Neurocognitive impairment and functional outcome in cognitive-behavioral social skills training for schizophrenia

Abstract

Background:

Neurocognitive impairment is associated with low functional outcome in schizophrenia. Mixed findings exist regarding the impact of neurocognitive impairment on functional improvement following psychosocial treatments. Cognitive-Behavioral Social Skills Training (CBSST) is an intervention aimed at improving functioning. The current study examined whether baseline neurocognitive impairment moderated functional outcome in CBSST relative to a goal-focused supportive contact condition (GFSC).

Study design:

This is a secondary analysis of two single-blind, randomized controlled trials conducted in individuals with schizophrenia or schizoaffective disorder (Trial 1: N = 64 (older adults M age = 55.0); Trial 2: N = 149 (M age = 41.4)). Participants received 36 weekly 2-h group sessions of either CBSST or GFSC. CBSST combined cognitive behavioral therapy with social skills training and problem solving to improve functioning. GFSC focused on setting and achieving functioning goals, with minimal therapist guidance. Regression models examined whether multiple domains of baseline neuropsychological impairment moderated functional outcome at 9- or 12- month follow-up in CBSST relative to GFSC.

Study results:

Executive function impairment, but not any other neuropsychological domain, was a significant moderator of treatment outcome. Lower executive functioning was associated with better functioning at follow-up in CBSST than in GFSC, whereas higher executive functioning was associated with similar functional outcome in CBSST and GFSC.

Conclusions:

Individuals with cognitive impairment can benefit from CBSST and may even benefit more than individuals with intact cognition. The thought challenging, cognitive flexibility and problem-solving skills trained in CBSST may be compensatory for individuals with schizophrenia with executive function deficits.

1. Introduction

Neuropsychological impairment is a core feature of schizophrenia (Dickinson et al., 2007), and plays an important role in determining many functional outcomes (Bowie et al., 2006; Slade and Salkever, 2001). Antipsychotic medication reduces psychotic symptoms but has limited impact on social and/or cognitive aspects of the disease (Carpenter and Koenig, 2008; Gray and Roth, 2007; Palmer et al., 2009). Cognitive deficits and negative symptoms most strongly impact social disability in schizophrenia (Green, 1996; Kurtz, 2006). Thus, there is significant need to develop psychotherapy treatments to improve cognition, negative symptoms and functioning in schizophrenia (Bellack and Mueser, 1993; Liberman, 1994; Penn and Mueser, 1996). Two such interventions garnering significant empirical support are cognitive behavioral therapy (CBT) and social skills training (SST) (Turner et al., 2014, 2018).

Considering the amount of resources expended and the cost of these interventions, it is important to identify predictors of treatment outcome. Although there is considerable evidence that cognitive impairment predicts poor functioning, there is inconsistent evidence that cognitive impairment predicts treatment outcome in psychosocial interventions (Bowen et al., 1994; Green et al., 2000; Kern et al., 1992; McKee et al., 1997; Wykes and Reeder, 2005). A review (Kurtz, 2011) of 20 studies of evidence-based behavioral treatments found that cognitive deficits predicted poorer treatment outcome in SST, work and comprehensive rehabilitation interventions in schizophrenia. Interestingly, however, severity of neuropsychological impairment was not found to predict poorer outcome in CBT. In fact, one study from this review reported paradoxical effects of cognition on outcome, in that poorer executive functioning predicted a better response to CBT (Garety et al., 1997). We also found that poorer executive functioning predicted better outcome in cognitive behavioral social skills training (CBSST) in older adults with schizophrenia (Rajji et al., 2022).

One possible explanation for this paradoxical finding that poorer executive functioning predicted better outcome in CBT interventions is that patients with poorer executive functioning are most in need of interventions like CBT that teach problem-solving, hypothesis-testing and cognitive flexibility skills, because they are less able to learn these skills on their own. Thus, while cognitive impairments like attention and memory deficits may interfere with SST skill learning, CBT techniques such as cognitive restructuring that teach patients to gather evidence to test hypotheses about beliefs and increase cognitive flexibility may be compensatory for individuals with poorer executive functions. Individuals with better executive functions show less improvement in therapy because they can already do this on their own. These techniques may aid individuals with more severe cognitive impairment in ways that SST alone cannot. Given this possibility and the uncertainty about the impact of cognitive impairment on CBT and SST outcomes in schizophrenia, this study examined the impact of cognitive impairment on CBSST (Granholm et al., 2016).

