Culture, Medicine, and Psychiatry

, Volume 33, Issue 1, pp 41–60

Expressed Emotion, Human Agency, and Schizophrenia: Toward a New Model for the EE-Relapse Association

Authors

    • Department of PsychiatryYale University
  • Steven R. López
    • Department of PsychologyUniversity of Southern California
  • Keith H. Nuechterlein
    • Departments of Psychiatry and PsychologyUniversity of California, Los Angeles
Original Paper

DOI: 10.1007/s11013-008-9119-x

Cite this article as:
Breitborde, N.J.K., López, S.R. & Nuechterlein, K.H. Cult Med Psychiatry (2009) 33: 41. doi:10.1007/s11013-008-9119-x

Abstract

Although there is a clear statistical association between expressed emotion (EE) and schizophrenic relapse, our understanding of the mechanisms underlying this relationship is underdeveloped. This study tested a theoretical model in which caregivers’ perceptions of their ill relative’s agency underlie the EE-relapse association. To evaluate this model, we completed qualitative and quantitative analyses of narratives provided by individuals caring for a relative with schizophrenia. The results indicate that high-EE caregivers perceive the expression of symptoms as stemming from their ill relative’s agency more frequently than low-EE caregivers. This was true for both high-EE-criticism and high-EE-emotional overinvolvement caregivers. High-EE and low-EE caregivers did not differ in their perceptions of the role of their ill relative’s agency with regard to the control of symptoms. The findings suggest that EE may be a proxy risk factor for caregivers’ perceptions of their ill relative’s agency.

Keywords

SchizophreniaExpressed emotionHuman agencyNarrative

Introduction

It is clear from past research that family factors influence the course of schizophrenia. For example, studies of expressed emotion (EE) have repeatedly found that the presence of high levels of criticism, emotional over-involvement (EOI), or hostility within the familial environment is associated with an increased risk of schizophrenic relapse (Butzlaff and Hooley 1998). Given the success of this line of research, EE has become one of the most studied psychosocial constructs in psychiatric literature to date (Jenkins and Karno 1992).

The study of EE, however, suffers from a major limitation. While there is a strong statistical association between EE and the course of schizophrenia, our understanding of the mechanisms underlying this association is underdeveloped (Hooley and Gotlib 2000). Recent research has suggested that caregivers’ perception of their ill relative’s agency may account for this association. For example, several studies examining a similar construct (i.e., attributions of controllability) have found that certain high-EE caregivers (i.e., those displaying high levels of criticism) perceive the negative aspects of schizophrenia as being under the volitional control of their ill relative (e.g., López et al. 1999). Although there is some evidence suggesting that this perception is predictive of the occurrence of future relapse, to date investigations of this hypothesis have produced mixed results (Barrowclough and Hooley 2003). Ultimately, clarifying the relationship between EE, relapse, and caregivers’ perceptions of their ill relative’s agency may help to inform the development and adaptation of effective treatments for schizophrenia.

Agency and Schizophrenia

Scholars have long recognized that the onset of schizophrenia is associated with a disruption in an ill individual’s sense of agency. For instance, included among Schneider’s (1959) first-rank symptoms of schizophrenia is the perception that one’s actions, thoughts, and/or emotions are controlled by an outside force.1 More recently, several experimental studies have demonstrated that individuals with schizophrenia have greater difficulty differentiating between self- and other-generated action than individuals without schizophrenia (Frith et al. 2000). Moreover, individuals with schizophrenia have identified reestablishing a coherent sense of agency as a critical component of the recovery process (Davidson 2003).

Unfortunately, clear definitions of human agency are lacking within the schizophrenia literature and are limited in social science research in general (Emirbayer and Mische 1998). Consequently, the concept of human agency has become an “abstraction greatly underspecified, often misused, [and] much fetishized these days by social scientists” (Comaroff and Comaroff 1997, p. 37).2 As such, it is important to begin this study with a clear, though brief, discussion of what is human agency.

Fundamentally, human agency is the ability to intentionally complete a desired act (Ahern 2001a). However, there are four important caveats with regard to this definition. First, it is important not to situate agency solely within individual agents. Rather, human agency stems from the interaction of individuals and elements of their local moral worlds (Giddens 1984) ranging from qualities of the physical environment to one’s social relationships (Coyne et al. 1988; Ward 1907). Ultimately, this interaction is recursive in that human action both reproduces and transforms the structures that influence human action (Emirbayer and Mische 1998). Second, scholars have long recognized that agency has a moral component (Guignon 2004). More specifically, this line of research suggests that human agency is measured not only by the ability to complete an act, but also by the quality of the acts that one completes. Third, the definition of human agency as the ability to intentionally complete a desired act does not clearly highlight the importance of temporality in human agency. Although human action is fundamentally grounded in the present, our actions in the present are often interpreted based on events in the past and are completed, in part, to achieve an envisioned future (Desjarlais 1996). Thus, complete understandings of human agency require exploration of present actions as well as how such actions fit into and influence the temporal flow of an agent’s experience.

