A person centered analysis of student well being and academic achievement using PISA 2022 data from Kazakhstan

This study aimed to identify heterogeneous profiles of adolescent well-being in Kazakhstan and examine their associations with school belonging, attendance, academic achievement, and structural factors such as socio-economic status (SES) and school type. Grounded in multidimensional models of student well-being and school climate, the research adopted a person-centred analytical perspective to capture heterogeneity among students. A quantitative, cross-sectional design was employed using nationally representative PISA 2022 data from Kazakhstan (N = 19,359). Latent Profile Analysis (LPA) was conducted on standardized scores for school belonging, attendance, and life satisfaction using the mclust and tidyLPA packages in R. Model selection was guided by multiple fit indices (BIC, BLRT, entropy, APP, and interpretability). Weighted multinomial and binary logistic regression models (svyglm) estimated predictors of profile membership, while weighted ANCOVA with plausible values and pooled results via Rubin’s Rules (MIcombine) assessed achievement differences across profiles. Survey weights were applied to ensure representativeness. A six-profile model best represented student well-being (BIC = 148,310; entropy = 0.87), comprising Normative (65%), Low Attendance (16%), High-Risk (9%), Super Social (5%), Disengaged (3%), and Outsiders (1%) groups. Higher SES reduced the likelihood of belonging to the High-Risk profile and increased odds of Super Social membership. Students in specialized schools were more likely to be High-Risk but less likely to be Disengaged or Low Attendance. Academic achievement differed significantly across profiles, with High-Risk and Super Social students outperforming Disengaged peers despite lower emotional well-being. Limitations include the cross-sectional design, single-item indicators, and lack of replicate weights. Overall, the study demonstrates that student well-being in Kazakhstan is heterogeneous and structurally patterned, underscoring the need for differentiated, data-driven interventions and the integration of well-being metrics alongside academic goals.