Mental Health Status by Race, Ethnicity and Socioeconomic Status among Young Adults in Texas during COVID-19

Abstract

Background

Differences in symptoms of depression and anxiety by race/ethnicity and socioeconomic status (SES) among a diverse cohort of young adults during the COVID-19 pandemic (Spring 2020-Fall 2021) have not been examined.

Method

We analyzed four waves of biannual, panel data from n = 2629 emerging adults (16–25 years old) from the Texas Adolescent Tobacco and Marketing Surveillance study (TATAMS). We conducted a series of mixed effects ordinal logistic regression models to compare the independent and joint effects of race/ethnicity and SES on symptoms of (a) depression and (b) anxiety, adjusting for sex, cohort, and time.

Results

Symptoms of depression (aOR range: 1.54 – 2.19; 95% CI: 1.02 – 3.08) and anxiety (aOR range: 1.64 – 2.19; 95% CI: 1.22 – 2.79) were elevated among low SES young adults, across all racial/ethnic groups. Across SES groups, symptoms of depression were lower among non-Hispanic Blacks compared to non-Hispanic Whites (aOR range: 0.33 – 0.41; 95% CI: 0.18 – 0.62) and Hispanics /Latinos (aOR range: 0.33 – 0.38; 95% CI: 0.20 – 0.57); similarly, symptoms of anxiety were lower among non-Hispanic Blacks compared to non-Hispanic Whites (aOR range: 0.44; 95% CI: 25 – 0.77) and Hispanics/Latinos (aOR range: 0.47 – 0.56; 95% CI: 0.29 – 0.83). No significant interaction (joint effect) was observed between SES and race/ethnicity during this period.

Conclusion

Low SES was persistently related to poor mental health. Lower odds of symptoms of anxiety and depression among non-Hispanic Black young adults may reflect the ‘mental health paradox’. Overall, mental health policies should prioritize lower SES young adults regardless of race and ethnicity.

Introduction

Depression and anxiety are increasingly prevalent among young people in the United States [1, 2]. As of 2021, 19% of young adults (ages 18–25) reported a major depressive episode (MDE) in the past 12-months, more than any other category of adults in the US [3]. Young adults who reported a past-year MDE also reported the highest prevalence of severe impairment at 13.3% [3]. While other adult age groups (26–49, 50 +) maintained consistently low past-year prevalence estimates of MDE from 2005–2020 at 7% and 4% respectively, past-year MDE nearly doubled among young adults [4]. Prevalence of anxiety among 18–25 year-olds has also significantly increased over the last decade, from 8% in 2008 to 15% in 2018 [2]. Depression has several negative impacts, including poor role functioning, job loss, and lack of educational attainment while anxiety increases psychosocial and emotional impairment [5, 6]. Earlier age of onset for depression and anxiety leads to poorer prognosis during the life course with increased intensity of disease and recurrence of episodes [7, 8]. Epidemiological investigation of depression and anxiety is essential given the increasing prevalence and negative ramifications of these conditions on young people.

The COVID-19 pandemic exacerbated poor mental health outcomes among young adults as they dealt with loss of family members, unstable employment, and increased isolation and loneliness due to social restrictions [9-13]. Prevalence estimates of anxiety and depression among the 18–29 age group were 49% in 2020, increasing to 57% in early 2021 [14]. During COVID-19, young adults aged 18–24 had significantly higher odds (adjusted odds ratio (aOR): 1.56–7.73, p-value < 0.005) of depressive or anxious symptoms compared to all other age groups separately [15]. However, other research has also indicated negligible changes in mental health symptomology among young people across the COVID-19 pandemic [16]. Thus, this literature is mixed.

Prior to the COVID-19 pandemic, negative mental health outcomes were also disproportionately experienced by racial and ethnic minorities compared to non-Hispanic White individuals. Racial and ethnic minorities are more likely to experience discrimination both overall and at an earlier age than non-Hispanic Whites, potentially explaining higher rates of anxiety and depression among these populations [17-21]. Longitudinal studies show that racial and ethnic minorities, particularly non-Hispanic Black, non-Hispanic Asian, and Hispanic, tend not only to experience elevated depressive symptoms during late adolescence and early adulthood but also tend to have a steeper trajectory, attaining more severe symptomology more quickly than their non-Hispanic White peers [19-22].

At the peak of the COVID-19 pandemic, Hispanic and non-Hispanic Black reported the highest prevalence of depression and anxiety [23, 24]. Hispanic and non-Hispanic Black young adults in particular reported higher levels of victimization distress, race-related discrimination, and job insecurity compared to non-Hispanic White young adults [25, 26]. Therefore, investigation on trajectories of depressive and anxious symptomology as experienced during the COVID-19 pandemic is critical.

Poor mental health outcomes are also disproportionately experienced by young adults of low socioeconomic status (SES) based on area deprivation and limited resources [20]. Young adults of lower SES often experience increased social stress, usually in the form of poverty-related discrimination [27]. Low SES in adolescents and young adults is also associated with poor social relationships later in life based on chronicity of poverty-related stressors [27, 28]. They not only experience an increased number of stressful life events due to low income, related social disadvantage, and poor parental support but also experience these stressors cumulatively, often leading to poorer resilience and increasingly worse mental health symptomology over the life course [27, 28].

During the height of the COVID-19 pandemic, low SES was significantly associated with increased depressive and anxious symptomology among young adults, as many of them faced financial uncertainty and educational disruptions while also experiencing higher SARS-CoV-2 infection and mortality rates than their higher SES counterparts [11, 28, 29]. Economic disparities in mental health may have been heightened during the COVID-19 pandemic due to inequitable access to mitigation strategies such as social distancing and healthcare treatment. For example, young adults of lower economic status were more likely working jobs that could not be done remotely, resulting in elevated risk for infection and spread of the COVID-19 virus [30, 31]. As a result, there is also considerable need to investigate the impact of lower socioeconomic status on symptoms of depression and anxiety as experienced during the COVID-19 pandemic.

