|Home | About | Journals | Submit | Contact Us | Français|
Deficits in social/communications skills have been documented in schizophrenia, but it is unclear how these deficits relate to cognitive deficits and to everyday functioning. In the current study, social/communication skills performance was measured in 29 schizophrenia patients with a history of good vocational functioning (GVF) and 26 with a history of poor vocational functioning (PVF) using a role-play–based social skills assessment, the Maryland Assessment of Social Competence (MASC). A battery of standard cognitive tasks was also administered. MASC-indexed social skills were significantly impaired in PVF relative to GVF patients (odds ratio = 3.61, P < .001). Although MASC social skills performance was significantly associated with cognitive performance in domains of verbal ability, processing speed, and memory, the MASC nevertheless remained an independent predictor of vocational functioning even after controlling for cognitive performance. Social/communications skills predict vocational functioning history independently of cognitive performance, and social skills measures should be considered for inclusion in test batteries designed to predict everyday functioning in schizophrenia.
Even with recommended treatments, schizophrenia is characterized by poor functioning in self-care, social, and vocational roles. Consequently, much recent research has focused on the determinants of everyday functioning and on how to design new pharmacological and psychosocial interventions to improve it. Cognition is one important predictor of community role functioning in schizophrenia and a target for treatment development.1–6 There is also considerable evidence that social dysfunction contributes to poor community functioning in schizophrenia.7–12 Consensus about the nature of this multidimensional social deficit is emerging slowly. The “social problem–solving” framework provides a useful organizational heuristic. Although it is clear that real-time social interactions do not unfold in a neatly ordered sequence,13 this framework posits 3 overlapping stages, referred to as: “receiving,” “processing,” and “sending” skills, respectively.7,14–16 First, emotion and social cue perception skills help an individual recognize relevant social cues, such as a conversational partner's facial expressions and body language.10,11,17 Second, various cognitive processes—including basic cognitive capacities, such as verbal memory and reasoning, and more specifically social attributes, such as “theory of mind” and attributional style11—shape an individual's interpretation of incoming communications and the generation of response alternatives. Finally, various social or communication skills (eg, the selection of appropriate conversational content, the use of nonverbal communication strategies, and the sensitivity to appropriate conversational balance) are necessary for effective interpersonal performance and attainment of social goals.18,19
The social problem–solving framework helps to clarify the distinction between 2 terms used in the schizophrenia social dysfunction literature: “social cognition,” on the one hand, and “social competence,” on the other. Social cognition typically refers to the receiving and processing stages outlined above; related research has focused on emotion and social cue perception and theory of mind (reviewed in Couture et al11). Social competence refers more to sending skills, ie, the verbal and nonverbal communication skills that allow successful execution of interpersonal interactions.10,19 While much research has focused on the relationship of social cognition to community functioning in schizophrenia,11,17 we have argued that deficits in social competence (ie, poor communication or sending skills) are an equally significant impediment to effective community functioning.18 The importance of this facet of social dysfunction in schizophrenia has been explored in a number of recent studies.10,20
A variety of measurement strategies are available to assess aspects of social cognition.11,17 Options for assessment of the social/communication skills that underlie social competence are more limited. Performance-based measures have been most widely used, particularly “role-play tests” (RPTs).7,9,15,20–23 RPTs consist of observational ratings based on standardized, simulated conversations or role-plays, which allow direct assessment of the verbal and nonverbal behaviors that determine the effectiveness of social interactions. RPTs have been employed in numerous empirical investigations over the past 25 years (see Bellack et al21), including assessment of the effects of pharmacological and psychosocial treatments.24 However, a great deal remains to be learned about the nature and correlates of social/communication skill deficits in schizophrenia. For example, although there is evidence of specific social skill associations with cognitive measures (ie, correlations with IQ, verbal ability, verbal memory, and cognitive flexibility but not with working memory7,9,20), the full pattern and extent of these associations have not been investigated. Further, while it is often assumed that effective performance in the simulated social interactions used by RPTs will correspond to effectiveness in aspects of actual everyday role performance in the community, the evidence supporting this link is quite limited.1,25
