School-level administrative data is associated with childhood and young adult mental health: Evidence from a UK longitudinal cohort

School-level administrative data is associated with childhood and young adult mental health: Evidence from a UK longitudinal cohort | Research Square Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Help Center Sign In Submit a Preprint Cite Share Download PDF Article School-level administrative data is associated with childhood and young adult mental health: Evidence from a UK longitudinal cohort Rebecca Ferdinand, Anna Suarez, Agnieszka Musial, Kathleen Rastle, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8681038/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Reported rates of mental health problems among children and young people in the United Kingdom have been increasing. While studies have shown that measures of the perceived school environment are linked to pupil mental health, research using school-level administrative measures remains limited. In this study, we linked longitudinal data from 6,398 participants in the Twins Early Development Study (TEDS) to national pupil and school databases and investigated the associations between administrative measures of the school environment and developmental psychopathology from ages 12 to 21. We found that administrative measures of the school environment were cumulatively associated with psychopathology burden, explaining 0.3%-3.9% of the variance. The associations were attenuated, but effects remained significant after controlling for genetic predisposition to mental health problems using psychiatric polygenic scores, mental health at age 7, and family socioeconomic status. Additionally, we found evidence of small bidirectional links (β= 0.02-0.07) using cross-lagged panel analyses. Our findings highlight the importance of considering school-level administrative measures for mental health outcomes and demonstrate the value of linking administrative data with mental health outcomes to guide school-level intervention. Social science/Psychology/Human behaviour Social science/Education Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Over the past decade, the United Kingdom has seen an increase in the reported rate of psychopathology in children and young adults 1–3 . This trend has been observed across the range of psychiatric severity; one in five young people aged 8-25 was found to have a probable mental health disorder in 2023, up from 1 in 9 in 2017 4 , and the number of children admitted to acute medical wards for mental health-related concerns increased by 65% between 2012 and 2022 5 . Though the causes of this increase are not fully understood 6–9 , the resulting demand for mental health support in the UK now exceeds available resources; as of March 2024, 28% (270,300) of children referred to child and adolescent mental health services from April 2022 to April 2023 were still awaiting access to support 10 . The annual Good Childhood Report has examined trends in the mental health of UK children since 2005 and has highlighted the need for increased attention on how school experiences impact children's mental health. In 2024, a larger proportion of 10 to 17-year-olds reported being unhappy with school (14.3%) than with any other aspect of life, including family (6.8%), friends (7.3%), and their physical appearance (11.9%) 11 . Responses over the past decade also demonstrate that the percentage of young people reporting unhappiness with school has increased over time 12 . Understanding and addressing the school-related factors that contribute to poor mental health in children could, therefore, be crucial in alleviating the current burden on child and adolescent mental health services. However, while there is substantial evidence on how family and peer contexts influence children's mental health, research on the impact of the school environment remains limited. Large-scale studies in the UK find that differences between schools account for 1.4-4.5% of the variance in child psychopathology 13,14 , with differences in the perceived school environment explaining a large proportion (29.5-48.8%) of this variance 13 . These ratings include pupil and teacher reports of school engagement, pupil behaviour and safety, and pupil-teacher relationships. Although these UK-based studies support extensive international research showing a clear association between the school environment and pupil mental health 13–15 , their cross-sectional design limits the capacity to determine the direction of the relationship and establish whether the perceived school environment has a causal effect on student psychopathology. A study of 20 English secondary schools aimed to address the limitations of cross-sectional research by tracking both the school environment and student mental health over a three-year period 16 . The study initially found an association between pupils’ ratings of the school environment and their mental health over time. However, after controlling for baseline mental health, these associations were no longer significant. This could indicate that the school environment has no direct impact on child mental health. Alternatively, this attenuation may reflect over-control, as baseline mental health was measured at age 11, and may already have been influenced by school experiences. However, this finding could also reflect the influence of children’s mental health on how they perceive and report their school environment. Young people with poorer mental health may perceive their environment more negatively, which could inflate the observed association between school environment ratings and mental health. The fact that this association disappears after accounting for baseline mental health raises the possibility that students' reported school environment may be, at least in part, a reflection of individual mental health differences rather than the objective characteristics of the school environment. In our previous work, we explored this possibility using longitudinal twin data 17 . Consistent with earlier research 16 , we found that perceived school environment measures, including teacher quality, classroom chaos and school engagement, at ages 7, 9, and 16 were associated with student psychopathology over time, cumulatively explaining 26-56% of the variance. We demonstrated that these effects were attenuated, but remained significant, when we controlled for baseline mental health, socioeconomic status and genetic predisposition to mental health difficulties. Using twin modelling methods, we also showed that shared genetic factors explained a majority (~71%) of the associations between the perceived school environment and psychopathology across all ages. Our findings suggest that the genetically driven individual differences that influence psychopathology may also play a considerable role in shaping perceived measures of the school environment. Differences in genetic predisposition towards psychopathology may therefore result in pupils experiencing and reporting the same school environment differently. Evidence on the influence of schools on child mental health could therefore be strengthened by examining whether schools influence mental health through objective structural features, beyond individual perceptions and pre-existing risk. A number of recent studies have shown that structural measures of the school environment are associated with psychopathology in children and adolescents, both cross-sectionally 13,14 and over time 18 . Higher levels of pupil deprivation in a school have been associated with individual mental health difficulties, as has urban location of the school, even after controlling for individual-level factors such as family socioeconomic status 13,14,18 . However, the observed effects of such measures were small, so findings have often been interpreted as evidence that structural features of schools are of limited relevance to mental health, particularly relative to pupils’ subjective experiences 13,18 . Notably, these studies focused on selected measures and tended to assess the significance of each one in isolation, rather than considering the cumulative effect these administrative measures might have on mental health over childhood and adolescence. Additionally, previous research has generally not