Part Study 1: The predictive value of EE, social support and DUP/DUI in relation to outcome in first-episode psychosis 75 Chapter 7: 7.1 Method for Study Aims for Study Study had two main aims. 1. To determine whether the patient’s perceived level of EE in the family was related to his/her condition after one year and the extent to which the patient’s social network moderated this adaptation. 2. To determine whether the length of duration of untreated psychosis has an adverse effect on Expressed Emotion and on outcome. 7.2 1. Hypotheses for Study The family environment that is evidenced by lower EE will be associated with better patient outcome at one year as measured by lower symptoms, better social functioning and better quality of life than those families evidenced by higher EE. 2. Patients with a larger and more satisfying social network will have better outcomes (as above) than those with smaller, less satisfying networks. 3. Longer DUP and/or DUI will be associated with poorer outcome in the patient and with higher levels of EE. 7.3 Participants The participants for this study were patients of the Early Psychosis Intervention Programme (EPIP) at the Institute of Mental Health, Woodbridge Hospital, Singapore. Patients, between the ages of 15 and 40 with first-episode psychosis were referred to this programme. Patients with a diagnosis of psychosis 76 Table 7.1 Demographic details of patient sample (n=161) Variable Sex Educational level Employment status Marital status Race Religion No. of patients (%) Male 85 (53) Female 76 (47) Primary 18 (11) Secondary 48 (30) Tertiary 40 (25) Pre-University 19 (12) Vocational 17 (11) No education (0.6) Others (4) Employed 31 (19.1) Unemployed 63 (38.8) Student 22 (13.6) Homemaker 15 (9.3) National Service 12 (7.4) Single 126 (78) Married 29 (18) Divorced/separated/widowed (3.6) Chinese 135 (84) Malay 20 (12) Indian (3) Buddhism 59 (36) Christianity 29 (18) Islam 20 (12) Hinduism (1) Taoism 14 (9) Others 26 (16) 77 due to active substance abuse, or secondary to mental retardation, neurological disorder or any other medical or physical illness were excluded. Only patients who lived at home with family were included. Ethics approval for the study was obtained from the National Healthcare Group (NHG) Domain Specific Review Board and the Clinical Research Committee of Woodbridge Hospital. The sample consisted of 161 patients (mean age: 27 years, SD = 6.5, range: 16-40). Of the 335 patients accepted as part of the EPIP programme and available for the study, 42 (12.5%) were deemed unsuitable by the case manager, one died and 37 (11%) declined to participate. A further 95 (28%) patients were not considered stable or were not available within the 12 week time period. The remaining 161 patients (48%) formed the population for this study. The 11% refusal rate is in line with other studies, for example 3% (Barrelet et al., 1990), 12% (Vaughn & Leff, 1976) and 15% (Leff et al., 1987). Exclusion due to other causes in this study was 137 (41%), this is also similar to other studies, 30% (Macmillan et al, 1986) and 52% (Parker et al., 1988). Demographic details are given in Table 7.1. All clinicians complete the Structured Clinical Interview for DSM IV (SCID) to arrive at the diagnosis; the primary clinician in charge of the patient was responsible for completing the SCID (a breakdown of diagnoses is shown in Table 7.2). 7.4 Procedure Data collection took place between January 2001 and November 2004. When the case manager and psychiatrist considered the patient to be stable, the author was informed and subsequently invited the patient to participate in this study. Patients were seen within 12 weeks of first referral of psychosis. This was based on the fact that this has been shown to be the most practical time period (personal communication with Dr. Patterson of the Early Intervention Service in Birmingham, England). 78 Table 7.2 Patient DSM IV diagnoses DSM (IV) Diagnosis No. of patients (%) Schizophrenia type disorder 98 (65) Brief psychotic Disorder 13 (9) Schizoaffective Disorder (4) Delusional Disorder (3) Bipolar Disorder 20 (12) Depressive Disorder 10 (6) Psychotic disorder (not otherwise specified) (4) Patients were monitored regularly by case managers to check for any decline in psychiatric health status between check up appointments. If any significant change in status was reported, the patient was invited to make an appointment to see the Psychiatrist. If a patient missed any appointments they were also contacted by the case manager and if necessary a home visit was made. At six months and one year after initial contact, the patients were assessed by the psychiatrist using the same measures as at baseline. The psychiatrist obtained baseline measures on the patient, these included the Positive and Negative Syndrome scale (PANSS; Kay, Fiszbein & Opler, 1987), The Global Assessment of Functioning (GAF; American Psychiatric Association 1994), The World Health Organisation Quality of Life Assessment (WHOQOL-Bref; WHOQOL Group 1998), Duration of untreated psychosis (DUP) and Duration of untreated illness (DUI). The psychiatrists were blind to the EE status of the patient. 