A quasi-experimental controlled study of a school-based mental health programme to improve the self-esteem of primary school children

The study addresses whether a school-based programme grounded in Harter’s self-esteem enhancement theory can improve self-esteem among primary school children, particularly those with low baseline self-esteem. Self-esteem is closely linked to child and adolescent mental health; low self-esteem is associated with eating disorders, depression, anxiety, and adverse outcomes into adulthood, whereas positive self-esteem relates to well-being and adjustment. Prior self-esteem interventions often lacked strong theoretical bases or rigorous evaluations and focused on secondary school students or specific domains (e.g., appearance). Given evidence that self-esteem trajectories in Japan decline from primary to junior high school and that Japanese self-esteem levels are comparatively low, the authors developed and quantitatively evaluated the Treasure File Programme (TFP) as a school-based approach to bolster self-esteem in primary school children.

Meta-analytic and programmatic reviews indicate that theory-driven interventions are more effective, yet few self-esteem programmes have been explicitly theory-based. Past programmes like Enhancing Self-Esteem (cognitive theory) and Comparison of self-esteem enhancement programmes (social learning theory) offered systematic procedures but were limited by lack of experimental evaluation and reliance on highly skilled therapists. Harter’s model posits two key determinants of self-esteem: perceived competence in personally important domains and social support from parents, peers, and teachers, plus emphasis on inner qualities (kindness, morality). Programmes incorporating Harter’s ideas have primarily targeted eating disorder prevention in older students (e.g., Everybody is Different; Every Body is a Somebody; Straight Talking About Self-esteem and Resilience; Beautiful from the Inside Out). These showed mixed efficacy and limitations (e.g., age restrictions, female-only samples, lack of control groups), and typically focused on appearance, neglecting other domains personally important for younger children. The transition from primary to secondary school also alters curriculum and pedagogy, underscoring the need for a novel, primary school-targeted, theory-based intervention.

Design: Quasi-experimental controlled study with intervention and control groups (no randomization due to headteacher consent limitations). Assessments occurred at baseline (April 2016) and after intervention (March 2017), approximately 11 months apart. Participants: 2297 students screened from nine public primary schools across six cities in central Japan; five schools (n=1151) chose intervention and four schools (n=1146) chose control. Consented and analyzable samples after exclusions: 794 intervention (387 boys, 407 girls) and 592 control (299 boys, 293 girls), grades 2–6 (ages ~7–11). Procedure/Intervention: The Treasure File Programme (TFP) comprised seven sessions integrated into the school year, delivered by homeroom teachers trained via a standardized manual and two-part training (2 h each). Sessions spanned May–July (1–3), October–December (4–6), and March (7). Key components: goal setting in personally important domains; identifying strengths and virtues; creating a personalized “Treasure File” folder; exchanging positive messages among classmates; receiving positive messages from teachers and family; positive self-evaluation; and end-of-year goal review. Children maintained records of own performance and positive feedback within their folder to facilitate self-referential comparisons over time. Measurement: Primary outcome was self-esteem via the Japanese primary school version of KINDL-R (validated; reliability confirmed). KINDL-R includes six domains—physical well-being, emotional well-being, self-esteem, family, friends, and school—each with four items (24 items total), five-point Likert responses. Domain and total scores converted to 0–100 scale (higher is better). Reliability in this sample: overall α=0.772 (subscales 0.707–0.746). Statistical analysis: Participants were stratified by baseline self-esteem into low (< mean − 1 SD), middle (within 1 SD of mean), and high (> mean + 1 SD) groups to test primary and secondary effects (low: n=137 intervention, n=94 control; middle: n=516 intervention, n=394 control; high: n=141 intervention, n=104 control). Repeated-measures ANCOVA tested group × time interactions within each baseline level, controlling for grade, gender, school, and baseline scores (p<0.05). Effect sizes reported as partial η² (0.010 small, 0.059 medium, 0.138 large). Multilevel models accounted for the nesting (students within classes; class random effect, school fixed effect) after examining ICC and design effect (Deff) thresholds (ICC<0.1, Deff<2.0). Additional t-tests compared mean changes and 95% CIs between groups. Analyses used SPSS v27 and R v4.1.0.