examined the bidirectional nature of these effects. As a result, it remains unclear whether the school environment influences pupil mental health or if mental health difficulties also shape the school environment. While the studies discussed account for individual-level confounders in the associations between administrative school environment measures and student mental health, they do not consider the potential confounding role of genetic factors. Research indicates that more objective environmental measures, such as third-party family environment observations or documented life events, show some heritability, though to a lesser degree than subjectively reported measures 19–21 . This suggests that genetic influences on environmental measures, including school environment measures, may go beyond shaping subjective perceptions. Children are not randomly assigned to schools - factors such as family background and parental involvement, which are themselves influenced by genetics, play an important role in school selection. Consequently, genetic predispositions towards mental health difficulties might not only impact how children experience the same environment but also influence the types of school environments they are exposed to. It is therefore essential to consider the role that genetic predispositions to mental health difficulties may play in shaping the relationship between school-level administrative measures and pupil mental health. In this study, we use UK registry data from the National Pupil Database (NPD) and Ofsted linked to longitudinal data on self-, parent-, and teacher-reported psychopathology from the Twins Early Development Study (TEDS), to investigate how school-level administrative measures influence young people's psychopathology over time. Using all available psychopathology measures, we used confirmatory factor analysis to create a ‘p-factor’ of psychopathology at each age, giving a single measure of each child’s total psychopathology burden at that time. We examined the cumulative variance explained in the p-factor at ages 12, 16 and 21 by a range of administrative school environment measures, including pupil community characteristics, quality ratings and school and individual-level absences. To control for genetic confounding, we used psychiatric polygenic scores (PGS), which combine the effects of DNA variants identified by genome-wide association studies (GWAS) to predict an individual’s propensity to mental health difficulties 22 . We also controlled for family socioeconomic status and baseline mental health difficulties at age 7. Additionally, we used the longitudinal data to assess the direction of associations between objective school environment measures and mental health across adolescence. Finally, we used self-, parent-, and teacher-reported school environment data from TEDS to determine whether the administrative measures explained variance in psychopathology beyond that explained by perceived measures alone. Our research study was preregistered in the Open Science Framework (OSF; https://osf.io/3c5rk/). Results Descriptive statistics and gender differences We calculated descriptive statistics (means, standard deviations, skew and kurtosis) for all mental health measures, and for school-level and perceived school measures (see Supplementary Tables 1-3). Variables with skewness beyond ±1 were transformed to achieve normality. For positively skewed variables, we applied a log 10 transformation. For negatively skewed variables, we applied a reflection (reversing the scale) and then a log 10 transformation. Descriptive statistics were also calculated for males and females separately. We conducted univariate analyses of variance (ANOVA) to explore sex differences across all measures. Because significant sex differences emerged, explaining 0.2-10% of the variance (see Supplementary Tables 4-6), the variables were adjusted for sex and age using linear regression. Sex- and age-adjusted standardised variables were then used in all downstream analyses. We used confirmatory factor analysis (CFA) to construct a cross-rater general factor of psychopathology (p-factor) using the mental health measures at each age group (See Methods section for detailed factor construction procedures).  Fit statistics for each p-factor are presented in Supplementary Table 7. Correlations between objective school environment measures and mental health Figures 1a and 1b show contemporaneous associations between the p-factor and school-level administrative measures at 12, and the p-factor and school-level administrative measures at 16. We used all measures available across our data sources (the National Pupil Database and Ofsted). Across both ages, we observed significant but weak-to-modest positive correlations (r = -0.05 – 0.16) between most school-level measures and p-factors at ages 12 and 16.  At age 12, community characteristic-related measures show the strongest associations with mental health (percentage of pupils with special educational needs (SEN) statements; r = 0.14, percentage of pupils having free school meals (FSM); r = 0.11). At age 16, individual absence measures showed the strongest associations with mental health (percentage of unauthorised absences; r = 0.16, percentage of overall absences; r= 0.14). Higher p-factor scores indicate a greater burden of psychopathology, so these correlations suggest that greater levels of these measures are associated with higher levels of psychopathology. Figures 2a and 2b show the time-lagged associations between school environment measures at 12 and 16 and the p-factor at 21. Fewer measures across 12 and 16 are associated with the p-factor at age 21; however, those that are significantly associated show a similar magnitude of effects to previous contemporaneous and longitudinal associations (r = -0.04 – 0.14). In particular, the individual overall absence percentage measure shows the same association with the p-factor at 16 as with the p-factor at age 21 (r = 0.14). Contemporaneous and longitudinal models to predict psychopathology (p-factor) To assess the extent to which all school-level measures explained variance in child psychopathology, both contemporaneously and longitudinally. We conducted exploratory and confirmatory factor analysis (EFA and CFA) on all school-level administrative measures. This allowed us to reduce data dimensionality, use Full Information Maximum Likelihood (FIML) to retain participants with partial data, and address multicollinearity when using all measures in a multiple linear regression. See the Methods section for full exploratory and confirmatory factor analysis procedures. This resulted in four latent factors of the administrative measures: individual absence, school-level characteristics (including school quality ratings and school-level absenteeism), pupil community characteristics (including the proportion of children receiving free school meals and with special educational needs) and headcount (including number of pupils and proportion of male and female students) and extracted factor scores for each participant. We then ran a series of contemporaneous models using the latent factors as predictors of the p-factor at ages 12 and 16, as shown in Table 1. In the unadjusted models, contemporaneous latent factors explained 2% of the variance in psychopathology at age 12 and 4% at age 16 – small but significant effects at both ages. We then sequentially adjusted for potential confounders to make the strongest possible case for causality. After adjusting for 27 polygenic scores, the latent factors explained 0.8% of the variance in the p-factor at both 12 and 16. We additionally adjusted for family socioeconomic status, after which the latent factors explained 0.3% of the variance in the p-factor at 12, and 0.5% at 16. Finally, we additionally adjusted for baseline mental health at age 7, after which latent factors explained 0.3% of the variance at age 12 and 0.4% at age 16. Table 1: Variance in general factor of psychopathology (p-factor) at age 12 explained by latent factors of school-level administrative measures at ages 12, and general factor of psychopathology (p-factor) at age 16 explained by latent factors of school-level administrative measures at ages 16. Age Model Adjusted R- squared p-value Age 12 Unadjusted 0.022 <0.001 Adjusted for PGS 0.008 <0.001 Adjusted for PGS, SES 0.003 <0.001 Adjusted for PGS, SES, age 7 P-factor 0.003 <0.001 Age 16 Unadjusted 0.039 <0.001 Adjusted for PGS 0.008 <0.001 Adjusted for PGS, SES 0.005 <0.001 Adjusted for PGS, SES, age 7 P-factor 0.004 <0.001 Notes: P – p-factor, E- school environment, PGS – polygenic scores, SES – socioeconomic status, P7 – p-factor at age 7. Table 2 shows the series of longitudinal models we ran using the latent factors of the administrative measures as predictors, and includes school-level factors from earlier waves (i.e. at ages 16 and 21, we incorporate all school-level factors from ages 12 and 16). In the unadjusted models, latent factors at 12 and 16 explained 3.7% of the variance in psychopathology at age 16 and 3.9% at age 21 – small but significant amounts at both ages. We then sequentially adjusted for potential confounders to make the strongest possible case for causality. After adjusting for 27 polygenic scores, the latent factors explained 0.9% of the variance in the p-factor at 16 and 1.5% at age 21. We additionally adjusted for family socioeconomic status, after which the latent factors explained 0.5% of the variance in the p-factor at 16, and 0.9% at 16. Finally, we additionally adjusted for baseline mental health at age 7, after which latent factors explained 0.4% of the variance at age 16 and 0.9% at age 21. Table 2: Variance in general factor of psychopathology (p-factor) at age 16 and age 21 explained by latent factors of school-level administrative measures at ages 12 and 16 Age Model Adjusted R- squared p-value 16 Unadjusted 0.037 <0.001 Adjusted for 27 PGS 0.009 <0.001 Adjusted for PGS, SES 0.005 <0.001 Adjusted for PGS, SES, age 7 P-factor 0.004 0.004 21 Unadjusted 0.039 <0.001 Adjusted for 27 PGS 0.015 <0.001 Adjusted for PGS, SES 0.009 <0.001 Adjusted for PGS, SES, age 7 P-factor 0.009 <0.001 Notes: P – p-factor, E- school environment, PGS – polygenic scores, SES – socioeconomic status, P7 – p-factor at age 7. Robustness checks To assess robustness, we compared these findings with to complete cases multiple linear regression models run without performing factor analysis on the school-level administrative measures (See Supplementary Table 8-11 for results) and elastic net models, which use cross-validation to limit overfitting (See Supplementary Table 12-13 for results). Variance explained (R²) was similar across all three methodological approaches, although the robustness models did not consistently reach significance unlike the latent factor linear regression models. Sample sizes were considerably smaller for the complete-case linear regression and elastic net due to listwise deletion. Differences across methods, therefore, likely reflect noise and instability due to low sample size rather than modelling strategy. Cross-lagged relationships between the objective school environment and mental health We used cross-lagged panel analyses to examine stability and change in the relationship between the school environment and mental health across adolescence. Figure 3 shows the standardised path estimates for the associations between p-factor scores and each of the four objective school environment factor scores at ages 12 and 16 (see the Methods section for full exploratory and confirmatory factor analysis procedures). For each objective school environment factor, we first estimated an unadjusted cross-lagged panel model. To make the strongest possible case for causality, we then sequentially introduced covariates (adding mental health polygenic scores, followed by socioeconomic status, and finally the p-factor at age 7) until the cross-lagged paths no longer reached significance. We found that the stability of pupil community and headcount factors were very high (β = 0.90, 95%CI 0.89, 0.91; β = 0.95, 95%CI 0.94, 0.96) compared to school-level characteristics (β = 0.61, 95%CI 0.60, 0.93) and individual absences (β = 0.26, 95%CI 0.24, 0.29). In the individual absence model, the cross-lagged paths were significant after controlling for the stability and cross-sectional correlations. The effect sizes were small, and the p-factor effect on individual absence (β = 0.07, 95%CI 0.04, 0.10) was comparable to the individual absence effect on the p-factor (β = 0.04, 95%CI 0.01, 0.06). We found that only the individual absences model had a significant residual correlation at age 16 (β = 0.06, 95%CI 0.04, 0.09). We also observed that after adjusting the model for 27 psychiatric polygenic scores only, all correlations between individual absences and mental health were non-significant (see Supplementary Table 14 for full statistics after adjustment). In the pupil community characteristics model, we observed a small bidirectional relationship between the p-factor at 12 and pupil community characteristics at 16 (β = 0.02, 95%CI 0.01, 0.03) and pupil community characteristics at 12 and the p-factor at 16 (β = 0.05, 95%CI 0.02, 0.07) after controlling for stability and their cross-sectional correlations. After adjusting for 27 psychiatric polygenic scores, the cross-lagged pathways remained significant; however, after adjusting for both polygenic scores and socioeconomic status, the cross-lagged pathways were fully attenuated (see Supplementary Tables 15-16). In the school-level characteristics model (Figure 3c), we observed a small bidirectional relationship between the p-factor at 12 and school-level characteristics at 16 (β = 0.03, 95%CI 0.01, 0.06) and school-level characteristics at 12 and the p-factor at 16 (β = 0.03, 95%CI 0.01, 0.06). The cross-lagged pathway from the school-level characteristics at 12 to the p-factor at 16 remained statistically significant after jointly adjusting for all three confounders (see Supplementary Tables 17-19 for adjusted model statistics). In the headcount model (Figure 3d) there were no significant cross-lagged paths, or residual cross-sectional path at age 16. Model fit statistics for cross-lagged models are presented in Supplementary Table 20. Table 3 shows the total variance explained in the p-factor at ages 12, 16, and 21 by latent factors of both perceived school environment measures and school-level administrative measures. We conducted exploratory and confirmatory factor analysis (EFA and CFA) on perceived school environment measures, as we did with the school-level administrative measures. Again, this allowed us to reduce the dimensionality of the data, use Full Information Maximum Likelihood (FIML) to retain participants with partial data, and address issues of multicollinearity when using all measures in a multiple linear regression. See the Methods section for full exploratory and confirmatory factor analyses procedures. We constructed four factors of the subjective school environment at age 9: classroom environment (child-report), classroom environment (parent report), feelings about school (parent- and child-report) and classroom attitude (teacher-report) and six factors at age 16: teacher quality, classroom chaos, school satisfaction, school engagement, classroom peer relationships and subject attitudes (all self-report) and extracted factor scores for each participant . We performed contemporaneous reduced models – using just the perceived school environment factors as predictors of psychopathology. We then performed contemporaneous full models – using both the perceived and school-level factors as predictors and calculated the delta R 2 to establish the extent to school-level administrative measures over and above perceived school environment measures. At age 12, perceived school environment factors at age 9 (the closest available perceived measures in TEDS) 21.8%. The full model, including the school-level administrative factors, explained 22.3% of the variance, with a delta R 2 of 0.01. A likelihood ratio test confirmed that the full model explained significantly more variance than the reduced model: χ 2 (4) =7.25, p < 0.001, N = 1,808). At age 16, the perceived school environment explained 17.3% of the variance in the age 16 p-factor. The full model, including the school-level administrative factors, explained 18.8% of the variance, with a delta R 2 of 0.02. This is significantly more variance than the reduced model (χ 2 (4) =8.19, p < 0.001, N = 1,561). At age 21, factors of the perceived school