A final caveat to consider is the cultural nature of agency. Although the enactment of human agency involves the intentional completion of a desired act, the specific actions and events that are used in the completion of this act possess no intrinsic meaning by themselves (Mink 1978). Searle (1984) has addressed this point in his discussion of the structure of human action:

…One and the same set of human bodily movements might constitute a dance, or signaling, or exercising, or testing one’s muscles, or none of the above. Furthermore, just as one and the same set of types of physical movements can constitute completely different kinds of actions, so one type of action can be performed by a vastly different number of types of movements. (p. 57)

Ultimately, then, human agency involves a process of meaning making—both by the individual completing the act, who must determine what are the appropriate actions needed to complete the desired act, and (in many instances) the entity toward whom the desired act is directed, who must infer what meaning or goal the agent was attempting to convey or accomplish based on the series of acts completed by the agent. In this regard, human agency may be conceptualized as a fundamentally cultural process. More specifically, agents may often draw on shared meaning systems and methods of communication when selecting the actions that they complete in their attempt to accomplish a desired act with the goal of increasing the likelihood that their “audience” will come to a similar understanding of the desire or goal underlying the series of actions that they have completed (Alexander 2006). Audiences, too, may often rely on shared meaning systems when interpreting the meaning of acts within their local moral worlds. (Alexander 2006). In this regard, human agency may best be viewed as situated within the intersubjective nexus between individuals and their local moral worlds (see Kleinman and Joan Kleinman 1991 for a more thorough discussion of the intersubjective nature of human experience).

Although first-hand accounts of schizophrenia have suggested that re-establishing a coherent sense of agency is an important component in the recovery process (Deegan 2001), researchers have devoted limited attention to examining how human agency may be associated with the occurrence of relapse specifically. However, studies of self-efficacy, a central component of individuals’ perception of agency (Bandura 2000), suggest that ill individuals’ agency could be an influential factor in the course of the disorder. More specifically, self-efficacy is associated with several known risk factors for triggering a schizophrenic relapse: strong physiological reactions to stressful events (Nuechterlein and Dawson 1984) and poor adherence to psychiatric and psychological treatment programs (Zygmunt et al. 2002). For instance, outside of the schizophrenia literature, studies have demonstrated that individuals with lower confidence in their ability to complete a difficult task experience greater physiological stress reactions when completing that task (O’Leary and Brown 1995). It is important to note, though, that while strong physiological reactions to stressful events and poor adherence to psychiatric and psychological treatment program have been presented within the medical literature as “universal” risk factors for relapse for all individuals with schizophrenia, explorations of whether these relationships may vary across cultural settings are limited.

Agency in the Context of Schizophrenia: The Role of Caregivers’ Narratives

The onset of chronic illness is often associated with a profound and traumatic disruption in one’s life (Charmaz 1983). Within this experience, many of the basic components of everyday functioning are disrupted—including one’s perception of what one can or cannot do (Bury 1982). In order to address this disruption, ill individuals may often draw on models within their local moral worlds in order to recreate a coherent sense of self within the context of their illness (Kleinman 1980).

One model that individuals may draw on to address this disruption is narratives created by their caregiving relatives. Caregivers’ narratives are a powerful socializing device (Ochs and Capps 2001), and individuals suffering from a chronic illness may incorporate information contained within these narratives into their definition of self within the context of their illness (Capps and Ochs 1995). Of note, within caregivers’ narratives (and narratives in general), questions of agency (e.g., Who did what and why did they do it?) are a central component (Harré 1995). Thus, if individuals with schizophrenia draw on their caregivers’ narratives in constructing a new sense of self in the context of chronic illness, caregivers’ perceptions of their ill relative’s agency may be prominently featured.

Expressed Emotion as a Proxy Risk Factor for Caregivers’ Perception of Their Ill Relative’s Agency

In developing a hypothesized model for the relationship between EE, relapse, and caregivers’ perceptions of their ill relative’s agency, it is important to specify the nature of the respective interrelationships among these variables. The information reviewed above suggests that caregivers’ perception of their ill relative’s agency may predict the occurrence of future schizophrenic relapse. Additionally, following Weiner’s (1993) attribution theory, we would speculate that caregivers’ perceptions of their ill relative’s agency lead to the development of the behaviors and emotions associated with EE. Studies outside of the EE literature have provided evidence supporting the hypothesized direction of the relationship between perceptions of whether an action is under the volitional control of an individual and the resulting emotions and behavior (e.g., Dijker and Koomen 2003; Reyna and Weiner 2001). Finally, using the terminology provided by Kraemer and colleagues (2001), we hypothesize that EE may be a proxy risk factor for the influence of caregivers’ perception of their ill relative’s agency on the course of schizophrenia. Kraemer and colleagues define a proxy risk factor as a variable that is correlated with a strong risk factor for a given outcome. Kraemer et al. note that in proxy risk factor relationships, the proxy risk factor may appear to be a causal predictor of the outcome variable, “even though the only connection between the [proxy risk factor] and the outcome lies in the strong risk factor correlated with both” (p. 851). Thus, we suggest that the association between EE and relapse may not stem from the direct influence of EE on the course of schizophrenia but, rather, occurs due to the association of a strong risk factor (i.e., caregivers’ perception of their ill relative’s agency) with both EE (i.e., the proxy risk factor) and relapse (i.e., the outcome of interest), respectively. This hypothesized relationship among EE, relapse, and caregivers’ perceptions of their ill relative’s agency is depicted in Fig. 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs11013-008-9119-x/MediaObjects/11013_2008_9119_Fig1_HTML.gif
Fig. 1