79 7.5 Instruments completed by the patient: 7.5.1 Expressed Emotion Expressed emotion was measured by the Level of Expressed Emotion scale (LEE; Cole & Kazarian 1988; Gerlsma, van del Lubbe, & van Nieuwenhuizen, 1992; Gerlsma & Hale, 1997). The original LEE of Cole and Kazarian (1988) was a 60 item self-report questionnaire consisting of four a priori sub scales: intrusiveness, attitude versus illness, expectancy/tolerance and emotional responsibility. Gerlsma et al. (1992) examined the factorial structure of the scale using exploratory factor analysis based on the original author’s suggestion that three of the four sub scales showed considerable intercorrelation and might be reflecting only one factor. Due to low item-total correlations 15 items were deleted. The remaining items were factor analyzed. Three factors emerged which Gerlsma et al. labeled: Lack of emotional support (19 items), intrusiveness (7 items) and irritability (7 items); reliability using Cronbach’s alpha was high at .89, .78, .79 respectively and .91 for the total scale. The factor analysis was confirmed by Startup (1999). One concern regarding lack of items explicitly referring to perceived criticism was addressed in a later study (Gerlsma & Hale, 1997) where a five item perceived criticism scale was included. In this study they found good internal consistency of the four factors using Cronbach’s alpha (lack of emotional support (LES) .92; irritability .88; intrusiveness .84, criticism .72, total scale .93). The items of the scale are listed in Appendix A, Table A2. Regarding predictive validity, the LEE has predicted depression improvement at month follow up (Gerlsma & Hale, 1997); and in relation to schizophrenia, relapse (Kopelowicz et al., 2002) and rehospitalisation (Cole & Kazarian, 1993; Donat,1996). Regarding construct validity, a strong relationship with relational dissatisfaction as measured by the Maudsley Marital Questionnaire (Arrundell et al., 1983 cited in Gerlsma & Hale, 1997) was found. Respondents are given the instruction, ‘the following statements describe the 80 ways in which a person may act towards you. Please think of your caregiver and indicate whether they have acted towards you in these ways during the past months. Respondents are then asked to answer on a point scale: Always Untrue, Sometimes untrue, Sometimes true, and Always True. After reverse scoring for several of the items, the scores are summed for each of the subscales; these are then summed to give a total score on perceived EE. This scale was chosen for this study for several reasons. It is short and easy to administer, an important consideration to avoid over-taxing respondents. It does not require lengthy training and has good reliability and validity. It also dispenses with the dichotomous high/low measure of the CFI that has previously been criticised (Hatfield, Spaniol, & Zipple, 1987). A further issue is that emotional overinvolvement (EOI) has been described as a ‘mixed concept’ (Chambless, Bryan, Aiken, Steketee, & Hooley, 1999, p. 73), comprising tendencies in the relative to be controlling and intrusive, in combination with positive aspects such as concern. Chambless et al. (1999) suggest that it may be beneficial to look at EOI as second order factor with warmth, intrusiveness, etc, used as first order factors, the existence of the Intrusiveness sub-scale in the LEE therefore, is a step towards this goal. This scale was formed using the original CFI measures as its base and has been found to correspond well with the CFI, having significant correlations with the Warmth and Critical comments scales of the CFI (Kazarian, Cole, Malla, & Baker, 1990). It is designed primarily to be given to the patient, shifting the focus to perceived EE rather than observed EE. Support for this idea comes from Hooley and Teasdale (1989) who used a straightforward question “How critical is your spouse of you?” and found that this accounted for more variance in relapse than EE measured by the CFI. The LEE has not been used in an Asian country before therefore a lengthy validation procedure was undertaken which is described later in this section. The scale which resulted from the process was labeled the LEE(S) to distinguish it from 81 that of Gerlsma et al. Briefly, the LEE(S) has 26 items and has retained the four factor structure of the LEE. Internal consistency scores as measured by Cronbach’s α were for the most part good (whole scale .88; LES .89; Criticism .80; Intrusiveness .76) except for the sub-scale Irritability (.46). This could have been due to the small number of items in this sub-scale (3). 7.5.2 Social Support Social support was measured using the Short Form Social Support Questionnaire (SSQ6; Sarason et al., 1987). This is a six-item short form version of the original 27 item SSQ (Sarason, Levine, Basham & Sarason, 1983). The short form has been found to correlate highly with the longer version (Sarason et al., 1987). Each item has two parts. The first asses the number of available others the individual feels he or she can turn to in times of need. For example, ‘Whom can you really count on to distract you from your worries when you feel under stress?’ The respondent has the option of answering ‘no one’ or providing the initials and relationship of up to nine individuals (e.g. IAG (sister)). The second part measures the degree of satisfaction with the perceived available support in the particular situation referred to. The respondents indicate their satisfaction on a 6-point likert scale from ‘very dissatisfied’ to ‘very satisfied’. The number of supports for each individual are then summed (range 0-54) to give a number of supports score. The satisfaction scores are also scored (range 6-36) to give a satisfaction score. The authors report high internal consistency for both number and satisfaction scales (alpha = 0.90-0.93), high test-retest reliability. This scale has been used in Singapore and has been found to be a reliable measure for this population. In a study of Singaporean parents of a sick child, Tan (1998) reported internal consistency for the number scale was .62 and for the satisfaction scale .43, and in a study with the Singapore police force, Tong et al., (2004) report internal 82 consistencies of .85 for number and .86 for satisfaction. In addition, the SSQ was also scored for number of supports listed as family members and, and number of supports listed as non-family. Tong et al., (2004) in their study on the Singapore police force found alpha scores for these two factors to be .85 and .89 respectively. This scale was chosen for this study because it has been successfully used with a population of this type before and has been found to be appropriate having good internal consistency, reliability, factor validity and construct validity (Furukawa, Harai, Hirai, Kitamura & Takahashi, 1999). Further in a review of social support assessment instruments, the SSQ was considered ‘most adequate’ (McDowell & Newell, 1987). 