Psychiatry Research 217 (2014) 171–176
Contents lists available at ScienceDirect
Psychiatry Research journal homepage: www.elsevier.com/locate/psychres
Examining the cognitive model of caregiving – A structural equation modelling approach Klaus Hesse n, Stefan Klingberg Department of Psychiatry and Psychotherapy, University Hospital Tuebingen, Calwerstr. 14, 72076 Tuebingen, Germany
art ic l e i nf o
a b s t r a c t
Article history: Received 8 March 2013 Received in revised form 5 February 2014 Accepted 8 February 2014 Available online 22 February 2014
The study tests assumptions from the “cognitive model of caregiving” (Kuipers et al., 2010), which aims to inform interventions for carers of people with psychoses. The sample comprised 61 relatives of patients with schizophrenia. Standardized psychological assessments were conducted twice within 6 months including Involvement Evaluation Questionnaire (IEQ-EU), a short form of the Symptom Checklist 90-R (SCL K9), the Family Questionnaire (FQ), scales measuring control attributions of the Illness Perception Questionnaire for Schizophrenia (IPQS-R) and emotions toward the ill relative. Structural equation modelling was used to analyse data. We identified two pathways: (a) from “attributing control to relatives” to distress, intermediated by anxiety for the patient and emotional overinvolvement (EOI), and (b) from “attribution control to patient” to distress, intermediated by anger about the patient and criticism. The model provided a good fit to the data and was successfully replicated at a second point in time. We were able to find supporting evidence for a cognitive model of caregiving. Control attributions and emotions of informal caregivers are important when interventions are planned reducing expressed emotion and burden of caregivers. & 2014 Elsevier Ireland Ltd. All rights reserved.
Keywords: Schizophrenia Expressed emotion Caregivers
1. Introduction The research on cognitive models plays an important role in the development of psychotherapeutic interventions. The current “cognitive model of caregiving” includes cognitive and affective variables leading to expressed emotion (EE) and burden of care (Kuipers et al., 2010). The model highlights the influence of appraisals like attributions and emotions such as fear and anger on the relationship between caregivers and the service user as well as on the burden of caregiving. In the 1980s and 1990s, high expressed emotion (EE) was seen as responsible for relapses in psychoses (Butzlaff and Hooley, 1998) and more recent findings also indicate a causal influence of EE on relapse (Hooley, 2007). However, the majority of recent findings on the role of EE suggest rather an interactive than a causal interpretation of EE (King, 2000; van Os et al., 2001; Lobban et al., 2006). Increased stress and depression among relatives are associated with high EE (Kuipers et al., 2006). Family interventions initially were conceptualised to reduce EE (Lobban et al., 2013) but there is mixed evidence regarding the effects of family intervention on EE (Gleeson et al., 2010). In order to reduce relapse
n
Corresponding author. Tel.: þ 49 7071 29 82288; fax: þ 49 7071 294141. E-mail address: [email protected] (K. Hesse).
http://dx.doi.org/10.1016/j.psychres.2014.02.023 0165-1781/& 2014 Elsevier Ireland Ltd. All rights reserved.
decreasing EE and improving communication patterns are not necessary in all cases (Bellack et al., 2000; Lenior et al., 2002). The deconstruction of EE and a cognitive view led to a better understanding of the role of psychosocial stress among the patients and their families (Barrowclough et al., 2003; AlvarezJimenez et al., 2010). The development and the consequences of marked criticism and emotional overinvolvement have different paths. One approach for understanding EE consists in an attributional view inspired by the theory of Weiner (1993). Attributions are cognitions about the perceived control of defined events or actions. One important conclusion of the attributional model of EE research is that relatives who judged patients to be responsible for their symptoms are more critical (Barrowclough and Hooley, 2003). Growing evidence suggests that critical comments (CC) are more predictive for the course of the illness than EOI (Marom et al., 2005). Furthermore, relatives with marked emotional overinvolvement (EOI) attribute positive events more often to the patients (Grice et al., 2009), show more self blame (Peterson and Docherty, 2004) and have poorer health (Breitborde et al., 2010). In contrast to the EE research, caregiver burden research usually refers to the stress-model of Lazarus as the theoretical frame-work (Lazarus and Folkman, 1984; Lazarus, 1991). Lazarus proposed a cognitive theory of emotions with appraisal and coping processes influencing emotions. A primary appraisal (e.g. to be concerned about psychotic symptoms) is correlated with burden (Quinn et al., 2003). In addition, secondary appraisals like attributions of control
172
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
are correlated with negative emotions such as anxiety, anger and depression of relatives (Hazel et al., 2004; Fortune et al., 2005). The concept of attributions has clear implications for treatment as attributions have a close relation to emotions and can be influenced by interventions. Some variables used in research of burden and EE but not highlighted in the “cognitive model of caregiving” (Kuipers et al., 2010) are not included in our model as illustrated in Fig. 1. These variables have either no reliable correlation to burden, or they cannot be changed by means of psychotherapy. Only few researchers use path models or structure equation models examining the burden of caregivers. Mostly hierarchical regressions are used and the included variables have only little relation to theory (e.g. Grandon et al., 2008). Socio-demography and psychopathology of service users are less important in models for family interventions, because they provide no new starting points for family interventions. Further, the explained variance for burden in studies using patient variables and socio-demographic variables from both relatives (Roick et al., 2006) and patients (Grandon et al., 2008) as predictors is rather small, so they can be excluded. Objective burdens such as financial loss are equally only minor predictors of subjective burden (Mory et al., 2002). In some studies the severity of symptoms is no predictor for distress of caregivers (Harvey et al., 2001). In addition, the diagnosis of the patient, regardless of whether he has an affective or a schizophrenic disorder, is not associated with increased burden (Ostman and Hansson, 2004). Hence psychopathology is not the key for understanding the burdens of caregiving. Secondary appraisals like attributions and coping are more powerful in explaining burden (Roick et al., 2006) and EE (Raune et al., 2004). The aim of this study is to test assumptions derived from the cognitive model of caregiving (Kuipers et al. 2010) which integrates views of the attributional theory of Weiner (1993) and the stress theory of Lazarus (Lazarus and Folkman, 1984; Lazarus, 1991). For this purpose we propose a working model (Fig. 1) which includes empirically testable hypotheses on the relationship between burden of caregivers, expressed emotions, attributions and emotions.
2. Method
having a minimum negative symptoms measured with a modified version of the PANSS negative factor. The interview was conducted by trained raters. The inclusion criteria for the relatives were (1) age over 18 years, (2) command of the German language sufficient to understand instructions and provide informed consent, and (3) at least 2 h face-to-face or telephone contact with the patient per week. Written informed consent, conforming to the local ethics committee guidelines, was obtained from all subjects. The protocol was approved by the local ethic committee (Project 289/2006). The relatives were contacted by the study therapists of the service user. After the service user and the relatives had given informed consent they were asked to complete the questionnaire (t0) and to bring it to an assessment with the therapist. Six months later they were asked by phone to complete the questionnaire a second time and to return it by mail or to bring it to a second conversation (t1). Only participants with data for both assessments were included in the analysis. N ¼ 75 relatives entered the study at t0, in t1 N¼ 61 relatives completed the second assessment, thus the loss to follow-up was 19%.
2.2. Measures Distress was measured by two indicators with different scopes as a latent variable. The SCL K9 measures symptoms in relatives. It is a nine item short form of the Symptom Checklist 90R containing the most selective items of the 90 Item version. Its reliability in the general population is good and correlation to the original version is high (Klaghofer and Brähler, 2001). The Involvement Evaluation Questionnaire “IEQ-EU” is a self-report instrument measuring burden with proven psychometric properties in six European languages (van Wijngaarden et al., 2000). The characteristics of the German version are comparable to the other languages (Bernert et al., 2001). As proposed for research, the total burden of caregivers was calculated as the mean of the 27 core-items of the IEQ-EU. EE was measured by the family questionnaire (FQ), the scales of which align well with the subcategories of the Camberwell Family Interview (Wiedemann et al., 2002). The reliability of the scales was good and the correlation of the two subscales was R ¼0.19. Emotions are measured by three visual analogue scales per emotion. The relatives were asked what they feel by thinking of the ill relative. For anxiety about the patient the emotional words anxiety, worry and suspension built a scale with α ¼ 0.74. For anger about the patient, anger, rage and wrath were chosen, the internal consistency of the scale being α ¼ 0.87. Attributions were measured with two five item scales extracted from the “illness perception questionnaire for schizophrenia – relatives version” (Lobban et al., 2005). The scales were translated, reversed items were replaced by positive formulations and an additional item was added to increase reliability. For example the item ‘Nothing I do will affect their mental health problems at all’ was replaced by‚ ‘What I do will affect their mental Health problems’ to avoid double negative. The item ‘It is important not to criticize them, so that they don’t relapse.’ has been added to the ‘control relative’ scale. The internal consistency of the internal control attribution “control relative” scale was α ¼ 0.74, and in the original version it had been 0.63. The external control attribution scale “control patient” has an internal consistency of α ¼ 0.68, and in the original version it had been α ¼ 0.66.