The ongoing “Measurement and Treatment Research to Improve Cognition in Schizophrenia” program (MATRICS)5 has encouraged further research into the predictors of everyday role functioning in this disorder. This initiative was undertaken to help schizophrenia researchers reach a broad consensus on assessment and treatment options for cognitive and functional impairment. From a measurement standpoint, a particular issue is the relationship between traditional cognitive measures and measures of social cognition and competence, particularly the degree of independence and overlap among these variables in predicting everyday community performance.10,12,26,27 Recognizing the conceptual and psychometric challenges of directly measuring community performance, the MATRICS evaluated 2 RPTs and 2 interview-based rating scales as “proxy measures of functional capacity” in the proposed MATRICS treatment assessment battery.28,29 The MATRICS Neurocognition Committee concluded that all the measures performed “reasonably well” but declined to recommend a single measure for the battery, preferring to wait for further research and developments in the field.29
One of the RPTs evaluated in the MATRICS program was the Maryland Assessment of Social Competence (MASC),21 an empirically developed, standardized approach that has been employed in a number of psychometric, mechanism, and clinical trials. The context for current analyses was a study of factors contributing to good vocational functioning (GVF) and poor vocational functioning (PVF) in schizophrenia (J.M.G., Principal Investigator). Chronic vocational disability is among the most serious consequences of the illness. Rates of competitive employment in this population are generally less than 15%, with obvious implications for economic self-sufficiency, self-esteem, and social relationships, as well as for families and social service systems.30 In the parent study, 39 schizophrenia patients with PVF were compared with 38 patients with GVF and 44 healthy controls on a range of neuropsychological, experimental, clinical, and outcome variables. The focus of the current article is on the relationship between social/communication (ie, sending) skills, measured with the MASC, and cognition, measured with traditional neuropsychological instruments, in predicting vocational functioning in schizophrenia. Based on the literature reviewed above, we predicted that the data would show strong associations between the MASC and neuropsychological measures. However, because social skills performance depends on emotional, personality, and life experience factors that are not exclusively cognitive in nature, we hypothesized that the MASC would demonstrate an independent relationship to vocational functioning, beyond that accounted for by the neuropsychological measures alone.
The parent study was designed to determine the cognitive, symptomatic, and brain structural features that are associated with vocational functioning in schizophrenia. Seventy-seven patients were recruited and divided into 2 groups: those with GVF , defined as being competitively employed for 20 or more hours a week, 75% of the time over the previous 2 years, and those with PVF, who were persistently unemployed over the previous 2 years. Individuals with intermediate employment status, as well as homemakers and students, were excluded from the study. All patients had been ill for a minimum of 5 years and met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria for schizophrenia or schizoaffective disorder as determined by review of medical records, consultation with treating clinicians, and structured clinical interview. All were free of a history of mental retardation, comorbid neurological or medical illness, or current substance dependence or abuse. Patients with any past diagnosis of substance dependence were also excluded. Patients were recruited from 2 outpatient clinics at the Maryland Psychiatric Research Center (n = 58) and from local mental health centers and private practitioners (n = 19). Patients were clinically stable and on stable medication for at least 12 weeks (no change in medication type or dose) prior to participation. One patient was clinically stable without antipsychotic medication. (A comparison group of healthy control subjects was also recruited. Data for this group were not considered in current analyses.)
All participants in the parent study underwent a lengthy assessment protocol. Twenty-nine of 38 GVFs and 26 of 39 PVFs also completed the MASC social skills RPT. There were a number of reasons for the reduced number of MASC completers relative to the entire sample. One (PVF) subject did not receive the MASC due to equipment failure. However, in most cases noncompletion of the MASC was due to fatigue and/or patient refusal (9 GVFs and 12 PVFs) at the conclusion of a lengthy assessment protocol. The subsample that completed the MASC did not differ from the full sample on age, education, of any of the Wechsler Adult Intelligence Scale-III or Wechsler Memory Scale-III (WAIS-III or WMS-III) index scores. This subsample was the focus of current analyses.