- Sample and reliability: 794 intervention and 592 control participants completed both assessments; overall Cronbach’s α=0.772 (subscales 0.707–0.746). - Baseline comparability: No significant baseline differences between groups for physical well-being (p=0.678), emotional well-being (p=0.321), self-esteem (p=0.243), or school (p=0.704); family (p=0.011) and friends (p<0.001) differed but school effects were controlled as covariates. - Primary outcome (self-esteem): Intervention group increased from 52.99 to 58.16 (+5.17; 95% CI 3.35, 6.99); control changed 54.58 to 54.27 (−0.32; 95% CI −2.28, 1.66). Difference in change = 5.49 (SE 1.38; 95% CI 2.78–8.20), p<0.001. - By baseline level (ANCOVA, group × time): • Low-level self-esteem: Significant improvement with TFP vs control (F=7.618, p=0.006, partial η²=0.033). • Middle-level: Significant improvement (F=8.919, p=0.003, partial η²=0.010). • High-level: No significant effect (F=0.634, p=0.427, partial η²=0.003); both groups showed declines over time. - Multilevel models: Self-esteem domain showed a significant effect favoring TFP (F=8.148, p=0.006, partial η²=0.005). No significant effects in other domains: physical well-being (p=0.598), emotional well-being (p=0.313), family (p=0.337), friends (p=0.689), school (p=0.224), total score (p=0.145). ICCs were <0.10 but Deff >2 in several domains, justifying multilevel modeling. - Secondary domains (t-tests of change): No significant between-group differences in physical well-being, emotional well-being, family, school, or total scores. A friends domain difference in change (2.80; 95% CI 0.70–4.89; p=0.009) was observed by t-test, but not supported in the multilevel analysis; authors conclude no robust intervention effect in non-self-esteem domains.

The findings support the hypothesis that a Harter theory-based school programme can increase self-esteem among primary school children with low and middle baseline self-esteem. Likely mechanisms include: (1) enhanced social support via structured positive feedback from peers, teachers, and family, which is known to bolster self-esteem; (2) a skill-building approach that does not constrain focus to a single domain (e.g., appearance), allowing alignment with each child’s personally important areas (academic, arts, social) and enabling self-referential growth; and (3) sustained access to personalized records (the Treasure File) that document progress and positive messages, reinforcing self-appraisals over time. High baseline self-esteem children showed declines in both groups, possibly reflecting developmental or stability effects; the authors note conceptual distinctions between high stable self-esteem and high unstable self-esteem (narcissism-related), and speculate that parental warmth and peer approval components may enhance self-esteem without fostering narcissism, though narcissism was not measured here. Improving self-esteem in primary school may help mitigate later mental, social, and physical problems linked to low self-esteem. However, given the small effect sizes and quasi-experimental design, results should be interpreted cautiously.

The Treasure File Programme, a school-based, theory-driven intervention, significantly improved self-esteem among primary school children with low and middle baseline self-esteem, with no robust effects observed in other quality-of-life domains. The programme may serve as an educational option to promote child mental health through self-esteem enhancement. Future research should employ randomized controlled trials, include mid-term assessments, measure potential mediators and related constructs (e.g., narcissism), better control for personal and contextual covariates, and standardize and evaluate teacher implementation to validate and refine the programme.

Key limitations include: (1) non-randomized allocation at the school level based on headteacher choice, potentially introducing selection and expectancy biases; (2) lack of blinding—students and staff knew group assignments and programme intent; (3) inability to collect or control for personal/family covariates (e.g., parental education, income, child anxiety/depression); (4) only pre–post measurement over a year without mid-term assessments; (5) listwise exclusion of cases with missing KINDL-R items (13.3% intervention, 15.2% control), with residual risk of bias despite sensitivity checks; (6) potential variability in teacher competence despite standardized training and limited fidelity checks; and (7) limited, non-systematic teacher feedback. These factors may affect internal validity, generalizability, and precision of effect estimates.