environment at 9 and 16 explained 13.4% of the variance in the p-factor. The full model, also including the objective environment at ages 12 and 16, explained 14.3% of variance, with a delta R 2 of 0.01. This was also significantly more variance than the reduced model (χ 2 (8) = 2.82, p = 0.004, N = 1,333). Table 3: Variance explained in the general factor of psychopathology (p-factor) at ages 12, 16, and 21 by latent factors of the perceived school environment measures (reduced models) and the perceived school environment measures and school-level administrative measures (full models) Age Model Adjusted R 2 Adjusted delta R 2 P-value 12 Reduced 0.21 0.01 <.001 Full 0.22 16 Reduced 0.17 0.02 <.001 Full 0.19 21 Reduced 0.13 0.01 <.01 Full 0.14 Robustness checks To assess robustness, we compared these findings to complete cases multiple linear regression models run without performing factor analysis on the school-level measures (Supplementary Table 21). Variance explained (R²) was similar across both methodological approaches, although we observed that at age 21, the school-level administrative factors did not explain significantly more variance in psychopathology than the perceived school environment alone. Sample sizes were much smaller for the complete cases linear regression due to listwise deletion (N=500, compared to N=1333, for the models using latent factors). Differences across the two methods likely reflect noise and instability due to low sample size rather than the modelling strategy. Discussion Despite concern about growing rates of child and adolescent mental health difficulties, evidence on how schools may contribute to this trend remains limited. Previous studies, including our own, 17 have found associations between perceived school environment measures and psychopathology, but such findings leave open the possibility that these associations are partly driven by individual differences in perception. Our study aimed to address this by examining the associations between psychopathology and school-level administrative measures from childhood into early adulthood. Using linked data from TEDS, the National Pupil Database, and Ofsted, we found that administrative indicators of the school environment were associated with mental health across compulsory schooling and into early adulthood. Associations were strongest for indicators reflecting pupils’ everyday school experiences, including individual absence, compositional features of the pupil community (such as proportions eligible for free school meals and with special educational needs), and overall school-level absence. Accountability measures from Ofsted inspections showed weaker associations, which mostly did not persist into early adulthood. This pattern suggests that these structural and relational features of schools may be more relevant to psychopathology and its development than externally assessed indicators of institutional performance. We performed factor analysis on the administrative measures, resulting in four latent factors: individual absence, school-level characteristics, pupil community characteristics and headcount. These school-level administrative factors explained a small but significant amount of variance in psychopathology into emerging adulthood, suggesting that these factors contribute cumulatively to a small increase in mental health risk. We also found small bidirectional associations across adolescence for all administrative factors except headcount. This pattern suggests that while administrative indicators may reflect systemic features of schools that shape pupil mental health, changes in pupils’ mental health may also influence these indicators over time. This may occur through processes such as selection; for example, children with mental health difficulties may be more likely to be absent or require SEN support, thereby altering school-level compositions over time. The cumulative associations remained statistically significant after adjustment for genetic liability to psychiatric disorders, family socioeconomic status, and psychopathology at age 7, although effect sizes were very small. Bidirectional associations between the school-level characteristics factor and psychopathology at ages 12 and 16 were also not fully attenuated by these adjustments. This reflects a deliberately conservative analytic strategy aimed at minimising confounding and strengthening causal interpretation in an associational design. Although polygenic scores capture only a fraction of genetic liability for psychiatric disorders, their inclusion nonetheless helps to account for stable heritable risk and further contributes to the conservative nature of our estimates. However, this approach may have led to some degree of overcontrol. Given that children in the UK can begin school from age four, psychopathology measured at age seven may already reflect prior exposure to the school environment. Moreover, family socioeconomic status and genetic predispositions are not merely background characteristics of students, but are structurally entangled with school environments, potentially shaping classroom practices, peer dynamics, and institutional responses. While adjusting for these covariates allows us to make the strongest possible case for a causal contribution of school-level administrative factors to psychopathology, we believe this strategy likely yields conservative estimates of effect. The weaker associations observed for administrative school environment measures are theoretically expected. In line with previous studies 13,14,18 , our findings show that the associations between school-level administrative measures and psychopathology are smaller than those observed for perceived school environment measures. This pattern is consistent with theoretical accounts emphasising the importance of proximity in environmental influences on mental health, in which more distal, structural features of environments exert weaker effects than experiences that are directly perceived and encountered 23 . Administrative measures such as absenteeism rates, pupil community characteristics, and school characteristics may therefore influence mental health indirectly, through their downstream effects on classroom climate, peer relations, and access to support, rather than through immediate subjective experience. Although these effects are small at the individual level, they may nonetheless have substantial population-level implications, given that approximately nine million children attend schools in the UK 24 . In the context of overburdened child and adolescent mental health services, even modest reductions in risk associated with school-level administrative measures could translate into meaningful public health gains 25 . From a policy perspective, this suggests that targeted interventions, such as directing additional mental health resources to schools with higher absenteeism or greater concentrations of disadvantaged and special educational needs pupils, may represent a pragmatic approach to reducing the overall burden of psychopathology among young people. We also recognise that our study identified these small effects using school-level data routinely collected through national administrative sources – data that were not designed or collected with student mental health in mind. This makes our results more meaningful, as these measures may not fully capture aspects of the school environment that are likely more pertinent to child and adolescent mental health. For example, school environment factors such as behaviour management policies, break durations, and academic pressure - elements that are known to vary between schools and have been linked to pupil mental health in public discourse 26–29 - are neither systematically nor objectively monitored by the UK government. Given that the school environment is often seen as a crucial setting for addressing the rise in mental health issues among children and adolescents 6,13,18,30 , this lack of systemic data on key contextual factors presents a challenge. Without a comprehensive understanding of how various aspects of school life impact mental health, developing effective school-based interventions remains difficult. This underscores the urgent need for broader and more systematic data collection on the school environment