2.1. Subjects and procedure 2.3. Statistics The sample comprised 61 relatives of outpatients who participated in randomised controlled trial (RCT) for the treatment of negative symptoms with cognitive behavioural therapy in three German University Hospitals (TONES-study, ISRCTN25455020). The DSM-IV diagnosis of schizophrenia was confirmed by a structured clinical interview (SCID-I). The design and the characteristics of the patient sample of the TONES-study is described in detail elsewhere (Klingberg et al., 2009). In order to assess psychiatric symptomatology the Positive and Negative Syndrome Scale (PANSS) was used (Kay et al., 1987). The main inclusion criterion for the patients was
Control Relative
Based on the theoretical assumptions outlined in Section 1, a structural equation model was created, focussing on the role of attributions and emotions in explaining burden and EE. In contrast to multiple regressions, structural equation modelling is not limited to a single outcome and can be used to evaluate relations among variables that are free of measurement error. With the exception of “burden/distress” we abstained from constructing latent variables in order to reduce complexity.
Anxiety for Patient
Symptoms Caregivers Emotional Overinvolvement
Distress Criticism
Control Patient
Anger about Patient
Burden of Caregivers
Fig. 1. Working model derived from the cognitive model of caregiving (Kuipers et al., 2010).
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176 The statistical analysis was carried out with AMOS 17.0, estimating model parameters and testing the adequacy of the proposed model (Arbuckle, 2008). We used structure equation modelling because it allows to estimate the direct and indirect effects of multiple predictors on two outcome variables (burden and symptoms relatives). As in multiple regression analysis, each association between variables is adjusted for other associations between variables in the model. For these purposes, measures of absolute model fit (e.g., non-significant χ²) and parsimony have to be considered (Byrne, 2001). For a variety of such fit measures certain criteria have to be met in order for the structural equation model to be acceptable as a plausible explanatory model for the empirical data. The root meansquare error of approximation (RMSEA) should be under 0.05 for close and under 0.08 for acceptable fit (Arbuckle, 2008). The Tucker–Lewis Index (TLI) and the Comparative Fit Index (CFI) indicate a close fit when near or above 0.95 (Hu and Bentler, 1999). Hair et al., (1998) argued that the minimum sample size for structural equation modelling (SEM) must be greater than the minimum ratio of at least five respondents for each estimated parameter. Other authors contend that under conditions of normality a ratio of sample size to parameters of 2:1 can be sufficient (Herzog et al., 2007). In our model, the number of distinct parameters to be estimated is 20. We reach the demanded sample size with three respondents per parameter. However, with smaller N lower goodness of fit indices and higher standard errors have to be expected (Jackson, 2003). Since the generalisability of the findings with smaller N is limited, we checked the model at t1 again.
173
Table 1 Sample characteristics. Relatives (n¼61) Age: mean (S.D.) 54.7 (10.8) Gender males: n (%) 23 (38) Parents: n (%) 43 (71) More than 9 h direct contact a week: n (%) 32 (65) Living with patient: n (%) 28 (46) IEQ-EU Overall burden: mean (S.D.) 0.679 (0.438) SCL K9: mean (S.D.) 0.645 (0.608) Control relative: mean (S.D.) 2.780 (0.587) Control patient: mean (S.D.) 2.416 (0.596) Anxiety for the patient: mean (S.D.) 5.902 (2.255) Anger about patient: mean (S.D.) 2.115 (2.372) PANSS total: mean (S.D.) PANSS positive scale PANSS modified negative scale Years of illness: mean (S.D.)
Patients (n¼ 44)
34.318 (9.7) 15 (34)
56.475 10.787 20.377 9.045
(13.016) (2.950) (5.373) (7.572)
IEQ-EU: Involvement Evaluation Questionnaire; SCl K9: Symptom Checklist short form 9 items; PANSS: Positive and Negative Symptom Scale.
3. Results Table 1 gives demographic data and symptom scores of relatives and patients. As in most studies with relatives the majority are parents, especially mothers. About two-thirds have more than 9 h direct contact. Half of the relatives live with the service user. There are only a few first-episode patients and all suffer from negativesymptoms –this is one of the inclusion criteria for the psychotherapy study. Some but not all patients also suffer from persistent positive symptoms. Our sample consisted of a high percentage of female patients not representative for the whole sample of the TONES-study. The characteristics of the sample are presented in Table 1. First we tested the model displayed in Fig. 1. Although the key associations reached significance, the model did not fit the data well (χ² ¼33.59; d.f. ¼18; p ¼0.014; TLI ¼0.780; CFI ¼0.859; RMSEA ¼0.120). The modification indices of AMOS indicated covariances in the data not specified in the model. Next we added a direct path from anxiety to distress yielding a better, though not optimal, fit (χ² ¼24.68; d.f. ¼ 17; p¼ 0.102; TLI¼0.885; CFI ¼ 0.930; RMSEA ¼0.089). Adding a further path from “control patient” to EOI led to a well fitting model without any modification indices (χ² ¼16.47; d.f. ¼16; p¼ 0.421; TLI¼0.993; CFI ¼0.996; RMSEA ¼ 0.022). This final model is illustrated in Fig. 2. We checked the stability of the final model by applying it to the data obtained from the second time of measurement (t1). The model is fitting the data in t1 acceptable in all parameters, the RMSEA indicates only acceptable fit whereas all others indicated good fit as viewed in table 2. The model proved to be stable over time, regression sizes varied across time as seen in Fig. 2. While in t0 the pathways from “control relative” through anxiety and EOI to distress is much stronger than the pathways from “control patient” through anger and criticism to burden, in t1 the difference between the two paths is lower. All regressions are significant at both points in time except the path between criticism and burden at t0.
4. Discussion The major result of this study is a strong association between attributions and emotions of relatives. Two kinds of attributions led to different emotional consequences. The variable “control patient” describes that relatives see the control of symptoms by the patient. The attribution “control relative” measures the amount of control of patients' symptoms relatives attribute to themselves. There are several studies testing parts of the
presented model with different foci and methods. One major benefit of our structure equation model is the good presentability of complex results. With respect to previous research four aspects seem important. Firstly, we found clear evidence for an association between “control relative” and relatives' anxiety at both assessments. A similar result has been reported using a mixed measurement of anxiety and depression (Fortune et al., 2005). Secondly, the path between “control patient” and relatives’ anger has been confirmed in t0 and t1. In a meta-analysis on help giving behaviour the path between control of symptoms by the help seeker and anger about him was the strongest and most reliable across different models (Rudolph et al., 2004). Earlier studies found an association between hostility towards patient and perceived controllability of patients based on assessments with the Camberwell Family Interview, for a review see Barrowclough and Hooley (2003). Thirdly, we confirmed the paths between emotions and expressed emotion (EE). As hypothesised, in our results the path between anger about the patient and criticism was significant. We could not find any existing literature testing this association. However, this finding is consistent with the strong association of hostility and critical comments in EE research (Barrowclough and Hooley, 2003). The path between anxiety about the patient and emotional overinvolvement (EOI) has been confirmed, although the direction of these paths could not be conclusively deduced from theory. Neither the theory of Lazarus (1991) nor the article of Kuipers et al. (2010) gives a causal interpretation. Hence these paths could go in the other direction as well. Earlier studies found an association between self-criticism and EE (Docherty et al., 1998). Although not identical, self-criticism and internal attributions of symptoms have a great deal in common. However, EOI was not only associated with anxiety about the patient but also with “control patient” at t0 and t1, which was not part of our hypotheses. This finding could be interpreted as an emotional reaction to behaviour of the service user. The association between EE and controllability has been found by other researchers using a variety of methods (Hooley and Campbell, 2002). Finally, we found an association between EE and distress, although the regression from criticism to distress is not significant at t0. The published evidence on this topic is heterogeneous. When assessed with self-report instruments EE is more closely related to burden than in other studies (King et al., 2003). A cross-sectional study has shown that both EOI and CC are correlated to burden, but that only EOI is a longitudinal predictor for burden (Alvarez-
174
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
Control Relative
0.30/0.28
Anxiety for Patient
0.45/0.36 0.53/0.51 0.44/0.27
Emotional Overinvolvement
0.30/0.37
0.55/0.59
Distress
0.31/0.45 0.37/0.45
Criticism
0.18/0.43 0.70/0.59
Control Patient
0.33/0.33
Symptoms Caregivers
Anger about Patient
Burden of Caregivers
0.43/0.74
Fig. 2. Structure equation model explaining distress with attributions emotions and expressed emotion. Cross-sectional model on t0/ t1. The first coefficient on the arrows refers to t0, the second coefficient refers to t1. Rectangles indicate observed indicator variables. Ovals indicate unobserved latent variables. Numbers on single-headed arrows indicate standardized regression weights, on double-headed arrows standardized covariances between the unique variables of the observations. Bold paths represent the main objectives of the study. The italicised parameter shows a non-significant estimate. Error variables are omitted for reasons of clarity. There were no undefined matrices and no constrained parameters. The overall model fit in t0 was χ² ¼16.469, d.f. ¼16, p o 0.421. Fit indices: CFI ¼0.996, TLI¼ 0.993, RMSEA (LO–HI 90) ¼0.022 (0.000–0.122) and the fit in t1 was χ²¼ 21.249, p o 0.169; d.f. ¼ 16. Fit indices: CFI ¼0.968, TLI¼ 0.945, RMSEA (LO–HI 90) ¼0.074 (0.000–0.149).