The MASC was derived from an empirically developed social problem–solving procedure that has proven to be reliable and valid in studies with the seriously mentally ill.9,21 It employs a series of 3-minute simulated conversations with a live confederate who portrays the protagonist in the situation (eg, an employer). In each case, the subject is presented with a short written description of the scenario and assigned a particular role. Then the situation giving rise to the role-play is described on an audiotape, after which the confederate delivers a prompt line to begin the interaction. Four MASC scenes were administered, all with a vocational context. Two scenes required the subject to initiate conversations to meet new people in the workplace, and 2 required assertion skills to ask the boss for a promotion or a second chance in a job training program after violating the rules. Patient responses in each scene were videotaped for subsequent coding on 3 performance dimensions: conversational content, nonverbal content, and overall effectiveness. Conversational content involved what the subject said. Nonverbal content involved voice inflection, facial expression, and eye contact. Overall effectiveness was an integrative category that reflected the subject's ability to achieve the goal in the interaction (eg, to make friends or get a job promotion). Each category was rated on a 5-point Likert scale by raters who were blind to patient vocational functioning designation. Interrater agreement (ICC) was high for each scale: conversational content ICC = 0.89, nonverbal content ICC = 0.93, and overall effectiveness ICC = 0.88.21 We should note that while MASC data were collected for healthy control subjects in the context of the parent study, they are not analyzed in the current article. The MASC was developed for use with the seriously mentally ill,9,31 and healthy controls perform at near-ceiling levels on the measure with little variance among subjects.
The cognitive battery was centered on the widely used WAIS-III and WMS-III.32,33 The WAIS-III yields composite indexes for verbal comprehension, perceptual organization, working memory, and processing speed. From the WMS-III, composite auditory and visual memory domain scores were created by combining the immediate and delayed subtest scores within each modality rather than using the standard cross-modality “immediate” and “delayed” memory index scores. This was done to create more homogeneous constructs and supported by preliminary analyses demonstrating very high correlations between immediate and delayed subtest pairs (average Spearman r = .85, range = .72–.94). These 6 Wechsler composites tap widely recognized domains of cognition. Their reliability and validity in schizophrenia populations are well established,34–37 as is their connection with functional and vocational status.1–3,38 However, a number of other cognitive domains that have received special attention in schizophrenia functional outcome research are not directly represented by the Wechsler indexes. Therefore, the Wechsler batteries were supplemented with additional cognitive measures addressing these dimensions of cognition. For example, the Wisconsin Card Sorting Test (WCST)39 has been the prototype test of cognitive flexibility and problem solving in relation to community functioning in schizophrenia4 and has emerged as an important correlate of vocational functioning.38,40,41 Similarly, crystallized verbal ability (eg, Wide Range Achievement Test, Reading [WRAT-R]42), verbal fluency, and the WAIS Comprehension subtest have also been associated with community functioning and, particularly, with vocational performance.1,3
As indicated, a total of 77 schizophrenia patients participated in the parent study, of whom 55 (29 GVF and 26 PVF) completed the MASC. Demographic characteristics of the GVF and PVF groups within this subsample are reported in table 1. The GVF and PVF subgroups were matched on age, gender, and race. GVF subjects had completed more education than PVF subjects (t(53) = 3.06, P = .003) and had higher full-scale intelligence quotient (FSIQ) scores (t(53) = 2.58, P = .013). For both groups, the great majority of patients were diagnosed with schizophrenia rather than schizoaffective disorder. The difference between groups in age of illness onset approached significance (t(53) = 1.93, P = .059). Table 1 also presents results of the study groups on the MASC total score and subscales. The between-groups difference for each of these MASC variables was significant (all P < .001).
