Reducing participant burden of ESM-based monitoring of mood disorder onset through variable sample size EWMA

The exponentially weighted moving average (EWMA) procedure is a promising statistical process control technique for monitoring experience sampling method (ESM) data to detect imminent depressive episodes in real time, allowing for timely interventions. However, asking participants to complete multiple ESM questionnaires per day - relevant studies used a daily sample size of five or ten questionnaires - over extended periods is burdensome. We propose a variable sample size (VSS) EWMA approach in which the daily sample size depends on the extremity of participants' ESM scores on the previous days. VSS EWMA allows us to keep the daily sample size and thus participant burden low when the individual is doing well. We evaluated VSS EWMA performance through a simulation study and an application to empirical ESM data. Our findings indicate that VSS EWMA improves the balance between detection speed and participant burden. Based on these results, we provide recommendations on how to tune VSS EWMA to different priorities and outline directions for future research.