Application of neural network paradigms to a study of nursing burnout

MA-Thesis / Master von Felix Ladstätter, Eva Garrosa

Abstract:

Nursing is considered as a risk profession with eminent levels of stress and burnout, which are probably increasing. Objectives: The first was to implement artificial neural networks which could be used for the prediction of burnout in nursing. The burnout-model that was implemented included socio-demographic variables, job stressors, and personal vulnerability, or resistance (hardy personality). The second was to assess whether artificial neural network paradigms offered greater predictive accuracy than did conventional burnout methodologies (hierarchical stepwise multiple regression).

Design: Two different types of artificial neural networks were implemented. A three-layer feed-forward network and a radial basis function network. The sample used in this Master thesis was of 473 nurses and student nurses in preparation from three hospitals in Madrid (Spain) who filled in the ‘Nursing Burnout Scale’ survey. The data were analyzed using descriptive statistics, pre-processed, and used for the training and validation of the implemented artificial neural networks.

Results: The network paradigms were better suited for the analysis of burnout than hierarchical stepwise multiple regression. Both could capture nonlinear relationships that are relevant for theory development. At predicting the three burnout sub-dimensions emotional exhaustion, depersonalization, and lack of personal accomplishment however, the radial basis function network was slightly better than the three-layer feed-forward network.

Conclusions: The usage of artificial neural networks in the field of burnout has been justified due to their predictive efficacy.

Inhaltsverzeichnis:

	Abstract	iii
	Acknowledgements	ix
	Contents	xiii
	List of Figures	xvii
	List of Tables	xxi
1.	Burnout	1
1.1	The Origin of Burnout	1
1.1.1	The Uncovering of Burnout	2
1.2	Burnout as a Global Problem	3
1.3	Why is Burnout increasing?	4
1.4	Understanding Burnout	7
1.4.1	Definitions	8
1.4.2	Possible Symptoms	10
1.4.3	Burnout vs. Job Stress	13
1.4.4	Burnout vs. Depression	14
1.4.5	Burnout vs. Chronic Fatigue	14
1.5	Assessment and Prevalence	15
1.5.1	Assessment Tools	15
1.5.2	Reliability and Validity	16
1.5.3	Self-report Measures of Burnout	18
1.5.4	How often does Burnout occur?	21
1.6	Correlates, Causes and Consequences	22
1.6.1	Possible Antecedents of Burnout	24
1.6.2	Possible Consequences of Burnout	28
1.7	Theoretical Approaches to Explain Burnout	30
1.7.1	An Integrative Model	31
1.8	Prevention and Intervention of Burnout	33
1.8.1	Classification	33
1.8.2	Individual Level Interventions	35
1.8.3	Individual/Organizational Level Interventions	38
1.8.4	Organizational Level Interventions	42
2.	Artificial Neural Networks	47
2.1	Introduction to Neurocomputing	47
2.1.1	Biological Motivation	48
2.1.2	Evolution of Artificial Neural Networks	50
2.1.3	Categorization of Artificial Neural Networks	52
2.2	Artificial Neuron Model	53
2.2.1	Notation and Terminology	53
2.2.2	Single-Input Neuron	54
2.3	Basic Transfer Functions	55
2.3.1	Hard Limit Transfer Function	56
2.3.2	Linear Transfer Function	57
2.3.3	Sigmoid Transfer Function	57
2.3.4	Hyperbolic Tangent Sigmoid Transfer Function	58
2.3.5	Radial Basis Transfer Function (Gaussian Function)	59
2.4	Multiple-Input Neuron	60
2.5	Training Algorithms	61
2.6	Network Architectures	63
2.6.1	A Single Layer of Neurons	63
2.6.2	Multiple Layers of Neurons	64
2.7	Perceptron	66
2.7.1	Perceptron Learning Rule	68
2.7.2	The Perceptron Training Algorithm	69
2.7.3	Limitations of the Perceptron	70
2.8	Self-Organizing Map (SOM)	71
2.8.1	Competitive Learning	72
2.8.2	Kohonen Training Algorithm	78
2.8.3	Example of the Kohonen Algorithm	79
2.8.4	Problems with the Kohonen Algorithm	80
2.9	Multi-layer Feed-forward Networks	82
2.9.1	Hidden-Neurons	84
2.9.2	Back-propagation	85
2.9.3	Back-propagation Training Algorithm	91
2.9.4	Problems with Back-propagation	99
2.10	Radial Basis Function (RBF) Network	107
2.10.1	Functioning of the Radial Basis Network	111
2.10.2	The Pseudo Inverse (PI) RBF Training Algorithm	113
2.10.3	Example of the PI RBF Algorithm	116
2.10.4	The Hybrid RBF Training Algorithm	118
2.10.5	Example of the Hybrid RBF Training Algorithm	124
2.10.6	Problems with Radial Basis Function Networks	128
3.	Application of ANNs to Burnout Data	131
3.1	Introduction	132
3.1.1	The Nursing Profession	132
3.1.2	Burnout in Nurses	133
3.1.3	Objective	136
3.2	Data	137
3.2.1	Participants	137
3.2.2	Measures	138
3.2.3	Statistical Data Analysis	139
3.2.4	Variables used for the Development of the ANNs	139
3.3	Implementation of the NuBuNet (Nursing Burnout Network)	140
3.3.1	Self-Organizing Map (SOM)	141
3.3.2	Three-layer Feed-forward Back-propagation Network	143
3.3.3	Radial Basis Function Network	145
3.4	Processing the Data	146
3.4.1	Data Preparation (Pre-Processing)	146
3.4.2	Network Preparation and Training	149
3.4.3	Post-Processing	153
3.5	Results	153
3.5.1	Three-layer Feed-forward Back-propagation Network	154
3.5.2	Radial Basis Function Network (PI Algorithm)	169
3.5.3	Radial Basis Function Network (Hybrid Algorithm)	180
3.5.4	Comparison of the Results	198
3.6	Discussion	200
4.	References	209
4.1	Burnout	209
4.1.1	Internet Directions	220
4.2	Artificial Neural Networks	221
4.2.1	Internet Directions	227

