intention to quit and use of painkillers (Ibid.). The greatest negative health
correlations have been found with ‘Personal’ burnout, while ‘Work-Related’ burnout
has had the greatest correlation with long term sickness absence (>9 days) (Borritz
2006; Borritz 2010). Of note, changes in burnout scores across time have also
predicted changes in sickness absence (Borritz 2006).
Dementia Knowledge Questionnaire
The Dementia Knowledge Questionnaire (DKQ) is a psychometric instrument
designed to test the knowledge that carers of people with dementia have about the
condition (Graham 1997). It consists of 4 sections relating to ‘Rudimentary
Knoweldge’, ‘Epidemiology’, ‘Aetiology’ and ‘Symptoms’, with maximum potential
scores of 3, 2, 6 and 8, respectively (Ibid.). The total score out of 19 can be sub-
divided into ‘Irrational Beliefs’ (for scores of 0 or 1 out of 3 on ‘Rudimentary
Knowledge’) and ‘General Knowledge’ (a total of the three other categories, out of
16) (Ibid.).
In interviews with 109 informal carers, higher scores on the ‘General Knowledge’
section of the DKQ was associated with, “lower levels of depression but… higher
rates of anxiety” (Graham 1997b, p. 934). Greater knowledge was also associated
with carer confidence and feelings of competence in care-giving (Graham 1997b).
Other significant carer attributes included having, ‘reduced expectations’ and making,
‘positive comparisons’ of the person with dementia (Graham 1997b, p. 933-934).
33
Approaches to Dementia Questionnaire
The Approaches to Dementia Questionnaire (ADQ) is a psychometric instrument
designed to explore the attitudes and behaviour of care staff towards people with
dementia (Lintern 2001). It consists of 2 scales, derived through factor analysis,
termed ‘Hope’ and ‘Personhood’, which require agreement/disagreement with
statements relating to dementia care, on a 5-point likert scale, scored 1 to 5 (Ibid.).
Total scores range from 19 to 95 and greater scores have been associated with
person-centred attitudes to dementia care style (Ibid.).
The ‘Hope Subscale’ consists of 8 statements (representing optimistic attitudes) and
the ‘Personhood Subscale’ consists of 11 statements (representing respect for
‘individual agency’), both with good internal reliability (Cronbach’s alpha of 0.76 and
0.85, respectively; total score 0.83) but with negative skewness i.e. greater numbers
of higher scores (Personhood > Hope) (Ibid.). Greater scores on the ‘Hope
Subscale’ have been linked with greater social engagement between staff and
residents with dementia, including, “purposeful activities” and “qualitatively better
physical care interventions” (Lintern 2001, p. 15).
There have also been associations between the ADQ and greater dementia
knowledge in staff, as well as observations of greater engagement with residents
(‘Hope Subscale’) and physical care (‘Personhood Subscale’) (Lintern 2001).
34
The ADQ questions were devised in consultation with, ‘experts in the field’ and were
piloted with 20 nurses/care assistants on a dementia care NHS ward before use with
123 care staff, of varying grade and experience, from 5 care homes across the U.K.
(Ibid.). This population is directly comparable to that under examination in Cardiff for
this survey and should therefore have good applicability.
Copenhagen Psychosocial Questionnaire
The Copenhagen Psychosocial Questionnaire II (COPSOQ) is a psychometric
instrument designed to explore the, ‘working conditions, health and well-being’ of
employees. It was developed and validated through surveying ‘representative’
samples of 1858 and 3517 Danish workers, respectively (Kristensen et al. 2002;
Pejtersen et al. 2010). The COPSOQ (Short version) consists of 40 questions,
covering 23 ‘dimensions’ that investigate job stress and satisfaction, and was
produced through reviews of existing questionnaires, theoretical discussions and
statistical analysis (Kristensen et al. 2002).
The questions typically consist of a 5 point likert scale, with responses scoring
between 0 and 4, with some scoring inverted. The dimensions were typically made
up of 2 questions on a particular subject, giving them a range of 0 to 8 points,
however the resulting scores for the 23 dimensions cannot be combined to make a
total score. Cronbach’s alpha scores for the dimensions were not provided for the
short version of the COPSOQ, apart from the work ‘predictability’ scale (0.74) (Ibid.).
35
COPSOQ has been used to explore sickness absence across 8000 randomly
selected Danish residents, with ‘Emotional Demands’ and ‘Role Conflicts’ at work
predicting greater incidence of annual sick leave of 3 or more weeks (Rugulies et al.
2010). Other studies using COPSOQ, in eldercare workers and nurses, have
shown that low scores on ‘Commitment to the Workplace’, ‘Meaning of Work’ and
greater ‘Emotional Demands’ are associated with greater long-term sickness and
intention to leave (Clausen et al. 2010; Li et al. 2010).
Data Analysis
Data collection, input and analysis were completed using SPSS Statistics package
versions 18 and 20.