7.5.3 Measures to assess Construct Validity In addition to the above scales, a subset of patients completed the Family relationships index (FRI) (n=78) and the Chinese Parenting scale (n=34) to allow for testing of construct validity of the Level of Expressed Emotion scale. The Family Relationships Index (FRI: Holahan & Moos, 1981, 1982) The FRI comprises three subscales of the Family Environment Scale (FES: Moos & Moos, 1981, 1986) which is a 27-item index of the quality of family relationships. The three subscales form the relationship domain: ‘Cohesion’ which is the degree of commitment, help and support family members provide for one another, (e.g. ‘Families members really help and support one another’); ‘Expressiveness’ which is the extent to which family members are encouraged to act openly and to express their feelings directly, (e.g. ‘We say anything we want to around home’) and ‘Conflict’ which is the amount of openly expressed anger, aggression and conflict among family members. (e.g. ‘Family members often criticize each other’). The respondents answer True (1) or False (2) to the statements. The 83 responses for each of the sub-scales are then summed to produce scores. It has been shown to have good internal consistency (alpha = .89), moderate association between the three scales (r= .43) and good construct validity (Holahan & Moos, 1981). The Chinese Parenting Scale (Koh & Chang, 2002) The Chinese parenting scale comprises five sub-scales that can be subsumed into two super-ordinate factors: Control and Warmth. The Control factor includes the sub-scales of Indulgence (reversed) (e.g. ‘My parents will buy things that I like or request’) and Control/discipline (e.g. ‘My parents decide what I in my free time’). The Warmth factor includes the sub-scales of Teach (e.g. ‘As a response to bad grades my parents encourage me to try harder’), Physical nurturance (e.g. ‘When the weather turns cold, my parents make sure that I am kept warm’ and Noconcern (reversed) (e.g. ‘My parents not care what I in my daily life’). The respondents answer on a point scale from (strongly disagree) to (strongly agree). Once the relevant scores have been reversed, the item scores are totaled for each sub scale and summed to produce scores for Control and Warmth. This scale was developed for use in Singapore and has been used in several studies; the authors report good reliability and validity (personal communication with J. Koh). 7.6 Baseline and follow-up patient assessments conducted by the clinician (Psychiatrist) 7.6.1 Symptoms Symptoms were measured by The Positive and Negative Syndrome scale (PANSS; Kay et al., 1987). This is a widely used instrument for measuring severe psychopathology in adults with schizophrenia. The scale comprises 30 items and was designed to assess three main domains: the positive sub-scale (7 items), the 84 8.9.2 Model 2: DUI and LEE(S) sub-scales as predictors of outcome Although Model provided excellent fit to the data, it is highly recommended to test alternative plausible models (Kline, 1995). As it has been suggested (King, Ricard, Rochon, Steiger & Nelis, 2003) that it may be useful to examine the component parts of expressed emotion, this was tested in the next model (figure 8.3). Looking at the correlations of the sub-scales of the LEE (Table 8.11), whilst Criticism, Irritability and Intrusiveness all seemed to correlate very strongly with each other, Lack of Emotional Support (LES) seemed to be less so. In line with Koenigsberg & Handley’s (1986 ) suggestion that EE may be in fact bi-modal rather than unimodal, Criticism, Irritability and Intrusiveness were combined to form ‘Negative EE’. Table 8.11 Correlations between sub-scales of LEE(S) (n=109) LES LES Criticism Irritability Intrusiveness .10 .31* -.01 .57 .53* .40* Criticism Irritability Intrusiveness * p < .01 level (one-tailed) The literature also suggests that warmth may be an important factor in outcome (Bertrando et al, 1992; Ivanovic, Vuletic & Bebbington et al., 1994) acting as a buffer against the negative aspects of criticism, irritability etc. The sub-scale Lack of Emotional Support (LES) can be seen to be Lack of Warmth based on results of the small focus groups discussed in Appendix B, and the correlations with the 108 Warmth scale of the Chinese Parenting Scale. For ease of interpretation, the subscale LES was reversed to reflect emotional support (warmth). It was hypothesized that warmth would act as a buffer against the other sub-scales (Negative EE). Examination of Table 8.10 shows that the model provided an excellent fit to the data. However, the modification indices did suggest an additional path from Warmth to General Symptoms. In order to determine whether this additional path would provide a substantial improvement to the model, a χ2 difference test ( ∆χ2) was