Deconstruction of the psychosocial concept of expressed emotion

In 1966, expressed emotion (EE) was conceived as a measure of the family emotional climate, the premise being that the expression of emotion in families had an impact on the member with psychiatric illness, namely schizophrenia. The Social Psychiatry Unit of the UK Medical Research Council created six family measures with construct validity: warmth, number of positive comments, severity of criticism, number of critical remarks, dissatisfaction and hostility.^{Reference Leff and Vaughn2} The final measure consisted of critical comments, hostility and emotional overinvolvement (EOI). These terms were considered normative at the time, but they may now sound culturally deaf and outdated. A more culturally appropriate and normative concept/language would be family emotional expressiveness/climate or family cohesion. In this article, I will continue to use the original term EOI in describing research findings that used the Camberwell Family Interview and the EOI concept.

The measurement of warmth was dropped because of its complex relationship with the other variables; however, the original researchers made several comments about warmth noted later in this article. A second study^{Reference Brown, Birley and Wing3} clarified that high EE was present when there was ‘hostility’ and seven or more critical comments. Using these parameters, patients with schizophrenia in high-EE families had a relapse rate of 58%, compared with 16% for those in low-EE families. The key relative's initial EE score was considered the best single predictor of relapse 9 months post-discharge. EE was the first psychosocial construct that assessed the family emotional climate using a semi-structured clinical interview, the Camberwell Family Interview, rather than a checklist of behaviours.^{Reference Brown and Rutter4}

The next step was to assess EE outside of south-east England. In 1984, Los Angeles (n = 69) was chosen as the next site, but the samples were required to be similar; male patients with schizophrenia, age 17–50 and ‘Caucasian in ethnic origin’. Similar results were found: 53% of family members scored high on critical comments, which was associated with higher patient relapse.^{Reference Vaughn, Snyder, Jones, Freeman and Falloon5}

Looking further afield, a sample of patients and their families extracted from the World Health Organization's Determinants of Outcome of Severe Mental Disorder study found that patients in Nigeria, India and Columbia had a more favourable course than patients living in Czechoslovakia, Denmark, Taiwan, the UK, the USA and the USSR, with researchers speculating that this was due to ‘differences in intensity of family bonds, and the type of family structure’.^{Reference Sartorius, Jablensky, Ernberg, Leff, Korten, Gulbinat, Hafner, Gattaz and Janzarik6} Further studies in Australia, Denmark, Poland, India, Egypt, Israel, China, Japan and Iran also found substantial EE variation on patient outcome.^{Reference Hashemi and Cochrane7}

In 1998, American researchers Butzlaff & Hooley^{Reference Butzlaff and Hooley8} summarised their findings on EE in a meta-analysis of 27 studies. They confirmed that EE was a valid predictor of relapse for people with schizophrenia, and called for an end to research designed to replicate the EE–relapse association. In 2017, these findings were reconfirmed, using the same samples plus two more samples^{Reference Weintraub, Hall, Carbonella, Weisman de Mamani and Hooley9} and employing p-curve analysis (www.p-curve.com/app3). Despite their call for an end to EE–relapse research, Butzlaff & Hooley found that the mean effect sizes (P = 0.004) across locations were significant when the studies were grouped into geographical areas: UK, Northern Europe, Southern Europe, Eastern Europe, North America, Australia and Asia. They recommended further examination of geographical location and the magnitude of the EE–relapse link. The early EE researchers summed it up best: ‘the nature of EE is clearly grounded in cultural conventions; it is culture specific […] but not ethnocentric or culture bound’.^{Reference Jenkins and Karno10} Bhugra & McKenzie^{Reference Bhugra and McKenzie11} opined that EE must be seen in cultural context and embedded in the normative data of the population before being considered in the pathogenesis of relapse.

What does EE actually measure?