Hypothesized relationship among expressed emotion (EE), schizophrenic relapse, and caregivers’ perception of their ill relative’s agency

Study Goals

We conducted two studies in order to explore the relationship between caregivers’ perception of their ill relative’s agency, EE, and the course of schizophrenia. In the first study, we completed a qualitative analysis of narratives elicited from individuals caring for relatives with schizophrenia to examine whether there was a relationship between caregivers’ EE status (high EE vs. low EE) and their perception of their ill relative’s agency. In the second study, we completed a quantitative analysis to test the generalizability of the findings of Study 1 across caregivers and to examine whether EE was a proxy risk factor.

Study 1

Method

Participants and Measures

Participants in this study are 21 Caucasian individuals with schizophrenia and their key relative. These participants were selected from two past studies of family factors and schizophrenia (Nuechterlein et al. 1992; Vaughn et al. 1984). Participants from the Vaughn et al. study (n = 17) were selected to mirror the characteristics of participants in a similar study completed in the United Kingdom which was operationalized as including only individuals identified by the investigators as being white. As such, it is likely that within this sample, there was considerable ethnocultural variation. Among the key relatives included from the Vaughn et al. study, there were 12 mothers, 4 fathers, and 1 wife. These key relatives were selected from a larger pool of subjects who were from all levels of socioeconomic status as defined by Hollingshead and Redlich (1953). The pool of participants included from the Nuechterlein et al. study (n = 4) was comprised of three mothers and one father, and they were selected from a larger sample of participants who were from all levels of socioeconomic status as defined by Hollingshead and Redlich. There was little available information with regard to the ethnicity of the participants included from this study other than that they were identified as “Anglo-American” and “non-Hispanic” by the investigators. Diagnosis of schizophrenia was confirmed using the CATEGO algorithm associated with the Present State Examination (PSE; Wing et al. 1974) for participants included from the Vaughn et al. study and using Research Design Criteria (Spitzer et al. 1978) and the PSE for participants included from the Nuechterlein et al. study. Key relatives were defined as individuals who maintained considerable face-to-face contact with the ill individual.

An abbreviated version of the Camberwell Family Interview (CFI; Vaughn and Leff 1976) was used to assess key relatives’ EE. The CFI is a 1.5- to 2-h-long semistructured interview. Raters were trained to code the interviews for the level of EOI (0–5) and hostility (0–3) as well as the number of critical comments. Following the protocol proposed by Leff and Vaughn (1985), ratings of EOI are based on reports of self-sacrificing, devoted, or extremely overprotective behavior elicited by the interview as well as key relatives’ behavior during the administration of the interview such as displays of emotional distress and dramatization of minor events. Ratings of critical comments were based on a frequency count of the number of instances in which a caregiver expressed disapproval with regard to an aspect or behavior of his or her ill relative. Finally, caregivers received a categorical score based on the quality of hostility displayed during the interview toward their ill relative: generalized criticism (1), rejection (2), generalized criticism and rejection (3), and no hostility displayed (0). Following the traditional criteria (Leff and Vaughn 1985), high EOI was defined as a score of 4 or 5, and high criticism was defined as six or more critical comments over the course of the interview. Key relatives who were high in EOI or criticism or displayed any amount of hostility were rated as high EE. All other key relatives were rated as low EE.

Participants selected from the Vaughn et al. (1984) study fell into the following three groups based on their EE scores: (a) high EE due to EOI (n = 3), (b) high EE due to criticism (n = 7), and (c) low EE (n = 7). Selection of participants for each group was done randomly from the available pool of subjects except for the high-EE-EOI group, for which all eligible participants were included. An additional four high-EE-EOI participants were selected from the Nuechterlein et al. (1992) study, bringing the total number of participants in that group to seven.

Analyses

The author conducted open-ended analyses of the narratives of all the 21 family caregivers to identify the key relatives’ perception of their ill relatives’ agency. For this analysis, a systematic and thorough reading of the narratives was completed that focused on identifying how key relatives describe their ill relative’s agency through the language and structure of their narrative. It is important to note that the individual who completed the qualitative analysis was not blind to the EE status of the key relatives.

