We analyzed the electronic health records (EHR) at Clinica San Juan de Dios in Manizales, Colombia to identify predictors for the transition from Major Depressive Disorder (MDD) to Bipolar Disorder (BD).
Through a multivariate Cox regression model, we pinpointed several predictors of BD conversion, including the severity of the initial MDD episode, psychosis, hospitalization at first episode, family history of mood or psychotic disorders, and female gender. We also found associations with certain medications and clinical features like delusions, suicide attempts, and substance use, identified using natural language processing (NLP) of clinical notes. These factors together predicted the conversion to BD within five years of the initial MDD diagnosis with significant accuracy.
This study not only validates known risk factors but also uncovers new ones, such as suicidal ideation and attempts, highlighting the potential of EHR data in improving early diagnosis and understanding of BD.