Introduction
Adolescence is a critical developmental period marked by profound cognitive, emotional, and social changes, making it a pivotal stage for both educational achievement and subsequent mental health trajectories (Baird et al., Reference Baird, Choonara, Azzopardi, Banati, Bessant, Biermann, Capon, Claeson, Collins, De Wet-Billings, Dogra, Dong, Francis, Gebrekristos, Groves, Hay, Imbago-Jácome, Jenkins, Kabiru and Viner2025; McGorry et al., Reference McGorry, Mei, Dalal, Alvarez-Jimenez, Blakemore, Browne, Dooley, Hickie, Jones, McDaid, Mihalopoulos, Wood, El Azzouzi, Fazio, Gow, Hanjabam, Hayes, Morris, Pang and Killackey2024; Sawyer, Azzopardi, Wickremarathne, & Patton, Reference Sawyer, Azzopardi, Wickremarathne and Patton2018). A large body of literature has established an association between poor school achievement during comprehensive education (i.e., grades from 1 to 9) and a greater risk of developing a mental disorder later in life (Gyllenberg et al., Reference Gyllenberg, Ristikari, Kelleher, Kääriälä and Gissler2022; Kendler et al., Reference Kendler, Ohlsson, Mezuk, Sundquist and Sundquist2016b, Reference Kendler, Ohlsson, Keefe, Sundquist and Sundquist2018; Kendler, Ohlsson, Sundquist, & Sundquist, Reference Kendler, Ohlsson, Sundquist and Sundquist2017; MacCabe et al., Reference MacCabe, Lambe, Cnattingius, Sham, David, Reichenberg, Murray and Hultman2010; Pedersen, Østergaard, & Petersen, Reference Pedersen, Østergaard and Petersen2018; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023). The association is observed for different types of mental disorders such as schizophrenia and depression (Kendler et al., Reference Kendler, Ohlsson, Mezuk, Sundquist and Sundquist2016b; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023), with the strongest association typically detected with substance use disorders (Kendler et al., Reference Kendler, Ohlsson, Keefe, Sundquist and Sundquist2018; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023). In contrast to these findings, better school achievement has been associated with a greater risk of developing an eating disorder in some studies (Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023), but not in others (Sundquist et al., Reference Sundquist, Ohlsson, Winkleby, Sundquist and Crump2016).
Previous studies including information on relatives suggest that shared familial environmental or genetic factors explain only a relatively small portion of the association between school achievement and mental disorders (Kendler et al., Reference Kendler, Ohlsson, Mezuk, Sundquist and Sundquist2016a, Reference Kendler, Ohlsson, Keefe, Sundquist and Sundquist2018; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023). However, previous studies have only rarely used information on general upper secondary education after comprehensive education, leaving it unclear whether later school achievement is similarly associated with the risk of developing a mental disorder as school achievement during comprehensive education. Moreover, reduction in school performance has been suggested to be associated especially with the development of schizophrenia (Conejeros-Pavez et al., Reference Conejeros-Pavez, Vasquez, Diaz, Mena, Undurraga, Gonzalez-Valderrama, Claro, Undurraga and Crossley2025; Kahn & Keefe, Reference Kahn and Keefe2013), but to the best of our knowledge, this topic has not been examined in large-scale population-based studies.
In the present study, using nationwide register data from Finland, our primary aim was to examine the association between school achievement in upper secondary education and the subsequent risk of developing a mental disorder. The secondary aim was to investigate whether change in school achievement from comprehensive to upper secondary education was associated with the later risk of developing a mental disorder.
Methods
We linked data across nationwide Finnish registers using the unique identification number assigned to all Finnish residents. Our study cohort included all persons born in Finland between 1980 and 2001 who graduated from upper secondary education between 1999 and 2021. The follow-up began at the beginning of the month following graduation from upper secondary education. The final study cohort consisted of 579 781 individuals (the flowchart describing the construction of the final study cohort is shown in Supplementary Figure S1).
The study plan was approved by the ethics committee of the National Institute of Health and Welfare (THL/22/6.02.01/2019), and registers were linked with the permission of the National Institute of Health and Welfare and Statistics Finland (TK-53-1696-16). According to Finnish law, informed consent is not required for register-based studies.
