Social class discrimination as a predictor of first cigarette use and transition to nicotine use disorder in Black and White youth

Abstract

Purpose:

To characterize the association of social class discrimination with the timing of first cigarette use and progression to DSM-IV nicotine dependence (ND) in Black and White youth, examining variation by race, parent vs. youth experiences of discrimination, socioeconomic status (SES), and stage of smoking.

Methods:

Data were drawn from 1,461 youth (55.2% Black, 44.8% White; 50.2% female) and mothers in a high-risk family study of alcohol use disorder and related conditions. Cox proportional hazards regression analyses were conducted, using youth’s and mother’s social class discrimination to predict first cigarette use and progression to ND, stratifying by race. Interactions between discrimination and SES indicators (parental education and household income) were tested. Adjusted models included psychiatric covariates.

Results:

In the adjusted first cigarette use models, neither youth’s nor mother’s social class discrimination were significant predictors among Black youth, but mother’s discrimination was associated with increased risk (HR=1.53 [1.18-1.99]) among White youth. In the adjusted ND models, mother’s discrimination was associated with reduced ND risk for Black youth in middle income families (HR=0.29 [CI:0.13-0.63]), but neither youth’s nor mother’s discrimination predicted transition to ND among White youth.

Conclusions:

The observed race and smoking stage specific effects suggest that social class discrimination is more impactful on early stages of smoking for White youth and later stages for Black youth. The robustness of links with mother’s discrimination experiences further suggest the importance of considering family level effects and the need to explore possible mechanisms, such as socialization processes.

Introduction

Black youth are less likely than their White counterparts to initiate cigarette smoking [1] and those who do are at lower risk than White youth to progress to nicotine use disorder [2, 3]. However, smoking cessation rates are lower and tobacco related health conditions in adulthood are greater in Black than White individuals [4-7], underscoring the importance of elucidating pathways of risk for cigarette smoking in Black youth, including how they may vary from those in White youth. In the U.S., reflective of historical structural inequalities, the median household income in Black families is lower than that of White families [8] and fewer Black than White adults hold college degrees [9]. Low socioeconomic status (SES) has consistently been linked to increased risk for adolescent cigarette smoking [10-12], a seemingly contradictory finding, given the elevated likelihood of Black vs. White youth to live in low-SES households and the lower prevalence of smoking among Black adolescents. Few studies have investigated possible race differences in the link between SES and smoking; thus, these seemingly paradoxical findings may reflect the fact that etiological models of cigarette smoking are based in large part on majority White samples. In fact, one of the few studies to examine race differences revealed that among Black but not White adolescents, living in a high-income neighborhood was associated with elevated likelihood of smoking [13].

A related understudied risk factor for cigarette smoking is social class discrimination: marginalization or unfair treatment on a personal or systemic level because of one’s social class [14]. Social class discrimination – sometimes referred to as classism - has been tied to elevated risk for depression [15], sleep problems [16], overall poor mental health [17], and in prior studies based on the current study sample, risky behaviors [18] and alcohol use [19], but we are aware of only one study examining its relation to cigarette smoking. A study conducted in the Netherlands with individuals 16 to 90 years of age found an elevated prevalence of current regular smoking among those who endorsed classism, but the association did not remain significant after accounting for income [14]. The few investigations to consider moderation of social class discrimination by SES indicators found no evidence of variation in its impact on smoking [14], mental health [14], or sleep problems [16]. However, when assessed more broadly, i.e., without attributing discrimination to a specific characteristic, among Latinx and Asian adults, low education and low income, respectively, were found to exacerbate risk conferred by discrimination for substance use disorder, but notably, no evidence of moderation by these SES indicators was observed in Black adults [20]. Furthermore, despite higher rates of endorsement of social class discrimination among Black vs. White adults [15-17, 21], in one of the only known studies to examine its association with SES indicators separately by race, low SES was associated with endorsement of social class discrimination only among White adults [17]. However, in addition to the limited power for detecting an effect among Black participants, the challenge of differentiating racial from social class discrimination for many Black individuals [22] may have been a factor.

