AbstractBackgroundManic and depressive mood states in bipolar disorder (BD) may emerge from the non‐linear relations between constantly changing mood symptoms exhibited as a complex dynamic system. Dynamic Time Warp (DTW) is an algorithm that may capture symptom interactions from panel data with sparse observations over time.MethodsThe Young Mania Rating Scale and Quick Inventory of Depressive Symptomatology were repeatedly assessed in 141 individuals with BD, with on average 5.5 assessments per subject every 3–6 months. Dynamic Time Warp calculated the distance between each of the 27 × 27 pairs of standardized symptom scores. The changing profile of standardized symptom scores of BD participants was analyzed in individual subjects, yielding symptom dimensions in aggregated group‐level analyses. Using an asymmetric time‐window, symptom changes that preceded other symptom changes (i.e., Granger causality) yielded a directed network.ResultsThe mean age of the BD participants was 40.1 (SD 13.5) years old, and 60% were female participants. Idiographic symptom networks were highly variable between subjects. Yet, nomothetic analyses showed five symptom dimensions: core (hypo)mania (6 items), dysphoric mania (5 items), lethargy (7 items), somatic/suicidality (6 items), and sleep (3 items). Symptoms of the “Lethargy” dimension showed the highest out‐strength, and its changes preceded those of “somatic/suicidality,” while changes in “core (hypo)mania” preceded those of “dysphoric mania.”ConclusionDynamic Time Warp may help to capture meaningful BD symptom interactions from panel data with sparse observations. It may increase insight into the temporal dynamics of symptoms, as those with high out‐strength (rather than high in‐strength) could be promising targets for intervention.