to better inform efforts aimed at improving pupil mental health. Our study has some limitations. First, linking to the NPD and Ofsted databases required active consent from TEDS participants; however, not all participants provided this. We observed a small effect of socioeconomic status (SES), with participants who consented having, on average, a higher family SES than those who did not (See Methods section). However, given the modest effect size and the consistency of our findings with previous studies, we believe the impact on generalisability is limited. Nevertheless, future research should aim to include participants across the full spectrum of SES to ensure broader representativeness. Administrative measures were available only to students attending state-funded schools in the UK and did not include those attending private schools (7.12% of our total sample). This limitation restricts the generalisability of our findings to state school students. However, since only 5-7% of the UK population attends private schools 31 , our focus on state schools makes these findings more relevant to the majority of young people, whose school experiences are more predominantly shaped by government policies. Although the data for administrative school environment measures were relatively complete, there was substantial missingness in both the mental health and perceived school environment data at the item level - up to 90% for some measures. To address this, we used Full Information Maximum Likelihood (FIML), which enabled us to incorporate all available data by algorithmically estimating and imputing missing values. This approach allowed us to construct psychopathology factors at each age and address potential generalisability issues arising from missing data. Similarly, FIML was used to construct factors for the perceived and administrative measures of the school environment. Finally, we recognise that the participants in our study sample attended school between 2006 and 2012, which may limit the extent to which our findings reflect the current situation in UK schools. However, this perspective provided a unique opportunity to assess the long-term influence of the school environment on emerging adulthood, an area that previous studies have struggled to explore. Moreover, evidence since 2013 indicates that mental health issues among children and young people have increased 2,4 , with a larger proportion of children reporting unhappiness with school in 2024 compared to 2010 11 . Also, as of 2022, UK schools were rated lower than the European average for school belonging, school safety, and long-term absence 11 . Taken together, these trends suggest that our findings likely underestimate the current influence of the school environment on adolescent mental health and further highlight the need for robust, longitudinal research that examines how contemporary school environments influence mental health. In summary, administrative measures of the school environment show small but consistent associations with psychopathology across adolescence and into early adulthood. These associations were weaker than those observed for perceived school environment measures. Nonetheless, their persistence after stringent adjustment for genetic, familial, and early mental health factors suggests that administrative indicators contribute incrementally to mental health risk and are not fully explained by individual differences. Our findings also point to a dynamic interplay between schools and pupils, whereby school-level administrative characteristics may both influence and respond to pupil mental health, likely through more proximal features of school life such as classroom climate, levels of support and teacher– and peer–relationships. Our results highlight the value of integrating administrative and perceived indicators of the school environment in future research, and suggest that routinely collected non-perceptual data on mental health-relevant aspects of school life could help inform targeted mental health interventions in school settings. Methods Study design and participants Participants (N= 6,398, 54.1% female) are part of the Twins Early Development Study (TEDS), a population-based longitudinal study comprised of over 13,000 families with twins born between 1994 and 1996. More than 10,000 twin pairs remain actively involved. The families in TEDS are representative of the British population for their birth cohort in terms of socioeconomic status, ethnicity and parental occupation 32 . Here, we used data from a subsample of twins who had consented to the use of their data held on the National Pupil Database (NPD). Using ANOVA, we compared the family socioeconomic status (SES) at first contact (~1.5 years) for TEDS participants who had consented to the use of their NPD data (N = 11,982) with those who had not (N=11,702). We found that participants who consented had a significantly higher mean socioeconomic status (SES; M = 0.2, SD = 0.99) than those who did not (M = -0.2, SD = 0.97), although the effect was small (F(1, 23,682)= 963.9, p < 0.001, η 2 = 0.039). As twin pairs are likely to attend the same school, we randomly selected one twin from each pair for analyses to control for the non-independence of observations. Those who reported severe medical or neurological conditions were also excluded (N = 267). The sample size varied across time and measures, but all available data were used (see Supplementary Tables 1-3). Written informed consent was obtained from parents before data collection and from TEDS participants themselves for collections carried out after the twins had reached adulthood. King's College London's Ethics Committee approved the project for the Institute of Psychiatry, Psychology and Neuroscience PNM/09/10–104. Measures Psychopathology: TEDS provides extensive data on emotional and behavioural traits from childhood through to early adulthood. In the present study, we used psychopathology data collected at ages 7, 12, 16 and 21. Traits were assessed using a battery of questionnaires completed by parents, teachers, and participants themselves (see Supplementary Table 22 for a full list of measures). We used all mental health trait measures at each age point to calculate a theory-driven general factor of psychopathology 33,34 (or p-factor) for each age group using confirmatory factor analysis (CFA) and Full Information Maximum Likelihood (FIML) to account for missing data in lavaan for R 35 . Participants received a standardised score based on this p-factor, indicating their relative level of psychopathology (high scores indicated a higher burden of psychopathology). We used CFA as a technique for data reduction and to account for missing data in individual measures, rather than for studying the structure of mental health symptoms across development. All model fit statistics and factor loadings for p-factors at age 7, 12, 16 and 21 are presented in Supplementary Table 7. School-level administrative measures: Objective measures of the school environment come from two sources. The National Pupil Database (NPD) is an administrative dataset collected by the Department for Education (DfE) on state-educated pupils in England born since 1986. This study linked data on pupil-level absenteeism, school-level absenteeism, and school-level census data to participants in the TEDS sample (see Supplementary Table 23 for a full list of measures). Each participant was linked to the objective school environment data relevant to the years in which they were 12 and 16 (see Supplementary Note 1 for full details of the data linkage process). Additional measures are based on Ofsted inspection data collected between 2005 and 2015. Ofsted is the UK government office responsible for inspecting institutions that provide education and skills services, including all state-funded schools 36 . Multiple measures of school quality are publicly available, including ratings of teaching quality, student achievement and overall effectiveness. Inspectors rate each aspect of school quality on a 4-point scale (1= outstanding, 2= good, 3= satisfactory, 4= needs improvement), and given inspection ratings can determine the future governance of a school. The relevant Ofsted data for each TEDS participant at ages 12 and 16 were determined by selecting the