Overall, changes in mental health pre- and post-COVID have showed worsening mental health outcomes among adolescents and young adults [32]. While racial, ethnic and SES-related mental health disparities have been independently investigated to an extent both pre- and post-COVID-19, disparities in young adults based on the joint effects of race, ethnicity and SES on severity of mental illness during COVID-19 has not yet been investigated [10, 19-22, 33-35]. Whereas non-Hispanic White adults report more acute mental health episodes, people of color experience prolonged stressors (i.e. discrimination, financial instability) and chronicity that lead to more severe mental health outcomes earlier and throughout the life course [18, 20, 21, 34, 35]. Racial and ethnic minorities of low SES experience additional poverty-related stress in combination with discrimination-related stress and social isolation [19, 20]. Furthermore, racial and ethnic minorities (particularly Non-Hispanic Black and Hispanic) reported increased food insecurity, discrimination, and lack of safe housing options during the pandemic, increasing both exposure to chronic stress and risk of poor mental health [23-25]. Understanding the inter-relationship of race, ethnicity and SES across categories of mental health severity is critical, as those with severe symptoms of depression and anxiety have worse prognosis, impairment and psychiatric comorbidities [18, 36, 37]. Furthermore, while the COVID-19 pandemic had acute, negative impacts on young adult mental health, literature on sustained mental health outcomes during the pandemic by race, ethnicity, and SES among young adults is limited [10, 14, 15, 26].

Study Aims and Hypotheses

The primary research objective is to examine the independent and joint effects of differences in severity of symptoms of depression and anxiety by race, ethnicity and socioeconomic status (SES) among a diverse sample of young adults during the COVID-19 pandemic (Spring 2020-Fall 2021).

We hypothesize that i). independently, young adults across each racial and ethnic minority group and those who report low SES will have a higher likelihood of more severe symptoms of mental health outcomes compared to non-Hispanic White young adults and those who report high SES, respectively; ii). joint effects (i.e., interaction) of race and ethnicity and SES on severity of mental health outcomes will be higher in racial and ethnic minority and low SES groups; iii). severity of mental health symptoms will be higher in each racial and ethnic minority group compared to non-Hispanic White young adults across levels of SES; and iv). severity of mental health symptoms will be higher in those who report low SES compared to those who report high SES across all racial and ethnic groups [17-31].

Methods

Study Design and Sample

This study is a secondary data analysis using longitudinal, repeated measures data from the Texas Adolescent and Tobacco Marketing Surveillance (TATAMS) study, Waves 11–14 (Spring 2020-Fall 2021). The TATAMS study is a prospective cohort study of adolescents from major metropolitan counties in Texas (Bexar, Dallas, Harris, Tarrant, and Travis) that were in grades six, eight, and ten in Fall 2014 (Wave 1). Follow-up assessments were conducted every six months. At Wave 14, participants were 1, 3, and 5 years post-high school. Surveys were administered online and participants were provided a $25 Amazon gift card upon completion. Informed consent was obtained from all participants included in the study or from their parents if under the age of 18. The full, detailed methodological description of the setting, sampling frame, and recruitment strategies for TATAMS has been previously published; the study was approved by the University of Texas Health Science Center at Houston Institutional Review Board, number: HSC-SPH-13–0377 [38, 39].

Inclusion criteria were complete cases per wave (Waves 11 – 14). The TATAMS cohort comprised of n = 2,954 participants with response rates between 77%-81% across Waves 11–14; however, n = 325 (~ 11%) did not provide any data. Among complete cases, n = 2629 (89%) unique participants were included in this analysis; n = 2042 (77.67%) unique participants completed all four surveys; n = 291 (11.06%) completed three; n = 154 (5.86%) completed two; and n = 142 (5.40%) completed one. The final analytic sample included n = 9491 (80.32%) observations from n = 2629 (89%) unique participants over four time periods (Spring 2020, Fall 2020, Spring 2021, Fall 2021). All data were self-reported.

Measures

Outcomes

The outcomes are depression and anxiety symptom severity. Symptoms of depression were measured with the Patient Health Questionnaire-9 (PHQ-9) while symptoms of anxiety were measured with the General Anxiety Disorder-7 (GAD-7) [40, 41]. Both were measured at each time point (Waves 11–14) and were treated as categorical, ordinal outcome variables.

The PHQ-9 is a valid, reliable self-report 9-item instrument used to measure symptoms of depression. Participants were asked, “Over the last 2 weeks, how often have you been bothered by the following problems (e.g. little interest or pleasure in doing things; feeling down, depressed, or hopeless; feeling tired or having little energy)?” Scores were summed per participant; each item was scored from 0–3, with 0 (not at all), 1 (several days), 2 (more than half the days), and 3 (nearly every day). Summed scores were categorized based on validated cutoff values indicating increased symptom severity; 0–4 indicated minimal depression (referent), 5–9 mild, 10–14 moderate, 15–19 moderately severe, and 20 + severe.

The GAD-7 is a validated and reliable self-report 7-item questionnaire used to measure symptoms of anxiety. Participants were asked, “Over the last 2 weeks, how often have you been bothered by the following problems (e.g. feeling nervous, anxious or on edge; trouble relaxing; becoming easily annoyed or irritable)?” Each item was scored from 0–3 with 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day); scores were summed for each participant and were categorized using validated cutoff values to indicate levels of symptom severity; 0–4 indicated no anxiety (referent), 5–9 mild, 10–14 moderate, and 15–21 severe.

