Disorganized Symptoms and Executive Functioning Predict Impaired Social Functioning in Subjects at Risk for Psychosis

The Cognitive Assessment and Risk Evaluation (CARE) program at the University of California, San Diego, provides longitudinal assessment of individuals who are considered to be at risk for developing psychosis.⁹ For this study, we selected a subsample (N=22) of at-risk subjects from previously published reports^6,10,11 who had received the Social Adjustment Scale-Self-Report (SAS-SR) at 1-year follow-up.

All participants received a comprehensive clinical and neurocognitive battery at initial assessment.⁶ The clinical assessment included the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) and the Structured Interview for Prodromal Syndromes (SIPS).¹²

The SAS-SR was administered to all participants at baseline assessment and at 1-year follow-up. Domains related to work/school role and social/leisure time were examined. We did not examine the two SAS-SR family domains (family unit and family outside of the home) because of the young age of our sample.

The neurocognitive battery^6,11 included the Wisconsin Card-Sorting Test (WCST), the Stroop Color–Word Test, the Numerical Attention test, the Hopkins Verbal Learning Test (HVLT), Revised, the Spatial Span subtest of the Wechsler Memory Scale, 3rd Edition, as well as the Letter Number Sequencing, Block Design, and Vocabulary subtests of the WAIS, 3rd Edition.

Analyses were performed with SPSS. Change in symptom severity and social functioning over time was assessed via paired-samples t-tests. We performed correlations between follow-up social functioning measures and baseline clinical and neurocognitive batteries to determine which variables should be included in regression analyses (with significance at p<0.05). Backward multiple-regression analyses were then performed using baseline social functioning, clinical, and neurocognitive variables as predictors of follow-up social functioning. The assumptions of normality, linearity, and homoscedasticity of residuals were met. With the use of a p<0.001 criterion for Mahalanobis distance, no outliers among the cases were found. No cases had missing data, and no suppressor variables were found (N=22).

The baseline demographic characteristics and neurocognitive scores of our sample are presented in Table 1. The rate of transition to psychosis in our sample was 9.1%; two subjects converted to schizophrenia over the 1-year follow-up.

As shown in Table 1, paired-samples t-tests showed significant improvement between baseline and follow-up on the Global Assessment of Functioning (GAF) and all of the SIPS ratings. However, there were no significant changes over time on any of the SAS-SR ratings. The at-risk subjects remained impaired in their social functioning at follow-up.

Given our relatively small sample size, we limited the number of predictor variables that were used in the regression analyses to those that had significant correlations (p<0.05) with the social functioning variables at follow-up (r=0.45 to 0.67). Yet, for each dependent variable, the corresponding baseline variable was entered into the prediction model regardless of the magnitude of the correlation between them. Interestingly, the outcome social-functioning measures were not significantly correlated with baseline GAF, SIPS positive symptoms, or performance on the WCST. In order to avoid multicollinearity problems, we examined the correlations among the predictor variables. Those correlations were all small-to-moderate (r=0.43 to 0.59; p<0.05) except for the high association between SIPS Disorganized and SIPS Negative (r=0.71; p<0.001).

We decided to include both of those variables in the regression analyses because they represent different symptom domains. We performed three backward multiple-regression analyses, one for each SAS outcome variable.

When the four variables that were significantly correlated with follow-up, Overall SAS, in addition to baseline Overall SAS scores, were included in the first backward multiple regression; we observed a significant regression coefficient that accounted for 70% of the variance in overall functioning at outcome: F[5, 15]=7.18; p=0.001; R²=0.70 (at Step 1). After Stroop Color Naming and SIPS Negative were excluded from the model, baseline Overall SAS, SIPS Disorganized, and Stroop Color/Word Interference still accounted for 70% of the variance in follow-up Overall SAS: F[3, 17]=13.29; p<0.001; R²=0.70 (at Step 3). The most significant predictor of overall functioning was SIPS Disorganized (β=0.59), followed by Stroop Color/Word Interference (β = –0.44). Baseline SAS Overall was not a significant predictor (β=0.28; p=0.06).

When the four variables that were significantly correlated with follow-up SAS Social/Leisure, in addition to baseline SAS Social/Leisure, were included in the second backward multiple regression, we observed a significant regression coefficient that accounted for 64% of the variance in social role functioning at outcome: F[5, 15]=5.42; p=0.005; R²=0.64 (at Step 1). After SAS Social/Leisure and SIPS Negative were excluded from the model, the three-predictor model, including SIPS Disorganized, HVLT Total Recall, and Stroop Color/Word Interference, accounted for 61% of the variance in follow-up SAS Social/Leisure, F[3, 17]=8.78; p=0.001; R²=0.61 (at Step 3). The most significant predictor of social role functioning was SIPS Disorganized (β=0.49), followed by the Stroop Color/Word Interference (β = –0.37). HVLT's regression coefficient was not significant (β = –0.31; p=0.08).

When the three variables that were significantly correlated with follow-up SAS Work Role, in addition to baseline SAS Work Role, were included in the third backward multiple regression, we observed a significant regression coefficient that accounted for 81% of the variance in Work Role Functioning at outcome: F[4, 11]=11.62; p=0.001; R²=0.81 (at Step 1). After baseline SAS Work Role was excluded from the model, Stroop Color-Naming, Stroop Color/Word Interference, and Numerical Attention significantly contributed to the prediction of follow-up SAS Work Role, accounting for 79% of the variance: F[3, 12]=14.92; p<0.001; R²=0.79 (at Step 2). The most significant predictor of follow-up Work Role Functioning was Stroop Color-Naming (β = –0.50), followed by the Stroop Color/Word Interference (β = –0.38), and Numerical Attention (β=0.31).