conducted. This test can only be conducted on nested models, with a significant ∆χ2 indicating substantial improvement (Byrne, 2001, p115). The model including Warmth was labeled Model 2a. The results of this test show that comparison of Model (χ2(21) = 23.21) with Model 2a (χ2(20) = 18.15) gives a ∆χ2 (1) = 5.06, p = .03, showing that the additional path from Warmth to General Symptoms did substantially improve the model. The fit details are given in Table 8.10. Compared to model 1, model 2a did not account for as much of the variance in General Symptoms (10%) or Satisfaction (17%), but did account for more of the variance in Number (10%). Models and 2a explained the same amount of variance in QOL suggesting that it is in fact LES which is the salient factor here. Goodness-of-fit measures, as Byrne (2001) points out, not actually define a ‘good’ model alone. In addition to statistical measures, theoretical aspects must be considered. All the models presented here, as well as presenting good fit, are also theoretically sound. Clearly models and 2a both provide an excellent fit to the data, however comparison of the AIC scores shows that Model is the preferred model. For ease of reference, the squared multiple correlations for the three models are shown in Table 8.12. 109 .46 DUI † Social Functioning .24** -.68*** .10 -.18* Negative EE † General Symptoms -.16* † Warmth .12 .17 .32*** .25** † Satisfaction With Social Supports .21** .10 .25** Quality Of Life † Number Of Social Supports Figure 8.3: Model 2: DUI, Negative EE and Warmth as predictors of outcome at one year (without path from Warmth to General Symptoms) and Model 2a (including path from Warmth to General Symptoms) † *** ** * Indicates path in original model but shown to be non-significant Indicates significant path in original model Indicates path added as suggested by modification indices Squared Multiple Correlations Path coefficient significant at .001 Path coefficient significant at .01 Path coefficient significant at .05 110 Table 8.12 Squared Multiple Correlations for models Variable Models 2a Social Functioning .47 .47 .46 General Symptoms .12 .06 .10 Quality of Life .12 .12 .12 Satisfaction .18 .17 .17 Number .06 .10 .10 EE 00 00 00 111 Chapter 9: 9.1 Discussion of Study Expressed Emotion There were two main questions to be addressed concerning EE in this part of the study. The first concerned EE’s role as a predictor of outcome; in a first-episode, Singaporean population, and as a predictor of functional outcome and quality of life in addition to symptomatological outcome. The second question concerned EE as an emerging or existing concept. A further issue concerned the influence of cultural factors outlined in the introduction on EE. Regarding the first question, in this study it was hypothesized that patients from low EE environments would show better outcome at one year than patients from high EE environments. This hypothesis was supported in this study with lower EE being associated with better Social Functioning, and less General Symptoms at one year. In addition, significant differences between the high and low EE groups were found for General Symptoms, Quality of Life and Social Functioning. Multivariate analyses showed a direct effect of EE on General Symptoms and an indirect effect on Social Functioning (through General Symptoms). A small indirect effect was also noted of EE on QOL. Further analyses showed that the sub-scale LES was associated with General Symptoms and Social Functioning, the path coefficients being only slightly lower than those between EE and these variables. So, clearly, in this study EE seems to be a contributing factor to outcome at one year, not only symptomatological but also functional outcome. In addition, it is clear that EE was able to predict outcome in this first-episode, Singaporean population. The results regarding outcome are in line with those found in the literature, with the majority of studies showing that patients from high EE families relapsed or had poorer outcome (e.g. Moline et al., 1985; Tarrier et al., 1988). In this study EE was found to predict general symptoms, showing that high EE was associated with 112 higher general symptoms in the patient at one year. Whilst this cannot be compared directly to studies which have looked at EE in relation to relapse, the basic relationship is the same, high EE is not associated with good outcome. The important issue is that in this Singaporean population, EE has been shown to be a predictor of patient symptoms. Knowing this, interventions can be aimed at high EE families to help them in coping with, and understanding, the illness and hopefully improving the family atmosphere. EE was also able to predict social functioning which supports Huguelet et al.’s (1995) study which found a relationship between EE and psychosocial adaptation in a study of first-episode patients. The lack of a direct relationship between EE and Social Functioning does not support findings from the literature on chronic cases (Barrowclough & Tarrier, 1990). This may be because over time persistent EE would have a direct effect on social functioning. Another possibility is that these authors found that it was the EE component ‘Hostility’ that was related to lower social functioning particularly, rather than EE as a whole. From the model using the bi-modal version of EE (Negative EE and Warmth) it can be seen that it is Lack of Emotional Support that strongly underlies the predictive power of EE in this study. The negative aspects of criticism, irritability and intrusiveness (Negative EE) not hold much predictive power so this could explain the lack of relationship between EE and Social Functioning in this study. This suggests that what characterizes EE in this population is not the addition of negative behaviours (such as criticism) but rather the withdrawal of support or warmth. The fact that EE, in particular the aspects of Criticism Intrusiveness and Irritability was not seen to have great predictive power could be due to the fact that the Chinese not readily express their emotions. As discussed earlier, Ow and Katz (1999) report that the Chinese, in comparison to more individualistic cultures, are unwilling to disclose stressful feelings. Regarding the second question, is EE an emerging or existing concept? 