This study provides first evidence for the feasibility of EMA assessing temporal dynamics of SIB and loneliness in a high-risk PDD sample. Affect and loneliness ratings varied significantly within a few hours, i.e., on average three hours, while passive as well as active SI and their intensity seem to vary within longer periods, i.e. on average six hours. Variance in SB was low. Our pilot data reveal strong contemporaneous associations between affect, loneliness, and SIB, while temporal relations were mostly driven by autoregressive effects.

Few studies have used EMA for assessing SIB in clinical populations [10,11,12, 15] and to our knowledge, none of these studies have assessed SB. Additionally, EMA has not yet been used in PDD. Due to technical constraints of our EMA application (i.e., exclusive compatibility with Android operating systems), a considerable number of patients could not participate. In addition, we observed a relevant proportion of non-completers because of technical issues. Higher levels of depressive symptoms and loneliness might have been another factor for non-completion. There is little published evidence regarding participation and adherence to EMA studies in the field [8]. Our findings seem in line with other research reporting technical issues [16] and lower adherence in more severely depressed individuals [17]. Completers answered 81.3% of automated prompts, comparable to overall compliance in the field [8]. In contrast to other studies in inpatient settings [12, 15, 18], participants were not financially incentivized in the present study, speaking to the scalability of the presented approach.

While the present findings clearly support the need for assessing SIB and loneliness in short periods, i.e., hours rather than days, our findings regarding variability over time stand partly in contrast with findings from other studies reporting significant variability of SI over periods of one to few hours in patients with MDD [12, 15, 18]. This contrast may be explained by differences in size and characteristics of the respective clinical samples, e.g. the chronicity of depressive symptoms in PDD as well as a mostly stable clinical environment in the present sample. Inter-individual differences of the three individual cases showcased in Fig. 1 may demand an in-depth investigation of intra-individual effects in future studies.

In line with two previous studies, we observed contemporaneous associations of EMA items related to loneliness and SI [12, 15]. Having used sampling frequencies of 4 to 10 daily assessments the data suggest that loneliness and SIB co-occur simultaneously or peak within minutes apart from each other.

Limitations of the present study include the small sample size which limits generalizability and demands cautious interpretation of significant results; a short duration of EMA which may not have captured the full symptom range; and a fixed time resolution not capturing information on symptoms occurring between 8 p.m. and 8 a.m.

In conclusion, this study provides first evidence for the feasibility of EMA in PDD. Patterns of SIB and loneliness varied significantly within three to six hours, which may inform sampling schemes of future studies. Future studies may replicate and extend the present findings in larger samples. Those studies may further optimize both the single items of EMA as well as the temporal resolution of prompts based on findings from our pilot data. Further steps will then include developing just-in-time (digital) interventions, i.e. via smartphone, that can complement current interventions to improve early recognition and prevention of SIB using EMA data.