Measures
Exposure
School achievement at the end of upper secondary education was measured using grade data from the matriculation examination, which is organized twice a year and comprises several subject-specific exams. Although the number of mandatory and additional exams has varied over the years, currently all students are required to take five subject-specific exams, with mother tongue and literature being mandatory. Additional exams are chosen from mathematics, foreign languages, and humanities and sciences. The exams are graded according to the following seven verbal grades with Latin names: Laudatur (L), highest grade; Eximia cum laude approbatur (E); Magna cum laude approbatur (M); Cum laude approbatur (C); Lubenter approbatur (B); Approbatur (A), lowest grade for passing the exam; and Improbatur (I). Up to spring 2014, the grading followed a normal distribution each year, which resulted in 5% of students receiving L, 15% of students receiving E, 20% of students receiving M, C, and B, 15% of students receiving A, and 5% of students receiving I. Since spring 2014, the Average of Standardised Total Scores (ASTS) method, which takes into account the number of candidates who have participated in the specific exam, has been used (Matriculation Examination Board, 2024). All exams are first graded by local upper secondary school teachers and after that confirmed or re-graded by the national board. To calculate the matriculation examination GPA for each person, we converted the verbal exam grades to numerical grades (L = 7, E = 6, M = 5, C = 4, B = 3, A = 2, I = 1) and calculated the mean of all exam grades received before graduation. The numerical GPAs were rounded to the nearest integer and converted back to their corresponding verbal grades to represent the general performance of the student.
Outcome
Diagnoses of mental disorders were obtained from the Finnish Care Register for Health Care and the Register of Primary Health Care Visits, which contain data on all inpatient visits since 1970, outpatient visits in special healthcare since 1998, and primary care since 2011. Diagnoses were based on the ICD–8 from 1969 to 1986, the Finnish version of the ICD–9 from 1987 to 1995, and the ICD–10 from 1996 onwards. In some primary care facilities, ICPC–2 was used instead of ICD–10. The ICD–8, ICD–9, and ICPC–2 diagnoses were converted to corresponding ICD–10 sub-chapter categories (Suokas et al., Reference Suokas, Niemi, Gutvilig, McGrath, Komulainen, Suvisaari, Elovainio, Lumme, Pirkola and Hakulinen2025). Overlapping register entries were handled using a predefined algorithm (Suokas et al., Reference Suokas, Gutvilig, Lumme, Pirkola and Hakulinen2024).
Time to the first diagnosis of any mental disorder (ICD-10 diagnoses F00–F99) was used as the main outcome. Additionally, time to the following eight diagnostic sub-categories were analyzed: (1) substance use disorders (F10–F19), (2) schizophrenia spectrum disorders including schizophrenia, schizotypal, and delusional disorders (F20–F29), (3) bipolar disorder (F30–F31), (4) depression (F32–F33), (5) anxiety disorders including neurotic, stress-related, and somatoform disorders (F40–F48), (6) eating disorders (F50), (7) sleep disorders (F51), and (8) personality disorders (F60–F69).
Covariates
The following covariates were included in the analytical models: birth year; sex (0 = male, 1 = female); degree of urbanicity in residential location at the year prior to follow-up start (0 = missing, 1 = urban, 2 = semi-urban, 3 = rural), parental education level (1 = comprehensive, 2 = upper secondary, 3 = post-secondary or tertiary) and income level in quintiles at the year prior to follow-up start (0 = missing, 1 = 1st quintile, 2 = 2nd quintile, 3 = 3rd quintile, 4 = 4th quintile, 5 = 5th quintile); and parental mental health history before the date of follow-up start (0 = no psychiatric diagnosis, 1 = any psychiatric diagnosis [ICD-10 diagnoses F00–F99] before the follow-up period).
School achievement at the end of comprehensive education (year nine in comprehensive school) was measured as the average of grades from compulsory (e.g., mother tongue and literature) and optional (e.g., the second foreign language) subjects. This average grade, originally ranging from 4 to 10 (higher numbers indicating better school achievement), was standardized to a z-score (mean 0, SD 1) within each year of graduation for the analyses. The grades were obtained from the National Joint Application Register containing information on school grades of students applying to upper secondary education until 2007 and of all students in the final year of comprehensive school after 2007.
