Twenty-three RCTs (n = 1,599) were included. Eleven trials reported blinded outcome assessment.
The weighted mean effect size was large and significantly different from zero (indicating a significant positive effect of SST) for measures of understanding the training material (d = 1.2, 95% confidence interval (CI): 0.96, 1.43) and was moderate and significant for performance-based measures (d = 0.52, 95% CI: 0.34, 0.71) and measures of community and institutional functioning (d = 0.52, 95% CI: 0.31, 0.73). Heterogeneity was significant for measures of social and daily living skills.
For negative symptoms of schizophrenia, the effect size was moderate and significantly different from zero (d = 0.4, 95% CI: 0.19, 0.61). Heterogeneity was significant.
For overall symptoms of schizophrenia, effect size was not significantly different from zero (d = 0.15, 95% CI: -0.01, 0.31).
Results for effects of moderator variables on outcomes were reported in the paper. The fail-safe N calculation indicated that publication bias was unlikely.