Participants’ response to one question from the CFI was selected for analysis: “How did (ill relative) come to contact the hospital this most recent time?” This particular question was selected for two reasons. First, the answers provided for this question typically take on a “story-like” form, rather than a “yes” or “no” answer. Likewise, given that key relatives’ EE scores may be influenced by the severity of their ill relatives’ symptoms (King 2000), it was important to select narratives in which all caregivers would be describing an incident in which the severity of symptoms was as similar as possible across subjects (i.e., this question elicited narratives that described instances in which all of the ill relatives expressed symptoms severe enough to warrant hospitalization).

Results

To highlight the results of the qualitative analysis, excerpts from three narratives examined in this study are listed below. Two narratives were provided by high-EE key relatives (one high EE due to criticism and one high EE due to EOI), and one narrative was provided by a low-EE key relative.3

Narrative from a High-Expressed Emotion-Criticism Key Relative

My spouse and I both went to work and he was home and, uh, the sheriff called me at work and said they picked him up for drunkenness. Well this is early in the morning. I have no liquor in my house or anything else he couldn’t drink. And we didn’t give him any money because he would go and buy a bottle of something, a beer or whatever. It’s … the beer was no good for him. So I went directly up there and—and, uh, they told me then: “We picked him up. We had to picking him up for something. He was acting strange.” And here, I eventually found out he was practicing yoga on a parked car. And they said he had a drink. Whether he did or not, I don’t know, I don’t know where he got it. He didn’t have any money. Anyway, they took him from there to the hospital.

In this narrative from the high-EE-criticism key relative, there is a pattern with regard to the semantic case assigned to the ill relative and the expression/control of symptoms.4 Semantic case refers to the relationship between an object and a verb in a phrase (Fillmore 1968) and is a marker of agency within a narrative (Capps and Ochs 1995). For example, being assigned to the agentive case (e.g., I baked cookies) suggests that the object in the sentence (i.e., “I”) was the agentic force from which the action stemmed (i.e., baking cookies). Within the narrative elicited from the high-EE key relative, the key relative assigns the ill relative to the agentive case when describing the expression of symptoms and other problematic behavior (e.g., he was acting strange; he was practicing yoga on a parked car; he had a beer5). However, the ill relative is assigned to the affected case (i.e., who the action was done to) when the caregiver describes the control of the illness (e.g., he was taken in … by the sheriff; they would take him in to the hospital). This suggests that the key relative views other individuals as the agentic force responsible for these actions, whereas the ill relative is viewed as the recipient of these actions (i.e., these actions were done to the ill relative).

Narrative from a High-Expressed Emotion-Emotional Over Involvement Key Relative

He was taken in because by the sheriff because he was uh… he went a little berserk and he uh.… My car was in the lot, didn’t matter that it was in my car. It’s an old eighty-four car anyway you know. But it was the only one there and he was kicking it and taking out his rage and just kicking it and hurting himself and hard, you know, really, that he dented the whole car you know. But like I said it’s an old car. But he was really putting anger to into it and hate and everything and it went on, you know. It was still going on after ten minutes and I thought well gee something’s going to go wrong. He’s gonna hurt himself or something and—and yelling and really taking his frustrations and anger out on the car and so I called the sheriff. That’s the first time I ever did that. [Mm-hmm] And then they sai—they said that, uh, that they would take him in to the hospital or something. I don’t know what they were going to do.

Within the narrative of the high-EE-EOI key relative, a similar pattern emerges with regard to the key relative’s use of semantic case. Specifically, when describing the expression of symptoms, the key relative repeatedly assigns the ill relative to the agentive case (e.g., he went berserk, he was kicking it, [he was] taking out his rage, he dented the whole thing, etc.); however, when describing the control of symptoms, the ill individual is assigned to the affected case (e.g., [the sheriff] will take him [the ill relative] in to the hospital). Thus, consistent with the narrative from the high-EE-criticism key relative, the high-EE-EOI key relative’s use of semantic case portrays the ill individual as the agentic force behind the expression of symptoms, whereas others are depicted as the agentic force responsible for the control of the ill individual’s symptoms.

Narrative from a Low-Expressed Emotion Key Relative

He had a manic episode while he was out walking. And he trespassed. And the people called the sheriff and they picked him up and took him, you know, into the station. And I couldn’t contact him at all. They wouldn’t let me near him. They didn’t give him a phone call. Okay he went from that little manic episode and I don’t know what happened in jail but it apparently was very unpleasant. I mean more than usual. Especially since he was in that state of mind he was in. Apparently, it got worse while he was there. And um, he thought I was dead, that I died and his aunt was dead. And he went through some kind of a whole thing, you know. And from that, in fact that’s, I think, where he started actually hearing voices and going into this whole electronic spy thing that he was into, you know.