Statistical analysis
The study cohort was followed from the beginning of the month following upper secondary school graduation (1st of July or 1st of January) until the first mental disorder diagnosis, permanent emigration from Finland, death, or the end of follow-up on December 31st 2023, whichever came first. Associations between matriculation examination GPA at the end of upper secondary education and the subsequent risk of a mental disorder diagnosis were estimated using Cox proportional hazards models with age as the underlying time scale. Individuals with prevalent mental disorders at follow-up start were excluded from each disorder-specific analysis.
The following four separate analytical models were estimated: Model 1 was adjusted for sex and birth year, which, together with age as the time scale, adjusts for calendar time. Model 2 was adjusted for all covariates, i.e., sex, birth year, parental education level, parental income level, parental mental health history, and the degree of urbanicity. Model 3 was adjusted for the GPA at the end of comprehensive education and all covariates. Model 4 was adjusted for the GPA at the end of comprehensive education, all covariates, and estimated using stratified Cox proportional hazards models conducted within strata of full-sibling sets.
The stratified Cox proportional hazards models enabled us to account for unobserved confounding by shared family characteristics that could affect both school achievement and the incidence of mental disorders (Sjölander, Frisell, & Öberg, Reference Sjölander, Frisell and Öberg2022). In addition to these four models, we also estimated whether the GPA at the end of comprehensive education modified the association between school performance at the end of upper secondary and the later risk of a mental disorder using separate models where the GPA at the end of comprehensive education was divided into three groups (under −1 SD; between −1 SD and 1 SD; over 1 SD). Data management was conducted using Stata version 17.0 (Stata Corp, College Station, TX) and R version 4.4.1 (The R Foundation).
Results
Descriptives of the study cohort are reported in Table 1. During 7.1 million person-years of follow-up, 215 678 (37.2%) study cohort members were diagnosed with a mental disorder.
Descriptives of the study cohort stratified by mental disorder status at the end of follow-up
Table 1. Long description
The table presents descriptive statistics for a study cohort divided into two groups: Mental Disorder (N = 215,678) and No Mental Disorder (N = 364,103).
* Sex: In the Mental Disorder group, 67% are women and 33% are men. In the No Mental Disorder group, 51% are women and 49% are men.
* Birth Year: The median birth year is 1989 for the Mental Disorder group and 1990 for the No Mental Disorder group.
* Parental Education: For both groups, the majority of mothers and fathers have post-secondary or tertiary education (ranging from 46% to 57%).
* Urbanicity: Both groups show a similar distribution, with approximately 57-59% living in urban areas.
* Income Quintiles: Mother’s and father’s income quintiles are relatively evenly distributed at approximately 20% each across both groups, though the Mental Disorder group has a slightly higher concentration in the lowest quintile (21%).
* Parental Mental Health: 12% of mothers and 11% of fathers in the Mental Disorder group had a prior diagnosis, compared to 9% and 8% respectively in the No Mental Disorder group.
* Academic Performance: Median G P A in middle school is 8.50 for both groups. Matriculation Examination GPA distributions are nearly identical, with the largest segment in category C (33% for both groups).
The associations between upper secondary education GPA and the later risk of any mental disorder are shown in Figure 1 and Supplementary Table S1. No consistent association between upper secondary education GPA and later risk of any mental disorder was found when the analytical models were adjusted for sex, birth year, parental education level, parental income level, parental mental health history, and urbanicity (Models 1 and 2). When GPA at the end of comprehensive education was included as an additional covariate (Model 3), there was a clear gradient between upper secondary education GPA and the risk of any mental disorder, with greater GPA associated with a greater risk of any mental disorder. In the model adjusted for all covariates and estimated within strata of full-sibling sets (Model 4), no consistent association between upper secondary education GPA and later risk of any mental disorder was found.
The association between school achievement (Matriculation Examination GPA) and later risk of developing any mental disorder. Model 1 was adjusted for sex and year of birth. Model 2 was additionally adjusted for parental education level, parental income level, parental mental health history, and urbanicity. Model 3 was additionally adjusted for grade point average at the end of comprehensive education. Model 4 included the same adjustments as Model 3 but was stratified within strata of full-sibling sets. Grade L is the highest and grade A the lowest. Grade C is the reference category.
