Attrition analysis in the 2023 follow-up of the “Deliberate Self-Harm, Emotion Regulation, and Interpersonal Relations in Youth” Project

In this report, we analyse attrition at the two most recent data collection points of the longitudinal Swedish project “Deliberate self-harm, emotion regulation, and interpersonal relations in youth”, that collected survey data from the same cohort in 2007 (T1), 2008 (T2), 2017 (T3), and 2023 (T4). We compared T4 responders and non-responders using variables assessed at T1−T3, and T3 responders and non-responders using variables assessed at T4.

The results indicate that responding at T4 was significantly associated with responding at T3 (Cramer’s V = 0.40). Additionally, T4 responders were more likely to be women (V = 0.19), to have a non-foreign background (V = 0.11), have a higher socioeconomic status (V = 0.11−0.15), and score lower on measures of aggression and impulsivity (Cohen’s d = 0.15−0.28) at previous assessments. They were less likely to have encountered the death of someone important or physical assault between 2007−2017 (V = 0.10−0.13), but more likely to have experienced sexual assault within the same time frame (V = 0.10). Comparisons between T3 responders and non-responders on T4 variables revealed minimal differences, except that T3 responders were less likely to have been on extended sick leave (V = 0.12).

Overall, while most differences showed small effect sizes between responders and non-responders, these findings suggest that attrition is associated with specific demographic, behavioral, and experiential factors. These factors need to be considered when interpreting longitudinal data and understanding the limitations they may impose on generalizability.