Monitoring emotional intensity and variability to forecast depression recurrence in real time in remitted adults.

Abstract

Objective: Recurrent depressive episodes are preceded by changing mean levels of repeatedly assessed emotions (e.g., feeling restless), which can be detected in real time using statistical process control (SPC). This study investigated whether monitoring changes in the standard deviation (SD) of emotions and negative thinking improves the early detection of recurrent depression. Method: Formerly depressed adults (N = 41) monitored their emotions five times a day for 4 consecutive months. During the study, 22 individuals experienced recurrent depression. We used SPC to detect warning signs (i.e., changing means and SDs) of four emotions (positive and negative affect with high or low arousal) and negative thinking. Results: SD-based warning signs only preceded 23%-36% of recurrences, but almost never reflected a false alarm (0%-16%). Correspondingly, SD-based warnings had a high specificity (at the cost of sensitivity), while mean-based warnings had a higher sensitivity (but lower specificity). There was little overlap in mean-and SD-based warning signs. For the majority of emotions, monitoring for high SDs alongside monitoring changes in mean levels improved the detection of depression ( p < .015) compared to when only monitoring for changing mean levels. Conclusions: Warning signs for depression manifest not only in changing mean levels of emotions and cognitions but also in increasing SDs. These warnings could eventually be used to detect not just who is at increased risk for depression but also when risk is rising. Further research is needed to evaluate the clinical utility of depression SPC.What is the public health significance of this article? Changes in the mean levels of self-reported emotions precede recurrent depressive episodes and are therefore considered warning signals. This study shows that the early detection of recurrent depression improves when warning signals are not only tracked in the mean but also in the variability of emotions. Thus, monitoring for both changes in emotional intensity and variability may aid the early detection, and possibly the prevention, of depression.

Forecasting the onset of depression with limited baseline data only: A comparison of a person-specific and a multilevel modeling based exponentially weighted moving average approach.

“…There was a distinct Phase I period of 28 days during which the participants were in remission, after which (gradual) discontinuation of their antidepressants started. The remaining days (i.e., Days 29–122) were used for Phase II monitoring using EWMA, and this confirmed that the EWMA procedure may be useful for forecasting the onset of depression by monitoring mental states (Schreuder et al, 2024; Snippe et al, 2023).…”

“…It is also possible to monitor changes in other statistics than the mean by applying EWMA to multiple relevant day statistics, such as day standard deviations (Schat, Tuerlinckx, De Ketelaere, & Ceulemans, 2023). Research has shown that monitoring day standard deviations on top of day averages is beneficial for predicting depression (Schreuder et al, 2024). Future research can expand on the present study by monitoring additional types of day statistics on top of day averages.…”

“…Permutation Testing. To evaluate whether the observed MCC values were significantly larger than 0, permutation testing was used in Schreuder et al (2024). Specifically, the predicted transition values were randomly shuffled and compared to the observed transition, leading to a new MCC value.…”

“…Specifically, we used the person-specific estimates, multilevel estimates (leave-one-out case and historical case), and the known parameters of the 145 individuals in the target data (i.e., Data Set 6), as obtained in the previous section. Prior to explaining our simulation study, we describe the statistical procedure used in two recent articles applying EWMA to forecast recurrence of depression (Schreuder et al, 2024; Snippe et al, 2023), which we used as a basis for our simulation design.…”

“…As mentioned in the introduction, Snippe et al (2023) and Schreuder et al (2024) investigated whether the recurrence of depression can be forecasted by monitoring mental states using the EWMA procedure. Specifically, these studies analyzed data of 41 formerly depressed patients who participated in a 4-month ESM study while discontinuing their antidepressant medication.…”