Researchers developed a machine-learning model that uses only sleep-wake patterns to predict mood episodes in patients with mood disorders, achieving high accuracy. This approach simplifies real-world mental health monitoring using data from wearables or smartphones.