Table 2 Fit indices.
χ² (p; d.f.) TLI CFI RMSEA (LO–HI 90)
Model t0
Model t1
16.469 (0.421;16) 0.993 0.996 0.022 (0.000–0.122)
21.249 (0.169; 16) 0.945 0.968 0.074 (0.000–0.149)
p ¼probability value; d.f. ¼ degrees of freedom; TLI ¼ Tucker–Lewis index; CFI¼ Comparative Fit Index; RMSEA ¼ root mean square error of approximation.
Jimenez et al., 2010). There is evidence that EE and burden have common properties but the associations are not high enough to regard it as the same construct. According to our results, anxiety for the service user is directly connected to distress. It is a path between perceived emotions and distress not mediated by the expression of these emotions. Anxiety for a severely ill relative is part of a normal emotional reaction and not necessarily a pathological, over-involved reaction. From another point of view, anxiety is a common emotion in states of distress. The study conducted by Hazel et al. (2004) showed partly similar results. Hazel found correlations of anxiety and anger to distress. However the measured emotions were not specific for relatives of schizophrenic patients. Instead general emotional states and distress were measured with a depression scale. The axis “control relative” through anxiety and EOI to distress seems to be stronger than the axis from “control patient” through anger and criticism to distress. In t1 the difference between the two paths is lower. In research with emotions in everyday life, anxiety was also more strongly correlated with distress than with anger (Scherer et al., 2004).
4.1. Clinical implications The increasing evidence for an important role of attributions for relatives' behaviour and emotions provide a rationale for family interventions with the aim of reducing burden and EE. The change of control attributions should be one focus when planning family interventions since they can influence feelings of anxiety and anger in caregivers towards the service user. The link between attributions and negative emotions supports the use of reattribution techniques common in cognitive behaviour therapy. The aim of a “flexible attributional stance” (Lopez et al., 1999, p.313) for relatives has been confirmed. In our study the link between attributions and negative emotions has been highlighted. Specific interventions for reattribution could be applied
based on the predominant emotion and attribution style of the relative. For the “control relative”, anxiety and EOI path the overestimated influence of the relatives should be addressed. When anger about the service user is the relative's predominant emotion and there is a bias in perceiving too much controllability of symptoms by the patient, another direction of reattribution techniques might be required. A shift towards an attribution uncontrollable by persons may reduce anger and criticism. Studies like these stress the importance of considering the role of relatives for the study and treatment of psychosis. The burden of caregivers is relatively stable over time (Parabiaghi et al., 2007) and can have a long-term influence on patient's quality of life (Cotton et al., 2010). There is evidence that the care of families matters for the course of psychosis (Garety et al., 2008) and psychological mechanisms like EE (Alvarez-Jimenez et al., 2010), coping (Onwumere et al., 2011) and control attributions do have an influence on burden of caregivers and therefore on the ability to maintain the role as a caregiver. This seems more important when considering that reallocation has taken place in the last decades from inpatient treatment to outpatient care (Hjortsberg et al., 2011). More research about cognitive factors of burden is needed, when the support of caring families is planned. 4.2. Limitations and methodological issues One consequence of the small sample size is that the model tested had to be simple. More latent variables were not applicable, other promising variables like coping, need for closure, or psychopathology could not be included. Although the sample size is small, the hypothesised paths remained stable over 6 months. A random effect is thus unlikely. However, since the sample has been obtained from a therapy study of service users, the caregivers were highly motivated and optimistic as the patients began a new therapy. In our sample the proportion of female patients is higher than usual because male service consumers tended not to be in contact with relatives. This may be attributed to a better course of illness in females due to less social withdrawal and suspiciousness. The carers in our study had to have a contact time of at least 2 h per week to the patient, whereas in other studies the inclusion criterion was ten hours. In studies on family intervention more frequent contact is required to maximise the possible influence of the family on the patient. In contrast, for the purpose of testing assumptions of a cognitive model we chose criteria which lead to higher degrees of generalisability. The second limitation is the assessment of EE with a questionnaire. The CFI has been called the gold-standard for the
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
assessment of EE (Hooley and Parker, 2006). The most frequently used shortcut for assessing EE is the so called 5 min speech sample which has good concurrent validity with the CFI and the FQ used in our study (Leeb et al., 1991; Wiedemann et al., 2002). Structural equation models suggest causality although this has not been conclusively proven by data. Furthermore, according to the transactional stress model appraisals (in our model control attributions) and emotions interact reciprocally (Lazarus, 1999). For planning interventions with the aim of reducing EE and burden of caregivers it is useful to look at antecedents that can be modified and with the two sets of attribution–emotion pairs we have candidates.
Acknowledgements The study was supported by the German Research Foundation (Deutsche Forschungsgemeinschaft), Grant Kl 1179/3-1. The authors thank the families who have participated in the TONES-Study; Maria Memmou (University Hospital Düsseldorf), Sabine Kossow and Anne Horz (University Hospital Frankfurt) for access to families; and Tim Albrecht, Andreas Wittorf and Sabine Schneider for helpful comments on an earlier draft of the manuscript. References Alvarez-Jimenez, M., Gleeson, J.F., Cotton, S.M., Wade, D., Crisp, K., Yap, M.B.H., McGorry, P.D., 2010. Differential predictors of critical comments and emotional over-involvement in first-episode psychosis. Psychological Medicine 40, 63–72. Arbuckle, J.W., 2008. AMOS 17.0, User's Guide. SPSS, Chicago. Barrowclough, C., Hooley, J.M., 2003. Attributions and expressed emotion: a review. Clinical Psychology Review 23, 849–880. Barrowclough, C., Tarrier, N., Humphreys, L., Ward, J., Gregg, L., Andrews, B., 2003. Self-esteem in schizophrenia: relationships between self-evaluation, family attitudes, and symptomatology. Journal of Abnormal Psychology 112, 92–99. Bellack, A.S., Haas, G.L., Schooler, N.R., Flory, J.D., 2000. Effects of behavioural family management on family communication and patient outcomes in schizophrenia. British Journal of Psychiatry 177, 434–439. Bernert, S., Kilian, R., Matschinger, H., Mory, C., Roick, C., Angermeyer, M.C., 2001. The assessment of burden on relatives of mentally ill people: the German version of the involvement evaluation questionnaire (IEQ-EU). Psychiatrische Praxis 28 (Suppl. 2), 97–101. Breitborde, N.J.K., Lopez, S.R., Kopelowicz, A., 2010. Expressed emotion and health outcomes among Mexican-Americans with schizophrenia and their caregiving relatives. Journal of Nervous and Mental Disease 198, 105–109. Butzlaff, R.L., Hooley, J.M., 1998. Expressed emotion and psychiatric relapse. A meta-analysis. Archives of General Psychiatry 55, 547–552. Byrne, B.M., 2001. Structural Equation Modeling With AMOS: Basic Concepts, Applications and Programming. Lawrence Erlbaum Associates, New Jersey. Cotton, S.M., Gleeson, J.F.M., Alvarez-Jimenez, M., McGorry, P.D., 2010. Quality of life in patients who have remitted from their first episode of psychosis. Schizophrenia Research 121, 259–265. Docherty, N.M., Cutting, L.P., Bers, S.A., 1998. Expressed emotion and differentiation of self in the relatives of stable schizophrenia outpatients. Psychiatry 61, 269–278. Fortune, D.G., Smith, J.V., Garvey, K., 2005. Perceptions of psychosis, coping, appraisals, and psychological distress in the relatives of patients with schizophrenia: an exploration using self-regulation theory. British Journal of Clinical Psychology 44, 319–331. Garety, P.A., Fowler, D.G., Freeman, D., Bebbington, P., Dunn, G., Kuipers, E., 2008. Cognitive-behavioural therapy and family intervention for relapse prevention and symptom reduction in psychosis: randomised controlled trial. British Journal of Psychiatry 192, 412–423. Gleeson, J.F.M., Cotton, S.M., Alvarez-Jimenez, M., Wade, D., Crisp, K., Newman, B., Spiliotacopoulos, D., McGorry, P.D., 2010. Family outcomes from a randomized control trial of relapse prevention therapy in first-episode psychosis. Journal of Clinical Psychiatry 71, 475–483. Grandon, P., Jenaro, C., Lemos, S., 2008. Primary caregivers of schizophrenia outpatients: burden and predictor variables. Psychiatry Research 158, 335–343. Grice, S.J., Kuipers, E., Bebbington, P., Dunn, G., Fowler, D., Freeman, D., Garety, P., 2009. Carers' attributions about positive events in psychosis relate to expressed emotion. Behaviour Research and Therapy 47, 783–789. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data Analysis. 5. Prentice Hall, Upper Saddle River, NJ. Harvey, K., Burns, T., Fahy, T., Manley, C., Tattan, T., 2001. Relatives of patients with severe psychotic illness: factors that influence appraisal of caregiving and psychological distress. Social Psychiatry and Psychiatric Epidemiology 36, 456–461.
