Revisiting symptom dimensions in schizophrenia with exploratory graph analysis

The Positive and Negative Syndrome Scale (PANSS) groups symptoms into three categories, but factor analyses suggest five dimensions with superior psychometric properties. Exploratory graph analysis (EGA) offers a novel network-based approach that provides additional information such as connection of domains, bridge symptoms, and centrality measures. We applied EGA to PANSS data from 1213 participants with schizophrenia at baseline and after six weeks of treatment. We identified symptom clusters with community detection algorithms, item stability with bootstrapping, and the connecting role of individual items with bridge centrality metrics. EGA consistently revealed five communities: positive symptoms, negative symptoms, cognitive disorganization, depression/anxiety, and hostility/excitement. The latter was less stable and its symptom P6 (suspiciousness) moved to the positive symptom domain at endpoint. Other item assignments were stable, except for items P5 (grandiosity), and G8 (uncooperativeness). P4 (excitement), P6, G8, and G15 (preoccupation) emerged as key bridge symptoms. Overall, the replication of five factors with a different method underlines their clinical importance.