Figure 1. Long description
A forest plot titled Any mental disorder.
Axes and Legend:
* Y-axis: Labeled Grade, with categories from top to bottom: A, B, C (reference), M, E, and L.
* X-axis: Labeled Hazard ratio with 95% C I s, ranging from 0.9 to 1.3 with a vertical dashed reference line at 1.0.
* Legend: Model 1 (diamond), Model 2 (triangle), Model 3 (circle), and Model 4 (square).
Data Trends:
* Grade A: Models 1 and 2 are centered near 1.0. Model 3 shows a significant decrease (hazard ratio approximately 0.88). Model 4 has a wide confidence interval centered near 1.0.
* Grade B: Models 1 and 2 are near 1.0. Model 3 shows a decrease (hazard ratio approximately 0.93). Model 4 is centered near 1.0.
* Grade C: All models are fixed at the reference value of 1.0.
* Grade M: Models 1 and 2 show a slight increase (hazard ratio approximately 1.02). Model 3 shows a larger increase (hazard ratio approximately 1.09). Model 4 is centered near 1.05.
* Grade E: Models 1 and 2 show a slight increase (hazard ratio approximately 1.02). Model 3 shows a substantial increase (hazard ratio approximately 1.18). Model 4 is centered near 1.05.
* Grade L: Models 1 and 2 show a slight increase (hazard ratio approximately 1.02). Model 3 shows the highest risk (hazard ratio approximately 1.26). Model 4 is centered near 1.04 with a wide confidence interval.
The association between upper secondary education GPA and later risk of substance use disorders, schizophrenia spectrum disorders, bipolar disorder and depression is shown in Figure 2 and Supplementary Table S1. In the covariate-adjusted models (Model 1 and Model 2), there was a clear gradient in the association between upper secondary education GPA and substance use disorders, with higher upper secondary education GPA associated with a lower risk of such disorders. In the model additionally adjusted for the GPA at the end of comprehensive education (Model 3), the associations were almost completely diluted. In the covariate-adjusted sibling model (Model 4), higher upper secondary education GPA – except for average grade A – was associated with a lower risk of substance use disorder. The associations between upper secondary education GPA and later risk of schizophrenia spectrum disorders and bipolar disorder were mixed (Models 1–3). However, in the fully adjusted model estimated within strata of full-sibling sets, above average (reference category C) upper secondary education GPA was associated with a lower risk of schizophrenia spectrum disorders (Model 4). The association between upper secondary education GPA and depression was broadly consistent across all model specifications, with a higher upper secondary education GPA associated with a greater risk of depression.
The association between school achievement (Matriculation Examination GPA) and later risk of developing substance use disorder, schizophrenia spectrum disorder, bipolar disorder and depression. Model 1 was adjusted for sex and year of birth. Model 2 was additionally adjusted for parental education level, parental income level, parental mental health history, and urbanicity. Model 3 was additionally adjusted for grade point average at the end of comprehensive education. Model 4 included the same adjustments as Model 3 but was stratified within strata of full-sibling sets. Grade L is the highest and grade A the lowest. Grade C is the reference category.
Figure 2. Long description
A multi-panel forest plot with four panels arranged in a two-by-two grid. The Y-axis for all panels represents school grades from top to bottom: A, B, C (reference), M, E, and L. The X-axis represents the Hazard ratio with 95 percent C I s, ranging from 0.4 to 1.6, with a vertical dashed reference line at 1.0. A legend on the right identifies four markers: diamonds for Model 1, triangles for Model 2, circles for Model 3, and squares for Model 4.
* Top-left panel (Substance use disorders): Grades A and B generally show hazard ratios above 1.0, while grades M, E, and L show a linear decrease, with grade L reaching the lowest hazard ratio near 0.5 in Model 4.
* Top-right panel (Schizophrenia spectrum disorders): Data points are clustered around the 1.0 reference line with wide horizontal error bars, particularly for grade L in Model 4, which shows a significant drop below 1.0.