113 Looking at the scores for the LEE(S), it can be seen that of a possible range of 26112, the range of scores in this study was 26-92, as the scores were normally distributed it suggests that a substantial proportion could be considered to be high EE at this early stage of the illness. This therefore is evidence that EE is an existing concept, rather than one which emerges as the illness progresses. Knowing that high EE attitudes exist in families, rather than emerging as a result of the illness, has implications for the type and timing of interventions for these families. For instance, if high EE attitudes were found to be the result of the patient’s illness then educational interventions might prevent these attitudes from forming. If however they already exist in the caregivers then different approaches would need to be taken. Turning now to examine the extent of the influence of cultural factors in this population on EE. In most other studies, criticism has been shown to make the most contribution to relapse. Whilst not looking at relapse per se, criticism was not related to any of the outcome variables. In this population, lack of emotional support seemed to be the most important predictor of outcome. Why would this be the case? One reason comes from the small focus groups. The respondents suggested that if they were not shown love or concern then they would consider this criticism. So perhaps lack of emotional support is actually measuring criticism in this population. As Karanci and Inanduilar (2002) pointed out, in some cultures patients would expect their families to be protective or distressed and see these features as signs of caring – if they are absent then this would be stressful for the patient. It could also be, as mentioned in the introduction, that criticism in minor forms is an acceptable part of parenting in this population - it is seen as encouragement. Although certain emic items were added to the criticism sub-scale, it could still be the case that although criticism was measured, it is not viewed negatively and is not seen as a stressor. So, in this culture, emotional support may be particularly important, so lack of 114 it may be particularly difficult for these patients. This supports work done by Lopez, Nelson, Snyder & Mintz (1999) who found that for Mexican-Americans the lack of family warmth was the significant predictor of relapse. Lopez et al. suggest that for cultures such as these that emphasize close family ties, the importance of criticism and hostility may be dwarfed by the presence or absence of warmth. The concepts of individualism and collectivism may be able to shed some light on this. In Individualistic cultures (i.e. those that stress autonomy and independence) criticism may be seen as a personal attack and therefore holds great relevance for the person. However in collectivistic cultures where the group is more important that the self, lack of warmth could be seen to be going against the idea of the group, and group harmony (Kopelowicz et al., 2002). Intrusiveness was also not found to play a large role in predicting outcome. This could be because as mentioned previously, what is construed as intrusiveness or over-involvement in a Western culture is not necessarily the same for an Asian culture; the involvement of parents in the lives of their children is often seen as caring and concern, and not intrusiveness. In the small focus groups and student survey (reported in Appendix A), the respondents were clearly divided on what constituted intrusiveness. This would seem to reflect the mix of Chinese and Western influences seen in Singapore and could explain why this factor was not relevant here. Looking at the outcome in terms of recovery and non-recovery, the rates for this population compare well with those from Western studies. In this study, 78% met the criteria for symptom recovery, in Whitehorn et al.’s (2002) study of Canadian patients, 42% had achieved recovery at one year and Liberman et al (1993) found that 66% met the criteria. Following the criteria for recovery on Social functioning, 91.5% patients recovered, compared with 63% in Whitehorn’s study. The rates of recovery for this population are higher overall. Both the Whitehorn et al. and Liberman et al. studies were from Western cultures, and social support was not 115 measured. However, one explanation for the better rate of recovery seen in this population could be family support. It is clear from the literature that cultures where the extended family is the norm have better outcomes (e.g. Kurihara et al., 2000). Whilst the Singapore government is trying to encourage the extended family it is not clear how much progress in this regard has been made. However, this study did show that family support was an important factor and it could be that the notion of extended family is still relevant in this population. 