Primary Exposures

The primary exposures of interest are race, ethnicity, and socioeconomic status (SES). Race was gathered at baseline (Wave 1), with response choices: White, Black or African American, Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, or Other. Ethnicity was also gathered at baseline (Wave 1) and coded as a categorical variable; participants were asked, “Are you Hispanic or Latino?” with response choices: ‘no,’ ‘Yes, I am Mexican, Mexican American, or Chicano/a’ and ‘Yes, I am some other Hispanic or Latino/a ethnicity not listed here.’

The race and ethnicity variable was derived from the race and ethnicity measures. Racial and ethnic categories include non-Hispanic White, non-Hispanic Black, Hispanic/Latino, and non-Hispanic Other, with non-Hispanic White as the reference group. The non-Hispanic Other category collapsed individuals who reported their race as either Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, or Other due to sample size considerations.

SES was a categorical variable measured at each wave (Waves 11 – 14) and was treated as a time-varying variable. This variable was derived from a measure of perceptions of family income among adults (Kendall’s τ = 0.62) and has since been utilized in adolescent populations [42, 43]. Participants were asked “In terms of income, what best describes your family’s standard of living in the home where you live most of the time? Would you say your family is:” with response choices: ‘very well-off’, ‘living comfortably’, ‘just getting by’, ‘nearly poor’, and ‘poor’. Categories were dichotomized to ‘well-off’ (very well-off, living comfortably) and ‘poor’ (just getting by, nearly poor, poor) based on sparse data and model convergence.

Covariates

Sex, geographic region (i.e., metropolitan area of Texas), cohort (i.e., 1, 3 or 5 years post-high school at Wave 14), and time (i.e., Spring 2020-Fall 2021) were identified as potential confounders of the relationship between race and ethnicity, SES, and each mental health outcome [19, 20, 22]. Sex was coded as a binary, categorical variable gathered at Wave 1 with response choices: male or female (reference). Geographic region was a categorical, nominal variable specified as the county of the school in which the participant attended at baseline (Wave 1); response options were: Bexar County, Dallas County, Harris County, Tarrant County, or Travis County (reference). Cohort is a categorical, ordinal variable based on school-grade levels (1, 3 or 5 years post-high school) at Wave 14. This variable served as a proxy of distinct age groups based on grade level reported. Time was specified by survey wave; Wave 11 indicated Spring 2020, Wave 12 indicated Fall 2020, Wave 13 indicated Spring 2021, and Wave 14 indicated Fall 2021. Time was a continuous variable and centered at 0 (Wave 11, Spring 2020) in order to generate appropriate random intercepts for the mixed effects models.

Missing Data, Attrition, and Selection Bias

Among those who completed surveys, there were no missing data for race, ethnicity, sex, cohort, time, and county. Missing data were negligible (n = 17 observations, 0.18%) for symptoms of depression and anxiety and SES. These observations were excluded from the analytic sample.

Among those who did not complete any surveys, no significant differences in proportions of missingness were found for race and ethnicity, SES, county and cohort for each mental health outcome. However, 16% of observations from females compared to 23% of observations from males were missing for both mental health outcomes (z = −3.96, p < 0.001), indicating some selection bias.

Statistical Analyses

Descriptive statistics were used to describe the study sample (Table 1). Continuous variables were reported with means and standard deviations, while categorical variables were reported with frequency counts and percentages. Column percentages were reported in order to establish prevalence estimates of symptoms of depression and anxiety by each variable.

Table 1.

Descriptive statistics by mental health outcome among young adults (TATAMS, Waves 11–14 (n = 9491))^a,b