175
Hazel, N.A., McDonell, M.G., Short, R.A., Berry, C.M., Voss, W.D., Rodgers, M.L., Dyck, D.G., 2004. Impact of multiple-family groups for outpatients with schizophrenia on caregivers' distress and resources. Psychiatric Services 55, 35–41. Herzog, W., Boomsma, A., Reinecke, S., 2007. The model-size effect on traditional and modified tests of covariance structures. Structural Equation Modeling 14, 361–390. Hjortsberg, C., Helldin, L., Hjarthag, F., Lothgren, M., 2011. costs for patients with psychotic illness: differences depending upon state of remission. Journal of Mental Health Policy and Economics 14, 87–93. Hooley, J.M., 2007. Expressed emotion and relapse of psychopathology. Annual Review of Clinical Psychology 3, 329–352. Hooley, J.M., Campbell, C., 2002. Control and controllability: beliefs and behaviour in high and low expressed emotion relatives. Psychological Medicine 32, 1091–1099. Hooley, J.M., Parker, H.A., 2006. Measuring expressed emotion: an evaluation of the shortcuts. Journal of Family Psychology 20, 386–396. Hu, L.T., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling 6, 1–55. Jackson, D.L., 2003. Revisiting sample size and number of parameter estimates: some support for the N: q hypothesis. Structural Equation Modeling 10, 128–141. Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophrenia Bulletin 13, 261–276. King, S., 2000. Is expressed emotion cause or effect in the mothers of schizophrenic young adults? Schizophrenia Research 45, 65–78. King, S., Ricard, N., Rochon, V., Steiger, H., Nelis, S., 2003. Determinants of expressed emotion in mothers of schizophrenia patients. Psychiatry Research 117, 211–222. Klaghofer, R., Brähler, E, 2001. Konstruktion und teststatistische Prüfung einer Kurzform der SCL-90-R. Zeitschrift für Klinische Psychologie und Psychotherapie 49, 115–124. Klingberg, S., Wittorf, A., Herrlich, J., Wiedemann, G., Meisner, C., Buchkremer, G., Frommann, N., Wolwer, W., 2009. Cognitive behavioural treatment of negative symptoms in schizophrenia patients: study design of the TONES study, feasibility and safety of treatment. European Archives of Psychiatry and Clinical Neuroscience 259, 149–154. Kuipers, E., Bebbington, P., Dunn, G., Fowler, D., Freeman, D., Watson, P., Hardy, A., Garety, P., 2006. Influence of carer expressed emotion and affect on relapse in non-affective psychosis. British Journal of Psychiatry 188, 173–179. Kuipers, E., Onwumere, J., Bebbington, P., 2010. Cognitive model of caregiving in psychosis. British Journal of Psychiatry 196, 259–265. Lazarus, R.S., 1991. Progress on a cognitive–motivational–relational theory of emotion. American Psychologist 46, 819–834. Lazarus, R.S., 1999. Stress and Emotion: A New Synthesis. Springer, New York. Lazarus, R.S., Folkman, S., 1984. Stress, Appraisal and Coping. Springer, New York. Leeb, B., Hahlweg, K., Goldstein, M.J., Feinstein, E., Mueller, U., Dose, M., Maganaamato, A., 1991. Cross-national reliability, concurrent validity, and stability of a brief method for assessing expressed emotion. Psychiatry Research 39, 25–31. Lenior, M.E., Dingemans, P.M., Schene, A.H., Hart, A.A., Linszen, D.H., 2002. The course of parental expressed emotion and psychotic episodes after family intervention in recent-onset schizophrenia. A longitudinal study. Schizophrenia Research 57, 183–190. Lobban, F., Barrowclough, C., Jones, S., 2005. Assessing cognitive representations of mental health problems. II. The illness perception questionnaire for schizophrenia: Relatives' version. British Journal of Clinical Psychology 44, 163–179. Lobban, F., Barrowclough, C., Jones, S., 2006. Does expressed emotion need to be understood within a more systemic framework? An examination of discrepancies in appraisals between patients diagnosed with schizophrenia and their relatives. Social Psychiatry and Psychiatric Epidemiology 41, 50–55. Lobban, F., Postlethwaite, A., Glentworth, D., Pinfold, V., Wainwright, L., Dunn, G., Clancy, A., Haddock, G., 2013. A systematic review of randomised controlled trials of interventions reporting outcomes for relatives of people with psychosis. Clinical Psychology Review 33, 372–382. Lopez, S.R., Nelson, K.A., Snyder, K.S., Mintz, J., 1999. Attributions and affective reactions of family members and course of schizophrenia. Journal of Abnormal Psychology 108, 307–314. Marom, S., Munitz, H., Jones, P.B., Weizman, A., Hermesh, H., 2005. Expressed emotion: relevance to rehospitalization in schizophrenia over 7 years. Schizophrenia Bulletin 31, 751–758. Mory, C., Jungbauer, J., Angermeyer, M.C., 2002. Financial burden on caregivers of schizophrenic patients: a comparison between parents and spouses. Psychiatrische Praxis 29, 175–180. Onwumere, J., Kuipers, E., Bebbington, P., Dunn, G., Freeman, D., Fowler, D., Garety, P., 2011. Coping styles in carers of people with recent and long-term psychosis. Journal of Nervous and Mental Disease 199, 423–424. Ostman, M., Hansson, L., 2004. Appraisal of caregiving, burden and psychological distress in relatives of psychiatric inpatients. European Psychiatry 19, 402–407. Parabiaghi, A., Lasalvia, A., Bonetto, C., Cristofalo, D., Marrella, G., Tansella, M., Ruggeri, M., 2007. Predictors of changes in caregiving burden in people with schizophrenia: a 3-year follow-up study in a community mental health service. Acta Psychiatrica Scandinavica 116, 66–76. Peterson, E.C., Docherty, N.M., 2004. Expressed emotion, attribution, and control in parents of schizophrenic patients. Psychiatry 67, 197–207.
176
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
Quinn, J., Barrowclough, C., Tarrier, N., 2003. The Family Questionnaire (FQ): a scale for measuring symptom appraisal in relatives of schizophrenic patients. Acta Psychiatrica Scandinavica 108, 290–296. Raune, D., Kuipers, E., Bebbington, P.E., 2004. Expressed emotion at first-episode psychosis: investigating a carer appraisal model. British Journal of Psychiatry 184, 321–326. Roick, C., Heider, D., Toumi, M., Angermeyer, M.C., 2006. The impact of caregivers' characteristics, patients' conditions and regional differences on family burden in schizophrenia: a longitudinal analysis. Acta Psychiatrica Scandinavica 114, 363–374. Rudolph, U., Roesch, S.C., Greitemeyer, T., Weiner, B., 2004. A meta-analytic review of help giving and aggression from an attributional perspective: contributions to a general theory of motivation. Cognition and Emotion 18, 815–848. Scherer, K.R., Wranik, T., Sangsue, J., Tran, V., Scherer, U., 2004. Emotions in everyday life: probability of occurrence, risk factors, appraisal and reaction patterns. Social Science Information Sur les Sciences Sociales 43, 499–570.
van Os, J., Marcelis, M., Germeys, I., Graven, S., Delespaul, P., 2001. High expressed emotion: marker for a caring family? Comprehensive Psychiatry 42, 504–507. van Wijngaarden, B., Schene, A.H., Koeter, M., Vazquez-Barquero, J.L., Knudsen, H.C., Lasalvia, A., McCrone, P., 2000. Caregiving in schizophrenia: development, internal consistency and reliability of the Involvement Evaluation Questionnaire – European Version – EPSILON Study 4. British Journal of Psychiatry 177, 21–27. Weiner, B., 1993. On sin versus sickness. A theory of perceived responsibility and social motivation. American Psychologist 48, 957–965. Wiedemann, G., Rayki, O., Feinstein, E., Hahlweg, K., 2002. The Family Questionnaire: development and validation of a new self-report scale for assessing expressed emotion. Psychiatry Research 109, 265–279.