Investigating potential race differences in the link between social class discrimination and smoking behaviors is an important step in refining etiological models of smoking in Black individuals that lays the groundwork for tailored interventions, but this has yet to be undertaken. Several studies of racial discrimination and smoking have compared associations in White vs. racial/ethnicity minority participants, with some [23-25] but not all [26] finding associations across groups, but given the challenge of interpreting racial discrimination endorsement by White vs. historically oppressed racial/ethnic groups, a parallel with social class discrimination cannot be drawn. However, the finding from a study by Kendzor and colleagues that non-attribute specific discrimination was associated with elevated Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV [30] nicotine dependence (ND) risk across Black, Latinx, and White adults [27] suggests that social class discrimination may impact smoking in both Black and White populations. Possible distinctions by stage of smoking in the impact of social class discrimination - a key consideration in prevention efforts - also has yet to be investigated.

The current study was designed to characterize the role of social class discrimination and its potential variation by SES in the timing of first cigarette use and progression to nicotine use disorder (operationalized as DSM-IV ND) in Black and White youth. We examined both the youth’s and the mother’s discrimination experiences in relation to youth smoking outcomes to capture social class discrimination on a family level, given the evidence for intergenerational transmission of discrimination related risk for psychiatric outcomes in youth [28, 29]. We hypothesized a higher magnitude of effect in Black compared to White youth, as the elevated baseline level of discrimination among Black youth, who are universally exposed to racial discrimination, could have a compounding effect on smoking risk. We further anticipated that the impact of social class discrimination would be stronger among youth from low-income families or whose parents had relatively low levels of education. Finally, we expected that social class discrimination would be associated with both outcomes but, given that trying a cigarette is normative, the association with first cigarette use would be limited to early initiation.

Methods

Participants

The sample was composed of participants in the Missouri Family Study, a multi-wave high-risk family study of alcohol use disorder (AUD) and related conditions conducted from 2003-2012. Families at high risk for AUD were identified through two ascertainment strategies. In the first, families with a child aged 13, 15, 17, or 19 (index child) and one or more full siblings were identified through Missouri birth records. Biological mothers were contacted to complete a brief telephone screen to assess level of family risk for AUD based on the biological father’s drinking history. Families in which the mother reported that the father had a history of excessive drinking were classified as “high-risk” and all others were classified as “low-risk.” The second ascertainment strategy used data from Missouri driving records in combination with birth records to identify children in the target age range with biological fathers who had two or more drunk driving convictions. These families were classified as “very high-risk.” After mothers completed interviews that included assessment of the father’s DSM-IV AUD symptoms, a small number of families were reclassified as false positive if originally classified as high-risk but fathers did not meet AUD criteria, false negative if originally classified as low-risk but fathers met AUD criteria. (The distribution for family risk group is reported by race in Supplemental Table S1.)

After completing their interviews, mothers were asked permission to contact the index child and up to two full siblings. Fathers and offspring whose mothers granted permission were invited to participate in the study. Enrollment occurred over six years. By design, Black families were oversampled to provide a sufficient sample size to meet the overall study objective of investigating whether models of alcohol problem development common in the research literature that were based largely on White youth were equally pertinent to Black youth. The study’s aim to enroll 450 Black families and 300 White families (of any ethnicity) across the three risk levels was met, with 450 Black and 317 White families participating. Participants in three of the intake years completed up to four waves of data collection at two-year intervals; the remainder completed one or two. In total, 1,461 offspring (55.2% Black, 50.3% female; 659 index children, 802 siblings) completed at least one interview; 75% of offspring completed two or more. The mean age at first and last interviews were 17.6 (SD=3.8) and 21.4 (SD=4.3), respectively. (Detailed sample characteristics are reported by race in Supplemental Table S1).

Procedure and assessment battery

Data were collected via telephone by trained interviewers using a version of the Semi-Structured Assessment for the Genetics of Alcoholism [31, 32] modified for telephone administration. DSM-IV psychiatric disorders, substance use history, and related psychosocial domains (e.g., childhood maltreatment), demographic characteristics, and discrimination experiences were queried. The Washington University School of Medicine Human Research Protections Office and the Ethics Board of the State Department of Health and Senior Services approved the protocol. Informed consent was obtained from participants age 18 or older; both parental consent and offspring assent were obtained for participants under age 18.