Analytic Sample n (%)		Depressionc n (%)					Anxietyd n (%)
	n = 9491	Minimal (n = 5028, 52.98%)	Mild (n = 2021, 21.29%)	Moderate (n = 1172, 12.35%)	Moderately Severe (n = 744, 7.84%)	Severe (n = 526, 5.54%)	None (n = 5139, 54.15%)	Mild (n = 2225, 23.44%)	Moderate (n = 1202, 12.66%)	Severe (n = 925, 9.75%)
Age, range 16–25 (mean, ± SD)	20.22 (± 1.68)	20.23 (± 1.69)	20.25 (± 1.68)	20.18 (± 1.62)	20.22 (± 1.71)	20.14 (± 1.62)	20.24 (± 1.68)	20.19 (± 1.68)	20.20 (± 1.70)	20.21 (± 1.65)
Cohort (Grade at Wave 14)e
1 Year Post High School	2644 (27.86)	1408 (28.00)	541 (26.77)	326 (27.82)	216 (29.03)	153 (29.09)	1413 (27.50)	634 (28.49)	333 (27.70)	264 (28.54)
3 Years Post High School	3194 (33.65)	1719 (34.19)	689 (34.09)	389 (33.19)	225 (30.24)	172 (32.70)	1797 (34.97)	731 (32.85)	375 (31.20)	291 (31.46)
5 Years Post High School	3653 (38.49)	1901 (37.81)	791 (39.14)	457 (38.99)	303 (40.73)	201 (38.21)	1929 (37.54)	860 (38.65)	494 (41.10)	370 (40.00)
Age per Cohort (mean, ± SD)
1 Year Post High School	18.14 (± 0.73)	18.15 (± 0.73)	18.11 (± 0.75)	18.15 (± 0.71)	18.12 (± 0.74)	18.10 (± 0.69)	18.15 (± 0.73)	18.14 (± 0.73)	18.07 (± 0.73)	18.16 (± 0.72)
3 Years Post High School	20.17 (± 0.76)	20.17 (± 0.77)	20.17 (± 0.75)	20.16 (± 0.77)	20.16 (± 0.72)	20.23 (± 0.74)	20.18 (± 0.77)	20.14 (± 0.77)	20.11 (± 0.72)	20.23 (± 0.74)
5 Years Post High School	21.78 (± 0.92)	21.83 (± 0.93)	21.77 (± 0.93)	21.64 (± 0.94)	21.77 (± 0.91)	21.59 (± 0.88)	21.8 (± 0.93)	21.75 (± 0.91)	21.72 (± 0.94)	21.68 (± 0.90)
Race and Ethnicity
Hispanic/Latino	3630 (38.25)	1827 (36.34)	720 (35.63)	500 (42.66)	330 (44.35)	253 (48.10)	1913 (37.23)	830 (37.30)	491 (40.85)	396 (42.81)
Non-Hispanic White	2919 (30.76)	1507 (29.97)	716 (35.43)	356 (30.38)	198 (26.61)	142 (27.00)	1534 (29.85)	747 (33.57)	343 (28.54)	295 (31.89)
Non-Hispanic Black	1380 (14.54)	855 (17.00)	237 (11.73)	128 (10.92)	99 (13.31)	61 (11.60)	830 (16.15)	279 (12.54)	159 (13.23)	112 (12.11)
Non-Hispanic Otherf	1562 (16.46)	839 (16.69)	348 (17.22)	188 (16.04)	117 (15.73)	70 (13.31)	862 (16.77)	369 (16.58)	209 (17.39)	122 (13.19)
Sex
Female	5571 (58.70)	2594 (51.59)	1274 (63.04)	763 (65.10)	527 (70.83)	413 (78.52)	2624 (51.06)	1441 (64.76)	810 (67.39)	696 (75.24)
Male	3920 (41.30)	2434 (48.41)	747 (36.96)	409 (34.90)	217 (29.17)	113 (21.48)	2515 (48.94)	784 (35.24)	392 (32.61)	229 (24.76)
County
Bexar	308 (3.25)	146 (2.90)	68 (3.36)	51 (4.35)	23 (3.09)	20 (3.80)	152 (2.96)	71 (3.19)	49 (4.08)	36 (3.89)
Dallas	1202 (12.66)	665 (13.23)	241 (11.92)	134 (11.43)	92 (12.37)	70 (13.31)	677 (13.17)	270 (12.13)	145 (12.06)	110 (11.89)
Harris	2334 (24.59)	1252 (24.90)	482 (23.85)	283 (24.15)	193 (25.94)	124 (23.57)	1281 (24.93)	540 (24.27)	289 (24.04)	224 (24.22)
Tarrant	2649 (27.91)	1410 (28.04)	566 (28.01)	318 (27.13)	209 (28.09)	146 (27.76)	1422 (27.67)	619 (27.82)	351 (29.20)	257 (27.78)
Travis	2998 (31.59)	1555 (30.93)	664 (32.86)	386 (32.94)	227 (30.51)	166 (31.56)	1607 (31.27)	725 (32.58)	368 (30.62)	298 (32.22)
SESg
Well off	6950 (73.23)	3909 (77.74)	1493 (73.87)	819 (69.88)	454 (65.02)	275 (52.28)	3978 (77.41)	1621 (72.85)	805 (66.97)	546 (59.03)
Poor	2541 (26.77)	1119 (22.26)	528 (26.13)	353 (30.12)	290 (38.98)	251 (47.72)	1161 (22.59)	604 (27.15)	397 (33.03)	379 (40.97)

^aNumber of observations in the analytic sample from TATAMS Waves 11–14 among n = 2629 participants

^bSample frequencies and category percentages displayed unless otherwise stated

^cPatient Health Questionnaire-9 (PHQ-9): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Scores 0–4: minimal, 5–9: mild, 10–14: moderate, 15–19: moderately severe, 20 + : severe

^dGeneral Anxiety Disorder-7 (GAD-7): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Scores 0–4: none, 5–9: mild, 10–14: moderate, 15–21: severe

^eCohort is based on school-grade levels reported during Wave 14, indicating distinct two-year age cohorts among this sample

^fNon-Hispanic Other race and ethnicity includes Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, or Other

^gSES: Well-off includes ‘very well-off’ and ‘living comfortably’; Poor includes ‘just getting by’, ‘nearly poor’, and ‘poor’

Multivariate mixed effects ordinal logistic regression models were run to investigate main effects of race, ethnicity and SES on symptoms of depression and anxiety, respectively (Table 2) along with regression models for cohort on each outcome (not shown). Regression models were built using a change in estimate approach to assess for confounding; confounding was determined by whether point estimates for either race and ethnicity or SES changed by at least 10%, with the inclusion of potential confounders (i.e. sex, geographic region, cohort, and time). Bayesian Information Criteria (BIC) was used to assess model fit; models were considered a better fit if the BIC decreased by at least a ten-point margin. The proportional odds assumption for the mixed effects ordinal logistic regression models was tested using the Generalized Linear and Latent Mixed Models (GLLAMM) package in STATA; all models met the proportional odds assumption [44].

Table 2.