9.2 Social Support Patients with a larger and more satisfying social network were hypothesized to have better outcomes than those with smaller, less satisfying networks. The results showed that this hypothesis was partially supported with more satisfaction with social support being associated with better quality of life. In terms of the association between social support and the symptomatic and social functioning outcomes there was no relation. A beneficial effect of non-family social support on outcome has been noted in several first-episode studies (e.g. Corin & Lauzon, 1992) but this was not found here. In this study, patients had more family members in their social network and found greater satisfaction with those members than with nonfamily members. An important finding is that social support was associated with QOL in this first-episode population, supporting similar findings from chronic populations (e.g. Lam & Rosenheck, 2000). This is an important finding as community based interventions have been successful in improving patients’ social networks, Becker et al. (1998) found that increasing the social network in this way led to improved QOL in a group of patients with psychosis. So, there is scope here to improve the social networks and ultimately the QOL of the patients. In looking at why social support might be related to better QOL, it is useful to 116 return to the idea of Social Support as a main effect. As discussed earlier, one of the main effects of social support is that it increases self-esteem or mastery in individuals which leads to a better health outcome. This could be the case here as self-esteem, mastery and a positive self concept have all been found to be associated with better QOl in the mentally ill. (Rosenfeld, 1992; Zissi, Barry & Cochrane, 1998). It could be the case that great satisfaction with social support raises self-esteem in these patients and leads to better QOL. Caron et al (2005) found that the social support variables attachment and reassurance of worth were the strongest predictors of QOl in a study of patients with schizophrenia. These variables were not examined in the present study but a future study should take a broader measure of social support, looking in detail at aspects such as attachment, and should include a measure of self-esteem to ascertain exactly which aspect of the social support is contributing to the patient’s QOL. 9.2.1 Network Size Regarding Number of social supports specifically, the literature suggests that larger numbers in the social network are associated with better mental health outcomes (Cohen & Sokolovsky, 1978), this was not supported here. Contrary to studies which have shown associations between numbers and negative symptoms (Hamilton et al.,1989), number of social supports was not related to any of the outcome measures. Number of social supports was however directly related to Satisfaction with social support suggesting that more people in the network led to greater satisfaction. Support from family members accounted for 70% of all support and compared with support from non-family members, patients found social support from family members more satisfying. This highlights the importance of family support for this group of patients. This is also related to the cultural considerations mentioned in the 117 introduction. Typically, Singaporeans have been found to rely on close family as their social support when a situation is thought to involve stigma or loss of face. These close family ties as social support in Asian families has also been recognized elsewhere (Uba, 1994 cited in Kung, 2001). 9.2.2 Social Support and EE Those patients who perceived their families as higher in EE also had lower satisfaction with, and smaller numbers of social supports, particularly family support. This shows support for Barrera and Baca’s (1990) finding that lower satisfaction with social support was associated with conflict in a family. Clearly in this study, higher EE indicates a negative family atmosphere which is associated with lower satisfaction with social support. 9.2.3 Social Support – Main effect or buffer? Regarding social support’s role as a buffer, whilst the literature shows that there is clear evidence of a buffering effect on clinical outcomes, this was only partly shown in this study. Social support buffered the effect of EE on QOL, but not the other outcome variables. It may prove beneficial in future to look at the patient’s locus of control as Dalgard et al. (1995) found that social support only worked as a buffer if the patient had an external locus of control. This may explain why social support did not buffer the effects of EE on symptoms and social functioning as well as QOL. Given these findings, it may be necessary to consider social support’s role as a main effect. The Relationship Perspective on social support as outlined by Lakey and Cohen (2000) is seen as having a main, rather than a buffering effect. This perspective suggests that it is not possible to separate measures of support from closely related aspects of relationships, either positive (such as companionship, 118 intimacy, etc) or negative (such as conflict, hostility, etc.) (Lakey & Cohen, 2000). Some of the positive relationships, such as companionship, have been found to be stronger predictors of well being than social support (Rook, 1987). Similarly negative concepts such as conflict have been found to be better predictors of health than perceived social support (Fiore, Becker & Coppel, 1985; Pagel, Erdly & Becker, 1987; Rook, 1984). The mechanisms which are thought to be underlying these associations are similar to those found in the main effect model; the positive relationships would encourage a greater sense of self esteem in the patient, and