Item drift in the Youth Self-Report: IRT gradient analysis with suicidal ideation and self-harm items as illustrative examples

Recent increases in adolescent psychological problems have often been interpreted as evidence of a generalized deterioration in youth mental health. However, long-term trends in self-reported psychological distress have yielded a mixed picture and have not consistently paralleled increases in endorsement of specific symptoms, including suicide- and self-harm-related indicators. This discrepancy raises a measurement question: can item endorsement change even when overall latent distress remains relatively stable? We applied a time-anchored item response theory framework to Youth Self-Report (YSR) data from four repeated cross-sectional cohorts of Polish adolescents assessed in 2000, 2011, 2013, and 2018 (N = 3,920). Multiple-group graded response models were used to estimate cohort differences in latent distress and temporal changes in item functioning. Probability-gradient analyses were used to trace diffuse, severity-dependent changes in item functioning across the latent trait continuum. Latent levels of psychological distress increased modestly between 2000 and 2011 and then remained relatively stable through 2018. In contrast, item-level analyses revealed selective temporal drift in the relation between item endorsement and latent distress. Distinct patterns were observed, including severity-dependent drift, gradient reversal, and stable item functioning. Suicide- and self-harm-related YSR items showed higher endorsement probabilities at equivalent levels of underlying distress in later cohorts, especially at moderate-to-high levels of latent internalizing severity. Apparent stability in broad adolescent self-reported distress can conceal changes in how specific YSR items relate to underlying distress. Monitoring youth mental health may therefore benefit from item-level approaches that account for temporal changes in item functioning, while avoiding direct interpretation of single YSR items as full clinical measures of suicidality or self-harm.