* Bottom-left panel (Bipolar disorder): Most data points for grades A, B, M, and E hover near 1.0. Grade L shows higher variability, with Model 3 showing an increase above 1.0 and Model 4 showing a decrease below 1.0.
* Bottom-right panel (Depression): Shows a distinct inverse trend compared to substance use. Grades A and B are below 1.0, while grades M, E, and L show a linear increase in hazard ratio, peaking near 1.5 for grade L in Model 3.
The association between upper secondary education GPA and the later risk of anxiety, eating, sleep, and personality disorders is shown in Figure 3 and Supplementary Table S1. Although in models adjusted for all covariates (Model 3), a higher upper secondary education GPA was associated with a greater risk of anxiety disorder, the association was almost completely diluted in the sibling models (Model 4). Higher upper secondary education GPA was associated with a greater risk of eating disorders in a linear fashion in models adjusted for covariates (Models 1 and 2), and when additionally adjusting for GPA at the end of comprehensive education (Model 3). In the sibling model, these associations were to some degree diluted (Model 4). The associations between school achievement and sleep disorders were mixed, with above-average upper secondary education GPA associated with a smaller risk of sleep disorders in the covariate-adjusted models (Models 1, 2 and 3), but not consistently in the sibling-adjusted model (Model 4). Higher upper secondary education GPA was consistently associated with a greater risk of personality disorders in models adjusted for covariates (Models 1 and 2) and when additionally adjusting for GPA at the end of comprehensive education (Model 3). In the sibling models, these associations were mostly diluted.
The association between school achievement (Matriculation Examination GPA) and later risk of developing anxiety disorder, eating disorder, sleep disorder and personality disorder. Model 1 was adjusted for sex and year of birth. Model 2 was additionally adjusted for parental education level, parental income level, parental mental health history, and urbanicity. Model 3 was additionally adjusted for grade point average at the end of comprehensive education. Model 4 included the same adjustments as Model 3 but was stratified within strata of full-sibling sets. Grade L is the highest and grade A the lowest. Grade C is the reference category.
Figure 3. Long description
A four-panel forest plot grid. The x-axis for all panels represents Hazard ratio with 95 percent C Is, ranging from 0.5 to 2.5 with a vertical dashed reference line at 1.0. The y-axis lists school grades from top to bottom: A, B, C, M, E, and L. A legend on the right identifies four symbols: diamond for Model 1, triangle for Model 2, circle for Model 3, and square for Model 4.
* Top-left panel: Anxiety disorders. Data points for all grades cluster closely around the 1.0 reference line, showing minimal variation across models, though grade L shows a slight increase in hazard ratio for Model 3.
* Top-right panel: Eating disorders. Grades A and B show hazard ratios below 1.0. Grade C is the reference at 1.0. Grades M, E, and L show a significant linear increase in hazard ratio, with grade L reaching approximately 2.2 for Model 3. Model 4 squares consistently show wider confidence intervals and lower hazard ratios than Model 3 circles in the lower grades.
* Bottom-left panel: Sleep disorders. Grades A and B are slightly above 1.0. Grades M, E, and L show a slight downward trend below 1.0, particularly in Model 4 for grade L.
* Bottom-right panel: Personality disorders. Grades A and B are below 1.0. Grades M, E, and L show an increasing trend above 1.0. Grade L shows the highest risk, with Model 3 peaking near 2.0 and Model 4 showing a lower estimate with a very wide confidence interval.
The associations between GPA at the end of upper secondary education and the subsequent risk of any mental disorder and specific mental disorders across strata of comprehensive school GPA are shown in Table 2 and in Supplementary Figures S2–S10. Among persons who had lower than average GPA at the end of comprehensive education, the lowest and second lowest grade at the end of upper secondary education were associated with a lower risk of any mental disorder during the follow-up, whereas the above average grades at the end of upper secondary education were associated with a greater risk of any mental disorder during the follow-up. Among the specific mental disorders, reduction in school achievement between comprehensive education and upper secondary was associated with a greater risk of especially schizophrenia spectrum disorders. For example, higher than average GPA at the end of comprehensive education and then lowest or second lowest GPA at the end of upper secondary education were associated with a greater risk of schizophrenia spectrum disorder diagnosis. Conversely, individuals with lower than the average GPA at the end of comprehensive education who later had third or second highest grade at the end of upper secondary education had a greater risk of bipolar disorder, eating disorders, and personality disorders. Many estimates at the highest and lowest upper secondary GPA levels could not be estimated because of the small number of individuals.