promote active coping (Rook, 1987). Support is found in this study for the relationships perspective. Positive aspects of the relationship (in this case low EE or Warmth) are related to Social Support. Warmth is also related to health outcomes (in this case symptoms). Symptoms and social support are not related (see Figure 9.1). +ve Social Support -ve Symptoms Warmth Figure 9.1: Hypothesized relationships between Warmth and Social Support and Warmth and Health Outcome (symptoms) McNally & Newman (1999) state that whilst the associations concerning positive social support and health can be inconsistent, the associations with negative exchanges within relationships and health are much more robust. Expressed 119 Emotion is a prime example of a relationship defined by negative exchanges having a detrimental effect on health. Further support for the Relationships Perspective comes from the strong negative correlations found between family social support and EE. No correlations were found between non-family support and EE, suggesting that lower EE could be construed as a measure of family support. 9.3 DUP and DUI In this study the main questions concerning DUP and DUI were 1) whether either or both could predict outcome and 2) how either or both was related to EE and outcome. Regarding the first question, longer DUP and/or DUI was hypothesized to be associated with poorer outcome in the patient. This hypothesis was partially supported in this study as although DUP was not found to be associated with any of the outcome measures, DUI was correlated with Social Functioning, Positive Symptoms and General Symptoms. The path analysis showed direct effects of DUI on General Symptoms and QOL, with an indirect effect on Social Functioning (-.18, p = .02) through General Symptoms. This finding is in contrast to Hafner et al.’s (1998) finding that DUI was associated with negative rather than positive symptoms but also supports their findings that DUI predicted nonspecific symptoms and social functioning. It also lends support to the large amount of research which has found the relationship between DUI and social functioning (e.g. Keshavan et al., 2003) and Browne et al. (1996) who found a relationship between DUI and QOL. It is interesting that DUP was not significantly associated with the outcome variables given the large amount of evidence for this relationship from the literature; however this study does lend support to the studies which have also found no relationship between DUP and outcome (e.g. Craig et al 2000). The second question concerned the situation regarding the interaction 120 between DUI and EE and outcome. Mintz, Mintz and Goldstein (1987) suggest three possible interactions between these variables, none of which are supported by this study, see Figure 9.2. 1. DUI EE OUTCOME 2. EE DUI OUTCOME 3. DUI EE OUTCOME Figure 9.2: Hypothesized interactions of DUI, EE and outcome as outlined by Mintz et al. (1987) The present study on the other hand, suggests that DUI and EE are independent predictors of outcome, see Figure 9.3. DUI EE OUTCOME Figure 9.3: Hypothesized interaction of DUI, EE and Outcome as suggested by present study. As to which is the more important in terms of predictive ability, this study showed clear evidence for independent routes for the two variables in terms of 121 symptoms. Looking at the amount of variance accounted for in General Symptoms supports this; together with EE, DUI accounted for 12% of General Symptoms, without EE, DUI accounted for 6% of General Symptoms, this suggests that they are contributing equally. Both LEE and DUI had moderate negative effects on Social Functioning through General Symptoms, but DUI was clearly the better predictor of QOL. Given that so much emphasis has been put on DUP in the literature, it remains to be seen just how important this finding is regarding DUI. One reason that DUP attracts so much attention is that it is a relatively easy period of illness to target, the patient is showing frank symptoms which are easy to detect. The patient presents him or herself to the psychiatrist who can then begin treatment. DUI on the other hand is more difficult to target, this is a time when the patient is showing prodromal symptoms which, given their nature, are very difficult to detect. The patient is (in most cases) not aware that he or she is ill and no contact is made with the psychiatrist. To access these people, programmes must be targeted at those most ‘at risk’ such as those with a relative who already has the disorder. Many ethical issues surround this sort of approach, and often prevent clinicians from targeting this group. However, this prodromal phase in first episode patients has been shown to be a particularly useful time in which to intervene to prevent the transition to full psychosis, and any further brain toxicity (Yung & McGorry 1996). Some studies have already attempted to target this prodromal period. For instance, Falloon (1992, cited in McGlashan et al, 1996) undertook a large study in England between 1984 and 1989. General Practitioners were trained to recognize the prodromal symptoms and alerted Falloon of any potential cases who were then evaluated and given education, stress management, and minimal medication as required. Of 1,000 cases referred, 15 appeared to have prodromal symptoms. One patient was found to have acute psychosis but of the other 14, none went on to full 122 psychosis. This is clear evidence that early intervention at the prodromal stage is beneficial. Clearly both DUP and DUI are important factors. In this population, DUI seems to be more important in predicting outcome, although this should be replicated in a further study. 