Unlike the narratives from the high-EE caregivers, the relationship between the semantic case assigned to the ill relative and the expression/control of symptoms is less clear in the narrative elicited from the low-EE caregiver. At certain points within the narrative from the low-EE caregiver, the ill relative is assigned to the agentive case when the key relative describes the expression of symptoms (e.g., he trespassed; he thought I was dead; [he was] going into this electronic spy thing). However, this pattern is not consistent throughout the narrative. For instance, the caregiver also assigns the ill relative to semantic cases that imply that the ill relative was not the agentic force behind the expression of his symptoms (e.g., the affected case—“He had a manic episode,” “He was in that state of mind that he was in”—and the experiencer case—”He started actually hearing voices”). However, similar to the narrative from the high-EE caregiver, within the low-EE caregiver’s narrative, the ill relative is not assigned to the agentive case during instances in which the symptoms of schizophrenia are controlled (e.g., “[The sheriff] picked him up” and “[The sheriff] took him … into the station”). Rather, another individual (i.e., the sheriff) is portrayed as the agentic force responsible for these actions.

Discussion

The analysis described above suggests that there is an important difference between high-EE and low-EE caregivers with regard to their perception of their ill relative’s agency in the expression of symptoms. Specifically, the caregivers’ use of semantic case within the narratives suggests that high-EE caregivers view their ill relatives as the agentic force behind the expression of their schizophrenic symptoms and other problematic behavior, whereas low-EE caregivers may posses a more nuanced perception with regard to the role of their ill relative’s agency in the expression of symptoms. Although low EE caregivers describe their ill relative as the agentic force responsible for the expression of certain symptoms, in many instances low-EE caregivers portray the expression of symptoms as something that is passively experienced by their ill relative rather than enacted. A different picture is observed with regard to the caregiver’s view of the ill relative’s control of symptoms. Our analysis suggests that high-EE and low-EE caregivers do not differ in their view of their ill relative’s agency during the control of symptoms. Individuals suffering from a chronic illness may draw on caregivers’ narratives in order to reconstruct a coherent sense of self. Accordingly, the narrative models within the familial environment may influence the attitudes that the ill relative incorporates into the newly reconstructed sense of self. The incorporation of attitudes that the ill relative is the agentic force in the expression of symptoms, as is the case for individuals living in high-EE households may have deleterious effects on the course of the illness.

Study 2

The results of the qualitative analyses in Study 1 suggest that caregivers’ EE status may be associated with their perception of their ill relative’s agency. However, in Study 1, the relationship among EE, caregivers’ perception of their ill relative’s agency, and the course of schizophrenia (i.e., the occurrence of a future relapse) was not examined. In Study 2, quantitative analyses are used to explore the interrelationships among these three variables. As noted previously, we hypothesize that caregivers’ perception of their ill relative’s agency will underlie the relationship between EE and schizophrenic relapse. More specifically, we speculate that EE is a proxy risk factor for caregivers’ perception of their ill relative’s agency.

Method

Participants

The participants and narratives included in Study 1 were also used in Study 2.

Clinical Measures

Following the administration of the CFI, the caregiver-ill relative dyads from the Vaughn et al. (1984) study were tracked for 9 months, and dyads from the Nuechterlein et al. (1992) study were tracked for 12 months in order to assess whether a relapse occurred. During this period, participants were contacted on a regular basis in order to assess whether an exacerbation of symptoms had occurred (Nuechterlein et al., once every 2 weeks; Vaughn et al., once per month). If there was evidence that an exacerbation of symptoms had occurred, the Brief Psychiatric Rating Scale (BPRS; Overall and Gorham 1962; Nuechterlein et al. sample only) or Present State Examination (Vaughn et al. sample only) and Psychiatric Assessment Scale (PAS; Krawiecka et al. 1977; Vaughn et al. sample only) were administered to the ill relative in order to assess whether a relapse had occurred. Independent raters placed participants into one of three categories based on their scores on these measures: no relapse, relapse, nonremitting symptoms. Participants in the nonremitting symptoms category (n = 2) were included in all analyses except those in which relapse was a variable of interest, as it would be difficult to assess whether their clinical status changed during the course of the study. All raters were blind to the EE status of the participants and achieved acceptable levels of interrater reliability on the respective measures.

Procedure

A group of coders who were blind to the hypotheses of the study were trained to identify instances within the narratives in which the caregiver described the expression or control of schizophrenic symptoms. Expression of symptoms was defined as the expression of DSM-IV-TR (American Psychiatric Association 2000) criteria for schizophrenia as well as any other behavior that the caregiver saw as reflective of the disorder (e.g., nonspecific statements such as “he was acting weird”). The control of symptoms was defined as any action designed to reduce or eliminate the expression of a symptom or the act of successfully not expressing a symptom. Instances of expression or control were defined as the minimum number of words necessary to describe the expression or control of symptoms, with the conditions that, in each instance, the ill relative was assign a semantic case (either directly or implied) and there was at least one verb. For each identified instance of the expression or control of symptoms, the coders recorded (a) whether the instance describes the expression or control of symptoms and (b) what semantic case was assigned to the ill relative. Specifically, for each instance of expression or control of symptoms, coders noted whether or not the ill individual was assigned to the agentive case.