Contents lists available at ScienceDirect
Psychiatry Research journal homepage: www.elsevier.com/locate/psychres
Examining the cognitive model of caregiving – A structural equation modelling approach Klaus Hesse n, Stefan Klingberg Department of Psychiatry and Psychotherapy, University Hospital Tuebingen, Calwerstr. 14, 72076 Tuebingen, Germany
art ic l e i nf o
a b s t r a c t
Article history: Received 8 March 2013 Received in revised form 5 February 2014 Accepted 8 February 2014 Available online 22 February 2014
The study tests assumptions from the “cognitive model of caregiving” (Kuipers et al., 2010), which aims to inform interventions for carers of people with psychoses. The sample comprised 61 relatives of patients with schizophrenia. Standardized psychological assessments were conducted twice within 6 months including Involvement Evaluation Questionnaire (IEQ-EU), a short form of the Symptom Checklist 90-R (SCL K9), the Family Questionnaire (FQ), scales measuring control attributions of the Illness Perception Questionnaire for Schizophrenia (IPQS-R) and emotions toward the ill relative. Structural equation modelling was used to analyse data. We identified two pathways: (a) from “attributing control to relatives” to distress, intermediated by anxiety for the patient and emotional overinvolvement (EOI), and (b) from “attribution control to patient” to distress, intermediated by anger about the patient and criticism. The model provided a good fit to the data and was successfully replicated at a second point in time. We were able to find supporting evidence for a cognitive model of caregiving. Control attributions and emotions of informal caregivers are important when interventions are planned reducing expressed emotion and burden of caregivers. & 2014 Elsevier Ireland Ltd. All rights reserved.
Keywords: Schizophrenia Expressed emotion Caregivers
1. Introduction The research on cognitive models plays an important role in the development of psychotherapeutic interventions. The current “cognitive model of caregiving” includes cognitive and affective variables leading to expressed emotion (EE) and burden of care (Kuipers et al., 2010). The model highlights the influence of appraisals like attributions and emotions such as fear and anger on the relationship between caregivers and the service user as well as on the burden of caregiving. In the 1980s and 1990s, high expressed emotion (EE) was seen as responsible for relapses in psychoses (Butzlaff and Hooley, 1998) and more recent findings also indicate a causal influence of EE on relapse (Hooley, 2007). However, the majority of recent findings on the role of EE suggest rather an interactive than a causal interpretation of EE (King, 2000; van Os et al., 2001; Lobban et al., 2006). Increased stress and depression among relatives are associated with high EE (Kuipers et al., 2006). Family interventions initially were conceptualised to reduce EE (Lobban et al., 2013) but there is mixed evidence regarding the effects of family intervention on EE (Gleeson et al., 2010). In order to reduce relapse
n
Corresponding author. Tel.: þ 49 7071 29 82288; fax: þ 49 7071 294141. E-mail address: [email protected] (K. Hesse).
http://dx.doi.org/10.1016/j.psychres.2014.02.023 0165-1781/& 2014 Elsevier Ireland Ltd. All rights reserved.
decreasing EE and improving communication patterns are not necessary in all cases (Bellack et al., 2000; Lenior et al., 2002). The deconstruction of EE and a cognitive view led to a better understanding of the role of psychosocial stress among the patients and their families (Barrowclough et al., 2003; AlvarezJimenez et al., 2010). The development and the consequences of marked criticism and emotional overinvolvement have different paths. One approach for understanding EE consists in an attributional view inspired by the theory of Weiner (1993). Attributions are cognitions about the perceived control of defined events or actions. One important conclusion of the attributional model of EE research is that relatives who judged patients to be responsible for their symptoms are more critical (Barrowclough and Hooley, 2003). Growing evidence suggests that critical comments (CC) are more predictive for the course of the illness than EOI (Marom et al., 2005). Furthermore, relatives with marked emotional overinvolvement (EOI) attribute positive events more often to the patients (Grice et al., 2009), show more self blame (Peterson and Docherty, 2004) and have poorer health (Breitborde et al., 2010). In contrast to the EE research, caregiver burden research usually refers to the stress-model of Lazarus as the theoretical frame-work (Lazarus and Folkman, 1984; Lazarus, 1991). Lazarus proposed a cognitive theory of emotions with appraisal and coping processes influencing emotions. A primary appraisal (e.g. to be concerned about psychotic symptoms) is correlated with burden (Quinn et al., 2003). In addition, secondary appraisals like attributions of control
172
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
are correlated with negative emotions such as anxiety, anger and depression of relatives (Hazel et al., 2004; Fortune et al., 2005). The concept of attributions has clear implications for treatment as attributions have a close relation to emotions and can be influenced by interventions. Some variables used in research of burden and EE but not highlighted in the “cognitive model of caregiving” (Kuipers et al., 2010) are not included in our model as illustrated in Fig. 1. These variables have either no reliable correlation to burden, or they cannot be changed by means of psychotherapy. Only few researchers use path models or structure equation models examining the burden of caregivers. Mostly hierarchical regressions are used and the included variables have only little relation to theory (e.g. Grandon et al., 2008). Socio-demography and psychopathology of service users are less important in models for family interventions, because they provide no new starting points for family interventions. Further, the explained variance for burden in studies using patient variables and socio-demographic variables from both relatives (Roick et al., 2006) and patients (Grandon et al., 2008) as predictors is rather small, so they can be excluded. Objective burdens such as financial loss are equally only minor predictors of subjective burden (Mory et al., 2002). In some studies the severity of symptoms is no predictor for distress of caregivers (Harvey et al., 2001). In addition, the diagnosis of the patient, regardless of whether he has an affective or a schizophrenic disorder, is not associated with increased burden (Ostman and Hansson, 2004). Hence psychopathology is not the key for understanding the burdens of caregiving. Secondary appraisals like attributions and coping are more powerful in explaining burden (Roick et al., 2006) and EE (Raune et al., 2004). The aim of this study is to test assumptions derived from the cognitive model of caregiving (Kuipers et al. 2010) which integrates views of the attributional theory of Weiner (1993) and the stress theory of Lazarus (Lazarus and Folkman, 1984; Lazarus, 1991). For this purpose we propose a working model (Fig. 1) which includes empirically testable hypotheses on the relationship between burden of caregivers, expressed emotions, attributions and emotions.
2. Method
having a minimum negative symptoms measured with a modified version of the PANSS negative factor. The interview was conducted by trained raters. The inclusion criteria for the relatives were (1) age over 18 years, (2) command of the German language sufficient to understand instructions and provide informed consent, and (3) at least 2 h face-to-face or telephone contact with the patient per week. Written informed consent, conforming to the local ethics committee guidelines, was obtained from all subjects. The protocol was approved by the local ethic committee (Project 289/2006). The relatives were contacted by the study therapists of the service user. After the service user and the relatives had given informed consent they were asked to complete the questionnaire (t0) and to bring it to an assessment with the therapist. Six months later they were asked by phone to complete the questionnaire a second time and to return it by mail or to bring it to a second conversation (t1). Only participants with data for both assessments were included in the analysis. N ¼ 75 relatives entered the study at t0, in t1 N¼ 61 relatives completed the second assessment, thus the loss to follow-up was 19%.
2.2. Measures Distress was measured by two indicators with different scopes as a latent variable. The SCL K9 measures symptoms in relatives. It is a nine item short form of the Symptom Checklist 90R containing the most selective items of the 90 Item version. Its reliability in the general population is good and correlation to the original version is high (Klaghofer and Brähler, 2001). The Involvement Evaluation Questionnaire “IEQ-EU” is a self-report instrument measuring burden with proven psychometric properties in six European languages (van Wijngaarden et al., 2000). The characteristics of the German version are comparable to the other languages (Bernert et al., 2001). As proposed for research, the total burden of caregivers was calculated as the mean of the 27 core-items of the IEQ-EU. EE was measured by the family questionnaire (FQ), the scales of which align well with the subcategories of the Camberwell Family Interview (Wiedemann et al., 2002). The reliability of the scales was good and the correlation of the two subscales was R ¼0.19. Emotions are measured by three visual analogue scales per emotion. The relatives were asked what they feel by thinking of the ill relative. For anxiety about the patient the emotional words anxiety, worry and suspension built a scale with α ¼ 0.74. For anger about the patient, anger, rage and wrath were chosen, the internal consistency of the scale being α ¼ 0.87. Attributions were measured with two five item scales extracted from the “illness perception questionnaire for schizophrenia – relatives version” (Lobban et al., 2005). The scales were translated, reversed items were replaced by positive formulations and an additional item was added to increase reliability. For example the item ‘Nothing I do will affect their mental health problems at all’ was replaced by‚ ‘What I do will affect their mental Health problems’ to avoid double negative. The item ‘It is important not to criticize them, so that they don’t relapse.’ has been added to the ‘control relative’ scale. The internal consistency of the internal control attribution “control relative” scale was α ¼ 0.74, and in the original version it had been 0.63. The external control attribution scale “control patient” has an internal consistency of α ¼ 0.68, and in the original version it had been α ¼ 0.66.