9.4 Summary To sum up, EE and DUI are clear independent predictors of outcome. Further there is evidence that EE is an existing, rather than an emerging concept. The evidence for social support’s role as a buffer is limited, and an explanation involving social support’s role as a main effect is preferred with the suggestion that low EE is in fact a measure of family support. The importance of DUI rather than DUP in this population is also highlighted. A number of cultural factors are identified here. The most important would seem to be the significance of lack of emotional support rather than criticism for these patients. This suggests that it is the withdrawal of a positive attitude rather than the addition of a negative one that is crucial for this population. Collectivism is also suggested as being a factor in the role of emotional support. The importance of family for this group is also highlighted. 123 [...]... amount of variance in the model: 18% of satisfaction with social support, 6% of number of social supports, 12% of QOL, 12% of General Symptoms and 47% of Social Functioning Tables showing the decomposition of standardized effects for all the models are given in Appendix C Table 8.10 Goodness -of- Fit statistics for models 1, 2, 2a 2 (p) 2/ df CFI RMSEA (90% CI) Model1 7.57 (.87) 0.58 1.000 Model 2 23 .21 ... a significant ∆ 2 indicating substantial improvement (Byrne, 20 01, p115) The model including Warmth was labeled Model 2a The results of this test show that comparison of Model 2 ( 2( 21) = 23 .21 ) with Model 2a ( 2( 20) = 18.15) gives a ∆ 2 (1) = 5.06, p = 03, showing that the additional path from Warmth to General Symptoms did substantially improve the model The fit details are given in Table 8.10 Compared... studies were conducted The first involved discussion of the concepts in the sub-scales of the LEE with small focus groups to determine their relevance in Singapore and to elicit any emic items The second involved further examination of the relevance of the concepts using a survey of students These two studies gave good evidence for the existence of the concepts of the LEE in Singapore and the focus groups... would indicate good functioning in all areas, 71-80 no more than 85 a slight impairment in social functioning, 61-70 some difficulty in social functioning, 51-60 moderate difficulty and 41-50 serious difficulty in social functioning A score of 31-40 would indicate major impairment in several areas of social functioning, i.e social, family, work, etc and 21 -30 would indicate inability to function in almost... to be no rating greater than ‘3’ on any of the individual symptom items of the PANSS The criterion for recovery in terms of social functioning was defined as a GAF score equal to, or greater than 50 Quality of life was not explored in this way as a meaningful cut off point was not available Following these criteria, for symptom control, 100 patients (78%) recovered and 29 (22 %) were defined as non-recovery... Scores of 11 -20 and 1 -20 would indicate occasional and persistent inability to maintain personal hygiene respectively It has good reliability and validity (Whitehorn et al., 20 02) 7.6.3 DUP and DUI DUP and DUI are measured by the clinician DUP is defined as the onset of hallucinations and/or delusions, disorganized thinking and/or behavior to the time of appropriate treatment This is ascertained from... Number of social supports to Satisfaction with social supports (see Figure 8 .2) 105 .47 † Social Functioning DUI 26 ** -.68*** 12 00 Expressed Emotion -.18* † General Symptoms 24 ** † -.33*** - .24 * 18 Satisfaction With Social Support 28 ** 12 † 06 29 ** Quality Of Life † Number Of Social Supports Figure 8 .2: Model 1: DUI and EE as predictors of outcome at one year † *** ** * Indicates path in original model... therefore it had to be translated into Mandarin and validated for use in this population The translation and validation process of the LEE incorporated a series of steps which are outlined below along with a summary of the findings; full details of this process and findings are given in Appendix A 7.8.1 Conceptual and construct operationalisation equivalence In order to fully examine these concepts two studies... completing the scale Good reliability (Guttman split half: 80-.90) and validity (Intra Class Correlation; 87-.94) have also been reported in a study covering 15 centres worldwide, with a total sample of 4,804 respondents (Saxena, Carlson, Billington & Orley, 20 01) 87 Table 7.3 Facets incorporated within the domains of the WHOQOL-Bref* Domain Facets incorporated within domains Physical Health Activities of. .. model 1, model 2a did not account for as much of the variance in General Symptoms (10%) or Satisfaction (17%), but did account for more of the variance in Number (10%) Models 1 and 2a explained the same amount of variance in QOL suggesting that it is in fact LES which is the salient factor here Goodness -of- fit measures, as Byrne (20 01) points out, do not actually define a ‘good’ model alone In addition . family, work, etc. and 21 -30 would indicate inability to function in almost all areas. Scores of 11 -20 and 1 -20 would indicate occasional and persistent inability to maintain personal hygiene. involved further examination of the relevance of the concepts using a survey of students. These two studies gave good evidence for the existence of the concepts of the LEE in Singapore and the. difficulty in social functioning, 51-60 moderate difficulty and 41-50 serious difficulty in social functioning. A score of 31-40 would indicate major impairment in several areas of social functioning,