Operationalization of key constructs and covariates

Social class discrimination

A version of the Experiences of Discrimination Scale [33], modified to include social class based discrimination in addition to racial discrimination [16, 17], was administered to youth and mothers, using a separate series of questions for each type of discrimination. Social class discrimination was assessed by asking: “Because of your social class (that is, your social or economic class) have you ever experienced discrimination (been prevented from doing something, or been hassled or made to feel inferior in any of the following situations)?” Frequency and distress level were queried for seven situations: at school, getting a job, at work, at home, getting medical care, on the street or in a public setting, and from the police or in the courts. Given our interest in any degree of exposure to discrimination and the nearly universal endorsement of some level of distress, social class discrimination was coded dichotomously as present if any experiences at any frequency were endorsed. (Correlations between mother’s and youth’s discrimination reports ranged from only r=0.11-0.15, indicating that they capture distinct constructs.)

Youth cigarette smoking outcomes

Lifetime cigarette use was defined as a “yes” response at any wave of data collection to the question, “Have you ever tried cigarette smoking, even a puff?” [34], consistent with our interest in capturing any level of cigarette use. Age at first cigarette was established by asking, “How old were you the very first time you smoked even a puff of a cigarette?”

Nicotine dependence was defined according to DSM-IV criteria (as data collection preceded publication of the DSM-5): endorsement of at least three of seven dependence symptoms occurring within the same 12-month period. Age at ND onset was defined as the first age that this clustering occurred.

In cases where participants reported an age at first cigarette or ND onset in more than one interview, we used the first report, assuming higher accuracy in closer proximity to the event.

Socioeconomic status indicators

SES was indexed by mother’s report of maternal and paternal education levels (<high school, high school, and >high school; with high school including GED equivalent) and household income (<$30,000 [low], $30,000-$75,000 [middle], and >$75,000 [high]).

Covariates

Lifetime youth-reported psychiatric and psychosocial risk factors and mother-reported history of maternal and paternal regular smoking (100 or more cigarettes over the lifetime [35]) were modeled as covariates. In addition to history of (any) alcohol and cannabis use, DSM-IV conduct disorder and major depressive disorder, two forms of child maltreatment (<age 16) were included. Childhood physical abuse/harsh punishment and neglect and sexual abuse were operationalized as endorsement of the event as described on a standard trauma checklist (e.g., physically abused, seriously neglected, raped) or as described in the early childhood experiences section of the interview (e.g., sexual contact with a family member <age 16, hurt on purpose by an adult). Youth reported age at first experience for all psychiatric and psychosocial factors. Consistent with our coding of smoking outcomes, if age at first experience was reported in more than one interview, we used the first report.

Data analysis

Analyses were stratified by race, given the substantial covariate imbalance between Black and White participants, which creates untrustworthy results in pooled analyses (see Imbens and Rubin [36]). Cox proportional hazards (PH) regression analyses were conducted to predict first cigarette use and progression from first cigarette to ND as a function of social class discrimination. This survival analysis approach is well suited to an adolescent/young adult sample, as it accounts for the possibility that participants have not passed through the period of risk. First cigarette was the starting point in the ND models, which adjusted for age at first cigarette by including dummy variables representing early (≤12) and late (≥17) age at first cigarette (reference group=13-16). Variables representing socioeconomic status and parental history of regular smoking were time-invariant. To ensure that only risk factors that preceded smoking outcomes were treated as predictors, psychiatric and psychosocial covariates were entered as time-varying. As age at first discrimination experience was not obtained, social class discrimination was coded, separately for first cigarette and ND models, as present only if it was endorsed (1) in a wave of data collection prior to the one in which the smoking outcome was first reported or (2) in the same wave, except for the baseline (lifetime) assessment, which covered too wide of a timeline to assume that discrimination experiences preceded smoking outcomes. If the reported age at first cigarette or ND onset was younger than the participant’s age at the time discrimination was first endorsed, it was recoded to absent.