CBSST combines CBT with SST and problem-solving training and has been found to improve functioning in individuals living with schizophrenia in multiple clinical trials (Granholm et al., 2005; Granholm et al., 2008; 2012, 2014; Rajji et al., 2022). Two prior clinical trials examined whether cognitive impairment was a moderator of functioning outcomes in CBSST in schizophrenia. In one study (Granholm et al., 2008), cognitive impairment predicted worse functional outcome for both treatment groups (CBSST and treatment as usual, or TAU), but was not a significant moderator (Granholm et al., 2008). That is, greater severity of cognitive impairment predicted worse functional outcome in general but did not predict worse outcome in CBSST relative to TAU. However, limited measures of executive functions were included (i.e., TrailMaking Test B, Controlled Oral Word Association Task – FAS Total, Wisconsin Card Sorting Test – Perseveration errors). In another recent clinical trial in older participants with schizophrenia (age >65) (Rajji et al., 2022), baseline executive dysfunction (i.e., the Executive Interview [EXIT]) was a significant moderator of functional outcome in CBSST relative to TAU, such that individuals with poorer executive functions demonstrated better functional outcome in CBSST than those with better executive functions or those with poorer executive functions in the TAU group.

To replicate and extend this previous research, the current study examined neuropsychological moderators of CBSST treatment outcomes relative to a goal-focused supportive contact (GFSC) control group, using a neuropsychological assessment battery with more comprehensive executive function measures and a sample representative of a broader age range. The primary study aim was to examine whether severity of baseline neuropsychological impairment, especially executive dysfunction, moderates functional outcome in CBSST relative to GFSC. This study was a secondary analysis of data pooled from two prior CBSST trials (Granholm et al., 2013, 2014). Although the specific moderator analyses were not prespecified in a published trial protocol, they were conceptually driven, aligned with the funded study aims, and designed to replicate the findings of Rajji et al. (2022). Consistent with Rajji et al. (2022), we hypothesized that individuals with lower baseline executive functioning would demonstrate greater improvements in functioning after CBSST relative to GFSC.

2. Methods

2.1. Design

This study is a secondary analysis of data from two prior randomized clinical trials (Granholm et al., 2013, 2014), both registered at ClinicalTrials.gov (NCT00237796; NCT00338975). The primary difference between the studies was that one focused on older adults with serious mental illness (M age → 55, SD → 6.6), while the other included a broader adult sample (M age = 41.3, SD = 9.8). The inclusion/exclusion criteria, 36-session CBSST group therapy intervention and GFSC control intervention were identical across trials. However, the study with younger participants included monthly booster sessions during follow-up (which were rarely attended), while the older adult study had a shorter post-treatment follow-up period (9 vs. 12 months; see below for more detail). The primary outcome variable, the Independent Living Skills Survey, was administered at all assessment points in both trials. Pooling the data yielded a more representative adult sample and greater statistical power to examine neuropsychological predictors of functional outcomes in this secondary hypothesis-driven analysis.

All study procedures were approved by the institutional review board of the University of California, San Diego, and the Veterans Affairs San Diego Healthcare System. Participants were adult Veterans and non-Veterans with a primary diagnosis of schizophrenia or schizoaffective disorder who were recruited through a variety of community settings throughout San Diego County, including board and care/assisted living homes, clubhouses/drop-in settings, outpatient psychiatry clinics, and other treatment settings in the UC San Diego Health System and VA San Diego Healthcare System. After meeting eligibility criteria, participants were randomly assigned to CBSST or goal-focused supportive contact (GFSC). For both studies, an independent statistician completed randomization using a computer-generated randomization list. Assessors were blind to treatment allocation. Each study had a coordinator, who was not involved in assessments or treatments, who contacted the statistician to ascertain treatment assignment. Participants were then treated for 9 months and followed for either 9 or 12 months after treatment. Both studies included assessments at baseline, 4.5-months (mid-treatment), and 9-months (end-of treatment). Follow-up assessments were conducted at 13.5 months and 18 months (Granholm et al., 2013) or 15 months and 21 months (Granholm et al., 2014). We used data from baseline and final follow-up assessments for this study. Data from the two studies were combined for each assessment time point.

