Depressive Symptoms in Adolescence and Young Adulthood

This cohort study examines the persistence of depressive symptoms and other mental health issues from age 18 years to age 21 to 22 years.


Introduction
Depressive symptoms have been increasing among US adolescents since approximately 2010.
National estimates of major depressive episode prevalence among adolescents increased from 8.1% in 2009 to 15.9% in 2019, with increases concentrated among females and ranging from 11.4% to 23.4%. 1 Approximately 1 in 4 adolescent females reported seriously considering suicide and making a plan for suicide in 2021, 2 an increase from approximately 14% and 11%, respectively, in 2009. 3spitalizations for intentional self-harm 4 and fatal suicide 5,6 also increased during the same time frame.][10][11] Together, these patterns have been labeled as a teen mental health crisis in the US.
An open question is the extent to which increases in adolescent depressive symptoms in recent birth cohorts are associated with subsequent increases in young adult mental health problems as those cohorts transition into adulthood.In a given person, depressive symptoms at 1 developmental time point are typically associated with depressive symptoms at subsequent time points 12,13 ; thus, the population-wide increase in the prevalence of adolescent depressive symptoms will, based on historical data, affect later developmental time points, increasing the overall population burden in young adulthood.Yet the rapid increases in depressive symptoms in the past decade have been unprecedented; thus, whether these symptoms tend to resolve more frequently than in previous birth cohorts is unknown.
A relevant epidemiological question can be framed around the interaction between birth cohort and adolescent depressive symptoms.If adolescent depressive symptoms are developmentally limited (ie, more likely to resolve before adulthood than in previous birth cohorts), then we could expect an inverse interaction between birth cohort and these symptoms.That is, if there is an inverse interaction, then the association between adolescent and young adult depressive symptoms would have lower magnitude in recent birth cohorts than in previous birth cohorts.If there is no interaction between birth cohort and adolescent depressive symptoms, then these symptoms would be as much of a factor in young adult symptoms in recent birth cohorts as in previous birth cohorts.In the context of the rising prevalence of adolescent depressive symptoms, no interaction would portend a greater burden of young adult symptoms in recent years.Another potential scenario is a direct interaction, whereby the association between adolescent and young adult depressive symptoms increases in magnitude in recent birth cohorts.A direct interaction would imply that an increased prevalence of adolescent depressive symptoms makes these symptoms more persistent into young adulthood in recent than in previous birth cohorts, suggesting a higher mental health burden in this population and warranting stronger interventions.
In the present study, we used nationally representative longitudinal data collected from adolescence through young adulthood across cohorts spanning the past 30 years to test the association between birth cohort and adolescent depressive symptoms at ages 18, 19 to 20, and 21 to 22 years and changes in these symptoms by cohort.We examined the extent to which increases in depressive symptoms and other measures of mental well-being in adolescence persist into young adulthood and whether this persistence changes across birth cohorts given the dynamic temporal patterns in recent years.

Methods
Data for analysis were drawn from the Monitoring the Future (MTF) study, a longitudinal panel cohort study wherein a new prospective cohort began to be followed every year starting at age 18 years, selected from a nationally representative sample of high school seniors.The University of Michigan Institutional Review Board (IRB) approved the study, and the Columbia University IRB approved the analysis protocols.Passive (letter sent to parents/caregivers) or active (written) informed consent was obtained, per school policy, at baseline from parents of participants younger than 18 years or The pool of eligible respondents included those who participated in nationally representative samples of approximately 15 000 US high school seniors (12th grade), and we included data from those surveyed annually since 1990. 14Among that pool, approximately 2450 students were randomly selected each year for longitudinal follow-up, with oversampling for students who reported drug use (statistical weights in analysis adjusted for oversampling). 15Students selected for follow-up were randomly assigned to begin assessments either We stratified by 5-year cohorts.Sample sizes by birth cohort and follow-up wave are provided in eTable 1 in Supplement 1.We included attrition weighting in all analyses that calculated individual weights as the inverse of the probability of participation based on time-invariant characteristics measured at 12th grade. 15,16Previous simulation analyses indicated that, while attrition varies by cohort, there is minimal resulting bias. 17