Analyses were conducted in Stata [37], applying the cluster sandwich estimator to account for the non-independence of observations among siblings. Listwise deletion was used, resulting in the deletion of a small percentage of cases (ranging from 0.50-6.85% across models), with no evidence of systemic differences by race. Violations of the PH assumption, which posits that over time, the hazard linked to a risk factor remains proportional, were identified using Schoenfeld residuals as computed by the Grambsch and Therneau test [38]. All violations were resolved by splitting the risk period and estimating hazard ratios for each period. In addition to those shown in the tables, adjustments for PH violations were made in the base first cigarette model for Black youth (gender), the base and adjusted ND models for Black youth (age at first cigarette), and the base ND model for White youth (age at last interview and age at first cigarette). With the exceptions of age at first cigarette and paternal regular smoking in the base and adjusted ND models respectively, for White youth and the interaction between youth discrimination and high maternal education in the adjusted ND model for Black youth, the separate hazard ratios generated from splitting risk periods differed significantly from each other.

Modeling was conducted in two steps, first establishing associations of social class discrimination with smoking outcomes without accounting for the potential influence of psychiatric and psychosocial covariates on this pathway (base model), then including those covariates in the model in order to estimate the independent contributions of social class discrimination to smoking outcomes (adjusted model). The base model included dichotomous variables representing youth’s and mother’s experience of social class discrimination in addition to dummy variables representing age at last interview (≤19 and ≥24; reference group=20-23), family risk group (false positive, high, false negative, and very high; reference group=low), household income (high and low; reference group=middle), and parental education levels (<high school and >high school; reference group=high school). Interactions between discrimination and SES indicator variables as well as gender were tested and significant interactions were retained. To enhance interpretability of interactions, separate hazard ratios were generated for each level of the variable in the significant interaction. (For example, if a significant interaction between youth’s social class discrimination and gender were observed, separate hazard ratios would be generated for female and male youth). The adjusted model included the covariates listed above in addition to all of the base model variables.

Results

Social class discrimination, SES indicators, and cigarette smoking outcomes by race

A significantly higher proportion of Black than White youth reported experiencing social class discrimination (26.3% vs. 15.6%; χ²₍₁₎=24.66; p<0.001), as did their mothers (35.1% vs. 16.3%; χ²₍₁₎=65.10; p<0.001). Cigarette use was significantly lower in Black than White youth (62.1% and 69.8%, respectively; χ²₍₁₎=9.38; p<0.01), but mean age at first cigarette did not differ significantly (14.5 [SD=3.4] vs. 14.1 [SD=3.5]; t₍₉₆₀₎₌1.73, p=0.08). The proportion of ever smokers meeting ND criteria was significantly lower for Black than White youth (32.6% vs. 42.0%; χ²₍₁₎=9.07; p<0.01) and mean age at ND onset was significantly later (18.7 [SD=3.3] vs. 17.3 [SD=2.8]; t_(315.9)=4.21, p<0.001), but mean years from first cigarette to ND did not differ significantly (4.4 [SD=3.8] vs. 4.5 [SD=3.3]; t₍₃₅₁₎=0.42, p=0.67). All three SES indicators were significantly lower for Black youth: household income (χ²₍₂₎=144.77, p<0.001), maternal education level (χ²₍₂₎=12.21, p<0.01), and paternal education level (χ²₍₂₎=27.27, p<0.001). (See Table 1.)

Table 1.

Social class discrimination, cigarette smoking outcomes, and sample characteristics by race

	Black	White	Test statistic
Social Class Discrimination
Youth**	26.3%	15.6%	χ2(1)=24.66
Mother**	35.1%	16.3%	χ2(1)=65.10
Youth Cigarette Smoking Outcomes
Initiation of cigarette smoking
Ever smoke a cigarette	62.1%	69.8%	χ2(1)=9.38
Age smoked first cigarette: Mean (SD)	14.5 (3.4)	14.1 (3.5)	t(960)=1.73
Nicotine dependence
DSM-IV criteria met (among ever smokers)*	32.6%	42.0%	χ2(1)=9.07
Age at onset: Mean (SD)**	18.7 (3.3)	17.3 (2.8)	t(315.9)=4.21
Years from first cigarette to nicotine dependence: Mean (SD)	4.4 (3.8)	4.5 (3.3)	t(351)=0.42
0 (same year)	13.0%	7.3%
1-2 years	25.9%	20.4%
3-5 years	27.8%	40.9%
6+ years	33.3%	31.4%
Socioeconomic Status Indicators
Household income**			χ2(2)=144.77
Low ($0-29,999)	54.0%	24.7%
Middle ($30,000-49,999)	19.5%	21.3%
High ($50,000 or higher)	26.5%	53.9%
Mother’s education level*			χ2(2)=12.21
< High school	10.8%	5.8%
High school	30.1%	29.8%
> High school	59.1%	64.4%
Father’s education level **			χ2(2)=27.27
< High school	17.5%	18.2%
High school	50.8%	37.8%
> High school	31.7%	44.0%