2.1. Subjects and procedure 2.3. Statistics The sample comprised 61 relatives of outpatients who participated in randomised controlled trial (RCT) for the treatment of negative symptoms with cognitive behavioural therapy in three German University Hospitals (TONES-study, ISRCTN25455020). The DSM-IV diagnosis of schizophrenia was confirmed by a structured clinical interview (SCID-I). The design and the characteristics of the patient sample of the TONES-study is described in detail elsewhere (Klingberg et al., 2009). In order to assess psychiatric symptomatology the Positive and Negative Syndrome Scale (PANSS) was used (Kay et al., 1987). The main inclusion criterion for the patients was
Control Relative
Based on the theoretical assumptions outlined in Section 1, a structural equation model was created, focussing on the role of attributions and emotions in explaining burden and EE. In contrast to multiple regressions, structural equation modelling is not limited to a single outcome and can be used to evaluate relations among variables that are free of measurement error. With the exception of “burden/distress” we abstained from constructing latent variables in order to reduce complexity.
Anxiety for Patient
Symptoms Caregivers Emotional Overinvolvement
Distress Criticism
Control Patient
Anger about Patient
Burden of Caregivers
Fig. 1. Working model derived from the cognitive model of caregiving (Kuipers et al., 2010).
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176 The statistical analysis was carried out with AMOS 17.0, estimating model parameters and testing the adequacy of the proposed model (Arbuckle, 2008). We used structure equation modelling because it allows to estimate the direct and indirect effects of multiple predictors on two outcome variables (burden and symptoms relatives). As in multiple regression analysis, each association between variables is adjusted for other associations between variables in the model. For these purposes, measures of absolute model fit (e.g., non-significant χ²) and parsimony have to be considered (Byrne, 2001). For a variety of such fit measures certain criteria have to be met in order for the structural equation model to be acceptable as a plausible explanatory model for the empirical data. The root meansquare error of approximation (RMSEA) should be under 0.05 for close and under 0.08 for acceptable fit (Arbuckle, 2008). The Tucker–Lewis Index (TLI) and the Comparative Fit Index (CFI) indicate a close fit when near or above 0.95 (Hu and Bentler, 1999). Hair et al., (1998) argued that the minimum sample size for structural equation modelling (SEM) must be greater than the minimum ratio of at least five respondents for each estimated parameter. Other authors contend that under conditions of normality a ratio of sample size to parameters of 2:1 can be sufficient (Herzog et al., 2007). In our model, the number of distinct parameters to be estimated is 20. We reach the demanded sample size with three respondents per parameter. However, with smaller N lower goodness of fit indices and higher standard errors have to be expected (Jackson, 2003). Since the generalisability of the findings with smaller N is limited, we checked the model at t1 again.
173
Table 1 Sample characteristics. Relatives (n¼61) Age: mean (S.D.) 54.7 (10.8) Gender males: n (%) 23 (38) Parents: n (%) 43 (71) More than 9 h direct contact a week: n (%) 32 (65) Living with patient: n (%) 28 (46) IEQ-EU Overall burden: mean (S.D.) 0.679 (0.438) SCL K9: mean (S.D.) 0.645 (0.608) Control relative: mean (S.D.) 2.780 (0.587) Control patient: mean (S.D.) 2.416 (0.596) Anxiety for the patient: mean (S.D.) 5.902 (2.255) Anger about patient: mean (S.D.) 2.115 (2.372) PANSS total: mean (S.D.) PANSS positive scale PANSS modified negative scale Years of illness: mean (S.D.)
Patients (n¼ 44)
34.318 (9.7) 15 (34)
56.475 10.787 20.377 9.045
(13.016) (2.950) (5.373) (7.572)
IEQ-EU: Involvement Evaluation Questionnaire; SCl K9: Symptom Checklist short form 9 items; PANSS: Positive and Negative Symptom Scale.
3. Results Table 1 gives demographic data and symptom scores of relatives and patients. As in most studies with relatives the majority are parents, especially mothers. About two-thirds have more than 9 h direct contact. Half of the relatives live with the service user. There are only a few first-episode patients and all suffer from negativesymptoms –this is one of the inclusion criteria for the psychotherapy study. Some but not all patients also suffer from persistent positive symptoms. Our sample consisted of a high percentage of female patients not representative for the whole sample of the TONES-study. The characteristics of the sample are presented in Table 1. First we tested the model displayed in Fig. 1. Although the key associations reached significance, the model did not fit the data well (χ² ¼33.59; d.f. ¼18; p ¼0.014; TLI ¼0.780; CFI ¼0.859; RMSEA ¼0.120). The modification indices of AMOS indicated covariances in the data not specified in the model. Next we added a direct path from anxiety to distress yielding a better, though not optimal, fit (χ² ¼24.68; d.f. ¼ 17; p¼ 0.102; TLI¼0.885; CFI ¼ 0.930; RMSEA ¼0.089). Adding a further path from “control patient” to EOI led to a well fitting model without any modification indices (χ² ¼16.47; d.f. ¼16; p¼ 0.421; TLI¼0.993; CFI ¼0.996; RMSEA ¼ 0.022). This final model is illustrated in Fig. 2. We checked the stability of the final model by applying it to the data obtained from the second time of measurement (t1). The model is fitting the data in t1 acceptable in all parameters, the RMSEA indicates only acceptable fit whereas all others indicated good fit as viewed in table 2. The model proved to be stable over time, regression sizes varied across time as seen in Fig. 2. While in t0 the pathways from “control relative” through anxiety and EOI to distress is much stronger than the pathways from “control patient” through anger and criticism to burden, in t1 the difference between the two paths is lower. All regressions are significant at both points in time except the path between criticism and burden at t0.
4. Discussion The major result of this study is a strong association between attributions and emotions of relatives. Two kinds of attributions led to different emotional consequences. The variable “control patient” describes that relatives see the control of symptoms by the patient. The attribution “control relative” measures the amount of control of patients' symptoms relatives attribute to themselves. There are several studies testing parts of the
presented model with different foci and methods. One major benefit of our structure equation model is the good presentability of complex results. With respect to previous research four aspects seem important. Firstly, we found clear evidence for an association between “control relative” and relatives' anxiety at both assessments. A similar result has been reported using a mixed measurement of anxiety and depression (Fortune et al., 2005). Secondly, the path between “control patient” and relatives’ anger has been confirmed in t0 and t1. In a meta-analysis on help giving behaviour the path between control of symptoms by the help seeker and anger about him was the strongest and most reliable across different models (Rudolph et al., 2004). Earlier studies found an association between hostility towards patient and perceived controllability of patients based on assessments with the Camberwell Family Interview, for a review see Barrowclough and Hooley (2003). Thirdly, we confirmed the paths between emotions and expressed emotion (EE). As hypothesised, in our results the path between anger about the patient and criticism was significant. We could not find any existing literature testing this association. However, this finding is consistent with the strong association of hostility and critical comments in EE research (Barrowclough and Hooley, 2003). The path between anxiety about the patient and emotional overinvolvement (EOI) has been confirmed, although the direction of these paths could not be conclusively deduced from theory. Neither the theory of Lazarus (1991) nor the article of Kuipers et al. (2010) gives a causal interpretation. Hence these paths could go in the other direction as well. Earlier studies found an association between self-criticism and EE (Docherty et al., 1998). Although not identical, self-criticism and internal attributions of symptoms have a great deal in common. However, EOI was not only associated with anxiety about the patient but also with “control patient” at t0 and t1, which was not part of our hypotheses. This finding could be interpreted as an emotional reaction to behaviour of the service user. The association between EE and controllability has been found by other researchers using a variety of methods (Hooley and Campbell, 2002). Finally, we found an association between EE and distress, although the regression from criticism to distress is not significant at t0. The published evidence on this topic is heterogeneous. When assessed with self-report instruments EE is more closely related to burden than in other studies (King et al., 2003). A cross-sectional study has shown that both EOI and CC are correlated to burden, but that only EOI is a longitudinal predictor for burden (Alvarez-
174
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
Control Relative
0.30/0.28
Anxiety for Patient
0.45/0.36 0.53/0.51 0.44/0.27
Emotional Overinvolvement
0.30/0.37
0.55/0.59
Distress
0.31/0.45 0.37/0.45
Criticism
0.18/0.43 0.70/0.59
Control Patient
0.33/0.33
Symptoms Caregivers
Anger about Patient
Burden of Caregivers
0.43/0.74
Fig. 2. Structure equation model explaining distress with attributions emotions and expressed emotion. Cross-sectional model on t0/ t1. The first coefficient on the arrows refers to t0, the second coefficient refers to t1. Rectangles indicate observed indicator variables. Ovals indicate unobserved latent variables. Numbers on single-headed arrows indicate standardized regression weights, on double-headed arrows standardized covariances between the unique variables of the observations. Bold paths represent the main objectives of the study. The italicised parameter shows a non-significant estimate. Error variables are omitted for reasons of clarity. There were no undefined matrices and no constrained parameters. The overall model fit in t0 was χ² ¼16.469, d.f. ¼16, p o 0.421. Fit indices: CFI ¼0.996, TLI¼ 0.993, RMSEA (LO–HI 90) ¼0.022 (0.000–0.122) and the fit in t1 was χ²¼ 21.249, p o 0.169; d.f. ¼ 16. Fit indices: CFI ¼0.968, TLI¼ 0.945, RMSEA (LO–HI 90) ¼0.074 (0.000–0.149).