All coders completed a minimum of 10 training narratives, and their codes for these narratives were compared to a set of master codes completed by the first author using a kappa statistic. Using the criteria suggested by Fleiss (1975), coders were considered reliable when they achieved a κ > 0.75 with regard to their ability (a) to identify codable instances with the narrative, (b) to determine if a given instance described the expression or control of symptoms, and (c) to correctly identify the semantic case assigned to the ill relative.

Each narrative included in the analyses was rated by three coders. Consensus ratings for each narrative were generated by identifying instances in which at least two of the three agreed on a rating. However, there was one exception to this rule. In instances in which two or more raters coded a specific instance but disagreed with regard to whether the instance described the expression or control of symptoms or the appropriate semantic case, a fourth reliable rater provided a “tie-breaking” decision. Of the total combined number of ratings for expression/control and semantic case included in the analyses (n = 1,518), <7% were derived using this tie-breaking method.

Statistical Analyses

Analysis 1

The relationship between EE and caregivers’ perceptions of agency was examined using a two-step process.

Step 1: First, caregivers’ EE status (high EE vs. low EE) was entered into an independent-samples t-test. The percentage of instances in which the ill relative was assigned to the agentive case in the caregiver’s narrative served as the dependent variable. As certain caregivers provided longer narratives than others, percentages were used instead of frequency counts to control for variability in the number of coded items across the caregivers’ narratives. This analysis was completed twice—once for the percentage of instances that described the expression of symptoms (AG-EX%) and once for the percentage of instances that described the control of symptoms (AG-CO%).

Step 2: Caregivers’ specific EE profile (high EE-criticism, high EE-EOI, low EE) was entered into an ANOVA with the percentage of instances in which the ill relative was assigned to the agentic case in the key relative’s narrative as the dependent variable. Again, this analysis was completed twice to assess how specific EE profiles may be associated with AG-EX% and AG-CO%, respectively. If the results of these analyses suggested that there was a meaningful relationship between specific EE profiles and caregivers’ perceptions of their ill relative’s agency, independent-samples t-tests with a Bonferroni correction to maintain an α of 0.05 were used to compare the specific EE profiles.

Analysis 2

To assess whether EE is a proxy risk factor for caregivers’ perceptions of their ill relatives’ agency, the z′ test for mediation developed by MacKinnon and colleagues (2002) was used. Although mediation and proxy risk factors (also referred to as confounding relationships) are conceptually different, the statistical tests for these relationships are the same (MacKinnon et al. 2000). A simulation study of the various tests of mediation suggests that z′ outperforms other tests with regard to Type I error rates and statistical power (MacKinnon et al. 2002). Regression coefficients and standard errors obtained through logistic regression were standardized prior to the calculation of z′ (MacKinnon and Dwyer 1993).

If EE is a proxy risk factor for caregivers’ perception of their ill relative’s agency, five expectations need to be met (MacKinnon et al. 2000, 2002).

Step 1: When entered into a logistic regression equation by itself, EE will predict relapse.

Step 2: Caregivers’ perception of their ill relative’s agency will be a significant predictor of relapse when entered into a logistic regression equation by itself.

Step 3: When both EE and caregivers’ perception of their ill relative’s agency are entered into the same logistic regression equation, only caregivers’ perception of their ill relative’s agency will be a statistically significant predictor of relapse.

Step 4: Caregivers’ perception of their ill relative’s agency will predict caregivers’ EE status.

Step 5: The results of the z′ test will suggest that caregivers’ perception of their ill relative’s agency is a statistically significant confounding variable in the relationship between EE and relapse.

Results

Expressed Emotion and Expression and Control of Symptoms

Analysis 1

Step 1: With regard to the expression of symptoms, the percentage of instances in which ill relatives were assigned to the agentive case was lower among narratives from low-EE caregivers (M = 26.30, SD = 14.57) than narratives from high-EE key relatives (M = 53.87, SD = 18.57; t[19] = 3.42; p < 0.01). However, when describing the control of symptoms, there was no difference between narratives from low-EE key relatives (M = 20.94, SD = 12.62) and high-EE key relatives (M = 21.26, SD = 24.21) with regard to the percentage of instances in which the ill relative was assigned to the agentive case (t[18] = 0.03; p = 0.98).6

Step 2: There was a statistically significant relationship between specific EE profiles and AG-EX% (F[2,18] = 5.81; p = 0.01). Specifically, when describing the expression of symptoms, the percentage of instances in which ill relatives were assigned to the agentive case was lower among narratives from low-EE key relatives (M = 26.30, SD = 14.57) than narratives from high-EE-criticism key relatives (M = 56.60, SD = 24.69; t[12] = 2.80; p = 0.016) and high-EE-EOI key relatives (M = 51.15, SD = 10.95; t[12] = 3.61; p < 0.01), respectively. There was no difference in the percentage of instances in which ill relatives were assigned to the agentive case among high-EE-criticism and high-EE-EOI key relatives (t[12] = −0.53; p = 0.60).