Spearman correlations were calculated to explore the pattern of associations between MASC variables and selected cognitive variables, across the schizophrenia groups, with a Bonferroni adjustment of the significance criterion to P < .005 to correct for multiple comparisons. We considered the associations of MASC variables with WAIS-III FSIQ and the 6 Wechsler composite indexes (verbal comprehension, perceptual organization, working memory, processing speed, auditory memory, and visual memory) and with a selection of tests representing cognitive domains that have been found to be related to community functioning in schizophrenia: the WCST categories score39 (executive functioning), WRAT-R subtest42 (crystallized verbal ability), the WAIS-III Comprehension subtest (social judgment), and Controlled Oral Word Association43 (COWA; verbal fluency). To assist in interpretation of the Wechsler index associations, correlations are also reported for the following Wechsler subtests: WAIS-III Digit Symbol, WMS-III Logical Memory I, WMS-III Verbal Paired Associates I, WMS-III Family Pictures I, and WMS-III Facial Recognition I. Correlations of these variables with the MASC total score are provided in table 2.
Only 2 of the Wechsler composite indexes, verbal comprehension and processing speed, correlated significantly with the MASC total score. However, near significant correlations with FSIQ, perceptual organization, and auditory memory (P < .05) and significant associations with individual subtests clarify that the MASC association to cognition was considerably more general. Indeed, only the working memory index showed a negligible association. The highest individual subtest correlation with the MASC was for WAIS-III Digit Symbol Coding, a core processing speed measure. Although the composite indexes for auditory and visual memory were not reliably associated with the MASC, WMS-III Logical Memory, tapping auditory memory, and WMS-III Family Pictures, tapping visual memory was. The correlation of executive problem solving (WCST categories) with MASC performance approached but did not meet current significance criteria (P = .052).
We regressed the MASC on the full set of cognitive variables to determine the extent of the multivariate correlation. The regression included the 6 Wechsler index scores plus WAIS-III Comprehension, COWA, WCST Categories, and WRAT-R. FSIQ, Digit Symbol, and the individual Wechsler memory subtests were excluded as redundant with Wechsler index scores. The regression model was significant (F10,43 = 4.77; P < .001) and showed that the group of cognitive variables was associated with approximately half of the variance in MASC performance (R2 = .53, adjusted R2 = .42). However, the magnitudes of the individual variable correlations make it clear that no single variable shared more than 22% of its variance with the MASC (eg, for Digit Symbol r = .47, r2 = .22).
We also examined whether these associations varied as a function of the 2 different MASC scene types or the 3 MASC performance dimensions. The correlations between the cognitive variables and the MASC scenes requiring conflict resolution were consistently larger than those with the scenes requiring conversation initiation, but the magnitude of the difference was small across all variables. Among the 3 MASC performance dimensions, conversation content scores correlated consistently more with the cognitive variables than either the nonverbal content or overall effectiveness scores. Again, however, these differences were small.
The difference between the GVF and PVF groups on the MASC total score was significant (t(53) = 4.32, P < .001). We conducted 2 profile analyses using SAS PROC GLM to determine if group differences varied as a function of MASC variables. There was no main effect for MASC scene type (F1,53 = 0.60; P = .443), indicating that, across the schizophrenia groups, MASC scores did not differ depending on whether the scene required conversation initiation skills or assertiveness skills. There was a main effect for performance dimension (F2,106 = 22.82; P < .0001). Mean contrasts (comparing each dimension to the average of the other 2) revealed that MASC conversation content scores were higher for both GVFs and PVFs than were nonverbal and overall effectiveness scores (F1,53 = 45.79; P < .0001). However, vocational group did not interact either with MASC scene type (F1,53 = 0.07; P = .80) or with MASC performance dimension (F2,106 = 1.76; P = .18). Given that no MASC variable discriminated between vocational groups significantly better than others, all further analyses were conducted using the MASC total score (summed across scene types and performance dimensions) as the most reliable index of overall measure performance.
For present purposes, we conducted logistic regression analyses to determine the association of the MASC total score with our binary vocational functioning classification (see table 3). To achieve common scaling across tests, we standardized each measure using SDs from the full 55-person schizophrenia sample. Analyses showed that the MASC was a strong univariate predictor of vocational functioning, with an odds ratio = 3.61. Given the standardized scores, this means that a 1 SD difference in the MASC total score was consistent with more than a 3-fold increase in the probability that a subject would be categorized as a GVF. Table 4 presents the percentages of individuals correctly classified into vocational groupings using the MASC total score. This variable alone permitted nearly 73% accurate group classification.