closest inspections to 2005-08 (when participants were 12 years old) and 2010-2013 (when participants were 16 years old). See Supplementary Table 23 for a full list of Ofsted measures. Given the large number of objective school environment measures, we conducted exploratory and confirmatory factor analysis (EFA and CFA) to reduce the dimensionality of the data and address potential issues with multicollinearity. We performed EFA using psych for R on the objective environmental measures available at age 16 in our sample. Based on the EFA results,  we extracted and created latent factors of the objective school environment at ages 12 and 16 using confirmatory factor analysis (CFA) in lavaan for R. This resulted in the creation of four latent factors measuring features of the objective school environment: individual absence, school-level characteristics (including school quality ratings and school-level absenteeism), pupil community characteristics (including the proportion of children receiving free school meals and with special educational needs) and headcount (including number of pupils and proportion of male and female students). Full Information Maximum Likelihood (FIML) was used to account for data missingness (see Supplementary Tables 24-25 for full factor loadings and model fit statistics). Each participant was assigned a standardised score for each factor, indexing the individual administrative school environment they had experienced. Perceived school environment TEDS provides multiple measures of the school environment, including classroom environment, relationship with teachers, and homework, rated by parents, teachers, and participants (see Supplementary Table 26 for a full list of subjective school environment measures). Measures were collected at ages 7, 9 and 16, but the measures differ across data collection waves. As with the administrative measures of school environment, we also conducted exploratory and confirmatory factor analysis (EFA and CFA) to reduce the dimensionality of the data and address potential issues with multicollinearity. Based on the results of the EFA, we constructed four factors of the subjective school environment at age 9 (classroom environment (child-report), classroom environment (parent report), feelings about school (parent and child-report) and classroom attitude (teacher-report)) and six factors at age 16 (teacher quality, classroom chaos, school satisfaction, school engagement, classroom peer relationships and subject attitudes). FIML was used to account for data missingness (see Supplementary Tables 27-28 for full factor loadings and model fit statistics). Each participant was assigned a standardised score based on these factors, indexing the subjective school environment reported by their parent, teacher, and themselves. Polygenic scores Genome-wide polygenic scores (PGS) were calculated for each participant to estimate individual genetic predisposition to psychiatric problems. Genotyping Genotyping was performed using one of two DNA microarrays (Affymetrix GeneChip 6.0 or Illumina HumanOmniExpressExome chips) according to standard protocols. DNA samples have been collected from 12,500 TEDS participants. Following rigorous quality control procedures, 10,346 samples were available for genomic analyses. Of them, 7,026 belonged to unrelated individuals, and 3,320 to additional dizygotic co-twins, with 7,289 individuals genotyped using Illumina arrays and 3,057 individuals genotyped with Affymetrix arrays. Further information on genotyping and quality control is detailed elsewhere 37,38 . Polygenic scores PGS were calculated to index genetic vulnerability to psychiatric problems using LDPred 39 . Details of the methods and analysis used are presented in earlier studies 34,38 . To construct the PGS, we used the genome-wide association summary statistics for 27 adult psychiatric disorders. We selected all mental health and mental health-related polygenic scores available in TEDS that were constructed using well-powered genome-wide association studies. These were: ADHD (attention deficit hyperactivity disorder) 40 , anxiety 41 , anxious feeling/worrier 41 , ASD (autism spectrum disorder) 42 , bipolar disorder 43 , multiple disorders: total child psychiatric problems 44 , daytime sleepiness 45 , depressive symptoms 46 , externalising problems 47 , fed-up feelings 48 , guilty feelings 48 , insomnia 49 , irritability 48 , major depressive disorder 50 , miserableness 48 , mood swings 48 , nervous feelings 48 , neuroticism 51 , OCD (obsessive compulsive disorder) 52 , PTSD (post-traumatic stress disorder) 53 , risk-taking behaviour 48 , risk tolerance 54 , schizophrenia 55 , sensitivity/hurt feelings 48 , sleep disorders 48 , subjective wellbeing 46 , Tourette’s syndrome 56 . We used all PGS available in TEDS related to psychopathology. Socioeconomic status Family SES was assessed at first contact (when the twins were ~1.5 years old). Mothers and fathers reported their educational level, occupational status and mothers’ age at the birth of their first child. The composite of SES was computed as the mean of the standardised scores of these measures. Socioeconomic status (SES) was taken at first contact as this wave offers minimal missing data, and SES has been shown to be developmentally stable, with age-to-age correlations of ~0.70 57 . Statistical analysis Analyses were performed in R version 4.4.0. Descriptive statistics and data preparation: we calculated descriptive statistics for all measures across all age collection waves. We conducted univariate analyses of variance (ANOVA) to explore phenotypic sex differences across measures. As significant sex differences emerged, all variables were adjusted for sex and age using linear regression. Age- and sex-adjusted standardised variables were used in all subsequent analyses. General factor of psychopathology (p-factor): we calculated a p-factor, i.e., a general factor for psychopathology, for each age group, using confirmatory factor analyses (CFA; See Measures). Correlations: Pearson’s correlations were used to evaluate the contemporaneous associations between objective school environment measures and psychopathology for each participant at ages 12 and 16, as well as the associations over time between the objective school environment measures at 12 and mental health at 16, and objective school environment measures at 12 and 16 and mental health at 21. We applied Benjamini-Hochberg false discovery rate correction (FDR) to account for multiple testing. Exploratory and Confirmatory Factor Analysis: We examine applied exploratory factor analysis (EFA) to the school-level administrative measures and the perceived school environment measures. Based on the results we calculated latent factors of the school-level administrative measures and the perceived school environment measures using confirmatory factor analysis (CFA; See Measures). Contemporaneous models to predict psychopathology (p-factor) outcomes: We wanted to explore how well all school-level administrative measures combined explain psychopathology in childhood. We used multiple linear regression to determine the variance explained by the latent factors of the school-level administrative measures at age 12 in child psychopathology (p-factor) at age 12. We then used multiple linear regression to determine the variance explained by latent factors of the school-level administrative measures at age 16 in child psychopathology (p-factor) at age 16. We took a conservative approach and controlled both models for possible confounders in stages. First, we controlled for possible genetic confounding used 27 adult psychiatric polygenic scores (PGS). Then, we additionally controlled for family socioeconomic status (SES). Finally, in the fully adjusted model, we also controlled for baseline psychopathology (p-factor) at age 7. We reported the adjusted R 2 for all adjusted and unadjusted models. We conducted a series of additional analyses to check for robustness: We repeated all multiple linear regression models using the school-level administrative measures without performing EFA and CFA. We also repeated the unadjusted models using penalised elastic net regression, a regularisation technique that addresses issues with multicollinearity and overfitting 58 . Longitudinal models to