Measures
We examined 4 separate mental health scales: depressive affect, self-derogation, low self-esteem, and loneliness.Four items each were used to measure depressive affect, self-derogation, and low self-esteem, while 3 items were used to measure loneliness based on the extent to which respondents agreed or disagreed with certain statements.Example statements were as follows: "Life often seems meaningless" (depressive affect), "I take a positive attitude toward myself" (low selfesteem when reverse-coded), "I feel I do not have much to be proud of" (self-derogation), and "A lot of times I feel lonely" (loneliness).Response options for each item ranged from 1 (disagree) to 5 (agree).Items for each scale were summed to yield scores ranging from 4 to 20 (or from 3 to 15 for the loneliness scale), with higher scores indicating higher depressive affect, self-derogation, and loneliness, whereas a lower score on the self-esteem scale indicated worse self-esteem.9][20] These scales have been operationalized and implemented similarly in other analyses of MTF data. 21,22e to symptom score non-normality, we created binary indicators.

JAMA Network Open | Psychiatry
Depressive Symptoms in Adolescence and Young Adulthood

Covariates
Due to potential confounding, participants were queried about their sex (male vs female), race and ethnicity, and parental educational level at baseline (at least 1 parent was a college graduate vs no college).Until 2005, respondents selected 1 racial and ethnic category, including American Indian (Native American Indian), Asian American, Black or African American, Cuban American, Mexican American or Chicano, Other Latin American, Puerto Rican, White or Caucasian, or other.From 2005 onward, respondents could select multiple categories from the following: American Indian or Alaska Native, Asian American, Black or African American, Cuban American, Mexican American or Chicano, Native Hawaiian or Other Pacific Islander, Other Hispanic or Latino, Puerto Rican, or White.Across years, we harmonized the categories as follows: American Indian or Alaska Native, Asian or Pacific Islander, Black, Hispanic or Latino, non-Hispanic multiracial (based on reporting multiple racial identities), and White.For respondents surveyed before 2005, we also categorized as other those who selected that option.To account for potential response differences due to the COVID-19 pandemic, we included a dichotomous variable for data collection in 2020 vs all other years.

Statistical Analysis
Analyses included descriptive graphs and logistic regression models conducted from April to October 2023 using survey procedures in SAS 9.4 (SAS Institute Inc).Descriptive prevalence estimates included sampling weights.Logistic regression models estimated the association between mental health at age 18 years and mental health at ages 19 to 20 and 21 to 22 years, incorporating attrition weights and controlling for race and ethnicity, parental educational level, birth cohort, and an indicator for data collection in 2020 vs previously.Interactions with birth cohort and sex were also tested.Two-sided P < .05indicated statistical significance.
Additionally, we estimated the population attributable fraction (PAF), which measures the total proportion of outcomes that can be attributed to the exposure and is calculated as follows: where p is the exposure prevalence, and RR (relative risk) is the relationship between the exposure and outcome. 23In this study, the exposure was depressive symptoms at age 18 years, and the outcomes were depressive symptoms at ages 19 to 20 and 21 to 22 years.
Population attributable fractions are community impact measures that change based on both the prevalence of exposure or independent variables and the magnitude of the association between exposure or independent variables and outcomes.The PAF increases as exposure prevalence increases, holding RR constant, and increases as RR increases, holding prevalence constant.

Prevalence of High Depressive Symptoms and Change in Depressive Symptoms
Of the 36 552 respondents who were followed-up longitudinally, 18 597 (50.5%) were females and 17 687 (48.8%) were males; 1.1% identified as American Indian or Alaska Native, 3.3% as Asian or Pacific Islander, 10.5% as Black, 14.0% as Hispanic or Latino, 66.0% as White, 1.8% as non-Hispanic multiracial, and 1.7% as other race and ethnicity; and 44.8% had a parent who graduated from college (eTable 2 in Supplement 1).
Increases among recent birth cohorts were also evident across other domains of mental health and well-being (eTable 3 in Supplement 1), including self-derogation, low self-esteem, and (to a lesser extent) loneliness.

Longitudinal Associations and Interactions
Table 2 shows the overall longitudinal associations of depressive symptoms across follow-up waves.
Among males, those with high depressive symptoms at baseline had 10.