Multivariate ordinal logistic regression analyses of race and ethnicity, SES, and mental health outcomes among young adults (TATAMS, Waves 11–14 (n = 9491))^a,b

	Depressionc		Anxietyd
	aOR (95% CI)	p value	aOR (95% CI)	p value
Race and Ethnicitye
Non-Hispanic White	(Ref.)	–	(Ref.)	–
Hispanic/Latino	1.10 (0.85, 1.43)	0.475	0.85 (0.66, 1.11)	0.232
Non-Hispanic Black	0.40 (0.28, 0.57)	< 0.0001*	0.44 (0.31, 0.63)	< 0.0001*
Non-Hispanic Other	0.87 (0.62, 1.21)	0.414	0.78 (0.56, 1.08)	0.1333
SESf
Well-off	(Ref.)	–	(Ref.)	–
Poor	1.90 (1.62, 2.23)	< 0.0001*	1.92 (1.64, 2.25)	< 0.0001*

^aNumber of observations in the analytic sample from TATAMS Waves 11–14 among n = 2629 participants

^bMultivariate ordinal logistic regression models for symptoms of depression and anxiety outcomes also adjusted for sex, time (i.e. survey wave) and cohort (i.e. grade at Wave 14). Individual was specified as the random intercept for each multivariate model. Odds ratios were modeled over lower severity

^cPatient Health Questionnaire-9 (PHQ-9): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Categories 0–4: minimal, 5–9: mild, 10–14: moderate, 15–19: moderately severe, 20 + : severe

^dGeneral Anxiety Disorder-7 (GAD-7): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Scores 0–4: none, 5–9: mild, 10–14: moderate, 15–21: severe

^eNon-Hispanic Other race and ethnicity includes Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, or Other

^fSES: Well-off includes ‘very well-off’ and ‘living comfortably’; Poor includes ‘just getting by’, ‘nearly poor’, and ‘poor’

(Ref.): Reference category

aOR: adjusted odds ratio

95% CI: 95% Wald confidence intervals

^*p value < 0.05, statistically significant

To test for joint effects, interactions between race and ethnicity and SES were investigated for both mental health outcomes (Table 3) along with separate interaction terms for race and ethnicity, cohort, and SES by time (not shown). Stratified regression analyses for race and ethnicity and each mental health outcome by SES are found in Table 4, while stratified regression analyses for SES and each mental health outcome by race and ethnicity are found in Table 5.

Table 3.

Multivariate ordinal logistic regression analyses of race and ethnicity by SES interaction terms and mental health outcomes among young adults (TATAMS, Waves 11–14 (n = 9491))^a,b

	Depressionc		Anxietyd
	aOR (95% CI)	p value	aOR (95% CI)	p value
Race and Ethnicitye
Non-Hispanic White	(Ref.)	–	(Ref.)	–
Hispanic/Latino	1.10 (0.83, 1.46)	0.489	0.80 (0.61, 1.05)	0.110
Non-Hispanic Black	0.42 (0.29, 0.63)	< 0.0001*	0.46 (0.32, 0.68)	< 0.0001*
Non-Hispanic Other	0.94 (0.66, 1.33)	0.710	0.78 (0.55, 1.10)	0.162
SESf
Well-off	(Ref.)		(Ref.)	–
Poor	2.11 (1.52, 2.95)	< 0.0001*	1.72 (1.22, 2.41)	0.002*
Race and Ethnicity*SES
Non-Hispanic White*Well-off	(Ref.)		(Ref.)
Hispanic*Poor	0.94 (0.63, 1.41)	0.773	1.29 (0.85, 1.95)	0.227
Non-Hispanic Black*Poor	0.81 (0.49, 1.36)	0.429	0.96 (0.57, 1.95)	0.862
Non-Hispanic Other*Poor	0.73 (0.43, 1.24)	0.244	1.03 (0.61, 1.74)	0.914

^aNumber of observations in the analytic sample from TATAMS Waves 11–14 among n = 2629 participants

^bMultivariate ordinal logistic regression models for symptoms of depression and anxiety outcomes also adjusted for sex, time (i.e. survey wave) and cohort (i.e. grade at Wave 14). Individual was specified as the random intercept for each multivariate model. Odds ratios were modeled over lower severity

^cPatient Health Questionnaire-9 (PHQ-9): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Categories 0–4: minimal, 5–9: mild, 10–14: moderate, 15–19: moderately severe, 20 + : severe

^dGeneral Anxiety Disorder-7 (GAD-7): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Scores 0–4: none, 5–9: mild, 10–14: moderate, 15–21: severe

^eNon-Hispanic Other race and ethnicity includes Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, or Other

^fSES: Well-off includes ‘very well-off’ and ‘living comfortably’; Poor includes ‘just getting by’, ‘nearly poor’, and ‘poor’

(Ref.): Reference category

aOR: adjusted odds ratio

95% CI: 95% Wald confidence intervals

^*p value < 0.05, statistically significant

Table 4.

Multivariate ordinal logistic regression analyses of race and ethnicity and mental health outcomes among young adults stratified by SES (TATAMS, Waves 11–14 (n = 9491))^a,b

	Depressionc		Anxietyd
SESd	Well-off	Poor	Well-off	Poor
	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)
Race and Ethnicitye
Non-Hispanic White	(Ref.)	(Ref.)	(Ref.)	(Ref.)
Hispanic/Latino	1.09 (0.82, 1.46)	0.99 (0.59, 1.65)	0.78 (0.59, 1.04)	0.93 (0.58, 1.51)
Non-Hispanic Black	0.41 (0.27, 0.62)*	0.33 (0.18, 0.61)*	0.44 (0.29, 0.65)*	0.44 (0.25, 0.77)*
Non-Hispanic Other	0.89 (0.62, 1.29)	0.56 (0.29, 1.08)	0.80 (0.56, 1.14)	0.58 (0.31, 1.07)

^aNumber of observations in the analytic sample from TATAMS Waves 11–14 among n = 2629 participants

^bMultivariate ordinal logistic regression models for symptoms of depression and anxiety outcomes adjusted for sex, time (i.e. survey wave) and cohort (i.e. grade at Wave 14). Individual was specified as the random intercept for each multivariate model. Odds ratios were modeled over lower severity

^cPatient Health Questionnaire-9 (PHQ-9): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Categories 0–4: minimal, 5–9: mild, 10–14: moderate, 15–19: moderately severe, 20 + : severe

^dGeneral Anxiety Disorder-7 (GAD-7): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Scores 0–4: none, 5–9: mild, 10–14: moderate, 15–21: severe

^eSES: Well-off includes ‘very well-off’ and ‘living comfortably’; Poor includes ‘just getting by’, ‘nearly poor’, and ‘poor’

^fNon-Hispanic Other race and ethnicity includes Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, or Other

(Ref.): Reference category

aOR: adjusted odds ratio

95% CI: 95% Wald confidence intervals

^*p value < 0.05, statistically significant

Table 5.