2.2. Interventions

Participants in both treatment conditions were offered a total of 36 weekly group therapy sessions (9 months), which were 2 h with a lunch or snack break mid-way. Participants in the Granholm et al. (2014) trial were offered monthly booster sessions after the 9-month treatment phase, but these were rarely attended. Aside from the length of follow-up period and booster sessions, the interventions for the two RCTs were identical. Each group session was co-facilitated by two master’s- or doctoral-level therapists with a minimum of two years of CBT experience. Fidelity ratings indicated adequate adherence to intervention protocols in both trials and are described in greater detail in the prior publications.

2.2.1. Cognitive-behavioral social skills training

CBSST (Granholm et al., 2016) was delivered in 3, 6-session modules that were intended to be completed twice, for a total of 36 weekly sessions. The 3 CBSST modules were Cognitive Skills, Social Skills, and Problem-Solving Skills. Training thought challenging skills was the exclusive focus of the Cognitive Skills Module, but thought challenging was also used throughout the other 2 modules. Group members were introduced to the general concepts of CBT, including the relationship between thoughts, behaviors, and feelings (generic cognitive model), automatic thoughts, and thought challenging through behavioral experiments, examining evidence for and against beliefs, and mistakes in thinking. Thought challenging skills were used to address symptoms and challenge defeatist beliefs that interfere with functioning behaviors. The primary thought challenging skill trained was the 3C’s: Catch It, Check It, Change It (“It” is referring to an unhelpful thought).

The primary goal of the Social Skills Module was to improve communication skills through behavioral role plays, including active listening, expressing positive and negative feelings, and making positive requests. Important role plays included assertive interactions with co-workers, friends and family, making new friends, and effectively interacting with case managers, doctors, and other support persons.

Basic problem-solving skills were trained in the Problem-Solving Skills Module using the acronym, SCALE – Specify the problem, Consider all possible solutions, Assess the best solution, Lay out a plan, and Execute and Evaluate the outcome. The focus was on developing specific, feasible plans to solve real-world problems, including scheduling pleasant activities, improving living situations, finances, using public transportation, finding a volunteer or paid job, and enrolling in classes.

2.2.2. Goal-focused supportive contact

The GFSC intervention was a semi-structured enhanced supportive contact control condition with a primary focus, like CBSST, on setting and achieving functioning goals. Sessions included components of psychoeducation, empathy, and nondirective reinforcement of health, coping, and symptom management behaviors, which grew out of group discussions about goals, with minimal therapist guidance.

2.3. Participants

The inclusion/exclusion criteria were the same across studies except for age. Inclusion criteria were: (1) Age >18 (Granholm et al., 2014) or Age >45 (Granholm et al., 2013), (2) diagnosis of schizophrenia (N = 170) or schizoaffective disorder (N = 43) based on the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (SCID) (First, 1997) and available medical re-cord review, and (3) capacity to provide informed consent. The minimal exclusion criteria consisted of: (1) Prior exposure to CBT or SST during the previous 5 years, and (2) level of care required at baseline that would interfere with participation in outpatient therapy groups or assessments (e.g., disabling medical problems, or current hospitalization for medical, psychiatric, or substance abuse problems). 213 participants from two prior studies (Granholm et al., 2013, 2014) were eligible and had baseline data; however, 107 participants had baseline and follow-up data on the outcome variable of interest, and were included in the current analyses. 59.8 % of the sample were male; 59.8 % were White/Non-Hispanic, 12 % were Black, 12.1 % were Latino or Hispanic, 1.9 % were Asian, 4.7 % were Native American, 6.5 % were Biracial, 0.9 % were Pacific Islander, and 1.9 % identified in the “Other” category. On average, participants were 48 years old (SD = 10.5) with 12 years of education (SD = 1.7). More detailed baseline sample characteristics are shown in Table 1.