Depressive Symptom Prevalence and PAFs
Figure 2 shows follow-up depressive symptoms stratified by whether adolescents had high depressive symptoms at baseline.Unadjusted regression models did not find significant interactions between baseline depressive symptoms and sex (eTable 5 in Supplement 1); thus, Figure 2 grouped males and females together.Results indicated that the presence of high depressive symptoms at each follow-up were concentrated among those with high depressive symptoms at age 18 years.By the most recent birth cohort, among those with high depressive symptoms at baseline,

Discussion
Depressive symptoms and other indicators of mental health and well-being are increasing in US adolescents and young adults.The magnitude of the association between having depressive symptoms at age 18 years and having depressive symptoms at ages 19 to 20 and 21 to 22 years has remained stable across birth cohorts, yet prevalence across all ages is increasing by cohort.Thus, the increases in youth mental health problems in the US are shifting to young adults as populations age.
By the most recent birth cohort, nearly half of high depressive symptoms at age 21 to 22 years were associated with high symptoms in adolescence; thus, early intervention and prevention are critical to reduce the burden on both adolescent and adult populations.
0][11] Ongoing studies often follow cohorts of similar age at baseline across time, and thus a broader historical lens is often not possible with existing cohorts.Data from other national longitudinal sources, such as the National Longitudinal Study of Adolescent to Adult Health, have shown that depressive symptoms typically decline after adolescence and into young adulthood 24 and then accelerate, particularly for females, at the start of midlife. 24Data from the present study were generally consistent with this pattern, especially for females, revealing decreases in depressive symptoms after age 18 years.However, this study adds to the literature by showing a substantial cohort variation in depressive symptom trajectory, with mean levels of depressive symptoms increasing in the most recent birth cohorts, across the age span of adolescence to young adulthood.
The PAFs indicate that, across all cohorts, adolescent depressive symptoms explain a substantial proportion of young adult depressive symptoms and that the proportions explained appear to be increasing for females.The PAF is an important component of evaluating the implications of increasing depressive symptoms by cohort 25,26 because it has 2 central components: magnitude of associations and prevalence.It increases as prevalence increases, even when the relationship between exposure and outcome remains steady.Given that we did not observe an interaction between birth cohort and depressive symptoms at age 18 years, the reason that the PAF increased for females was that prevalence at age 18 years was increasing.Thus, without considering PAFs, we might erroneously conclude that the public health burden of these symptoms is remaining stable for females since the magnitude of association is not changing.But the public health burden is increasing for females because the exposure prevalence is increasing, even while the magnitude of the relationship is stable across cohorts.In other words, in the case of adolescent to adult depressive symptoms across birth cohorts, the public health burden for females is increasing due to the increase in depressive symptoms at age 18 years.
Increases in adolescent, and now young adult, mental health symptoms indicate a growing need for additional infrastructure for treatment and services.Available data indicate that there are substantial gaps in availability of trained specialist mental health practitioners across the US 27 as well as access challenges even in areas in which practitioners are concentrated.Psychiatrists are less likely to accept insurance than other clinicians, 28 creating financial barriers for most US families.Financial barriers are among the most frequently cited reasons for unmet mental health care needs among people with psychiatric distress. 29Medicaid remains an important payer of mental health services in the US for individuals eligible for coverage, but from 2010 to 2015-a period of the Affordable Care Act implementation and Medicaid expansion in many states-the proportion of psychiatrists who accepted Medicaid decreased. 30Primary care physicians and pediatricians are increasingly a central source of mental health care for US adolescents and adults, 31  for adolescents to learn evidence-based stress-coping strategies are also critical to improving population mental health. 33e underlying factors associated with increases in depressive symptoms among adolescents remain under investigation.Substantial resources have been placed in identifying the role of digital technology and social media in accelerating mental health problems among youth, although thus far meta-analyses and systematic reviews have generally found small to null effect sizes, [34][35][36][37][38] although measurement challenges in identifying appropriate exposure constructs make it difficult to identify both the risks and benefits of digital and social media use. 39Other hypotheses have included interactions between historical patterns in biological risk factors (eg, earlier maturation and puberty) 40 and social risk factors (eg, political polarization and environmental instability limiting future outlooks). 41,42As new cohorts of adolescents emerge and existing cohorts transition into adulthood, it is unlikely that the value of these novel social factors will abate; thus, without investment into prevention and intervention, further acceleration of mental health problems may be expected.

