Application of Neural Network Paradigms to a Study of Nursing Burnout

Application of Neural Network Paradigms to a Study of Nursing Burnout

1. Application of Neural Network Paradigms to a Study of Nursing Burnout Felix Ladstätter Eva Garrosa

2. Content Objectives Artificial Neural Networks (ANN) Application of ANN to a burnout study

3. Objectives Development of an ANN based instrument for burnout analysis in nurses. To show that ANN are superior to hierarchical stepwise regression.

4. Artificial Neural Networks Motivation The human brain can perform complex tasks much faster than conventional computers . In the beginning, the field was heavily motivated by the discoveries of the neurophysiologic foundations of the brain. ANN

5. Illustration of a biological neuron ANN

6. ANN Illustration of an artificial neuron

7. Types of learning (1) Supervised learning Training samples are necessary. Goal: Assignation of a certain output pattern to a given input pattern. ANN

8. Types of learning ( 2) Unsupervised learning The learning task consists only of input pattern. The learning algorithm tries to group similar input pattern into the same cluster. ANN

9. Advantages of ANN Missing data and outlier resistance Simultaneous analysis of several output variables Analysis of non linear relationships ANN

10. ANN Network Architectures Multi Layer Perceptron Radial Basis Network Kohonen Network

11. ANN Multi Layer Perceptron

12. Back-propagation Initialization Propagation Error calculation Calculation of the new weights Check of the end criterion Either: Back to step 2 Or: End ANN

13. ANN Radial Basis Function Network

14. ANN Kohonen Network

15. Application of ANNs on Burnout

16. Data 473 nurses of hospitals in Madrid (Spain) 89,6 % female 10,4% male 35% professional nurses 65% student nurses NuBuNet

17. Measure NBS – 78 questions in 3 blocks (1-4) Burnout (24) Hardy personality (17) Job stressors (37) Socio demographic information Age Job status NuBuNet

18. Implemented Burnout Model NuBuNet

19. Data preparation NuBuNet

20. Training MLP Back-propagation RBF-Network Pseudo-Inverse algorithm Hybrid algorithm NuBuNet

21. Results (1) NuBuNet

22. Results (2) NuBuNet

23. Results (3) NuBuNet Method Data Burnout Dimension 1 2 3 M R 2 R 2 R 2 R 2 Hierarchical stepwise Regression Modeling 0,51 0,42 0,46 0,42 Validation 0,41 0,39 0,37 0,39 RBF (Hybrid) Training 0,69 0,62 0,60 0,63 Validation 0,50 0,38 0,48 0,45 MLP Training 0,67 0,58 0,66 0,63 Validation 0,50 0,40 0,40 0,43

24. Results (4) NuBuNet

25. Results (5) Both ANN architectures produced about 25% better results than the hierarchical stepwise regression. The RBF-Network is better suited for the modeling of burnout than the MLP. NuBuNet

26. Results (6) Ladstätter , F. , Garrosa, E. , Badea, C. and Moreno, B.(2010) 'Application of artificial neural networks to a study of nursing burnout', Ergonomics, 53: 9, 1085 — 1096 NuBuNet

27. Thanks for your attention