Text Sample:

Possible Antecedents of Burnout:

Possible causes of burnout can be classified into personality variables, work-related attitudes, and work and organizational variables. Besides above mentioned variables, Table 1.3 exhibits socio-demographic variables, even if they are no causes of burnout. However these characteristics may be linked to other factors, like gender to role taking, role expectations, or ‘feeling type’. Similarly, age is not a cause of burnout but it may be related to age-dependent factors like occupational socialization.

The number of minus or plus signs in Table 1.3 on page 23 indicates the strength and the direction of the correlation with burnout, founded on three subjective criterions: (1) the number of studies that found clear evidence for the relationship; (2) the methodological quality of these studies; (3) the consistency of the results across studies.

Socio-demographic variables:

The most consistently to burnout connected socio-demographic variable is the age (Maslach, Schaufeli, & Leiter, 2001) Younger employees experience a higher burnout rate than those aged over 30 or 40 years or in other words, it seems that burnout takes place rather at the beginning of the career. This confirms the observation that burnout is negatively related to work experience. Some authors interpret the higher rate of burnout among the younger and less experienced persons as a reality shock.

The other biographical characteristics do not show such clear relationships with burnout, although there are some studies showing that burnout takes place more frequently amongst woman than men. One explanation may be that, as a result of additional responsibilities at home, working woman experience higher overall workloads compared with working men, and workload is in turn positively related to burnout.

Personality variables:

It is somewhat difficult to interpret the meaning of correlations of burnout with personality features since persons interact with situations in complex ways. Even a high relationship of a particular personality characteristic does not necessarily involve causality. However, there are many studies which show that a ‘hardy personality’, characterized by participation in daily activities, a feeling of control over events, and openness to change, is consistently related to all three dimensions of the MBI. In other words, the more hardy a person is, the less burned-out he or she will be (Maslach et al., 2001). Another strong related personal characteristic is neuroticism, which includes trait anxiety, hostility, depression, self-consciousness and vulnerability. A neurotic person is emotionally unstable and she or he seems to be predisposed to experience burnout (Schaufeli & Enzmann, 1998).

A person’s control orientation may either be external or internal. Individuals with an external control orientation attribute events and achievements to powerful others or to chance, whereas those with an internal control orientation ascribe events and achievements to their own effort, ability, and willingness to risk. External control orientated persons are, compared with internal control orientated persons, more emotionally exhausted, depersonalized, and experience feelings of personal accomplishment (Glass & McKnight, 1996).

Another interesting and important relationship was found between a person’s coping style and burnout. Those individuals who are burned out cope with stressful events in a rather passive, defensive way, whereas an active and confronting coping style is used by less burned out persons.

Work and organizational variables:

Workload and time pressure are highly related to emotional exhaustion but, and this is striking, practically unrelated to personal accomplishment. Role stress, role conflict, and role ambiguity correlate fairly to substantially with burnout.

Role theory (e.g., Jackson & Schuler, 1985; Katz & Kahn, 1978) suggests that inter-role conflict and tension often results in individuals who find it increasingly difficult to successfully execute each of their roles because of constrained resources (e.g., energy, time) or the incompatibility among different roles (e.g., employee roles vs. family roles). Specifically, role stress emerges from the impact of the environment on an individual’s ability to fulfill role expectations (Beehr & Glazer, 2005). During the past decades, the number of (especially female) individuals having two or more jobs (due to economical reasons) has increased steadily. In the light of role theory, this development which implicates role stress since these individuals have to fulfill two or, when the family role is included three roles, is associated with negative consequences for the individuals and the organizations.