Limitations
This study is subject to limitations.Depressive symptoms were self-reported rather than ascertained using diagnostic instruments or clinical assessment.While self-reported depressive symptoms are typically reliable indicators of psychological distress, these measures have not been validated with clinical cut points for probable depression, and thus we do not know the extent to which the patterns represented here would generalize to clinical depression.We also cannot separate patterns in willingness to report depressive symptoms from true change; given that all depression diagnoses and symptoms were based on self-report (even to a clinician), there was no way to distinguish true change from willingness to report.However, the patterns observed in the MTF data were similar to those observed in other data sources, including those with diagnostic instruments 1 and administrative records of outcomes such as suicidal self-injury, 4 reducing the possibility that the patterns observed here were not meaningful for population mental health.Furthermore, the MTF panel has substantial attrition across follow-up waves, although statistical weights adjusted for the association of attrition with study estimates.Additionally, the baseline MTF survey is conducted only among adolescents in school, and thus we lacked data on individuals who left high school or were absent on the day of the survey.We anticipated that respondents who were not in high school likely had higher levels of depressive symptoms and other mental health problems; thus, the estimates here are likely to be conservative.

Conclusions
The findings of this cohort study suggest that the current increase in depressive symptoms among US adolescents is swiftly shifting to an increase in young adult depressive symptoms as adolescents move into adulthood.There was significant persistence in depressive symptoms from ages 18 years to 21 to 22 years, with the recent birth cohort reporting the most stable elevated symptoms across age compared with previous birth cohorts.Given this persistence, reducing the onset of depressive symptoms through primary prevention and mental health resources during the adolescent years is critical.

JAMA Network
The main factor was depressive symptoms score at age 18 years (>12 vs Յ12, with >12 representing the top decile across all years or cohorts).Additional factors included self-derogation symptoms (>14 vs Յ14), self-esteem (Յ11 vs >11), and loneliness symptoms (>13 vs Յ13) scores at age 18 years, with cutoff scores based on the top decile across all years or cohorts.Outcomes included the same mental health scale dichotomizations at ages 19 to 20 and 21 to 22 years.
24 (95% CI, 7.01-14.97)times the odds of high depressive symptoms at age 19 to 20 years and had 6.20 (95% CI, 3.93-9.78)times the odds of high depressive symptoms at age 21 to 22 years.Among females, those with high depressive symptoms at baseline had 9.16 (95% CI, 6.57-12.76)times the odds of high depressive symptoms at age 19 to 20 years and 7.28 (95% CI, 4.92-10.78)times the odds of high depressive symptoms at age 21 to 22 years.eTable 4 in Supplement 1 shows similarly poor mental health outcomes across follow-up waves as indicated by baseline depressive symptoms as well as baseline self-derogation, loneliness, and low self-esteem.
A critical question in this investigation was whether there were interactions between birth cohort and depressive symptoms at age 18 years; that is, whether the association between depressive symptoms at age 18 years and depressive symptoms at both age 19 to 20 years and age 21 to 22 years was changing across birth cohort.Table2shows no interaction among both females and males.There was no evidence of meaningful interactions between birth cohort and depressive symptoms at age 18 years compared with depressive symptoms at age 19 to 20 years or age 21 to 22 years.

Table 3
6% had high depressive symptoms at age 19 to 20 years and 46.3% had high depressive symptoms at age 21 to 22 years.Comparatively, in previous birth cohorts, the prevalence of high depressive symptoms at follow-up waves among those with high depressive symptoms at baseline ranged from 16.5% to 34.6%.Analyses for loneliness, self-derogation, and low self-esteem are available in eFigures 4, 5, and 6 in Supplement 1. shows the PAFs for the proportion of high depressive symptoms at ages 19 to 20 and 21 to 22 years attributable to high depressive symptoms at age 18 years.The results indicated that among these birth cohorts, the proportions of high depressive symptoms at ages 19 to 20 and 21 to 22 years were highly associated with depressive symptoms at age 18 years.In earlier cohorts, PAFs were typically larger for males; for later cohorts, PAFs were typically larger for females.While there is JAMA Network Open.2024;7(8):e2427748.doi:10.1001/jamanetworkopen.2024.27748(Reprinted) August 14, 2024 5/12 Downloaded from jamanetwork.comby guest on 08/31/2024 approximately 45.

