<p>Increasing evidence suggests that bipolar disorders are associated with mood instability even outside the context of mood episodes. Here we use data from the Prechter Longitudinal Study of Bipolar Disorder to identify subgroups of individuals with bipolar disorders based on mood instability, identify biopsychosocial predictors of mood instability and determine whether mood instability predicts future outcomes. In a total of 481 participants, mood was assessed every 2 months (Patient Health Questionnaire and Altman Self-Rating Mania Scale) over 5 years, and clinical and functioning outcomes were assessed in year 6. Low, moderate and high mood instability classes were identified. Neuroticism, sleep quality, childhood emotional neglect and physical abuse, stimulant abuse, hypomania age of onset and number of depressive episodes were the most influential predictors of mood instability. Being in the high instability class (based on mood from years 1 to 5) predicted greater suicidal ideation and functional impairment in year 6. In summary, we show that mood instability represents a core phenotype of bipolar disorder with distinct predictors and long-term implications. Routine assessment may improve personalization in bipolar disorder treatment and research.</p>