Table 1.

Baseline characteristics for participants with follow-up data available (N = 107).

	Total combined sample (n = 107)		CBSST group (n = 47)	GFSC group (n = 60)
	Mean (SD) or %	Range	Mean (SD) or %	Mean (SD) or %	t or χ2	df	p-Value	Effect Size (Cohen’s d)
Demographics
Age	48.5 (10.5)	18–78	47.9 (10.7)	49.0 (10.3)	−0.52	105	0.603	−0.10
Education	12.3 (1.7)	8–16	12.5 (1.7)	12.2 (1.7)	0.86	105	0.390	0.17
Gender (% male)	59.8	–	55.3	63.3	0.70	1	0.401	−.081
%White/non-Hispanic	59.8	–	55.3	63.3	0.70	1	0.401	−.081
%Black	12.1	–	12.8	11.7	–	–	–	–
%Asian	1.9	–	4.3	0	–	–	–	–
%Native American	4.7	–	6.4	3.3	–	–	–	–
%Pacific Islander	0.9	–	0	1.7	–	–	–	–
%Biracial	6.5	–	6.4	6.7	–	–	–	–
%Other	1.9	–	4.3	0	–	–	–	–
%Latino	12.1	–	10.6	13.3	–	–	–	–
Clinical Characteristics
%Schizophrenia Diagnosis	82.2	–	85.1	80	–	–	–	–
%Schizoaffective Diagnosis	17.8	–	14.9	20	–	–	–	–
Functional Measure
ILSS Composite	0.70 (0.11)	0.39–0.88	0.71 (0.09)	0.69 (0.12)	1.07	105	0.289	0.21
Cognitive Measures
Processing Speed	32.3 (9.3)	12–61	33.1 (9.5)	31.7 (9.2)	0.78	104	0.440	0.15
Working Memory	37.0 (9.5)	15–60	39.3 (9.8)	35.2 (9.0)	2.22	104	.029	0.44
Verbal Learning	33.6 (8.4)	17–64	34.7 (7.3)	32.7 (9.2)	1.24	104	0.216	0.24
Visual Learning	29.0 (12.9)	10–61	30.4 (13.8)	28.0 (12.1)	0.98	104	0.330	0.19
Verbal Memory	30.0 (11.4)	20–61	32.2 (12.0)	28.3 (10.8)	1.74	105	0.085	0.34
Visual Memory	29.1 (13.7)	19–65	31.4 (14.6)	27.3 (12.8)	1.56	104	0.122	0.31
Executive Functioning	38.2 (10.2)	20–62	39.0 (11.0)	37.5 (9.5)	0.79	105	0.429	0.16
Global NP	34.0 (8.0)	18–54	35.6 (8.4)	32.8 (7.6)	1.75	104	0.084	0.34

Note.

¹φ. Bold font denotes ps < 0.05. All cognitive variables are reported as t-scores; dfvaried due to missing data for some variables; ILSS = Independent Living Skills Survey. Global NP = Global neuropsychological performance, as determined by the average of the 12 cognitive tests’ t-scores.

2.4. Measures

2.4.1. Neurocognition (moderators)

Participants completed a comprehensive battery of neuropsychological tests that measured performance in the following cognitive domains: processing speed, executive functioning, working memory, and verbal and visual learning and memory. The cognitive domains assessed in the current study are aligned with those in the MATRICS Consensus Cognitive battery (MCCB (Nuechterlein et al., 2008);), a battery recommended for assessing cognition in schizophrenia. Additionally, most tests used for the current study (described below) are part of the MCCB but expanded the battery to include additional measures of executive functioning. The neuropsychological measures (and references to normative data used to convert raw scores to demographically-corrected T-scores) were as follows: for processing speed, Brief Assessment of Cognition in Schizophrenia (BACS) (Keefe et al., 2004): Symbol Coding and Category Fluency subtests, and Trail Making Test Part A) (Heaton et al., 1991; Reitan and Wolfson, 1993); for executive functioning (e.g., concept formation and abstract reasoning, problem solving, cognitive flexibility, inhibition), Delis-Kaplan Executive Function System (D-KEFS) (Delis et al., 2001): 20 Questions (total questions asked), Word Context (total consecutively correct), and Sorting (confirmed correct total sorts) subtests; for working memory, Wechsler Memory Scale- 3rd edition (WMS-III) (Wechsler, 1997) Spatial Span subtest, and Wechsler Adult Intelligence Scale- 3rd edition (WAIS-III) (Wechsler, 1997) Letter-Number Span subtest; for verbal learning and memory, Hopkins Verbal Learning Test- Revised (HVLT-R) (Benedict et al., 1998) — Trials 1–3 total recall and delayed free recall; and for visual learning and memory, Brief Visuospatial Memory Test Revised (BVMT-R) (Benedict et al., 1996) — Trials 1–3 total recall and delayed free recall. The mean score within each domain was computed and a global impairment T-score was computed as the average of all available T-scores. Global T-scores were calculated for participants with available data for at least ten of the neuropsychological test variables.