The association between school achievement (Matriculation Examination GPA) at the end of upper secondary education and the later risk of any mental disorder and specific mental disorders across GPA groups at the end of comprehensive education
Table 2. Long description
The table presents Hazard Ratios (H R) and 95 percent Confidence Intervals (C I) for various mental health diagnoses. The columns are divided into three Comprehensive Education G P A groups: S D less than minus 1, minus 1 less than or equal to S D less than or equal to 1, and S D greater than 1. Within each diagnosis, rows represent Matriculation Examination grades from A (lowest) to L (highest), with C as the reference group.
* Any mental disorder: H R values range from 0.87 for grade A in the lowest G P A group to 1.55 for grade L in the same group. Higher grades generally show higher H R across all G P A groups.
* Substance use disorders: Shows a reverse trend in the highest G P A group, where higher grades (M, E, L) have lower H R (0.71, 0.62, 0.65) compared to the reference.
* Schizophrenia spectrum disorders: Notable outlier for grade A in the highest G P A group with an H R of 8.28 (C I 3.06 to 22.42).
* Bipolar disorder: Grade E in the lowest G P A group shows a high H R of 2.56.
* Depression: H R increases with grade level across all G P A groups, peaking at 1.57 for grade E in the lowest G P A group.
* Anxiety disorders: Similar to depression, H R increases with grade, reaching 1.83 for grade L in the lowest G P A group.
* Eating disorders: Shows strong positive correlation with higher grades, particularly grade E (H R 2.68) and grade L (H R 2.49) in the first two G P A groups.
* Sleep disorders: Most H R values remain close to 1.00, showing less variation by grade.
* Personality disorders: Higher grades show significantly increased risk, with grade L reaching an H R of 3.08 in the lowest G P A group and 1.73 in the middle group.
Note: Some associations could not be estimated as there were too few persons in the category. Models were adjusted for sex, year of birth, parental education level, parental income level, parental mental health history, and urbanicity. Grade L is the highest and grade A the lowest.
Discussion
Using nationwide Finnish register data covering over half a million young adults, we found that associations between school achievement at the end of upper secondary education and mental disorders are largely disorder-specific. While better upper secondary school achievement was rather consistently associated with a smaller risk of developing substance use disorders – and to some extent schizophrenia-spectrum disorders – it was also associated with a greater risk of developing depression, anxiety, eating, and personality disorders. Results from the sibling analyses suggested that most of the associations, but not all, were to some degree attributable to unobserved shared family characteristics. Lastly, reduction in school achievement from comprehensive education to upper secondary education was associated with a considerably greater risk of schizophrenia spectrum disorders, whereas improvement in school achievement was associated with a greater risk of bipolar, eating, and personality disorders.
The present study findings are mostly in accordance with results from previous studies where lower school achievement in comprehensive education has been associated with a greater risk of substance use disorders (Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2017, Reference Kendler, Ohlsson, Keefe, Sundquist and Sundquist2018; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023). Whereas previous studies have typically shown that lower school achievement in comprehensive education is associated with a greater risk of schizophrenia spectrum disorders (Kendler et al., Reference Kendler, Ohlsson, Mezuk, Sundquist and Sundquist2016b; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023), in the present study only above average school achievement in upper secondary education was associated with a lower risk of schizophrenia spectrum disorders, and thus no dose–response association was found. On the contrary, whereas previous studies have also linked lower school achievement to a greater risk of depression and anxiety disorders (Gyllenberg et al., Reference Gyllenberg, Ristikari, Kelleher, Kääriälä and Gissler2022; Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2017, Reference Kendler, Ohlsson, Keefe, Sundquist and Sundquist2018; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023), in the present study better school achievement was associated with a greater risk of depression and anxiety disorders, although this association was to a large degree diluted in the siblings analyses.
