Retest improvements in self-reported anxiety may stem from inflated initial scores (initial elevation bias) rather than genuine symptom change. In this systematic review, we examined changes in anxiety scores reported by children and parents across repeated assessments in nonintervention control groups from randomized controlled trials across community, risk, and clinical samples, using four widely used anxiety measures for children and adolescents (Multidimensional Anxiety Scale for Children, Spence Children's Anxiety Scale, Revised Child Anxiety and Depression Scale, and Screen for Child Anxiety-Related Emotional Disorders). We searched BASE, MEDLINE, APA PsycInfo, Web of Science, and PubMed Central, alongside a cited reference search for seminal publications of the included measures. A random-effects model was used for meta-analysis. A total of 106 studies (N = 9,224 children) met inclusion criteria: 46 clinical samples, 37 risk samples, and 23 community samples. For child-rated anxiety, a weighted average effect of -0.25 (95% CI -0.30, -0.21) was found between the first and second assessments and -0.09 (95% CI -0.18, -0.01) between the second and third assessments. Metaregression showed that clinical and risk samples had larger decreases in anxiety scores, while the Screen for Child Anxiety-Related Emotional Disorders measure exhibited smaller decreases. For parent-rated child anxiety, the effect was -0.12 (95% CI -0.17, -0.06) between the first and second assessment. The findings support the presence of initial elevation bias mainly in child-reported anxiety. The results are unlikely to be explained by maturation, real improvements, or selective attrition, and they extend beyond the effects of regression to the mean as they appear even in community samples. Initial elevation bias complicates the interpretation of child anxiety ratings and has important implications for clinical practice and research. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
Meregalli et al. (Thu,) studied this question.