2.4.2. Functioning (primary outcome)

Everyday functioning was measured using the Independent Living Skills Survey (ILSS) (Wallace et al., 2000), a 51-item interview-based measure that assesses ten domains of functioning over the past 30 days. A composite score was computed as the average of scores (rated 1 = yes, performed; 0 = no, not performed; or not able to demonstrate) on 5 relevant functional domains (appearance and clothing, personal hygiene, health maintenance, transportation, and leisure and community activities; range = 0–1; higher scores indicate higher functioning).

2.5. Statistical analysis

All analyses were conducted using IBM SPSS Statistics (version 29) software. Normality was confirmed for variables of interest (i.e., skewness and kurtosis values were within an acceptable range). Independent sample t-tests and chi-square tests were used to compare the CBSST and GFSC groups on baseline characteristics (see Table 1; effect sizes are reported as Cohen’s d or Phi, where appropriate). Linear regression analyses were conducted to examine whether neurocognitive impairment at baseline moderated the association between group assignment (CBSST vs. GFSC) and functioning at follow-up. Each moderation model included group (coded as CBSST = 1; GFSC = 0), baseline functioning, baseline neurocognitive domain t-score, and the interaction between group and baseline neurocognitive domain T-score. To account for potential confounding factors associated with the different trials (e.g., different participant age ranges, rates of attrition, follow-up periods), we conducted additional moderation analyses including trial (study 1 vs. 2) and age as covariates to control for differences between studies. Interactions were further probed using the PROCESS macro in SPSS (Hayes, 2022). Specifically, the Johnson-Neyman (JN) technique (Johnson and Neyman, 1936) was used to identify the specific T-score value of the moderator (baseline cognitive variable) at which the association between group assignment and functional outcome at follow-up became significant. This JN-derived value was used for interpretation and visualization purposes only.

3. Results

213 participants were randomized at baseline across the two studies, but 106 individuals dropped out before the follow-up assessment, resulting in a total dropout rate of 50 %. Study completers (Mean age = 48, SD = 10.5) were significantly older than those who dropped out (Mean age = 42, SD = 10.5), (t(211) = 4.24, p < .001). There were no other significant group differences in any baseline demographic, clinical, functional, or cognitive measures between those who dropped out versus those who completed. CBSST and GFSC groups did not significantly differ on baseline characteristics (i.e., demographics, clinical diagnosis, functional measures, cognitive measures), except that those in the CBSST group had higher working memory at baseline (see Table 1). Covariate-adjusted models accounting for the different trials and age did not yield any significant difference in outcomes (see Supplemental Tables 1 and 2).

The regression analysis examining whether impairment in baseline executive functioning moderated the effect of group assignment on self-reported functional outcomes (ILSS) at follow-up was statistically significant overall (R² = 0.45, F(4,102) = 21.04, p < .001). The addition of the group by baseline executive functioning interaction resulted in a significant change to the model, F(1,102) = 6.75, p = .01, R² change = 0.04, and the group by baseline executive functioning interaction was significant (β = −0.78, p = .01; see Table 2). The Johnson-Neyman technique identified that, for individuals with an executive function domain score of T < 30.7 (representing 25.2 % of the sample), better follow-up functioning (ILSS) was found for CBSST relative to GFSC, but for individuals with higher executive functioning domain scores, functioning was similar at follow-up for participants in CBSST and GFSC (see Fig. 1). Including baseline working memory as a covariate did not significantly change the model results. Moderation analyses for other baseline cognitive domains were nonsignificant (ps > 0.05), indicating that baseline performance in other neurocognitive domains did not significantly impact associations between group assignment and functional outcomes at follow-up (Table 3).

