Problem: In recent years, mental health researchers and therapists have focused on finding ways to understanding the overlap of psychological disorders, such as anxiety and depression. Within the evidence-based assessment literature, this concern is conceptualized as co-morbidity. Addressing this concern, given its negative impact on the evidence-based assessment outcomes, remains essential for undertaking effective treatment or prevention strategies. For example, the overlap of symptoms essential for assigning a specific diagnosis causes a typical error that tends to occur “a false positive error.” An individual is assigned a specific diagnosis when the individual does not present the specific psychological disorder. Study Aim: The present study was undertaken to address the limitations of conventional statistical modeling strategies in the extant literature in addressing the co-morbidity concern. In particular, rather than conducting the conventional Pearson correlation or exploratory factor analyses, we used a data science Network Analytic modeling procedure. Empirical Network Analysis allows for (a) identifying the structure of complicated symptoms and (b) assessing how the essential symptoms of independent conditions or disorders overlap or interact. As a graphical model, a primary empirical network structure includes a random set of variables or symptoms (nodes) and the dependencies between the symptoms (edges).
Methods: Participants: Using scores on a checklist of trauma-related events, we formed two groups. Those participants who indicated not having experienced any traumatic event were assigned to the No-Trauma group (NT; N = 255; 125 men and 130 women). Those who self-reported one or more trauma-related experiences were assigned to the Trauma group (TR; N = 299; 81 men and 218 women). The groups didn’t differ significantly in age, t (552) = 1.23, p = 0.22.
Measures and Procedure: We received data sets from ongoing projects approved by an Institutional Research Board at the University of Texas at San Antonio. The self-report measures were selected to assess (a) at-risk conditions (negative affect, depression, anxiety, and interpersonal sensitivity) and (b) protective conditions (extrinsic motivation, social reassurance, positive future, and desire for belonging) for each group.
Results and Conclusions. First, we assessed and found estimates of internal consistency reliability for each self-report measure scores across the groups (coefficient-omega values > .70). Second, we used the bootstrap (non-parametric) method with the extended EBICglasso estimator to obtain the correct estimates for each network. We then, examined the network plots (thickness of the edges), the centrality plots, and the clustering plots to enhance the analysis’s interpretations. We list a few of several of the significant findings. First, the network structure of the risk and protective conditions was similar between groups. Second, for each group, there was no overlap among the risk or protective measures. Third, there was no overlap of the measures within each group. Fourth, for the No-Trauma group, the most vital links were among the protective factors of social reassurance, positive focus, and desire to belong. However, for the Trauma group, the most robust protective association was between social reassurance and positive focus. Results highlight the impact of examining a cluster of symptoms simultaneously.