We also found that reduction in school achievement from comprehensive education to upper secondary general education was associated with a greater risk of schizophrenia spectrum disorders. Although this topic has been examined in some previous studies (Conejeros-Pavez et al., Reference Conejeros-Pavez, Vasquez, Diaz, Mena, Undurraga, Gonzalez-Valderrama, Claro, Undurraga and Crossley2025; Kahn & Keefe, Reference Kahn and Keefe2013; Reichenberg & Kahn, Reference Reichenberg and Kahn2026), to our knowledge, the present study is the first large-scale population-based study to examine this question. Our results support the conclusion of a meta-analytic review, according to which poor academic achievement and cognitive performance early in life were associated with a greater risk of schizophrenia (Dickson et al., Reference Dickson, Hedges, Ma, Cullen, MacCabe, Kempton, Downs and Laurens2020), highlighting lower cognitive function as a potential risk factor for schizophrenia development (Owen & O’Donovan, Reference Owen and O’Donovan2024). However, it is worth noticing that sub-threshold prodromal symptoms (Striebel, Reference Striebel2025), including early cognitive decline, social withdrawal, and attenuated psychotic experiences, could plausibly contribute to the GPA reduction we observe prior to a schizophrenia spectrum disorder diagnosis (Harvey et al., Reference Harvey, Bosia, Cavallaro, Howes, Kahn, Leucht, Müller, Penadés and Vita2022). This would represent a form of reverse causation, operating through early expression of disease rather than through a causal pathway from changes in GPA to disorder onset.
Previous studies have provided partially inconclusive evidence regarding the association between school achievement and bipolar disorder. While lower school achievement in comprehensive education has generally been linked to a greater risk of bipolar disorders (Kendler et al., Reference Kendler, Ohlsson, Mezuk, Sundquist and Sundquist2016b; Weckström et al., Reference Weckström, Elovainio, Pulkki-Råback, Suokas, Komulainen, Mullola, Böckerman and Hakulinen2023), some studies have also suggested that excellent school performance in comprehensive education is associated with a greater risk (MacCabe et al., Reference MacCabe, Lambe, Cnattingius, Sham, David, Reichenberg, Murray and Hultman2010; Pedersen, Østergaard, & Petersen, Reference Pedersen, Østergaard and Petersen2018). In our study, a significant improvement in school performance was associated with a greater risk of bipolar disorder. To our knowledge, this has not been previously reported. There are a number of mechanisms that could potentially explain this finding. First, hypomanic symptoms, including elevated energy, reduced need for sleep, increased motivation, and heightened goal-directed behavior, may plausibly improve academic performance in the years preceding a bipolar disorder diagnosis. This would be consistent with a prodromal interpretation in which GPA improvement reflects an early expression of bipolar disorder rather than a causal risk factor, suggesting that the observed association may be explained by reverse causation. Second, previous studies have documented considerable overlap between creativity and bipolar disorder (Baas, Nijstad, Boot, & De Dreu, Reference Baas, Nijstad, Boot and De Dreu2016), which appears to be partially genetic (Greenwood, Reference Greenwood2020), indicating that the association may be explained by better performance in subjects that require creativity. Lastly, given that we observed similar associations also for eating disorders and personality disorders, these findings might not be diagnosis-specific and warrant further investigation and replications in different settings.