predict psychopathology (p-factor) outcomes: We wanted to explore how well all school-level administrative measures combined explain psychopathology across development. We used multiple linear regression to determine the variance explained by the latent factors of the school-level administrative measures at age 12 and 16 in child psychopathology (p-factor) at age 16. We then used multiple linear regression to determine the variance explained by latent factors of the school-level administrative measures at age 12 and 16 in child psychopathology (p-factor) at age 21. Again, took a conservative approach, controlling in stages for 27 adult psychiatric polygenic scores (PGS), family socioeconomic status (SES) and for baseline psychopathology (p-factor) at age 7. We reported the adjusted R 2 for all adjusted and unadjusted models. As with the contemporaneous models, we repeated all unadjusted and adjusted models using the school-level administrative measures without performing EFA and CFA. We also repeated unadjusted models using penalised elastic net regression, a regularisation technique that addresses issues with multicollinearity and overfitting 58 . Cross-lagged panel analyses: We explored the longitudinal relationships over time between the latent factors of school environment, indexed by the four latent factors, and psychopathology (p-factors) at ages 12 and 16 using cross-lagged models 59,60 (see Figure 4). A cross-lagged model examines relationships over time by specifying four types of pathways: (a) an initial cross-sectional pathway that assesses the relationship between two different variables contemporaneously, (b) two pathways that link different variables across multiple time points to assess cross-lagged effects, (c) pathways which connect the same variable across time to assess stability effects, and (d) residual cross-sectional pathways that assess the relationship between the two different variables at a later point in development 61,62 . Pathways are represented by partial regression coefficients, which are independently estimated and control for pre-existing relationships between variables (such as p-factor and an objective environment factor). The strength of these pathways was assessed using 95% confidence intervals. We performed four cross-lagged panel models, using each of the four latent factors of the school-level administrative measures and p-factors at ages 12 and 16, allowing for the comparison and interpretation of the developmental associations between the constructs. Contemporaneous models to assess the joint prediction of school-level administrative and perceived school environment factors: We wanted to explore how well school-level administrative measures predicted psychopathology across development, accounting for the prediction of perceived school environment measures. We performed three ‘reduced’ multiple linear regression models – these assessed the effects of the perceived school environment only: in model one we examined the variance in the p-factor at age 12 explained at by perceived school environment factors at age 9, as more contemporaneous measures were not available in TEDS. In model two, we assessed the variance in the p-factor at 16 explained by perceived school environment factors at age 16. In model three, we assessed the variance in the p-factor at 21 explained by perceived school environment factors at ages 9 and 16. We then re-ran each of the three models as ‘full’ models by additionally including school-level administrative factors (age 12 administrative factors to model one; age 16 administrative factors to model two; age 12 and 16 administrative factors to model three). We calculated delta R 2 values to compare variance explained by the perceived and administrative measures and performed likelihood ratio tests to establish whether the ‘full’ models provide a significantly better fit than the ‘reduced’ models. Declarations Acknowledgements We gratefully acknowledge the ongoing contribution of the participants in the Twins Early Development Study (TEDS) and their families. TEDS is supported by the UK Medical Research Council (MR/V012878/1 and previously MR/M021475/1). R.F. is supported by a university-funded studentship from Royal Holloway, University of London. K.R. was supported by a Sir Henry Wellcome Postdoctoral Fellowship and UK Medical Research Council Grant UKRI1503. This research was funded in part by the Wellcome Trust (213514/Z/18/Z). For the purpose of open access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. This work was undertaken in the Office for National Statistics Secure Research Service using data from ONS and other owners and does not imply the endorsement of the ONS or other data owners. Data availability Researchers can apply for access to the TEDS data through their data access mechanism: www.teds.ac.uk/researchers/teds-data-access-policy. Researchers can apply for access to the NPD data via the UK data service: https://ukdataservice.ac.uk/. The Ofsted data is made publicly available by the UK government: https://www.gov.uk/government/collections/maintained-schools-and-academies-inspections-and-outcomes-official-statistics Code availability All R code is available at OSF: https://osf.io/gsj6c/ Competing interests The authors declare no competing interests References Armitage, J. M. et al. Cross-cohort change in parent-reported emotional problem trajectories across childhood and adolescence in the UK. Lancet Psychiatry 10 , 509–517 (2023). Sadler, K. et al. Mental Health of Children and Young People in England, 2017: Trends and characteristics. in Mental Health of Children and Young People Surveys (2018). Collishaw, S. Annual research review: Secular trends in child and adolescent mental health. J. Child Psychol. Psychiatry 56 , 370–393 (2015). Newlove-Delgado, T. et al. Mental Health of Children and Young People in England, 2023 - Wave 4 Follow up to the 2017 Survey . https://digital.nhs.uk/data-and-information/publications/statistical/mental-health-of-children-and-young-people-in-england/2023-wave-4-follow-up (2023). Ward, J. L. et al. Admission to acute medical wards for mental health concerns among children and young people in England from 2012 to 2022: a cohort study. Lancet Child Adolesc. Health 9 , 112–120 (2025). Foulkes, L. & Andrews, J. L. Are mental health awareness efforts contributing to the rise in reported mental health problems? A call to test the prevalence inflation hypothesis. New Ideas Psychol. 69 , 101010 (2023). Patel, V. et al. Income inequality and depression: a systematic review and meta-analysis of the association and a scoping review of mechanisms. World Psychiatry Off. J. World Psychiatr. Assoc. WPA 17 , 76–89 (2018). Knapp, M. Mental health in an age of austerity. Evid. Based Ment. Health 15 , 54–55 (2012). Orben, A. Teenagers, screens and social media: a narrative review of reviews and key studies. Soc. Psychiatry Psychiatr. Epidemiol. 55 , 407–414 (2020). Children’s Mental Health Services 2022-23 . https://www.childrenscommissioner.gov.uk/resource/childrens-mental-health-services-2022-23/ (2024). The Good Childhood Report 2024 . https://www.childrenssociety.org.uk/good-childhood (2024). State of the Nation 2022: Children and Young People’s Wellbeing . https://www.gov.uk/government/publications/state-of-the-nation-2022-children-and-young-peoples-wellbeing (2022). Patalay, P., O’Neill, E., Deighton, J. & Fink, E. School characteristics and children’s mental health: A linked survey-administrative data study. Prev. Med. 141 , 106292 (2020). Ford, T. et al. The Role of Schools in Early Adolescents’ Mental Health: Findings From the MYRIAD Study. J. Am. Acad. Child Adolesc. Psychiatry 60 , 1467–1478 (2021). Aldridge, J. M. & McChesney, K. The relationships between school climate and adolescent mental health and wellbeing: A systematic literature review. Int. J. Educ. Res. 88 , 121–145 (2018). Leurent, B. et al. Is positive school climate associated with better adolescent mental health? Longitudinal study of young people in England. Ssm - Ment. Health 1 , None (2021). Rimfeld, K. et al. Longitudinal associations between school environment and mental health from childhood through early adulthood. Preprint at https://doi.org/10.21203/rs.3.rs-8626819/v1 (2026). Hinze, V. et al. Student- and School-Level Factors Associated With Mental Health and Well-Being in Early Adolescence. J. Am. Acad. Child Adolesc. Psychiatry 63 , 266–282 (2024). Kendler, K. S. & Baker, J. H. Genetic influences on measures of the environment: a systematic review. Psychol. Med. 37 , 615–626 (2007). Plomin, R. & Bergeman, C. S. The nature of nurture: Genetic influence on “environmental” measures. Behav. Brain Sci. 14 , 373–386 (1991). Plomin, R., Reiss, D., Hetherington, E. M. & Howe, G. W. Nature and nurture: Genetic contributions to measures of the family environment. Dev. Psychol. 