There was no statistically significant relationship between specific EE profiles and AG-CO% (F[2,17] = 0.24; p = 0.79).7 Specifically, when describing the control of symptoms, the percentage of instances in which ill relatives were assigned to the agentive case was similar among narratives from low-EE key relatives (M = 20.94, SD = 12.62), high-EE-criticism key relatives (M = 25.02, SD = 28.73), and high-EE-EOI key relatives (M = 16.87, SD = 19.32).

Expressed Emotion as a Proxy Risk Factor for Key Relatives' Perception of Ill Relative’s Agency

Analysis 2

Step 1: When entered into an exact logistic regression by itself, there was a statistically significant relationship between EE status and relapse (β = 2.03; odds ratio = 7.63; exact one-sided p < 0.04).

Step 2: Although there was no relationship between AG-CO% and schizophrenic relapse (β < 0.01; odds ratio = 1.00; exact one-sided p = 0.39), AG-EX% was a statistically significant predictor of relapse, where greater levels of AG-EX% were associated with a greater risk of relapse (β = 0.16; odds ratio = 1.18; exact one-sided p < 0.01).

Step 3: When both EE status and AG-EX% were entered into the same exact logistic regression equation, only AG-EX% remained a statistically significant predictor of relapse (EE, β < 0.01, odds ratio = 1.00, exact one-sided p > 0.49; AG-EX%, β = 0.14, odds ratio = 1.15, exact one-sided p < 0.01).

Step 4: AG-EX% was found to be a statistically significant predictor of EE (β = 0.18; odds ratio = 1.20; exact one-sided p < 0.01).

Step 5: Calculation of z′ suggests that EE is a proxy risk factor for AG-EX% (z′ = 1.65; p < 0.05).

General Discussion

The present results indicate that high-EE and low-EE caregivers have different perceptions with regard to their ill relative’s agency and that these perceptions are predictive of the course of illness. Although high-EE and low-EE caregivers do not differ in their perception of the role of their ill relative’s agency with regard to the control of symptoms, high-EE caregivers view the expression of symptoms as stemming from their ill relative’s agency to a greater extent than low-EE caregivers do. This finding is true for caregivers high in EE both due to criticism and due to EOI. The perception that an ill individual is the agentic force behind the expression of symptoms is positively associated with the occurrence of future schizophrenic relapse and may underlie the association between EE and relapse. More specifically, EE appears to be a proxy risk factor for caregivers’ perceptions of their ill relative’s agency.

The findings of this study add to a growing body of literature examining the role of human agency in the recovery process in schizophrenia. Noting the deleterious effects experienced by individuals with schizophrenia due to repeated exposure to attitudes that they have limited or no agency, multiple scholars have noted the importance of facilitating individuals’ recognition of themselves as active social agents as an important component in the recovery process in schizophrenia (for a review see Davidson 2003). This strategy has even been incorporated within psychological interventions for schizophrenia (e.g., Lysaker et al. 2007). Of note, our findings suggest an important caveat with regard to agency within the recovery process. Specifically, our study raises the possibility that exposure to attitudes in which an individual is portrayed as the agentic force responsible for the display of symptoms or other behaviors that may be viewed as problematic within one’s local moral world may increase the risk for future symptomatic exacerbations. One clinical implication of this findings may be that therapeutic interventions for schizophrenia may benefit by helping individuals with schizophrenia manage a complex self system in which they develop and/or maintain a sense of self as able to complete acts that are valued within their local moral world(s) but not necessarily responsible for the problematic and challenging behaviors that may stem from their illness. This challenge may represent another “conundrum of the recovery process” (Jenkins and Carpenter-Song 2005) in that the recovery process in schizophrenia may involve a simultaneous “restoring” and “relinquishing” of agency.8

One specific clinical intervention that may benefit from greater recognition of the “Catch-22” described above with regard to agency in the recovery process from schizophrenia is family psychoeducation. Family psychoeducation refers to a group of family interventions that attempt to provide family members with information about their relative’s illness as well as skills to cope with challenges that may occur within their relationship due to the illness (McFarlane et al. 2003). Multiple investigations of family psychoeducation have demonstrated that this intervention is effective in reducing rates of relapse and rehospitalization among individuals with schizophrenia (Baucom et al. 1998), and this intervention is currently recognized as an evidence-based treatment for schizophrenia (Dixon et al. 2001). While caregivers participating in psychoeducation programs may indirectly come to learn that their ill relatives are not the agentic force responsible for the expression of their symptoms, facilitating such an attitude among caregivers is not recognized as a key goal or principle of family psychoeducation for schizophrenia (see Froggatt et al. 2007 for the World Fellowship for Schizophrenia’s list of key goals and principles for family psychoeducation). The results of the current study raise the possibility that the efficacy of family psychoeducation programs may improve if such programs can more specifically educate caregivers with regard to the scope and limitations of their ill relative’s agency with regard to the expression and control of symptoms and other behavior associated with the illness.