Given the strong association of the MASC with overall cognitive performance and with various specific cognitive variables, further logistic regression analyses considered whether the MASC/vocational functioning association would remain significant after controlling for measures of cognitive performance. These analyses were done in a stepwise fashion, with cognitive variables entered as covariates in the first step and the MASC total score entered in the second step. WAIS-III Digit Symbol was the individual cognitive variable with the strongest bivariate association to the MASC total score (see table 2). Thus, one regression analysis controlled for Digit Symbol by entering it alone in the first step, followed in the second step by the MASC variable. Although the strength of prediction was attenuated relative to the univariate analysis, the MASC predicted vocational functioning independently of this measure (see table 3). A final analysis controlled for the full set of cognitive variables by entering them simultaneously in the first step. As before, this analysis included the 6 Wechsler index scores plus WAIS-III Comprehension, COWA, WCST Categories, and WRAT-R and excluded FSIQ, Digit Symbol, Logical Memory, and Family Pictures as redundant. The strength of prediction was again attenuated; however, this analysis demonstrated that MASC remained a significant, independent predictor of vocational functioning, even after control for a wide selection of intellectual and neuropsychological performance variables. (The outcome of this last analysis did not change in an alternative analysis in which Digit Symbol, Logical Memory, and Family Pictures were used in place of the 3 Wechsler index scores to which the tests contribute).
The classification accuracy findings in table 4 show that the MASC total score, by itself, enabled approximately 73% of all cases to be classified correctly. At approximately 80% classification accuracy, the entire set of cognitive variables was only incrementally better. The combination of MASC and cognitive variables allowed accurate classification of more than 85% of cases.
The current analyses explored the association of social/communication skills—that is, the specific verbal, nonverbal, and paralinguistic behaviors that comprise effective social responses—with a concrete marker of community functioning, namely competitive employment. Results support 2 main conclusions. First, performance on a measure of social/communication skills sharply distinguished GVF and PVF groups. Second, despite substantial associations of role-play performance with cognitive variables, social skills and cognition were, at least to a degree, independent predictors of vocational functioning.
Competent use of social/communication skills by people with schizophrenia is significantly related to actual role functioning in the community. In particular, patients with good social/communication skills, as measured by the MASC, were more likely than others to have been competitively employed over the previous 2 years. GVF patients demonstrated reliably greater social skill on the MASC, whether a test scene involved initiation of a conversation with a coworker or assertively pursuing a request to a workplace superior. They were more effective in their simulated conversations in terms of the actual verbal content they generated in the course of role-plays and in the nonverbal and paralinguistic behaviors that accompanied the verbal communications. The consistency of the advantage across the different test scenes and performance dimensions of the MASC indicated that the social skills tapped by this class of RPTs may be a fairly unitary set of behaviors, at least insofar as they relate to vocational functioning.
Our analyses clarified the relationship of RPT-indexed social skill with cognitive variables. The complete set of measures, including the cognitive domains most frequently associated with community functioning in schizophrenia, shared approximately 50% of its variance with the MASC total score. Correlations indicated that the relationship of role-play performance with cognition was broad, if not fully generalized. This is consistent with earlier findings of RPT associations with IQ, verbal ability, verbal memory, and cognitive flexibility but not working memory.7,9,20 In current analyses, role-play performance showed medium to high correlations with verbal comprehension and processing speed, among the Wechsler indexes. The association with verbal ability makes intuitive sense.7 Analogously to everyday social interactions, performance on the MASC and other RPTs require an individual to understand a situation and verbal interaction and to formulate context-appropriate verbal responses. The association of role-play performance with processing speed has not been shown in prior studies but has a degree of face validity. Competent social interaction depends on the ability to process and update details of an evolving interaction quickly. However, schizophrenia patients are often slow to respond in conversation, disrupting the flow of social exchange, and undermining communication objectives. Indeed, processing speed, indexed by Digit Symbol Coding, has been shown to be the most pronounced neuropsychological impairment in schizophrenia patients relative to healthy controls and strongly related to functional status.44 It is not wholly surprising, then, that this measure showed the largest correlation of any individual measure with social competence in current analyses.