30 , 32–43 (1994). Wray, N. R. et al. Research Review: Polygenic methods and their application to psychiatric traits. J. Child Psychol. Psychiatry 55 , 1068–1087 (2014). Kranzler, J. H., Floyd, R. G., Bray, M. A. & Demaray, M. K. Past, present, and future of research in school psychology: The biopsychosocial ecological model as an overarching framework. Sch. Psychol. Wash. DC 35 , 419–427 (2020). Funder, D. C. & Ozer, D. J. Evaluating Effect Size in Psychological Research: Sense and Nonsense. Adv. Methods Pract. Psychol. Sci. 2 , 156–168 (2019). Carey, E. G., Ridler, I., Ford, T. J. & Stringaris, A. Editorial Perspective: When is a ‘small effect’ actually large and impactful? J. Child Psychol. Psychiatry 64 , 1643–1647 (2023). Rainer, C., Le, H. & Abdinasir, K. Behaviour and Mental Health in Schools . https://cypmhc.org.uk/publications/behaviour-and-mental-health-in-schools-report/ (2023). Jones, R., Kreppner, J., Marsh, F. & Hartwell, B. Punitive behaviour management policies and practices in secondary schools: A systematic review of children and young people’s perceptions and experiences. Emot. Behav. Difficulties 28 , 182–197 (2023). Baines, E. & Blatchford, P. The decline in breaktimes and lunchtimes in primary and secondary schools in England: Results from three national surveys spanning 25 years. Br. Educ. Res. J. 49 , 925–946 (2023). Steare, T., Gutiérrez Muñoz, C., Sullivan, A. & Lewis, G. The association between academic pressure and adolescent mental health problems: A systematic review. J. Affect. Disord. 339 , 302–317 (2023). Cattan, S., Lereya, S. T., Yoon, Y., Gilbert, R. & Deighton, J. The impact of area level mental health interventions on outcomes for secondary school pupils: Evidence from the HeadStart programme in England. Econ. Educ. Rev. 96 , 102425 (2023). Elitist Britain 2019: The educational backgrounds of Britain’s leading people. GOV.UK https://www.gov.uk/government/publications/elitist-britain-2019/elitist-britain-2019-the-educational-backgrounds-of-britains-leading-people. Lockhart, C. et al. Twins Early Development Study (TEDS): A genetically sensitive investigation of mental health outcomes in the mid-twenties. JCPP Adv. 3 , e12154 (2023). Caspi, A. & Moffitt, T. E. All for One and One for All: Mental Disorders in One Dimension. Am. J. Psychiatry 175 , 831–844 (2018). Allegrini, A. G. et al. Genomic prediction of cognitive traits in childhood and adolescence. Mol. Psychiatry 24 , 819–827 (2019). Rosseel, Y. lavaan: An R Package for Structural Equation Modeling. J. Stat. Softw. 48 , 1–36 (2012). Ofsted. GOV.UK https://www.gov.uk/government/organisations/ofsted (2024). Rimfeld, K. et al. Twins Early Development Study: A Genetically Sensitive Investigation into Behavioral and Cognitive Development from Infancy to Emerging Adulthood. Twin Res. Hum. Genet. Off. J. Int. Soc. Twin Stud. 22 , 508–513 (2019). Selzam, S. et al. Evidence for gene-environment correlation in child feeding: Links between common genetic variation for BMI in children and parental feeding practices. PLOS Genet. 14 , e1007757 (2018). Vilhjálmsson, B. J. et al. Modeling Linkage Disequilibrium Increases Accuracy of Polygenic Risk Scores. Am. J. Hum. Genet. 97 , 576–592 (2015). Demontis, D. et al. Genome-wide analyses of ADHD identify 27 risk loci, refine the genetic architecture and implicate several cognitive domains. Nat. Genet. 1–11 (2023) doi:10.1038/s41588-022-01285-8. Watson, H. J. et al. Genome-wide association study identifies eight risk loci and implicates metabo-psychiatric origins for anorexia nervosa. Nat. Genet. 51 , 1207–1214 (2019). Grove, J. et al. Identification of common genetic risk variants for autism spectrum disorder. Nat. Genet. 51 , 431–444 (2019). Mullins, N. et al. Genome-wide association study of over 40,000 bipolar disorder cases provides new insights into the underlying biology. Nat. Genet. 53 , 817–829 (2021). Neumann, A. et al. A genome-wide association study of total child psychiatric problems scores. PLOS ONE 17 , e0273116 (2022). Lane, J. M. et al. Genome-wide association analyses of sleep disturbance traits identify new loci and highlight shared genetics with neuropsychiatric and metabolic traits. Nat. Genet. 49 , 274–281 (2017). Okbay, A. et al. Genetic variants associated with subjective well-being, depressive symptoms, and neuroticism identified through genome-wide analyses. Nat. Genet. 48 , 624–633 (2016). Karlsson Linnér, R. et al. Multivariate analysis of 1.5 million people identifies genetic associations with traits related to self-regulation and addiction. Nat. Neurosci. 24 , 1367–1376 (2021). UK Biobank. Neale lab http://www.nealelab.is/uk-biobank (2018). Jansen, P. R. et al. Genome-wide analysis of insomnia in 1,331,010 individuals identifies new risk loci and functional pathways. Nat. Genet. 51 , 394–403 (2019). Howard, D. M. et al. Genome-wide meta-analysis of depression identifies 102 independent variants and highlights the importance of the prefrontal brain regions. Nat. Neurosci. 22 , 343–352 (2019). Nagel, M. et al. Meta-analysis of genome-wide association studies for neuroticism in 449,484 individuals identifies novel genetic loci and pathways. Nat. Genet. 50 , 920–927 (2018). Arnold, P. D. et al. Revealing the complex genetic architecture of obsessive–compulsive disorder using meta-analysis. Mol. Psychiatry 23 , 1181–1188 (2018). Meier, S. M. et al. Genetic Variants Associated With Anxiety and Stress-Related Disorders: A Genome-Wide Association Study and Mouse-Model Study. JAMA Psychiatry 76 , 924–932 (2019). Karlsson Linnér, R. et al. Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences. Nat. Genet. 51 , 245–257 (2019). Trubetskoy, V. et al. Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature 604 , 502–508 (2022). Yu, D. et al. Interrogating the genetic determinants of Tourette syndrome and other tic disorders through genome-wide association studies. Am. J. Psychiatry 176 , 217–227 (2019). Hanscombe, K. B. et al. Socioeconomic Status (SES) and Children’s Intelligence (IQ): In a UK-Representative Sample SES Moderates the Environmental, Not Genetic, Effect on IQ. PLOS ONE 7 , e30320 (2012). Gidziela, A. et al. Explaining the influence of non-shared environment (NSE) on symptoms of behaviour problems from preschool to adulthood: mind the missing NSE gap. J. Child Psychol. Psychiatry 64 , 747–757 (2023). Voronin, I. TwinAnalysis. Boker, S. et al. OpenMx: An Open Source Extended Structural Equation Modeling Framework. Psychometrika 76 , 306–317 (2011). Gauld, C., Grasman, R. P. P. P. & Bailly, S. Usefulness of Cross-Lagged Panel Models for Clinical Research. CHEST 167 , 1537–1540 (2025). Lüdtke, O. & Robitzsch, A. A Critique of the Random Intercept Cross-Lagged Panel Model. Preprint at https://doi.org/10.31234/osf.io/6f85c (2021). Additional Declarations There is NO Competing Interest. Supplementary Files AdministrativeSchSupplementarymaterialsRF.docx Supplementary material SupplementarynotesRF.docx Supplementary note Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information

Related documents

• Prevalence of mental health, mental well-being, and school connectedness in UK secondary school students in 2021/2022: Results from the South West - School Health Research Network pilot study.#610370
• The Protective Paradox: Can School Connectedness Buffer Socioeconomic Disparities in Adolescent Mental Health?#351696
• Social disconnection and subsequent mental disorders: a population-based cohort study.#170509
• The 'UPIC' cohort: a nationwide prospective study of mental health among adolescents and young adults in Sweden - a study protocol.#1167217
• Optimizing Multicomponent Interventions to Improve Child Mental Health Following Reintegration From Institutions: Protocol for a Factorial Randomized Trial.#380666