Several mechanisms may potentially explain our findings. First, school performance in upper secondary education can be seen as a complex phenotype reflecting distinct underlying personality configurations and neurodevelopmental trajectories. It has a considerable overlap with intelligence (Madigan, Reference Madigan2019), and low intelligence has been associated with a greater risk of mental disorders (Fries et al., Reference Fries, Oberleiter, Bodensteiner, Fries and Pietschnig2025). At the same time, school performance also captures certain aspects of personality, such as perfectionism (Madigan, Reference Madigan2019), high conscientiousness (Mammadov, Reference Mammadov2022), and motivation (Jansen, Meyer, Wigfield, & Möller, Reference Jansen, Meyer, Wigfield and Möller2022). While high conscientiousness is associated with better mental health (Hakulinen et al., Reference Hakulinen, Elovainio, Pulkki-Råback, Virtanen, Kivimäki and Jokela2015; Kotov, Gamez, Schmidt, & Watson, Reference Kotov, Gamez, Schmidt and Watson2010), high perfectionism is linked to a greater risk of several mental disorders (Limburg, Watson, Hagger, & Egan, Reference Limburg, Watson, Hagger and Egan2017), and specifically to eating disorders (Bills, Greene, Stackpole, & Egan, Reference Bills, Greene, Stackpole and Egan2023). Taken together, these factors are consistent with our disorder-specific findings: traits associated with a lower risk of substance use disorders, such as high levels of perfectionism, may be associated with a higher risk of depression, anxiety, and eating disorders. Second, from the covariates included in our statistical models, parental education level, parental income level, parental mental health history, and the degree of urbanicity did not dilute the associations, indicating that these factors did not substantially attenuate the associations. However, the inclusion of school achievement at the end of comprehensive education significantly changed the results, suggesting that prior achievement partially accounts for the observed associations between later school achievement and mental health. Third, most of the associations, but not all, were attenuated in the sibling analyses, suggesting that shared family and genetic characteristics may account for part of the observed association between school achievement in upper secondary education and later mental health.
The main strength of this study was the use of high-quality register data that covered the entire Finnish population and included accurate details of school performance and mental disorders from both primary and secondary care. We were also able to account for major socioeconomic and intergenerational confounders, as well as unobserved shared family characteristics that can affect both school achievement and the subsequent incidence of mental disorders. Our study also has important limitations. First, we excluded individuals with mental disorder diagnoses prior to the onset of the follow-up period. Although this approach is standard for mitigating the concern for reverse causation, i.e., mental disorders influencing later school achievement (Li et al., Reference Li, Niemi, Gutvilig, Elovainio, Flykt and Hakulinen2026), it means that our results are not generalizable to those individuals who were diagnosed before the end of upper secondary education. Second, the possibility of unmeasured or residual confounding attributable to non-shared environmental factors or genetic influences cannot be excluded. Given that full siblings exhibit approximately 50% genetic concordance, confounding due to genetic influences remains possible. Similarly, exposure to non-shared environmental factors, such as traumatic life events, or individual-level factors, such as substance misuse, may both disrupt school achievement and increase psychiatric vulnerability (Daníelsdóttir et al., Reference Daníelsdóttir, Aspelund, Shen, Halldorsdottir, Jakobsdóttir, Song, Lu, Kuja-Halkola, Larsson, Fall, Magnusson, Fang, Bergstedt and Valdimarsdóttir2024; Hammerton et al., Reference Hammerton, Lewis, Heron, Fernandes, Hickman and Lewis2023; Perfect et al., Reference Perfect, Turley, Carlson, Yohanna and Saint Gilles2016; Thurston et al., Reference Thurston, Murray, Franchino-Olsen, Silima, Hemady and Meinck2025). Given the wide confidence intervals in some of the analyses, it is also possible that sibling analyses and comprehensive school GPA stratified analyses were in some cases statistically underpowered. Third, many of the observed associations are modest in magnitude, and the findings should therefore be interpreted cautiously with regard to their direct clinical utility. However, because GPA is unlikely to be used as a standalone screening tool, our results are better understood as large-scale evidence contributing to understanding how educational performance relates to mental health. Finally, in Finland, both education and health care are publicly provided and universally accessible, with minimal financial barriers. Consequently, the present findings may not generalize to high-income countries with different institutional settings.
Conclusion
Our results suggest that better school achievement in upper secondary education is associated with both higher and lower risk of specific mental health disorders. Reduction in GPA between comprehensive and upper secondary school was associated with a greater risk of schizophrenia spectrum disorders and improvement in school achievement with a greater risk of bipolar and personality disorders. These findings highlight the disorder-specific associations between school achievement in upper secondary education and subsequent mental disorders. Further large-scale studies are needed to evaluate our results in different populations.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291726105637.
Funding statement
This work was supported by the Research Council of Finland (354237 to C.H.; 370659 to K.K.) and the European Union (ERC, MENTALNET, 101040247). Views and opinions expressed are, however, those of the authors only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
Competing interests
The authors declare none.
Open access