SD=standard deviation

^**p<0.001

^*p<0.01

Social class discrimination as a predictor of cigarette smoking stages

Results of Cox PH regression analyses predicting timing of first cigarette in the Black and White subsamples are shown with 95% confidence intervals (CIs) in Tables 2 and 3, respectively. Results of analyses predicting transition from first cigarette to ND are shown in Tables 4 and 5. (Hazard ratios [HRs] reported below are from adjusted models unless otherwise specified and differences refer to statistically significant differences.)

Table 2.

Results of Cox proportional hazards (PH) regression analyses predicting timing of first cigarette as a function of youth’s and mother’s experience of social class discrimination: Black youth

	Baseline Model HR (95% CI)	Adjusted Model HR (95% CI)
Social Class Discrimination
Youth	1.04 (0.84-1.28)	0.91 (0.73-1.14)
Mother
x Low household income	1.28 (0.98-1.66)	1.30 (1.00-1.69)
x Middle to high household income	0.81 (0.62-1.06)	0.78 (0.59-1.04)
Socioeconomic Status Indicators
Household income a
Low	0.87 (0.67-1.13)	0.89 (0.67-1.18)
High	0.74 (0.57-0.97)	0.76 (0.57-1.02)
Mother’s education level b
< High school	0.93 (0.67-1.28)	0.87 (0.62-1.21)
> High school	0.87 (0.70-1.08)	0.95 (0.76-1.18)
Father’s education level b
< High school	1.31 (1.03-1.68)	1.29 (1.00 c -1.66)
> High school
1st cigarette ≤ age 17 d	1.07 (0.75-1.55)	1.13 (0.87-1.45) e
1st cigarette ≥ age 18 d	0.67 (0.42-1.07)
Psychosocial/Psychiatric Risk Factors and Family History
Youth alcohol use	-	1.62 (1.27-2.08)
Youth cannabis use		3.19 (2.56-3.98)
Youth major depressive disorder	-	1.29 (0.97-1.72)
Youth conduct disorder	-	0.90 (0.62-1.31)
Youth childhood physical abuse/harsh punishment or neglect	-	0.90 (0.72-1.11)
Youth childhood sexual abuse	-	1.44 (1.11-1.87)
Mother regular smoker (lifetime)		1.25 (1.03-1.52)
Father regular smoker (lifetime)	-	1.11 (0.90-1.37)

HR=hazard ratio; CI=confidence interval; bold indicates significant at p<0.05; x indicates that separate hazard ratios were estimated for different levels of a variable in a significant interaction

^acomparison group=middle household income

^bcomparison group=high school level education

^crounded down from 1.003

^drisk period-specific hazard ratios generated to adjust for violation of PH assumption

^eonly one value reported because PH assumption was not violated in this model. Both models included gender, age at last interview, and family risk group.

Table 3.

