Further development of a scale of perceived expressed emotion and its evaluation in a sample of patients with eating disorders

Abstract

High expressed emotion (EE) as measured by the Camberwell Family Interview (CFI) predicts the course of eating disorders (ED). Despite its important contribution to the field, the CFI has two major limitations; it is time-consuming and it does not consider the patient's perspective. Obtaining the patient's view may help shed light on the dyadic nature of caregiver's EE and the patient's illness course. The objectives of our study of 77 patients with ED were to develop further a brief measure to assess the patients' perceptions of their caregivers' EE, the patient version of the Brief Dyadic Scale of Expressed Emotion (BDSEE), and to evaluate its psychometric properties. Three clearly separate factors were identified: perceived criticism, perceived emotional overinvolvement, and perceived warmth. The BDSEE also demonstrated good levels of reliability and construct validity. The BDSEE subscales are significantly related to other measures of the perceived family emotional climate and to the CFI, the gold standard in the field of EE. The clinical implications of the ED patients' perceptions of their caregivers' EE are discussed. Findings support the utility of the expanded BDSEE for both research and clinical practise in assessing the perspective of patients with ED.

Introduction

Expressed emotion (EE) is one of the most well-established constructs in the study of families and psychopathology. EE reflects the quality of the relationship between the patient and the patient's closest caregiver (Brown and Rutter, 1966). EE is traditionally measured with the Camberwell Family Interview (CFI), an audiotaped semistructured interview that takes approximately 1½ hours to administer (Vaughn and Leff, 1976). The content of the interview and the tone of voice are used as a basis for rating the relatives' attitudes and emotions with respect to an ill family member. The five indices that comprise EE are: criticism (CC), hostility (H), emotional overinvolvement (EOI), warmth (W), and positive remarks (PR).

EE has been shown to be a good predictor of the outcomes of patients with eating disorders (ED) and their caregivers. Specifically, EE predicts treatment compliance, early treatment outcome, and long-term clinical outcomes for patients with ED (Wearden et al., 2000, Hooley, 2007). In fact, mothers' CC and EOI were better predictors of patients' poor outcome at 1-year follow-up than diagnosis, duration of illness, body mass index (BMI), age, gender, and other clinical variables (van Furth et al., 1996). Moreover, the mean effect size for EE and relapse was found to be greater in ED than in schizophrenia (Butzlaff and Hooley, 1998), though the number of ED studies included in the meta-analysis was much smaller for ED than schizophrenia. With regards to ED caregiver outcomes, high EE caregivers have greater levels of anxiety and negative/difficult caregivers behaviours than do low EE (Kyriacou et al., 2008).

Although EE as measured by the CFI is associated with the course of ED and other mental disorders, it has two major limitations. First, the CFI is a costly instrument to apply. Not only is it a lengthy interview but it requires coding which takes at least another 2 or 3 hours, as well as the many hours to train reliable coders. There have been several attempts to measure EE more easily and with fewer resources (Hooley and Parker, 2006). The Five Minute Speech Sample (FMSS; Magaña et al., 1986) and the Family Questionnaire (FQ; Wiedemann et al., 2002) were created as more efficient research-based alternatives. However, the FMSS has shown limited validity when compared to the criterion CFI in samples of patients with ED (van Furth et al., 1993, Rein et al., 2006). In addition, no conclusive validation data are available yet for the FQ in ED (Kyriacou et al., 2008). Therefore, an efficient and psychometrically sound alternative to the CFI in evaluating EE in caregivers of individuals with ED does not exist at this moment.

A second limitation of the CFI is that it only considers the caregivers perspective of the family's emotional climate. Many investigators agree that EE reflects a transactional process between the patient and the caregiver (Strachan et al., 1989), yet the primary measure of EE gives little voice to the patient's view of this transaction. There are some self-report measures, however, that do consider the patient's point of view. For example, Hooley and Teasdale (1989) developed the Perceived Criticism (PC) Scale, which consists of one question “How critical is your relative of you?” and a rating scale from 1 (not at all critical) to 10 (very critical). In a sample of depressed individuals and spouses, both PC and EE were simultaneous, independent predictors of higher rates of relapse 9 months after hospital discharge. Moreover, PC was the strongest predictor, accounting for 16% of the variance in relapse rates above and beyond both EE and relationship satisfaction. There have been numerous attempts to replicate, extend, and explain Hooley and Teasdale's (1989) striking findings with a variety of populations and methodologies (for review, see Renshaw, 2008). Although there is some evidence of the PC scale's reliability and validity (Hooley and Teasdale, 1989, Riso et al., 1996, White et al., 1998, Chambless and Steketee, 1999, Chambless et al., 1999, Chambless et al., 2001, Bachmann et al., 2006), it measures only one of the three main indices of EE (Leff and Vaughn, 1985) and has never been applied to the study of ED.

To date, the only available data regarding ED patients' perspective of their caregivers' EE comes from applying the Level of Expressed Emotion scale (LEE; Cole and Kazarian, 1988, Kazarian et al., 1990). The authors of the LEE drew on Vaughn and Leff's (1981) conception of criticism and EOI to derive four characteristic attitudes of significant others (intrusiveness, emotional implication, negative attitudes, and low tolerance/high expectations). Some researchers have applied the LEE with samples of persons with ED, but its validity with this clinical population is limited (Hooley and Parker, 2006). For example, available findings indicate a relationship of EE as assessed by the LEE with lower patient desire to involve their relatives in therapy (Perkins et al., 2005), and with negative experiences of caregiving (Winn et al., 2007). There is no evidence, however, of how perceived EE is related to the course of ED. Moreover, there is no measurement of perceived warmth.

In an attempt to develop a brief scale that measures the three main indices of EE, Lopez and colleagues built on Hooley and Teasdale's prior efforts by adding one item to the PC scale and by deriving two other subscales to assess perceived EOI (four items) and perceived warmth (two items). They also developed a patient version and a caregiver version to capture the views of both members of the dyad. Accordingly, they refer to the measure as the Brief Dyadic Scale of Expressed Emotion (BDSEE). Moderate levels of reliability and validity were observed in the initial study of 60 Mexican American families caring for an ill relative with schizophrenia (Medina et al., 2008, Keefe et al., submitted for publication). To improve the scale's reliability, the authors recommended increasing the number of items. Medina-Pradas and López added more items for the current study (see the Appendix) by drawing on multiple aspects of a given index of EE, particularly EOI. In addition, we were concerned about possible deficits in the patient's theory of mind (Russell et al., 2008, Harrison et al., 2009, Oldershaw et al., 2010, Medina-Pradas et al., submitted for publication). As a result, in generating new items, we followed Cutting et al.'s (2006) recommendation to phrase the items in a way that would not require respondents to take the point of view of their influential other but rather to simply respond with their own experience and opinion (items 7, 8, and 12).

The overall objective of the present study was to assess the psychometric properties of the expanded patient version of the BDSEE in a sample of persons with ED. With the expanded version we expected to replicate or improve upon the psychometric properties observed in the original study of the BDSEE, with particular attention to (a) the scale's factor structure, (b) the subscales' reliability, and (c) the subscales' validity, specifically their relation to other analogous measures of perceived family emotions and attitudes and to the field's gold standard—the CFI.