Several national registers containing demographic, health, and educational information were linked using the unique identification number assigned to all Finnish citizens. The study population consisted of persons born in Finland between 1980 and 1999 and who had two Finnish parents. Exclusion criteria from the sample were death, emigration, or a mental disorder diagnosis prior to the follow-up period. In the original data, 2.3% of the study population (N = 26,795) had missing data on education. The missing data could imply school drop-out, but also schooling abroad, as well as serious somatic illness, mental disability, or other significant barriers to finishing compulsory school. Due to the heterogeneity of the group without register-based educational information, we excluded them from the analyses. The remaining 1,070,880 individuals formed our final analytic sample and were followed from 1 August in the year they graduated from the nine-year compulsory education (around the age of 15 or 16) until the first diagnosed mental disorder, death, emigration, or the end of follow-up on 31 December 2017, whichever occurred first. Thus, those born in 1980 were followed from 1 August 1996 until 31 December 2017, and those born in 1999 were followed from 1 August 2015 until 31 December 2017 at the latest. The ethics committee of the National Institute of Health and Welfare (THL/22/6.02.01/2019) approved the study. Data were linked with the permission of the Statistics Finland (TK-53-1696-16) and the National Institute of Health and Welfare.

In Finland, education is free-of-charge from preschool to higher academic education. At the age of seven, children begin a nine-year compulsory comprehensive school, where the curriculum and grading are based on national guidelines that all schools are obliged to follow. After the ninth grade, children apply to secondary education such as a three-year general upper secondary education or vocational education and training.

The school achievement measure was the mean of the final ninth year grades from comprehensive school. The original grades were on a scale ranging from 4 to 10, where higher numbers indicate better school achievement (4 = fail, 5–6 = poor, 7–8 = good, 9–10 = excellent). The mean was standardized using a z-score transformation (mean 0, SD 1), and school achievement was also analyzed as a five-level categorical variable: −2SD (≤−2SD); −1SD (>−2SD – ≤−1SD); 0 (>−1SD – <1SD); 1SD (≥1SD – <2SD); 2SD (≥2SD).

The grades were obtained from the National Joint Application Register containing the information on school grades of students in the final year of comprehensive school between 1996 to 2015. Information on school grades from the years 1996–2007 was available only for those who applied for secondary education. From 2008 to 2015, information on school grades was available for all students in the final year of compulsory school attendance, including those who did not apply for secondary education. The school achievement measure was calculated based on all subjects including compulsory (e.g., mother tongue and literature (Finnish or Swedish), the first foreign language, mathematics, history and social studies, religion/ethics, physical education, mathematics, music, and visual arts) and optional (e.g., the second foreign language) subjects. To account for variation in the data collection method and possible differences in grading between years, the school achievement measures were standardized within each year of graduation.

Diagnoses of mental disorders were obtained from the Finnish National Hospital Discharge Register, which contains data on virtually all inpatient visits since 1970 and outpatient visits in special healthcare since 1998 and includes the ICD-10 diagnostic classification (or previous editions of the manual according to the year of visit) for the reason for each visit.

The first inpatient or outpatient secondary care contacts with any mental disorder (ICD-10 diagnoses F00–F99) was the main outcome. In addition, the following eight diagnostic sub-categories were analyzed: 1) mental and behavioral disorders due to psychoactive substance use (F10–F19), 2) schizophrenia spectrum disorders including schizophrenia, schizotypal, and delusional disorders (F20–F29), 3) bipolar disorder (F30–F31), 4) depression (F32–F33), 5) neurotic, stress-related, and somatoform disorders (F40–F48), 6) eating disorders (F50–F50.9), 7) nonorganic sleep disorders (F51), and 8) disorders of adult personality and behavior (F60–F69).

Demographic, socioeconomic, and intergenerational covariates included in the analyses were sex (0 = male, 1 = female), birth year, degree of urbanicity in residential location (0 = missing, 1 = urban, 2 = semi-urban, 3 = rural), parental education level at the time of their child’s graduation (0 = missing, 1 = comprehensive, 2 = upper secondary, 3 = lower tertiary, 4 = Bachelor’s or equivalent, 5 = Master’s or higher), parental income level in quintiles at time of their child’s graduation (calculated in relation to the study population) (0 = missing, 1 = 1st quintile, 2 = 2nd quintile, 3 = 3rd quintile, 4 = 4th quintile, 5 = 5th quintile), and parental mental health history (0=no psychiatric diagnosis, 1=any psychiatric diagnosis before the follow-up period). Sample characteristics are described in Supplementary Table S1.

The study population was followed from 1st August in the year that they graduated (around age 16) until the first psychiatric diagnosis, death, emigration, or the end of follow-up on 31th December 2017, whichever came first. We used Cox proportional hazard models to estimate the association between school grades at the age of 16 with the later risk of first diagnosed mental disorder. Model 1 evaluated the associations while adjusting for sex, the year of birth and time-varying calendar year period as covariates; Model 2 was adjusted for all covariates, i.e., sex, the year of birth, time-varying calendar year period, parental education level, parental income level, parental mental health history, and urbanicity; Model 3 was adjusted for all covariates and it was estimated using stratified Cox proportional hazard models conducted within strata of full-sibling sets, which enabled us to account for otherwise unobserved confounding by shared family characteristics that could affect both school achievement and the incidence of mental disorders [20]. We used the standardized school grade mean (>−1SD – <1SD) as the reference group for the estimation of the hazard ratios (HRs).

The cumulative incidences were estimated using competing risks regression [19], based on Fine and Gray’s proportional subhazards model, treating death and emigration as competing events. Psychiatric comorbidity was calculated as the count of separate mental disorders with which a person was diagnosed during the follow-up period. The associations of school achievement with comorbidity were assessed with multinomial logistic regression, using those without any mental disorders as the reference category. The analyses were conducted between January 2022 and March 2023, using Stata version 16.1 [21].