117UN MUNDO BIPOLAR
UN MUNDO BIPOLAR
5.3.1 Spearman Rank Correlation
The analysis of the individual questionnaires showed that there was variation in the total scores as reported by the carer. As one of the aims of this study is to assess for an association between variable and stress, it was felt necessary to initially perform a correlation between the stress score and the total scores of the individual questionnaires. This thesis is interested in understanding carer stress and the relationship with other variables. Thus, only questionnaires that were completed by the carer will be used in the correlation.
The PIP questionnaire will be used as a representation of the stress scores. This was chosen as the semi-structured stress questionnaire has not been used in published data, whilst the PIP is a stress questionnaire that has been previously validated and used as a measure of stress in other studies, see Table 11. It was therefore felt that the PIP provided the most valid representation of carer stress. The PIP generated two total scores, therefore, an average of the two scores was created for each carer that completed the questionnaire.
The analysis of the individual questionnaires showed that a normal distribution could not be assumed. Therefore, a Spearman Rank correlation was decided upon as it assumes a non-parametric distribution. A two-tail significance was decided as being most appropriate so to not assume a direction of probability.
The results of the Spearman Rank correlation can be seen below in Table 28, displaying the correlation coefficient and the significance of each value.
Table 28: Spearman Rank Correlation between Average PIP Score and the Total Score of Each Questionnaire
GHQ SDQ FSS Locus FNS Dys- cope Prob- cope Emo- cope TCR F-sat F- comm Correlation coefficient .460 .544 -.152 .184 .403 .392 .344 .320 -.115 -.198 -.111 Sig (2-tail) .000 .000 .262 .180 .002 .003 .010 .017 .406 .151 .422 N 56 56 50 55 56 55 55 55 54 54 54
Dys Cope= Dysfunctional Coping Strategies; Prob-cope= Problem-focused Coping Strategies; Emo-cope= Emotional-focused Coping Strategies; TCR= Total Circumplex Ratio;
The different colours correspond to the significance of the correlation coefficients. Those coloured red indicate that p<0.01 and those coloured blue indicate p<0.05. The strongest correlation was seen between the average PIP score and the SDQ score, r=0.544, p<0.01. From these results, it appears that the strongest association was between carer stress and the behaviour difficulties in a child, with more stress being seen as behaviour difficulties increase.
The next two strongest correlations were found between PIP and GHQ (r=0.460) and FNS (r=0.403). Both of these correlations were positive with a significance of p<0.01. However, the coefficient is still reasonably low, suggesting that there is a correlation but the association between increased carer stress and carer health or unmet needs are quite weak.
The three subgroups of the Brief COPE showed that the dysfunctional coping strategy (r=0.392, p<0.01) was the most associated to increased carer stress out of the three coping strategies that were measured. Problem-focused and emotional-focused coping strategies produced similar correlation coefficients but with a lower significance level p<0.05. This implies that there may be an association between more dysfunctional coping strategies being used and increased carer stress. However, the coefficient was low, intimating that the association is weak. The Spearman rank identified no correlation existed between stress scores and the total number of sources of support nor the locus of control.
Negative correlations were found between the average PIP score and the FSS, FACES total circumplex ratio (TCR), FACES family satisfaction (F-sat) and FACES family communication (F-comm). This trend in correlation suggests that as the number of sources of support available increases or the family functioning becomes more balanced with more satisfaction with the family and quality of communication, carer stress decreases.
However, all correlation coefficients were below 0.6, which indicates that the correlation being observed may not carry much meaning as ±0.6 is considered the cut- off value for a correlation to be considered a significant association (Conversation with statistician Dr Steven Lane). Therefore, a Mann Whitney U test will be performed to assess whether there are differences in scores between the carers that are classified as being stressed and those that are not.
5.3.2 Mann Whitney U Test
The Spearman Rank correlation suggested that there are some variables that may be associated with increased carer stress, notably, child behaviour, carer health and unmet needs. It was then decided to compare the total scores of the questionnaire that the carer completed between those that were categorised as being stressed or not stressed. As a normal distribution could not be assumed, a Mann Whitney U test was performed. Categorisation occurred using the average PIP score. A pre-determined cut-off value of 90 was decided upon, based on the mean of previous PIP scores found by studies that have used this measurement. Therefore, carers whose average PIP scores were <90, were classified as being not stressed, whilst those with scores >90 were classified as being stressed. This can be seen in Table 29.
Table 29: Categorisation of Carers Being Stressed Frequency Percentage (%)
Stressed (>90) 32 43
Not stressed (<90) 24 57
Using this categorisation, a Mann Whitney U Test was generated to indicate whether there are any significant differences between the stressed and not stressed groups when comparing the scores of the other variables that were measured via questionnaires. The results of the Mann Whitney U Test can be seen in Table 30.
The null hypothesis was that there is no difference between the questionnaire scores for those that are stressed and not stressed.
Using a significance level of 0.05, 4 questionnaires reported significant differences between carers that are stressed and those that are not. The questionnaires that indicated there was a difference were: General Health Questionnaire (GHQ-28), Strength and Difficulty questionnaire (SDQ), Family Needs Survey (FNS), Brief COPE Inventory subgroups emotional coping and dysfunctional coping. This indicates that these four questionnaires produce significantly different results depending upon whether or not the carer is stressed. These findings agree with the Spearman Rank correlation; that there may be an association between the scores of these questionnaires and the amount of stress being experienced by the carer.
The Mann Whitney U test identified that the most significant difference was with the dysfunctional coping strategy, closely followed by the SDQ and FNS. The findings from the Spearman Rank correlation along with the Mann Whitney U test, indicates that the questionnaires that will be of most use for future interventions in order to predict whether or not a carer is likely to be stressed, are the GHQ, FNS, SDQ and Brief COPE Inventory.
6.3.3 Binary Logistic Regression
In order to determine whether or not these questionnaires can be use in predicting carer stress, a binary logistic regression model was generated using the questionnaires that were significant within the Mann Whitney U test.
Using a binary logistic regression model, the dysfunctional coping strategy when combined with the total SDQ score, both remained significant (0.025 and 0.022 respectively). However, when the FNS total score was added into the formula, only the SDQ remained significant (0.035). This suggests that both the dysfunctional cope and the FNS variables are explaining similar variability in whether or not stress is being experienced by carers.
Therefore, two predictive models were found. When the dysfunctional coping variable was removed, the SDQ and FNS variables were added with either emotional coping strategy or with the GHQ variable. In both cases, the SDQ and FNS remained significant with the other two variables becoming insignificant. The similar findings were found when FNS was removed, and SDQ and dysfunctional coping were left in the equation. When the SDQ and dysfunctional coping variables were added with either emotional coping or the GHQ variables, only the SDQ and dysfunctional coping strategy remained significant. All workings can be seen in Appendix 4.
The binary regression suggests that a combination of either SDQ and dysfunctional coping variables or SDQ and FNS variables are most predictive in determining whether or not a carer is likely to be stressed. This suggests that the best model is based on SDQ and either FNS or dysfunctional coping. This is important to note as for future interventions, it appears that only the SDQ and one of the other questionnaires need to be given to the carer in order to predict carer stress.
Chapter Six
Discussion and Conclusion