Table 2 Fit indices.
χ² (p; d.f.) TLI CFI RMSEA (LO–HI 90)
Model t0
Model t1
16.469 (0.421;16) 0.993 0.996 0.022 (0.000–0.122)
21.249 (0.169; 16) 0.945 0.968 0.074 (0.000–0.149)
p ¼probability value; d.f. ¼ degrees of freedom; TLI ¼ Tucker–Lewis index; CFI¼ Comparative Fit Index; RMSEA ¼ root mean square error of approximation.
Jimenez et al., 2010). There is evidence that EE and burden have common properties but the associations are not high enough to regard it as the same construct. According to our results, anxiety for the service user is directly connected to distress. It is a path between perceived emotions and distress not mediated by the expression of these emotions. Anxiety for a severely ill relative is part of a normal emotional reaction and not necessarily a pathological, over-involved reaction. From another point of view, anxiety is a common emotion in states of distress. The study conducted by Hazel et al. (2004) showed partly similar results. Hazel found correlations of anxiety and anger to distress. However the measured emotions were not specific for relatives of schizophrenic patients. Instead general emotional states and distress were measured with a depression scale. The axis “control relative” through anxiety and EOI to distress seems to be stronger than the axis from “control patient” through anger and criticism to distress. In t1 the difference between the two paths is lower. In research with emotions in everyday life, anxiety was also more strongly correlated with distress than with anger (Scherer et al., 2004).
4.1. Clinical implications The increasing evidence for an important role of attributions for relatives' behaviour and emotions provide a rationale for family interventions with the aim of reducing burden and EE. The change of control attributions should be one focus when planning family interventions since they can influence feelings of anxiety and anger in caregivers towards the service user. The link between attributions and negative emotions supports the use of reattribution techniques common in cognitive behaviour therapy. The aim of a “flexible attributional stance” (Lopez et al., 1999, p.313) for relatives has been confirmed. In our study the link between attributions and negative emotions has been highlighted. Specific interventions for reattribution could be applied
based on the predominant emotion and attribution style of the relative. For the “control relative”, anxiety and EOI path the overestimated influence of the relatives should be addressed. When anger about the service user is the relative's predominant emotion and there is a bias in perceiving too much controllability of symptoms by the patient, another direction of reattribution techniques might be required. A shift towards an attribution uncontrollable by persons may reduce anger and criticism. Studies like these stress the importance of considering the role of relatives for the study and treatment of psychosis. The burden of caregivers is relatively stable over time (Parabiaghi et al., 2007) and can have a long-term influence on patient's quality of life (Cotton et al., 2010). There is evidence that the care of families matters for the course of psychosis (Garety et al., 2008) and psychological mechanisms like EE (Alvarez-Jimenez et al., 2010), coping (Onwumere et al., 2011) and control attributions do have an influence on burden of caregivers and therefore on the ability to maintain the role as a caregiver. This seems more important when considering that reallocation has taken place in the last decades from inpatient treatment to outpatient care (Hjortsberg et al., 2011). More research about cognitive factors of burden is needed, when the support of caring families is planned. 4.2. Limitations and methodological issues One consequence of the small sample size is that the model tested had to be simple. More latent variables were not applicable, other promising variables like coping, need for closure, or psychopathology could not be included. Although the sample size is small, the hypothesised paths remained stable over 6 months. A random effect is thus unlikely. However, since the sample has been obtained from a therapy study of service users, the caregivers were highly motivated and optimistic as the patients began a new therapy. In our sample the proportion of female patients is higher than usual because male service consumers tended not to be in contact with relatives. This may be attributed to a better course of illness in females due to less social withdrawal and suspiciousness. The carers in our study had to have a contact time of at least 2 h per week to the patient, whereas in other studies the inclusion criterion was ten hours. In studies on family intervention more frequent contact is required to maximise the possible influence of the family on the patient. In contrast, for the purpose of testing assumptions of a cognitive model we chose criteria which lead to higher degrees of generalisability. The second limitation is the assessment of EE with a questionnaire. The CFI has been called the gold-standard for the
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
assessment of EE (Hooley and Parker, 2006). The most frequently used shortcut for assessing EE is the so called 5 min speech sample which has good concurrent validity with the CFI and the FQ used in our study (Leeb et al., 1991; Wiedemann et al., 2002). Structural equation models suggest causality although this has not been conclusively proven by data. Furthermore, according to the transactional stress model appraisals (in our model control attributions) and emotions interact reciprocally (Lazarus, 1999). For planning interventions with the aim of reducing EE and burden of caregivers it is useful to look at antecedents that can be modified and with the two sets of attribution–emotion pairs we have candidates.
Acknowledgements The study was supported by the German Research Foundation (Deutsche Forschungsgemeinschaft), Grant Kl 1179/3-1. The authors thank the families who have participated in the TONES-Study; Maria Memmou (University Hospital Düsseldorf), Sabine Kossow and Anne Horz (University Hospital Frankfurt) for access to families; and Tim Albrecht, Andreas Wittorf and Sabine Schneider for helpful comments on an earlier draft of the manuscript. References Alvarez-Jimenez, M., Gleeson, J.F., Cotton, S.M., Wade, D., Crisp, K., Yap, M.B.H., McGorry, P.D., 2010. Differential predictors of critical comments and emotional over-involvement in first-episode psychosis. Psychological Medicine 40, 63–72. Arbuckle, J.W., 2008. AMOS 17.0, User's Guide. SPSS, Chicago. Barrowclough, C., Hooley, J.M., 2003. Attributions and expressed emotion: a review. Clinical Psychology Review 23, 849–880. Barrowclough, C., Tarrier, N., Humphreys, L., Ward, J., Gregg, L., Andrews, B., 2003. Self-esteem in schizophrenia: relationships between self-evaluation, family attitudes, and symptomatology. Journal of Abnormal Psychology 112, 92–99. Bellack, A.S., Haas, G.L., Schooler, N.R., Flory, J.D., 2000. Effects of behavioural family management on family communication and patient outcomes in schizophrenia. British Journal of Psychiatry 177, 434–439. Bernert, S., Kilian, R., Matschinger, H., Mory, C., Roick, C., Angermeyer, M.C., 2001. The assessment of burden on relatives of mentally ill people: the German version of the involvement evaluation questionnaire (IEQ-EU). Psychiatrische Praxis 28 (Suppl. 2), 97–101. Breitborde, N.J.K., Lopez, S.R., Kopelowicz, A., 2010. Expressed emotion and health outcomes among Mexican-Americans with schizophrenia and their caregiving relatives. Journal of Nervous and Mental Disease 198, 105–109. Butzlaff, R.L., Hooley, J.M., 1998. Expressed emotion and psychiatric relapse. A meta-analysis. Archives of General Psychiatry 55, 547–552. Byrne, B.M., 2001. Structural Equation Modeling With AMOS: Basic Concepts, Applications and Programming. Lawrence Erlbaum Associates, New Jersey. Cotton, S.M., Gleeson, J.F.M., Alvarez-Jimenez, M., McGorry, P.D., 2010. Quality of life in patients who have remitted from their first episode of psychosis. Schizophrenia Research 121, 259–265. Docherty, N.M., Cutting, L.P., Bers, S.A., 1998. Expressed emotion and differentiation of self in the relatives of stable schizophrenia outpatients. Psychiatry 61, 269–278. Fortune, D.G., Smith, J.V., Garvey, K., 2005. Perceptions of psychosis, coping, appraisals, and psychological distress in the relatives of patients with schizophrenia: an exploration using self-regulation theory. British Journal of Clinical Psychology 44, 319–331. Garety, P.A., Fowler, D.G., Freeman, D., Bebbington, P., Dunn, G., Kuipers, E., 2008. Cognitive-behavioural therapy and family intervention for relapse prevention and symptom reduction in psychosis: randomised controlled trial. British Journal of Psychiatry 192, 412–423. Gleeson, J.F.M., Cotton, S.M., Alvarez-Jimenez, M., Wade, D., Crisp, K., Newman, B., Spiliotacopoulos, D., McGorry, P.D., 2010. Family outcomes from a randomized control trial of relapse prevention therapy in first-episode psychosis. Journal of Clinical Psychiatry 71, 475–483. Grandon, P., Jenaro, C., Lemos, S., 2008. Primary caregivers of schizophrenia outpatients: burden and predictor variables. Psychiatry Research 158, 335–343. Grice, S.J., Kuipers, E., Bebbington, P., Dunn, G., Fowler, D., Freeman, D., Garety, P., 2009. Carers' attributions about positive events in psychosis relate to expressed emotion. Behaviour Research and Therapy 47, 783–789. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data Analysis. 5. Prentice Hall, Upper Saddle River, NJ. Harvey, K., Burns, T., Fahy, T., Manley, C., Tattan, T., 2001. Relatives of patients with severe psychotic illness: factors that influence appraisal of caregiving and psychological distress. Social Psychiatry and Psychiatric Epidemiology 36, 456–461.