Table 1 .
Adjusted Odds Ratios (AORs) Between Birth Cohort and High Depressive Symptoms at Ages 18, 19 to 20, and 21 to 22 Years a a Adjusted for race and ethnicity, parental educational level, and pre-or post-2020 indicator.b P < .05.

Table 2 .
Adjusted Odds Ratios (AORs) Between Baseline Mental Health Factors and Mental Health Outcomes at Ages 19 to 20 and 21 to 22 Years a test for trend in PAFs, we found that among females born from 1972 to 1976, 35.28% (95% CI, 19.42%-50.14%) of depressive symptoms at age 21 to 22 years were associated with depressive symptoms at age 18 years, whereas among females born from 1997 to 2001, 55.25% (95% CI, 38.11%-65.13%) of depressive symptoms at age 21 to 22 years were associated with depressive symptoms at age 18 years.Overall, in the most recent cohort born from 1997 to 2001, 39.38% (95% CI, 30.74%-46.37%) of high depressive symptoms at age 19 to 20 years were associated with depressive symptoms at age 18 years, and 43.44% (95% CI, 30.98%-52.44%) ofFigure 2. Prevalence of High Depressive Symptoms at Ages 19 to 20 and 21 to 22 Years, by Birth Cohort and Presence of High Depressive Symptoms at Age 18 Years a Adjusted for race and ethnicity, parental educational level, cohort, and pre-or post-2020 indicator.b Interaction between depressive symptoms by cohort and at age 18 y.JAMA Network Open | Psychiatry Depressive Symptoms in Adolescence and Young Adulthood JAMA Network Open.2024;7(8):e2427748.doi:10.1001/jamanetworkopen.2024.27748(Reprinted) August 14, 2024 6/12 Downloaded from jamanetwork.comby guest on 08/31/2024 no formal statistical

Table 3 .
Population Attributable Fractions (PAFs) for the Proportion of High Depressive Symptoms in Young Adults Associated With High Depressive Symptoms at Age 18 Years high depressive symptoms at age 21 to 22 years were associated with depressive symptoms at age 18 years.
32t the lack of specialized training for more complex mental health challenges may continue the treatment gaps.Rates of emergency Downloaded from jamanetwork.combyguest on 08/31/2024 department use for mental health care among Medicaid recipients differ up to 8-fold across US regions,32suggesting access gaps.While expansion of teletherapy and telehealth, as well as the launch of the 3-digit mental health crisis line, may be leveraged to expand mental health care access, it is clear that primary prevention, ideally before upcoming cohorts reach adulthood, is also urgently needed.Beyond the expansion of the psychiatric workforce, school-and family-based interventions

Open | Psychiatry SUPPLEMENT 1. eFigure 1.
Longitudinal Prevalence of High Loneliness by Birth Cohort and Sex Among US Young Adults Born 1972 Through 2001 and Measured From 1990 Through 2019 eFigure 2. Longitudinal Prevalence of High Self-Derogation by Birth Cohort and Sex Among US Young Adults Born 1972 Through 2001 and Measured From 1990 Through 2019 eFigure 3. Longitudinal Prevalence of Low Self-Esteem by Birth Cohort and Sex Among US Young Adults Born 1972 Through 2001 and Measured From 1990 Through 2019 eFigure 4. Prevalence of High Loneliness at Age 19/20 and Age 21-22 Stratified by Whether the Respondents Had High Depressive Symptoms at Age 18, by Birth Cohort eFigure 5. Prevalence of High Self-Derogation at Age 19/20 and Age 21-22 Stratified by Whether the Respondents Had High Depressive Symptoms at Age 18, by Birth Cohort eFigure 6. Prevalence of Low Self-Esteem at Age 19/20 and Age 21-22 Stratified by Whether the Respondents Had High Depressive Symptoms at Age 18, by Birth Cohort eTable 1. Sample Size by Cohort and Wave eTable 2. Demographics by Birth Cohort eTable 3. Adjusted Odds Ratios Between Birth Cohort and Mental Well-Being at Ages 18, 19/20, and 21/22 eTable 4. Adjusted Odds Ratios Between Baseline Mental Health Predictors and Mental Health Outcomes at Ages 19/20 and Age 21/22 eTable 5. Interaction P values for Age 18 Predictor by Sex