Identifying the bridge between depression and mania: A machine learning and network approach to bipolar disorder

Although the cyclic nature of bipolarity is almost by definition a network system, no research to date has attempted to scrutinize the relationship of the two bipolar poles using network psychometrics. We used state-of-the-art network and machine learning methodologies to identify symptoms, as well as relations thereof, that bridge depression and mania.

Observational study that used mental health data (12 symptoms for depression and 12 for mania) from a large, representative Canadian sample (the Canadian Community Health Survey of 2002). Complete data (N = 36,557; 54.6% female) were analysed using network psychometrics, in conjunction with a random forest algorithm, to examine the bidirectional interplay of depressive and manic symptoms.

Centrality analyses pointed to symptoms relating to emotionality and hyperactivity as being the most central aspects of depression and mania, respectively. The two syndromes were spatially segregated in the bipolar model and four symptoms appeared crucial in bridging them: sleep disturbances (insomnia and hypersomnia), anhedonia, suicidal ideation, and impulsivity. Our machine learning algorithm validated the clinical utility of central and bridge symptoms (in the prediction of lifetime episodes of mania and depression), and suggested that centrality, but not bridge, metrics map almost perfectly onto a data-driven measure of diagnostic utility.

Our results replicate key findings from past network studies on bipolar disorder, but also extend them by highlighting symptoms that bridge the two bipolar poles, while also demonstrating their clinical utility. If replicated, these endophenotypes could prove fruitful targets for prevention/intervention strategies for bipolar disorders.