Multivariate ordinal logistic regression analyses of SES and mental health outcomes among young adults stratified by race and ethnicity (TATAMS, Waves 11–14 (n = 9491))^a,b,c

	Depressionc				Anxietyd
Race and Ethnicitye	Non-Hispanic White	Hispanic	Non-Hispanic Black	Non-Hispanic Other	Non-Hispanic White	Hispanic	Non-Hispanic Black	Non-Hispanic Other
	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)
SESf
Well-off	(Ref.)	(Ref.)	(Ref.)	(Ref.)	(Ref.)	(Ref.)	(Ref.)	(Ref.)
Poor	2.19 (1.56, 3.08)*	1.98 1.57, (2.50)*	1.69 (1.14, 2.48)*	1.54 (1.02, 2.34)*	1.72 1.22, (2.43)*	2.19 1.73, (2.77)*	1.64 (1.12, 2.39)*	1.85 (1.22, 2.79)*

^aNumber of observations in the analytic sample from TATAMS Waves 11–14 among n = 2629 participants

^bMultivariate ordinal logistic regression models for symptoms of depression and anxiety outcomes adjusted for sex, time (i.e. survey wave) and cohort (i.e. grade at Wave 14). Individual was specified as the random intercept for each multivariate model. Odds ratios were modeled over lower severity

^cPatient Health Questionnaire-9 (PHQ-9): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Categories 0–4: minimal, 5–9: mild, 10–14: moderate, 15–19: moderately severe, 20 + : severe

^dGeneral Anxiety Disorder-7 (GAD-7): “Over the last 2 weeks, how often have you been bothered by the following problems?”; response options: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). Scores 0–4: none, 5–9: mild, 10–14: moderate, 15–21: severe

^eNon-Hispanic Other race and ethnicity includes Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, or Other

^fSES: Well-off includes ‘very well-off’ and ‘living comfortably’; Poor includes ‘just getting by’, ‘nearly poor’, and ‘poor’

(Ref.): Reference category

aOR: adjusted odds ratio

95% CI: 95% Wald confidence intervals

^*p value < 0.05, statistically significant

Individual was specified as a random effect to account for the nesting of responses within participant (i.e., repeated observations). All mixed effects models reported are random intercept models, as random-slope intercept models with time and individual random effects did not converge for either mental health outcome. Adjusted odds ratios (aOR) from the multivariate ordinal logistic regression models were reported for the main effect and stratified analyses while interaction aORs were reported for the interaction analyses. All hypothesis tests conducted were two-sided tests with 95% Wald confidence intervals (CI). Significant p values were reported at α < 0.05 for all regression outcomes. All final regression models adjusted for sex, time, and cohort as fixed effects, as described above; geographic region was dropped from the final models as the variable was not identified as a confounder during the model-building process. Additionally, as cohort and age are collinear variables, cohort was considered a more informative variable based on distinct age groups; thus, age was not considered as a potential confounder in the regression models. Analyses were conducted using STATA 17.0 SE [45].

Results

Sample Characteristics

Descriptive statistics across all observations from Waves 11–14 (Spring 2020 – Fall 2021) are displayed in Table 1. Participants were, on average, 20 years of age (20.22 ± 1.68, range: 16–25), mostly female (58.70%) and 38.25% Hispanic/Latino, 30.76% non-Hispanic White, 14.54% non-Hispanic Black, and 16.46% non-Hispanic Other. Within each cohort (i.e. grade level reported at Wave 14), average ages were 18.14 (± 0.72) for the 1 year post-high school cohort, 20.17 (± 0.76) 3 years post-high school cohort, and 21.78 (± 0.92) for the 5 year post-high school cohort. The majority of participants reported their SES as ‘well-off’ (73.23%) and their county of school attendance at Wave 1 as Travis (31.59%), Tarrant (27.91%), or Harris (24.59%) County.

The prevalence of severity of symptoms of depression across Waves 11–14 were 52.98% (minimal depression), 21.29% (mild depression), 12.35% (moderate depression), 7.84% (moderately severe depression), and 5.54% (severe depression). Likewise, the prevalence of severity of symptoms of anxiety across Waves 11–14 included 54.15% (no anxiety), 23.44% (mild anxiety), 12.66% (moderate anxiety), and 9.75% (severe anxiety).

Main Effects of Race and Ethnicity and SES on Mental Health Outcomes

After adjusting for SES, the odds of being in a more severe category of symptoms of depression, compared to the minimal category, was 0.40 (aOR = 0.40, 95% CI: 0.28, 0.57) times the odds among non-Hispanic Black young adults versus non-Hispanic Whites, after controlling for covariates. No other significant differences by race/ethnicity were noted. After adjusting for race and ethnicity, the odds of being in a more severe category of symptoms of depression, compared to the minimal category, was 1.90 (aOR = 1.90, 95% CI: 1.62, 2.23) times the odds among the poor SES category versus the well-off SES category, adjusting for covariates (Table 2).

