“…Adjustment was needed for many potential confounders associated with mood episodes; therefore, propensity scoring was employed to determine the hazard ratio (HR) for mood episodes in the multivariable model. The propensity score in relation to average LAN (≥3 lux vs. <3 lux) was calculated from the logistic regression model, which included the variables associated with mood episodes, such as age (per year), gender (female or male), employment status (yes or no), type of bipolar disorder (I or II), anxiety disorder (yes or no), substance use disorder (yes or no), depressive symptoms (MADRS score of ≥8 points), manic symptoms (high or low; categories were based on the median value), multiple mood episodes during the year before baseline assessment (yes or no), body mass index (per kg/m 2 ), bedtime (per hour), rising time (per hour), total sleep time (per min), sleep efficiency (per %), daytime physical activity (per counts/min), daytime illuminance (per lux), and day length (per min) 1,6,7,9,16‐21 …”