Missing Data
The data from covariates and any associated subscales were included in the
analysis if the participant had completed at least half of the questions for the relevant
scale, otherwise the data from that section was considered missing. Where data
was missing, but accounted for less than half of the questions, the missing data was
computed using the participant’s mean score for the other questions in that section.
36
Demographic Information
Descriptive statistics were obtained for the demographic information in terms of
numbers and percentages of respondents. The information was then examined and
responders categorised into 2 or 3 divisions according to the most pragmatic split of
the data. This was to enable a sufficient number of participants in each category for
later analysis.
Descriptive Statistics
Descriptive statistics (mean, 95% CI, S.D.) were produced for the dependent (CBI)
and independent (ADQ, DKQ and COPSOQ) variables according to the original
instrument models to enable comparison to results from the academic literature.
Exploratory Factor Analysis
The variables (CBI, ADQ, DKQ and COPSOQ) were examined using exploratory
factor analysis, having first checked assumptions relating to the data including
correlations between items and communalities. Communality refers to the proportion
of the item’s variance that is shared with other variables in the model, with 0
representing no sharing and 1 representing complete sharing (Field 2005, p. 630).
The method of extraction and rotation for the exploratory factor analysis was chosen
based on the anticipated degree of correlation between the resulting factors, based
on descriptions from previous research on the variables.
37
Principle axis extraction with oblimin rotation was chosen where there was assumed
to be a significant correlation between the factors (e.g. for ‘burnout’) and principle
components extraction with varimax rotation was used where no correlation was
assumed.
Factor items that were low or double loaded on the pattern matrix of the analysis
were removed from the factor models. The final models were assessed using
Kaider-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett’s Test of
Sphericity to ensure that the factors were a good fit for the data. KMO values vary
between 0 and 1, with scores above 0.9 considered, “superb”, reflecting, “distinct
and reliable factors” (Field 2005, p. 640). Bartlett’s Test assesses the hypothesis
that there is no correlation between the items within the factors (i.e. making them
unreliable), with significant scores rejecting this hypothesis (Field 2005, p. 652).
The factors produced through this process were examined to assess for a coherent
underlying concept with each factor given a label to approximate this. Items that
comprised an extracted factor were used at a later stage of the analysis by
calculating the mean of the component items that corresponded to each factor at a
loading of > 0.4. This was done, rather than using factors derived from items
weighted according to loadings, due to the desire to have clearly identified factors
influenced only by items that strongly correlate. This method does increase the
degree of correlation between the non-weighted factors, however as they are not
being used in analysis to explain variance between each other, the potential for bias
is reduced.
38
The non-weighted factors’ properties were further assessed using Cronbach’s Alpha
to check for internal consistency (>0.7 being considered acceptable) within and
Spearman’s rho to check correlations between the factors (Field 2005, p. 668). This
was to ensure that the factors were suitable for further analysis of this population.
The normality of the factors’ data distributions were checked using the Shapiro-Wilk
test (a significant result confirming non-normal data), directing the statistical testing
that could be used for further analysis i.e. parametric or non-parametric.
CBI Associations
The statistical associations between the CBI factors and the other variables were
explored using non-parametric methods, namely Mann-Whitney U test for 2
categorical variables and the Kruskall-Wallis test for 3 categorical variables.
For the demographic and ‘Offensive Behaviour’ variables, associations were
explored using the 3 CBI factors, however for the covariates, the CBI factors were
transformed into bivariate variables. This was through classification of the factors
into ‘low’ or ‘high’ burnout based on roughly even numbers of responders in each
category and also enabled assessment of the burnout models using bivariate logistic
regression.
39
Logistic Regression
The statistical characteristics of the 3 CBI factors suggested that an appropriate
method of multivariate analysis would be logistic regression.
As logistic regression makes uses bivariate categorical values as the dependent
variable, the significantly associated categorical variables were cross-tabulated to
ensure a sufficient number of responders in each subsection (low numbers results in
large standard errors in the analysis) (Field 2005, p. 264).
The models were refined through an iterative process of ‘Backward: Likelihood Ratio’
logistic regression analysis to produce the most parsimonious model available (Field
2005, p. 227). The process involved retaining components that significantly
explained variance in the model and removing the items that were least statistically
significant one at a time.
The resulting models have a ‘B’ value and S.E., which are the ‘log-odds’ and
standard error for predicting the dependent variable form the independent variable.
The Wald (Chi-square) value and significance (2-tailed p value) tests the ‘null
hypothesis’ that the variable has no effect on the model. Exp(B) is the odds ratio of
the predictor and where the 95% C.I. (Confidence Interval) crosses 1, the reliability
of the variable is in question and its’ generalizability is limited (Field 2005, p. 254).
Classification tables for the logistic regression were produced, giving the
percentages of the ‘High’, ‘Low’ and ‘Overall’ burnout scores correctly predicted from
their respective model.
40
Once the final model had been established, outlying responses for ‘Standardized
Residuals’ and ‘Leverage’ were noted and a boxplot produced to represent the range
of values.
‘Standardized Residuals’ outliers represent responses in the logistic regression that
were a poor fit for the model, with values above 3 being of concern and those above