After adjusting for SES, the odds of being in a more severe category of symptoms of anxiety, compared to no anxiety, was 0.44 (aOR = 0.44, 95% CI: 0.31, 0.63) times the odds among non-Hispanic Black young adults versus non-Hispanic Whites after controlling for covariates. Comparable point estimates were observed for non-Hispanic Blacks with Hispanic/Latino as the reference group, instead, for symptoms of depression and anxiety, respectively (aOR = 0.36, 95% CI: 0.26, 0.51; aOR = 0.52, 95% CI: 0.37, 0.72), after adjusting for SES and other covariates (results not shown in Table). Similarly, after adjusting for race and ethnicity, the odds of being in a more severe category of symptoms of anxiety, compared to no anxiety, was 1.92 (aOR = 1.92, 95% CI: 1.64, 2.25) times the odds among the poor SES category versus the well-off SES category, adjusting for covariates. No statistically significant result of cohort on symptoms of depression and anxiety emerged (results not shown in Table).

Joint Effects of Race and Ethnicity by SES for Mental Health Outcomes

Multivariate ordinal logistic regression models with interaction terms for race, ethnicity and SES for each mental health outcome are displayed in Table 3. No interaction term was statistically significant for either outcome with Non-Hispanic White or Hispanic/Latino (results not shown in Table) as the reference group. Likewise, no interaction terms for race, ethnicity, SES, or cohort by time were statistically significant (results not shown in Table).

Stratified Analyses of Race and Ethnicity by SES for Mental Health Outcomes

Stratified multivariate ordinal logistic regression models for race/ethnicity by SES for each mental health outcome are found in Table 4. Among the well-off SES category, non-Hispanic Black young adults had 0.41 (aOR = 0.41, 95% CI: 0.27, 0.62) times the odds of being in a more severe category of symptoms of depression compared with the minimal category of depression versus non-Hispanic Whites, after adjusting for covariates. Similarly, among the poor SES category, non-Hispanic Black young adults had 0.33 (aOR = 0.33, 95% CI: 0.18, 0.61) times the odds of being in a more severe category of symptoms of depression compared with the minimal category of depression versus non-Hispanic Whites, after controlling for covariates.

Among the well-off SES category, non-Hispanic Black young adults had 0.44 (aOR = 0.44, 95% CI: 0.29, 0.65) times the odds of being in a more severe category of symptoms of anxiety compared with no anxiety versus non-Hispanic Whites, after adjusting for covariates. Similarly, among the poor SES category, non-Hispanic Black young adults had 0.44 (aOR = 0.44, 95% CI: 0.25, 0.77) times the odds of being in a more severe category of symptoms of anxiety compared with no anxiety versus non-Hispanic Whites, after controlling for covariates.

Similar point estimates were observed for non-Hispanic Blacks with Hispanic/Latino as the reference group across categories of SES for symptoms of depression and anxiety, respectively (aOR range = (0.33 – 0.38, 95% CI range: 0.20, 0.57); aOR range = (0.47 – 0.56, 95% CI range: 0.29, 0.83)) after adjusting for covariates (results not shown in Table).

Stratified Analyses of SES by Race and Ethnicity for Mental Health Outcomes

Stratified multivariate ordinal logistic regression models for SES by race and ethnicity for each mental health outcome are found in Table 5. All models adjusted for sex, time, and cohort effects. Across all racial and ethnic groups, those in the poor SES category had 1.54 – 2.19 (aOR range: 1.54 – 2.19; 95% CI range: 1.02, 3.08) times the odds of being in a more severe category of symptoms of depression compared with the minimal category of symptoms of depression, versus those in the well-off SES group, after controlling for covariates. Similarly, across all racial and ethnic groups, those in the poor SES category had 1.64 – 2.19 (aOR range: 1.64 – 2.19; 95% CI range: 1.22, 2.79) times the odds of being in a more severe category of symptoms of anxiety, compared with no symptoms of anxiety, versus those in the well-off SES group, after adjusting for covariates.

Discussion

In this study of a diverse cohort of young adults during the COVID-19 pandemic, we found that non-Hispanic Black young adults had lower odds of reporting more severe symptoms of depression and anxiety than non-Hispanic White and Hispanic/Latino young adults. This finding contradicts previous studies reporting worse mental health symptoms among non-Hispanic Black young adults during the COVID-19 pandemic [25, 26]. We also found that young adults in the poor SES category had higher odds of reporting more severe symptoms of depression and anxiety compared to those who are well-off, consistent with prior findings of SES on symptoms of depression and anxiety [33]. Notably, although no significant interaction terms between race, ethnicity, and SES for each mental health outcome were identified, stratified analyses showed significantly lower odds (aOR range: 0.33 – 0.56; 95% CI: 0.18, 0.83) of reporting more severe symptoms of mental health outcomes among non-Hispanic Blacks compared to non-Hispanic Whites and Hispanics/Latinos across each level of SES. Stratified analyses also showed significantly higher odds (aOR range: 1.54 – 2.19; 95% CI: 1.02, 3.08) of reporting more severe symptoms of depression and anxiety among those in the poor SES category across all racial and ethnic groups.

Findings extend prior research by observing both independent and stratified effects between race, ethnicity, and SES and symptoms of depression/anxiety not only based on symptom severity but also based on a prolonged period of the COVID-19 pandemic [11, 14, 15, 19, 20, 33, 35, 46]. As there were no significant interaction effects of race and ethnicity or SES by time, results demonstrate a consistent impact of race, ethnicity and SES on severity of mental health outcomes across this time period. Additionally, to date, studies often operationalize measures of depression and anxiety as dichotomous, disallowing for severity of mental health outcomes [1, 2, 4, 14, 15, 46]. Similarly, longitudinal studies often assume linearity for depression and anxiety outcomes when, for psychometric scales, there is frequently clustering around scale extremities [16, 19, 20, 47, 48]. Therefore, these findings also demonstrate a different framework in which to investigate independent and joint effects of race, ethnicity and SES on mental health outcomes among young adults in a manner more congruent with psychometric properties.