As originally conceived, EE measures family emotional climate, an undefined interaction between family members. As mentioned above, the original measurement of EE is composed of three elements: critical comments, hostility and emotional overinvolvement (EOI). There is not a clear understanding of what is actually being measured; and initial investigators looked unsuccessfully for patient or relative traits. An initial hypothesis was that patients with more symptoms would evoke more family distress, but this was not found to be the case.^{Reference Heikkilä, Karlsson, Taiminen, Lauerma, Ilonen and Leinonen12,Reference Cutting, Aakre and Docherty13} No consistent relative variables were correlated with high EE, although these relatives were found to be more conventional, less satisfied with themselves, less flexible, less tolerant, lower in empathy,^{Reference Hooley and Hiller14} more conscientious,^{Reference King, Ricard, Rochon, Steiger and Nelis15} more burdened, more distressed and caring for a longer period of time.^{Reference Barrowclough and Parle16} Low-EE relatives were more fatalistic and less active in managing problems, whereas high-EE relatives were noted to be more active, more critical, disagreed with patients more and showed lower levels of accepting behaviour.^{Reference Hooley17} However, family beliefs about illness do show that high-EE relatives believed that patients could and ought to exert more control over their symptoms.^{Reference Hooley17} This finding led to the development of interventions based on family psychoeducation (FPE) that focus on illness education, resulting in reduced relapse rates.^{Reference McFarlane18} However, when FPE is applied across cultures, there is no single ‘universal’ successful FPE intervention, and McFarlane notes that the American version of FPE may mesh better with certain cultures than with other cultures.^{Reference McFarlane18} The differences in efficacy may lie in the quality of cultural adaptations, specific aspects of the studies’ design or cultural context.

Measuring EE in different cultural settings

As our understanding of the importance of sociocultural factors has grown, the importance of understanding the ‘what and how’ of measuring the family emotional climate in different cultures has also grown. The use of instruments and concepts in different sociocultural settings needs to be subjected to the following inquiry: (a) are we measuring the right concepts for the specific sociocultural context? (b) Once we decide on the correct concept, are we using the correct construct? If these are both correct, (c) do the instruments need cultural adaptation? (d) If we want to adapt instruments, what steps can we take to ensure fidelity to the original concept? (e) If we have an intervention based on the construct, is the intervention relevant for that cultural context? Specifically assessing EE, additional questions arise because EE is a complex concept with several internal components. Also, the measurement of EE using the Camberwell Family Instrument takes 2 h and the validity of shorter EE tools will require examination.

Is the reduction of EE culturally relevant?

Although it is often unfair to draw generalisations across cultures because this does not take into account within-group variation, some studies do suggest norms in specific cultural/ethnic settings. In African American families with a person with schizophrenia, high EE/high criticalness is associated with better patient outcomes.^{Reference Gurak and Weisman de Mamani19} On qualitative analysis, patients perceived high-EE family comments as ‘direct and expressive’, supportive and an expression of concern, whereas low-EE family comments were perceived as passive and uncertain. So, interventions to reduce EE will not be helpful for patients in African American families.

In Mexican American families with a person with schizophrenia, high EE/high EOI is associated with relapse, even when sampling (total n = 224) differed across time (1980 and 2000), recruitment sites (in-patient versus out-patient), years of illness, illness course and different interviewers/coders.^{Reference López, Ramírez García, Ullman, Kopelowicz, Jenkins and Nicholas20} In one sample, a subset of ‘Anglo-Americans’ (n = 54) had the same EE scores as Mexican Americans, but the EE score had different internal components: Anglo-Americans had 72% criticism/hostility and 8% EOI, whereas Mexican Americans who had 44% criticism/hostility and 40% EOI. An intervention focused on reducing criticalness will, at best, be irrelevant for families that have high EOI.

Interventions for Mexican Americans should therefore focus on reducing EOI. Kopelowitz and colleagues^{Reference Kopelowicz, Zarate, Wallace, Liberman, Lopez and Mintz21} successfully adapted an FPE intervention for Mexican American families. Before the intervention, a systematic assessment of each family enabled the clinical researchers to capture the cultural elements considered important for the adaptation. As a result, family members (n = 174) were successfully taught how to support medication adherence. Warmth and moderate EOI emerged as good prognostic factors.

Warmth

Warmth was an original component in the EE construct but was dropped because of its ‘complex relationships with the other components, which largely explained its relationship to relapse’ according to Leff & Vaughn^{Reference Leff and Vaughn2} (p. 23), who added that ‘if relatives showed considerable warmth in the absence of criticism and over-involvement, the patient tended to remain well’ (p. 83). The original observers of the importance of family emotional climate also predicted that if critical comments and EOI were controlled, then warmth would predict a better outcome.^{Reference Brown, Birley and Wing3} After 40 years, the importance of this research is now re-emerging.

In the Camberwell Family Interview, warmth was conceptualised as ‘the expression of positive feeling’, such as ‘the tone of voice including a display of enthusiasm, positive changes in manner and tone v. flatness, or the presence of coldness’, spontaneous expressions of warmth, expressions of sympathy, concern and empathy, especially when the relative is describing symptoms, and display of interest in the person and enthusiasm for their achievement. Evaluators were directed to pay attention to expressions of warmth about the patient, and warmth was measured on a 6-point scale (0, no warmth; 5, high warmth).