Table 2.

Predicting ILSS at follow-up from baseline executive functioning performance and group assignment (N = 107).

Predictor	b	β	t	p	95 % CI
Group	0.17	0.76	2.65	.009	0.04, 0.29
Baseline Executive  Functioning Performance	0.004	0.35	3.30	.001	0.002, 0.006
Baseline Executive  Functioning Performance x Group Assignment	−0.004	−0.78	−2.60	.011	−0.007, −0.001
Baseline Functioning (ILSS)	0.60	0.59	7.98	<.001	0.45, 0.75

Note: bold font indicates ps < 0.05; b = unstandardized coefficients; β = standardized regression coefficients; CI = Confidence Interval of unstandardized regression coefficients; Group (GFSC: 0; CBSST: 1); ILSS = Independent Living Skills Survey; higher scores on executive functioning and self-reported functioning are indicative of better cognitive performance and functioning, respectively.

Fig. 1.

Mean predicted ILSS at follow-up (N = 107)

Note: ILSS = Independent Living Skills Survey; CBSST = Cognitive-Behavioral Social Skills Training; GFSC = Goal-Focused Supportive Contact; EF = executive functioning; J-N = Johnson-Neyman technique (a t-score of 30.7 was calculated and used to dichotomize low versus high EF groups). The figure displays predicted ILSS follow-up scores based on a linear regression model including group, EF, the group × EF interaction, baseline ILSS, and baseline executive functioning. Bars represent covariate-adjusted means at the average values of baseline ILSS and EF, not raw group means.

Table 3.

Predicting ILSS at follow-up from baseline cognitive performance and group assignment; nonsignificant interaction results (N = 107).

Predictor	b	β	t	p	95 % CI
Baseline Processing Speed x Group Assignment	−0.001	−0.15	−0.49	0.622	−0.004, 0.003
Baseline Working Memory x Group Assignment	−0.001	−0.21	−0.62	0.538	−0.005, 0.002
Baseline Verbal Learning x Group Assignment	−0.002	−0.29	−0.82	0.416	−0.006, 0.002
Baseline Verbal Memory x Group Assignment	−0.000	−0.08	−0.32	0.746	−0.003, 0.002
Baseline Visual Learning x Group Assignment	−0.001	−0.13	−0.60	0.547	−0.003, 0.002
Baseline Visual Memory x Group Assignment	−0.001	−0.24	−1.17	0.247	−0.004, 0.001
Global NP Score x Group Assignment	−0.003	−0.48	−1.36	0.176	−0.007, 0.001

Note: β = unstandardized coefficients; 0 = standardized regression coefficients; CI = Confidence Interval of unstandardized regression coefficients; Group (GFSC: 0; CBSST; 1); ILSS = Independent Living Skills Survey; Global NP = Global neuropsychological performance, as determined by the average of the 12 cognitive tests’ t-scores. Higher scores on objective cognitive functioning and self-reported functioning are indicative of better cognitive performance and functioning, respectively. All models included group, a baseline cognitive domain, the interaction between group and cognitive functioning, and baseline ILSS as a covariate; only the nonsignificant interaction term results are reported for each model.

4. Discussion

The present study examined whether neurocognitive impairment moderated functional outcomes in CBSST relative to a robust control condition, goal-focused supportive contact (GFSC). Lower executive functioning was associated with better functioning at follow-up in CBSST than in GFSC, whereas higher executive functioning was associated with similar functional outcome in CBSST and GFSC. These findings suggest that CBSST may be particularly beneficial for individuals with schizophrenia and low executive functioning. CBSST may be compensatory for executive dysfunction, as it involves structured, skills-based practice of cognitive flexibility (e.g., via hypothesis testing and thought-checking) and problem solving. In contrast, individuals with more intact executive functioning in CBSST may have possessed the cognitive flexibility skills necessary to test beliefs and problem-solve independently, which could explain why they did not experience significant additional benefits from CBSST compared to GFSC. Individuals with low executive functioning in GFSC showed less functional improvement than those with intact executive functioning in GFSC, highlighting the added benefit of CBSST over a less-structured, general goal-oriented control group. Individuals with executive dysfunction in the GFSC group may have lacked the organizational and planning skills to work effectively toward their goals without structured support and skills training, whereas those with higher executive functioning may have been better equipped to navigate goal setting and planning.