Study results confirm hypotheses of the individual effects of poor SES on mental health outcomes among young adults during the COVID-19 pandemic, as they suffered the most financial instability, poor housing conditions, limited remote work opportunities, and increased COVID-19 infection rates [10, 13, 24, 33, 49, 50]. This finding is critical to consider; as mentioned earlier, persistently experienced poverty-related stressors lead to more severe depressive and anxious symptomology across the life course [27]. These stressors may have possibly been experienced persistently during the prolonged COVID-19 period analyzed in this study and might suggest a lack of built resilience while reinforcing the chronicity of more severe mental health outcomes experienced among this sample [10, 28, 29, 46]. However, SES was a time-varying, family-level variable measured at each wave, starting at Wave 1; therefore, perceptions of family standard of living could change and differentially impact severity of symptoms of depression and anxiety over time. That being said, since no SES by time interactions were significant, the impact of SES on severity of mental health outcomes was consistent throughout the duration of the study.

Study results contradict prior research concluding worse symptoms of depression and anxiety among non-Hispanic Black young adults during the COVID-19 pandemic [25, 26]. Although contradictory to our original hypotheses on the individual effect of race and ethnicity on mental health, results may reflect the ‘mental health paradox’ sometimes observed among racial and ethnic minorities; that is, although minorities experience more stressors associated with poor mental health such as race-related discrimination, hypervigilance, and social disadvantage, mental health is either better than or non-distinguishable from their non-Hispanic White counterparts [33, 34, 51-53]. One possible explanation for the ‘mental health paradox’ is that non-Hispanic Black individuals report higher self-esteem, perceived familial support, and strong socialization based on racial identity compared to non-Hispanic Whites despite also reporting higher stress exposures, indicating an underlying resilience [53-55]. However, this does not explain differences between non-Hispanic Black and Hispanic young adults observed in our study, which has not been thoroughly discussed within the context of the ‘mental health paradox’ [55].

Although racial and ethnic minority groups and young adults have demonstrated greater posttraumatic growth amidst adversity compared to their non-Hispanic White and older counterparts during the COVID-19 pandemic, posttraumatic growth among young adults during the pandemic remained low among those self-reporting depressive symptoms; furthermore, racial and ethnic differences in said growth are inconclusive [56, 57]. While the current study did not explore constructs of adaptive coping, future research should explore the role of posttraumatic growth on young adults of racial and ethnic minorities in how they manage self-reported symptoms of depression and anxiety. Moreover, non-Hispanic Black young adults reported the lowest unmet need for counseling or other mental healthcare treatment during the COVID-19 pandemic [11]. However, the literature is mixed on mental healthcare needs among racial and ethnic minorities, as non-Hispanic Black and Hispanic individuals have also indicated less access to mental healthcare during this time period [26]. Further research should consider mechanisms of adaptive coping as part of racial and ethnic identity as well as part of the prolonged COVID-19 pandemic as seen with other large-scale traumatic events while also investigating barriers to mental healthcare treatment and access among racial and ethnic minorities [11, 26, 58].

During COVID-19, minorities and young people experienced increased racial bias and housing instability while young Black people from low SES families were identified for heightened risk of negative mental health outcomes during the pandemic due to the joint nature of these identities [23, 24, 34, 59]. However, in our sample, although the interaction of race, ethnicity and SES was not significant, stratified analyses showed that for severity of symptoms of depression and anxiety, low income was highly predictive of increased odds of reporting both depressive and anxious symptoms for young adults regardless of racial and ethnic group. Stratifying by race and ethnicity unveiled a stronger relationship between low SES and negative mental health outcomes, while stratifying by SES showed a consistent relationship between racial and ethnic minorities and negative mental health outcomes. This may partially be due to the diverse cohort and Texas-representative sample in this study, as prior studies investigating these combined effects analyzed primarily Non-Hispanic White samples from less diverse regions; regardless, further investigation is needed to parse out joint effects [19, 20, 22].

The study’s limitations include the possibility of residual confounding due to unmeasured confounders like mental health stigma, knowledge, and attitudes, the potential for misclassification and recall bias due to self-reported data, and a large number of zero counts, all of which would bias results towards the null. SES was also dichotomized from five categories, indicating loss of information and thus, power. Additionally, SES was not assessed via a standard measure of household income but rather as a measure of perceptions of the participant’s family’s standard of living. Thus, misclassification of the exposure may also bias results towards the null. Generalizability is limited due to potentially informative attrition and selection bias based on male sex as well as the Texas-only sample. However, the sample is comparable to young adults aged 16–25 in Texas in 2021; nevertheless, findings may not be generalizable to less diverse regions, as Texas has the largest non-Hispanic Black population and second largest Hispanic population of any state in the United States [1, 3, 60-62]. Strengths of this study include its large, diverse, primarily non-White, population-based sample, and high retention rates.

Overall, this study sought to investigate how differences in mental health symptomatology were experienced across racial, ethnic and SES categories among a diverse, longitudinal cohort of young adults not only as a function of severity of illness but also as a function of the ongoing COVID-19 pandemic. Although racial and ethnic minority groups did not show significantly elevated outcomes as shown in initial COVID-19 research, negative mental health was persistently elevated among low SES youth throughout the prolonged COVID-19 period [23, 24]. While racial and ethnic mental health disparities among minority groups may not persist, poor SES status did persist across this period as highly predictive of negative mental health outcomes. Findings support direct efforts in prioritizing the mental health of youth reporting low SES, especially during large-scale public health crises. Mental health practitioners and policymakers should consider targeting constructs such as self-efficacy and familial support among low SES youth while also considering tailored collaborative care initiatives, as these have shown some effectiveness in ameliorating mental health issues among this population as they transition into emerging adulthood [63, 64].