It is important to note that this study does suffer from several limitations including the small sample, comprised entirely of Caucasians. Although conservative statistical analyses were used (i.e., exact statistics), the findings of this study should be interpreted with caution until large-scale confirmatory studies can be completed. Additionally, given that conceptualizations of human agency differ cross-culturally (Markus et al. 2006), the current findings may not be applicable to all individuals. Moreover, given that the label “Caucasian” could be applied to people from multiple cultural and ethnic backgrounds, we cannot say for certain that the differences between relatives high in EE and relatives low in EE do not stem from underlying cultural factors that we were unaware of within our pool of subjects. Likewise, as the statistical tests for a proxy risk factor and mediation are the same, our conclusions with regard to the direction of the relationship between EE and perceptions of agency are based on theory rather than statistical evidence. Future studies using structural equation modeling and longitudinal designs could evaluate the direction of this relationship with greater precision. Furthermore, the narrative selected for analysis in this study is a response to a question that may pull for a description of “negative” agency (“How did [ill relative] come to contact the hospital this most recent time?”). As such, the results of this study may not provide an accurate representation of caregivers’ perception of their ill relative’s agency in socially valued domains of functioning. However, the greatest limitation of this study is that it did not examine interactions between caregivers and their ill relatives. As the structure of narratives (and discourse in general) is influenced by the audience that they are directed toward (Breitborde 1983; Ochs 1997), it is unclear whether elements of a narrative elicited by an interview would also be present in caregivers’ discourse in the familial environment. Likewise, it is also unclear whether caregivers’ perception of their ill relative’s agency is, in fact, incorporated in their ill relative’s own understanding of their illness (i.e., the ill relative’s narrative). Analysis of everyday familial interactions could address both of these limitations (Capps and Ochs 1995).

Scholars have long been interested in examining the disruption in human agency associated with schizophrenia, and to date, several studies have highlighted the role of the family environment in shaping ill individual’s perceptions of agency (Davidson and Strauss 1992; Hall and Purdy 2000). However, the number of studies specifically examining this topic is relatively small. The results of the current study suggest that much can be learned through investigating what caregiving relatives may do to facilitate or hinder their ill relatives’ recognition of themselves as socially valued agents. As family members often serve as the primary caregiver for individuals with severe mental illness (Baronet 1999), developing a better understanding of their role in shaping their ill relative’s agency may provide valuable information with regard to facilitating the recovery process among individuals with schizophrenia.

Footnotes
1

It is important to note that reviews of Schneider’s first-rank symptoms have failed to confirm that these symptoms are unique to schizophrenia (e.g., Nordgaard et al. 2008).

 
2

We were first introduced to this quote by Ahern (2001b).

 
3

Identifying information in these narratives has been changed or removed in order to protect the confidentiality of the participants.

 
4

Although not specifically examining the use of the agentive case, Estroff (1993; Estroff et al. 1991) has noted the importance of semantic case with regard the experience of mental illness. Specifically, in examining narratives of individuals with schizophrenia, Estroff has argued that semantic case can provide insight into how individuals position themselves versus their illness. Estroff identifies two types of positioning: (a) “I am” illnesses, where there is little to no differentiation between self and illness or where the self may even be replaced by the illness (e.g., I am schizophrenic); and (b) “I have illnesses,” where illnesses are referred to as possessions of the self and there is greater separation between self and illness (e.g., I have cancer).

 
5

The verb “to have” can be interpreted in several ways. In some instances, the verb “to have” can be used idiomatically, where the subject of the verb is assigned to the agentive case (e.g., I had breakfast = I ate breakfast [Quirk et al. 1985]). In the narrative from the high-EE key relative, the key relative appears to use the verb “to have” in the phrase “he had a beer” as if it were synonymous with the verb “to drink.” Given that the subject in the phrase “he drank a beer” would be considered the agent, we considered the subject in the phrase “he had a beer” to be the agent as well.

 
6

One key relative did not describe any instances of the control of symptoms in his/her narrative. As such, this analysis was based on 20 participants only.

 
7

One key relative did not describe any instances of the control of symptoms in his or her narrative. As such, this analysis was based on 20 participants only.

 
8

We would like to thank an anonymous reviewer who noted the relevancy of this paper to our study.

 

Copyright information

© Springer Science+Business Media, LLC 2008