When family warmth is high, several international studies show that individuals with psychosis have a better outcome.^{Reference López, Ramírez García, Ullman, Kopelowicz, Jenkins and Nicholas20,Reference Kopelowicz, Zarate, Wallace, Liberman, Lopez and Mintz21} In Mexican American families, the best predictors of good patient outcome were high warmth and moderate EOI.^{Reference López, Ramírez García, Ullman, Kopelowicz, Jenkins and Nicholas20} In families from Barcelona (n = 21 dyads), high warmth as measured by the Camberwell Family Interview was correlated with fewer patient symptoms.^{Reference Medina-Pradas, Navarro, Pousa, Montero and Obiols22} In Puerto Rican children, high parental warmth was found to be protective against psychiatric disorders.^{Reference Santesteban-Echarri, Ramos-Olazagasti, Eisenberg, Wei, Bird and Canino23} In this study measurement of warmth included assessment of feelings of closeness, trust and anger towards the child. However, despite the increasing use of warmth as a predictor, the measurement of warmth still does not incorporate cultural/ethnic variation in the display of affection^{Reference Ruby, Falk, Heine, Villa and Silberstein24} or the cultural variation in the expression of warmth.^{Reference Lim25}

Cultural adaptation of interventions

Different methods have been proposed for cultural adaptation of interventions. Resnicow and colleagues^{Reference Resnicow, Baranowski, Ahluwalia and Braithwaite26} consider cultural adaptations as occurring across two dimensions: surface and deep. Surface adaptation of an intervention increases the receptivity and outward appeal of the intervention by alterations to language, visual representations, case scenarios, refinements of intervention materials and cultural sensitivity in recruitment and data collection, and clinical researchers often include treatment staff from the target group. This is fairly easy to do, and if done well does not affect the core of the intervention. On the other hand, deep structural adaptations revise core treatment components to align with the culture, social experiences and values of the group. Deep adaptations are more likely to threaten the fidelity of the intervention.

Wang-Schweig and colleagues^{Reference Wang-Schweig, Kviz, Altfeld, Miller and Miller27} suggest four steps for clinicians to ensure that an intervention is relevant and maintains fidelity. Step one examines the original theory underlying the intervention to determine whether the constructs are relevant to the population and align with the programme's core components. Step two uses a macro-level theory to identify and select key constructs most likely to be culturally dependent. Step three reviews the literature to determine whether there are cultural factors that might influence these constructs. Step four incorporates these culturally based constructs with the core constructs. This methodology assumes that there is culturally relevant literature available.

The FPE interventions of Kopelowitz and colleagues^{Reference Kopelowicz, Zarate, Wallace, Liberman, Lopez and Mintz21} for Mexican American families used Wang-Schweig et al's methodology by examining the underlying theory and selecting key constructs (moderate EOI and warmth) for implementation. Step three was met by their systematic assessment of each family to capture the cultural elements for the adaptation. Step four was met when the cultural elements were incorporated into the intervention.

Application of the concept of EE: recent examples

The extent of the current application of the concept of EE is illustrated in the following examples.

The use of the concept of EE is widespread in the medical and psychiatric literature, especially in the care of patients with chronic illness.^{Reference Wearden, Tarrier, Barrowclough, Zastowny and Rahill28} A southern California sample (n = 106) of individuals living with diabetes who had partners with high EE had poorer diabetes management.^{Reference Lister, Wilson, Fox, Herring, Simpson, Smith and Edwards29} However, in the analysis, warmth (P < 0.05), EOI (P < 0.05) and criticism (P < 0.05) were combined to create the factor of EE. EOI was later dropped because of a ‘quadratic relationship with outcome variables as well as significant multivariate outlier problems’. The ethnicity of the sample was approximately 50% White, 25% Hispanic, 13% African American and 8% Asian. There was no discussion of how ethnicity might be affecting the results.

In another example, neuroimagers stated that ‘high EE’ evoked activations in brain regions associated with aversive social information processing (amygdala, temporal pole, inferior frontal gyrus, anterior cingulate cortex and medial prefrontal cortex) and with right or bilateral language prosody areas.^{Reference Rylander, McKie, Elliott, Deakin and Tarrier30} The authors concluded that they had ‘confirmed the hypothesis for a distinct neural basis for high EE and that the pattern of activations offer a feasible biomarker approach for identifying neural substrates that mediate the influence of high EE on relapse’. The study was small (11 participants) and the conclusion seemed overreaching. However, the implication is that the high EE/criticalness paradigm is universally accepted.