Transmission of Mental Disorders in Adolescent Peer Networks

Key Points Question Is having peers with a mental disorder in the same social network during adolescence associated with later risk of mental disorder? Findings In this cohort study including more than 700 000 individuals in Finland, analysis of nationwide, interlinked registry data found that having classmates diagnosed with a mental disorder in the ninth grade of comprehensive school was associated with increased risk of receiving a mental disorder diagnosis later in life. Increased risk remained after adjusting for an array of parental, school-level, and area-level confounders. Meaning The findings of this study suggest that mental disorders might be socially transmitted within adolescent peer networks.

Number of any of the examined mental disorder diagnoses (ICD-10 F10-F50 or F90-F98) before follow-up (FU) shown for each year of the study period (i.e., years when the cohort members were on the ninth grade) together with the number of school classes without and with diagnosed cohort members (unexposed and exposed classes) and number of cohort members without and with diagnosed classmates (unexposed and exposed individuals).
eTable 2: Additional descriptive statistics of the study population.Incidence rates (IRs), hazard ratios (HRs) with 95% confidence intervals, and P-values shown for the associations in four follow-up time windows.IR was calculated per 100 000 person-years at risk.For the calculation of the HRs, the Cox models were adjusted for sex, birth year, school class size, school's ninth grade size, area-level urbanicity, area-level morbidity, area-level education level, area-level employment rate, parental education, parental income, and parental mental health, with a random intercept per school.Results of analyses with less than or equal to three participants are not reported due to data protection regulations.See also eFigure 2.
eFigure 2: Diagnosis-specific associations between having ninth grade classmates with a mental disorder diagnosis and later risk of being diagnosed with a mental disorder.
Hazard ratios with 95% confidence intervals shown for substance misuse, schizophrenia spectrum, eating, and behavioral and emotional disorders in four follow-up time windows.The arrow indicates that the confidence interval extends outside the range of the graph.The Cox models were adjusted for sex, birth year, school class size, school's ninth grade size, area-level urbanicity, area-level morbidity, area-level education level, arealevel employment rate, parental education, parental income, and parental mental health, with a random intercept per school.Note that the number of individuals who received a schizophrenia spectrum diagnosis during follow-up and who had more than one classmate with a schizophrenia spectrum diagnosis was too low to be reported due to data protection regulations (see eTable 2).For the diagnosis-specific associations for mood, anxiety, internalizing, and externalizing disorders, see Figure 2 of the main text.
eFigure 3: Diagnosis-specific associations between having ninth grade classmates with a mental disorder diagnosis and later risk of being diagnosed with a mental disorder using binary exposure.
Hazard ratios with 95% confidence intervals shown for all the examined diagnostic categories in four follow-up time windows.The arrow indicates that the confidence interval extends outside the range of the graph.The Cox models were adjusted for sex, birth year, school class size, school's ninth grade size, area-level urbanicity, area-level morbidity, area-level education level, area-level employment rate, parental education, parental income, and parental mental health, with a random intercept per school.Hazard ratios (and 95% confidence intervals) shown for the first year of follow-up and entire follow-up separately.
Model 1: adjusted for sex and birth year.
Model 2: adjusted for sex and birth year, with a random intercept per school.
Model 3: adjusted for sex, birth year, school class size, and school's ninth grade size, with a random intercept per school.
Model 4: adjusted for sex, birth year, parental education, parental income, and parental mental health, with a random intercept per school.
Model 5: adjusted for sex, birth year, area-level urbanicity, area-level morbidity, area-level education level, and area-level employment rate, with a random intercept per school.
Model 6: adjusted for sex, birth year, school class size, school grade size, parental education, parental income, parental mental health, area-level urbanicity, area-level morbidity, area-level education level, and area-level employment rate, with a random intercept per school.

specific associations between having ninth grade classmates with a mental disorder diagnosis and later risk of being diagnosed with a mental disorder.
eFigure 1: Schoenfeld residuals for the diagnosis categories.The graph depicts the combined coefficients of the exposure variable (i.e., one or more diagnosed classmates) as a function of follow-up time.P value from Schoenfeld test is shown in each plot (P < 0.05 indicates a failure to meet the proportional hazards assumption).The Cox models were adjusted for sex, birth year, school class size, school's ninth grade size, area-level urbanicity, area-level morbidity, area-level education level, arealevel employment rate, parental education, parental income, and parental mental health, with a random intercept per school.eTable3: Diagnosis-

eTable 4: Sensitivity analyses on the associations between having diagnosed ninth grade classmates and later risk of being diagnosed with a mental disorder. a Diagnoses in class (exposure) None (ref) Yes
Hazard ratios (HRs) with 95% confidence intervals (CIs) and P-values shown for the associations using binary exposure and limiting exposure diagnoses to three years preceding start of follow-up.Results are shown for the entire follow-up and separately in four shorter time windows.The Cox models were adjusted for sex, birth year, school class size, school's ninth grade size, area-level urbanicity, area-level morbidity, area-level education level, area-level employment rate, parental education, parental income, and parental mental health, with a random intercept per school.
a: Hazard ratios (HRs) with 95% confidence intervals (CIs) and P-values shown for the associations using binary exposure.Results are shown for the entire follow-up and separately in four shorter time windows.The Cox models were adjusted for sex, birth year, school class size, school's ninth grade size, area-level urbanicity, area-level morbidity, area-level education level, area-level employment rate, parental education, parental income, and parental mental health, with a random intercept per school.b: © 2024 .Alho J et al.JAMA Psychiatry.