For the auditory memory index, the MASC showed a high correlation with one underlying subtest but not with the other. As seen in table 2, the MASC was significantly associated with the Wechsler Logical Memory subtest, replicating our 1994 finding.9 This task measures the ability to remember details of a brief narrative. The story-like stimulus distinguishes logical memory from the other half of the Wechsler auditory memory index, a paired-associates learning task in which there is no similar contextual information to cue memory performance. One hypothesis is that the reliance of MASC performance on recollection of details from the audiotape presentation of the role-play narrative context for each MASC scene emphasizes complex, “contextualized” memory over arguably purer memory measures, such as Verbal Paired Associates, that preclude reliance on semantic or contextual cues. However, the selectivity of this association remains uncertain. Using a different role-play measure, Addington and colleagues found a strong association of social/communication skills with the full auditory index (ie, comprising both logical memory and verbal paired associates subtests).7
MASC associations with the fluid reasoning abilities measured by the perceptual organization tests and the problem-solving and feedback response tapped by the WCST did not meet stringent significance criteria used in this study. However, these associations approached significance in current analyses and reached significant levels in earlier analyses.7,45 Further investigation will be needed to test these relationships. On its face, MASC requires an individual to hold scene-relevant information in mind, to update this information with details of an evolving interaction, and to respond in real time. Thus, significant associations with the online storage and updating capacities tapped by working memory tasks might have been expected. However, the association of the WAIS working memory index with the MASC was both nonsignificant and trivial in absolute terms. Moreover, this finding is consistent with the only other study that has examined the relationship between MASC and working memory variables.9 In sum, the current analyses suggest that role-play–indexed social skill is strongly associated with cognitive measures tapping verbal and perceptual abilities, memory, problem solving, and information processing speed. Social skills performance was not significantly associated with working memory measures in these or earlier data.
Regarding independence and overlap between cognitive and social skills performance, results of current analyses were clear. MASC-indexed social skills discriminated between GVF and PVF groups significantly and, at least to a degree, independently of cognitive performance, despite the substantial correlations. The cognitive test battery analyzed here covered core domains of cognition, including the verbal memory, executive, and processing speed domains that have been most frequently linked with functional and vocational outcome in schizophrenia.1–4,6,38 Yet, taking this broad cognitive battery into account, performance on the MASC retained an independent association with vocational functioning. It has been suggested that social problem solving may mediate the effect of cognition on everyday functioning.10,46,47 Current analyses are not inconsistent with this hypothesis. However, they highlight that the social/communication skills indexed by the MASC and similar RPTs may also be conceptualized as a parallel dimension of functional assessment, partially overlapping with and partially distinct from cognitive performance.
It is not possible from these data to determine exactly what comprises the unique dimension of the relationship between social skills and functioning. A reasonable speculation is that it depends on emotional, personality, life experience, and other factors that are not directly cognitive in nature. An interesting question is whether tasks such as the MASC actually provide more useful information about an individual's potential functional level than direct community observation. Given the clinical assessment context and the simulated nature of the interactions, these tests are probably best seen as indicators of an individual's capacity for instrumental use of social/communications skills.22,25 A person's actual behavior in the community is influenced by numerous additional factors, including opportunity, experience, and temperament. As indexes of what an individual could do in the community given appropriate opportunity and support, however, RPTs such as the MASC may be more relevant from a psychiatric rehabilitation standpoint than actual community behavior.
It remains to be seen whether the degree of independence of social skills and cognitive performance in predicting vocational performance holds with respect to other concrete, everyday functional roles, such as social relationships and self-care. Further research will be needed to better understand the nature and breadth of the association between social skills and functioning. Still, current analyses suggest that RPTs may be valuable additions to assessment batteries designed to predict aspects of everyday community functioning.
This study was supported by US National Institute of Mental Health grants MH57749 (Gold) and MH67764 (Dickinson) and a US Veterans Administration Research Career Development Award (Dickinson).