175
Hazel, N.A., McDonell, M.G., Short, R.A., Berry, C.M., Voss, W.D., Rodgers, M.L., Dyck, D.G., 2004. Impact of multiple-family groups for outpatients with schizophrenia on caregivers' distress and resources. Psychiatric Services 55, 35–41. Herzog, W., Boomsma, A., Reinecke, S., 2007. The model-size effect on traditional and modified tests of covariance structures. Structural Equation Modeling 14, 361–390. Hjortsberg, C., Helldin, L., Hjarthag, F., Lothgren, M., 2011. costs for patients with psychotic illness: differences depending upon state of remission. Journal of Mental Health Policy and Economics 14, 87–93. Hooley, J.M., 2007. Expressed emotion and relapse of psychopathology. Annual Review of Clinical Psychology 3, 329–352. Hooley, J.M., Campbell, C., 2002. Control and controllability: beliefs and behaviour in high and low expressed emotion relatives. Psychological Medicine 32, 1091–1099. Hooley, J.M., Parker, H.A., 2006. Measuring expressed emotion: an evaluation of the shortcuts. Journal of Family Psychology 20, 386–396. Hu, L.T., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling 6, 1–55. Jackson, D.L., 2003. Revisiting sample size and number of parameter estimates: some support for the N: q hypothesis. Structural Equation Modeling 10, 128–141. Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophrenia Bulletin 13, 261–276. King, S., 2000. Is expressed emotion cause or effect in the mothers of schizophrenic young adults? Schizophrenia Research 45, 65–78. King, S., Ricard, N., Rochon, V., Steiger, H., Nelis, S., 2003. Determinants of expressed emotion in mothers of schizophrenia patients. Psychiatry Research 117, 211–222. Klaghofer, R., Brähler, E, 2001. Konstruktion und teststatistische Prüfung einer Kurzform der SCL-90-R. Zeitschrift für Klinische Psychologie und Psychotherapie 49, 115–124. Klingberg, S., Wittorf, A., Herrlich, J., Wiedemann, G., Meisner, C., Buchkremer, G., Frommann, N., Wolwer, W., 2009. Cognitive behavioural treatment of negative symptoms in schizophrenia patients: study design of the TONES study, feasibility and safety of treatment. European Archives of Psychiatry and Clinical Neuroscience 259, 149–154. Kuipers, E., Bebbington, P., Dunn, G., Fowler, D., Freeman, D., Watson, P., Hardy, A., Garety, P., 2006. Influence of carer expressed emotion and affect on relapse in non-affective psychosis. British Journal of Psychiatry 188, 173–179. Kuipers, E., Onwumere, J., Bebbington, P., 2010. Cognitive model of caregiving in psychosis. British Journal of Psychiatry 196, 259–265. Lazarus, R.S., 1991. Progress on a cognitive–motivational–relational theory of emotion. American Psychologist 46, 819–834. Lazarus, R.S., 1999. Stress and Emotion: A New Synthesis. Springer, New York. Lazarus, R.S., Folkman, S., 1984. Stress, Appraisal and Coping. Springer, New York. Leeb, B., Hahlweg, K., Goldstein, M.J., Feinstein, E., Mueller, U., Dose, M., Maganaamato, A., 1991. Cross-national reliability, concurrent validity, and stability of a brief method for assessing expressed emotion. Psychiatry Research 39, 25–31. Lenior, M.E., Dingemans, P.M., Schene, A.H., Hart, A.A., Linszen, D.H., 2002. The course of parental expressed emotion and psychotic episodes after family intervention in recent-onset schizophrenia. A longitudinal study. Schizophrenia Research 57, 183–190. Lobban, F., Barrowclough, C., Jones, S., 2005. Assessing cognitive representations of mental health problems. II. The illness perception questionnaire for schizophrenia: Relatives' version. British Journal of Clinical Psychology 44, 163–179. Lobban, F., Barrowclough, C., Jones, S., 2006. Does expressed emotion need to be understood within a more systemic framework? An examination of discrepancies in appraisals between patients diagnosed with schizophrenia and their relatives. Social Psychiatry and Psychiatric Epidemiology 41, 50–55. Lobban, F., Postlethwaite, A., Glentworth, D., Pinfold, V., Wainwright, L., Dunn, G., Clancy, A., Haddock, G., 2013. A systematic review of randomised controlled trials of interventions reporting outcomes for relatives of people with psychosis. Clinical Psychology Review 33, 372–382. Lopez, S.R., Nelson, K.A., Snyder, K.S., Mintz, J., 1999. Attributions and affective reactions of family members and course of schizophrenia. Journal of Abnormal Psychology 108, 307–314. Marom, S., Munitz, H., Jones, P.B., Weizman, A., Hermesh, H., 2005. Expressed emotion: relevance to rehospitalization in schizophrenia over 7 years. Schizophrenia Bulletin 31, 751–758. Mory, C., Jungbauer, J., Angermeyer, M.C., 2002. Financial burden on caregivers of schizophrenic patients: a comparison between parents and spouses. Psychiatrische Praxis 29, 175–180. Onwumere, J., Kuipers, E., Bebbington, P., Dunn, G., Freeman, D., Fowler, D., Garety, P., 2011. Coping styles in carers of people with recent and long-term psychosis. Journal of Nervous and Mental Disease 199, 423–424. Ostman, M., Hansson, L., 2004. Appraisal of caregiving, burden and psychological distress in relatives of psychiatric inpatients. European Psychiatry 19, 402–407. Parabiaghi, A., Lasalvia, A., Bonetto, C., Cristofalo, D., Marrella, G., Tansella, M., Ruggeri, M., 2007. Predictors of changes in caregiving burden in people with schizophrenia: a 3-year follow-up study in a community mental health service. Acta Psychiatrica Scandinavica 116, 66–76. Peterson, E.C., Docherty, N.M., 2004. Expressed emotion, attribution, and control in parents of schizophrenic patients. Psychiatry 67, 197–207.
176
K. Hesse, S. Klingberg / Psychiatry Research 217 (2014) 171–176
Quinn, J., Barrowclough, C., Tarrier, N., 2003. The Family Questionnaire (FQ): a scale for measuring symptom appraisal in relatives of schizophrenic patients. Acta Psychiatrica Scandinavica 108, 290–296. Raune, D., Kuipers, E., Bebbington, P.E., 2004. Expressed emotion at first-episode psychosis: investigating a carer appraisal model. British Journal of Psychiatry 184, 321–326. Roick, C., Heider, D., Toumi, M., Angermeyer, M.C., 2006. The impact of caregivers' characteristics, patients' conditions and regional differences on family burden in schizophrenia: a longitudinal analysis. Acta Psychiatrica Scandinavica 114, 363–374. Rudolph, U., Roesch, S.C., Greitemeyer, T., Weiner, B., 2004. A meta-analytic review of help giving and aggression from an attributional perspective: contributions to a general theory of motivation. Cognition and Emotion 18, 815–848. Scherer, K.R., Wranik, T., Sangsue, J., Tran, V., Scherer, U., 2004. Emotions in everyday life: probability of occurrence, risk factors, appraisal and reaction patterns. Social Science Information Sur les Sciences Sociales 43, 499–570.
van Os, J., Marcelis, M., Germeys, I., Graven, S., Delespaul, P., 2001. High expressed emotion: marker for a caring family? Comprehensive Psychiatry 42, 504–507. van Wijngaarden, B., Schene, A.H., Koeter, M., Vazquez-Barquero, J.L., Knudsen, H.C., Lasalvia, A., McCrone, P., 2000. Caregiving in schizophrenia: development, internal consistency and reliability of the Involvement Evaluation Questionnaire – European Version – EPSILON Study 4. British Journal of Psychiatry 177, 21–27. Weiner, B., 1993. On sin versus sickness. A theory of perceived responsibility and social motivation. American Psychologist 48, 957–965. Wiedemann, G., Rayki, O., Feinstein, E., Hahlweg, K., 2002. The Family Questionnaire: development and validation of a new self-report scale for assessing expressed emotion. Psychiatry Research 109, 265–279.
