M34. NETWORK STRUCTURE OF PROBAND-COLLATERAL DISCREPANCIES IN PSYCHOSIS SPECTRUM

Abstract Background Obtaining information from multiple informants (family members, parents and/or teachers) is considered ideal practice in clinical psychiatry and is an essential component of child and adolescent psychiatric care. Evidence consistently show reports from different informants differ which affects validity and reliability of the information obtained. Such discrepancies may be due the nature of psychopathology, sociodemographic factors, cultural factors and is dependent on the informant. Examining differences in informant reports can provide clinically pertinent information with the greatest discrepancy potentially providing critical information. This study aimed to quantify proband-collateral discrepancies in psychopathology symptoms and assess network structure of such discrepancies across psychopathology domains. Methods The sample (n = 5094) was extracted from the Philadelphia Neurodevelopmental Cohort, a community-based youth sample focusing on participants between 11–17 years of age with both proband and collateral ratings. The following analyses was conducted for this study. First, proband-collateral agreements were examined across psychopathology domains. Intra-class correlations, kappa’s, sensitivity, specificity, positive predictive values, negative predictive values were used to examine proband-collateral agreement. Second, for subjects who met criteria for psychosis spectrum, network structure was estimated to obtain multivariate structural associations (edge weights and strength) among disagreement scores calculated across psychopathology domains. Relative importance of a node in the network was evaluated by centrality indices. Robustness of findings were assessed by accuracy of edge weights and centrality indices. Results Correlations between overall psychosis symptom scores for proband and collateral report was low at 0.26, with probands (mean10.2 + 12.65) endorsing more and severe symptoms than collateral informants (mean 2.82 + 6.31). Intraclass correlation coefficient for psychosis spectrum domain was poor (0.23). Psychosis spectrum items had the lowest kappa’s, ranging from 0.03 to 0.25. Of the total sample, 273 subjects met criteria for psychosis spectrum. Across 14 nodes (psychopathology domains of ADD, OCD, conduct disorder, agoraphobia, social anxiety, panic disorder, depression, mania, PTSD, separation anxiety, ODD, psychosis spectrum, generalized anxiety disorder and general phobia) and 76 potential edges, a very sparse network structure was observed for psychosis spectrum. This network only had two connected positive edges- disagreement scores in depression and mania with the strongest positive edge indicating the greater the disagreements in depression scores, greater was the disagreements in mania reports. Network structure for subjects who reported symptoms of delusion (n=273) was also sparse but with 3 connected edges- Disagreements for ADD and ODD, general phobia and agoraphobia, and panic disorder and OCD. Network structure for subjects who reported hallucinations (n=586) showed a highly interconnected network with positive edges, the strongest edge was between disagreements in ADD and ODD. This seemed to mimic network structure for other psychopathology conditions. Discussion Psychosis spectrum appear to have distinct network patterns of disagreements between probands and collaterals. This likely point to differences in psychopathology for psychosis spectrum in comparison to other diagnostic conditions. Such discrepancies could potentially be leveraged for early identification.