Our results are consistent with previous CBSST trials that found individuals with schizophrenia with greater cognitive impairment are able to benefit, and may even benefit more, from CBSST (Granholm et al., 2008; Rajji et al., 2022). Previous research has also demonstrated that individuals with cognitive dysfunction may benefit from various other psychotherapies, such as CBT or SST. For example, one study (Mueser et al., 1991) found that individuals with memory deficits exhibited social skill improvement following a brief social skills training provided in an inpatient context, although at a slower rate than individuals with better memory performance. The results are also partially consistent with the review by Kurtz (2011) that found severity of neuropsychological impairment was not predictive of poorer CBT outcomes, and that poorer executive functioning predicted better CBT response in one study (Garety et al., 1997). However, this review also found that attention and verbal memory deficits were linked with poorer outcomes in SST interventions. It is possible that the current study’s integration of CBT with SST provided beneficial cognitive skills that helped to offset poorer treatment outcomes for those with cognitive difficulties. A growing body of literature in serious mental illness populations suggests that individuals with cognitive impairment benefit from various psychological treatments.

The current study has several strengths. The combination of data from two previous studies with similar methodologies (Granholm et al., 2013, 2014) allowed for examination of CBSST-related functional outcomes in a large sample of individuals with serious mental illness. Additionally, the sample is representative of various age ranges and participant characteristics (e.g., Veteran and non-Veteran). This study also included the use of a comprehensive neuropsychological assessment with sensitive neuropsychological measures, and the inclusion of a robust control group to provide nonspecific therapist contact of equivalent duration as compared to the CBSST group.

Despite these strengths, there were several limitations. Because we combined data from two previous studies, there were differences in follow-up periods. Additionally, our combined study dropout rate was high, with only 50 % of participants across the two combined samples with follow-up functioning data. Although completers and dropouts did not differ significantly on demographics, functioning (ILSS), global neuropsychological impairment, or executive functions at baseline, the potential for systematic differences or bias due to unmeasured confounders remains. Differential dropout may have introduced bias and/or regression to the mean could have contributed to the observed effects. To address these concerns, we conducted additional moderation analyses including covariates (trial, age) and the pattern of results remained consistent. The high attrition rate is a significant limitation and further research is needed to examine generalizability of the results, but the findings are suggestive of greater benefit for some individuals with greater executive function impairment, which is consistent with a prior report in an independent sample (Rajji et al., 2022). We also did not correct for multiple comparisons and correction would reduce the significance of the executive function interaction to a trend. However, we did have an a priori hypothesis that executive function impairment would moderate outcome based on our prior study (Rajji et al., 2024), and our aim was to explore the specificity of moderation effects by including multiple other cognitive domains, while not being overly conservative by sacrificing false negative to minimize false positive error. The secondary analyses in this report should be replicated in research with more conservative pre-specified analyses. For those with stronger executive functioning, full-length CBSST may not yield the same level of benefit seen in individuals with low executive functioning. Thus, future research should explore whether individuals with stronger executive functioning could benefit from less time- and resource-intensive options, such as fewer sessions of CBSST (e.g., targeted modules).

In summary, the findings suggest that individuals with schizophrenia with executive dysfunction should not be excluded from CBSST as they may be the most likely to benefit from treatment. CBSST may bolster cognitive flexibility and problem-solving skills via various strategies (e.g., hypothesis testing; fact checking automatic negative thoughts). CBSST strategies may help consumers with schizophrenia compensate for executive functioning deficits and contribute to improved functional outcomes as compared to individuals with intact executive functioning.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jpsychires.2025.10.002.