Results of Cox proportional hazards (PH) regression analyses predicting timing of first cigarette as a function of youth’s and mother’s experience of social class discrimination: White youth

	Baseline Model HR (95% CI)	Adjusted Model HR (95% CI)
Social Class Discrimination
Youth
1st cigarette ≤ age 13 a	1.57 (1.01-2.45)	0.85 (0.65-1.13) b
1st cigarette ≥ age 14 a	0.85 (0.61-1.17)
Mother	1.41 (1.09-1.83)	1.53 (1.18-1.99)
Socioeconomic Status Indicators
Household income c
Low
1st cigarette ≤ age 13 a	1.52 (1.03-2.23)	1.39 (0.94-2.04)
1st cigarette ≥ age 14 a	0.87 (0.61-1.23)	0.86 (0.61-1.22)
High	0.84 (0.63-1.13)	0.83 (0.63-1.10)
Mother’s education level d
< High school	1.31 (0.94-1.82)	1.32 (0.97-1.79)
> High school	0.85 (0.66-1.09)	0.76 (0.59-0.98)
Father’s education level d
< High school	1.03 (0.78-1.36)	1.04 (0.79-1.37)
> High school
1st cigarette ≤ age 15 a	0.51 (0.36-0.74)	0.54 (0.37-0.77)
1st cigarette ≥ age 16 a	1.17 (0.87-1.59)	1.25 (0.92-1.70)
Psychosocial/Psychiatric Risk Factors and Family History
Youth alcohol use	-	2.75 (2.09-3.63)
Youth cannabis use		2.48 (1.85-3.31)
Youth major depressive disorder	-	1.09 (0.79-1.51)
Youth conduct disorder	-	0.83 (0.45-1.54)
Youth childhood physical abuse/harsh punishment or neglect	-	1.13 (0.93-1.38)
Youth childhood sexual abuse	-	1.38 (1.00-1.90)
Mother regular smoker (lifetime)		1.11 (0.87-1.41)
Father regular smoker (lifetime)	-	0.98 (0.77-1.26)

HR=hazard ratio; CI=confidence interval; bold indicates significant at p<0.05

^arisk period-specific hazard ratios generated to adjust for violation of PH assumption

^bonly one value reported because PH assumption was not violated in this model

^ccomparison group=middle household income

^dcomparison group=high school level education. Both models included gender, age at last interview, and family risk group.

Table 4.

Results of Cox proportional hazards (PH) regression analyses predicting transition from first cigarette use to nicotine dependence onset as a function of youth’s and mother’s experience of social class discrimination: Black youth

	Baseline Model HR (95% CI)	Adjusted Model HR (95% CI)
Social Class Discrimination
Youth
x Mother’s education ≤ High school	0.69 (0.41-1.19)	1.15 (0.73-1.82)
x Mother’s education > High school
ND onset ≤ age 18 a	1.53 (1.03-2.26) b	0.31 (0.09-1.07)
ND onset ≥ age 19 a		0.72 (0.42-1.22)
Mother
x Low household income	1.36 (0.86-2.15)	1.21 (0.76-1.94)
x Middle household income	0.27 (0.12-0.59)	0.29 (0.13-0.63)
x High household income	1.20 (0.49-2.91)	0.90 (0.38-2.10)
Socioeconomic Status Indicators
Household income c
Low
ND onset ≤ age 18 a	0.87 (0.45-1.68)	0.63 (0.38-1.03) b
ND onset ≥ age 19 a	0.46 (0.26-0.79)
High	0.53 (0.30-0.96)	0.56 (0.30-1.06)
Mother’s education level d
< High school	0.86 (0.53-1.39)	1.05 (0.64-1.72)
> High school	0.72 (0.43-1.22)	0.82 (0.51-1.31)
Father’s education level d
< High school	0.88 (0.60-1.29)	0.74 (0.50-1.08)
> High school
ND onset ≤ age 17 a	0.59 (0.36-0.97) b	0.46 (0.23-0.90)
ND onset > age 18 a		0.92 (0.46-1.82)
Psychosocial/Psychiatric Risk Factors and Family History
Youth alcohol use	-	3.87 (2.14-7.00)
Youth cannabis use		2.85 (1.66-4.87)
Youth major depressive disorder	-	1.52 (1.02-2.26)
Youth conduct disorder	-	0.96 (0.64-1.44)
Youth childhood physical abuse/harsh punishment or neglect	-	1.30 (0.86-1.97)
Youth childhood sexual abuse	-	1.14 (0.74-1.76)
Mother regular smoker (lifetime)		0.95 (0.66-1.36)
Father regular smoker (lifetime)	-	1.23 (0.79-1.90)

HR=hazard ratio; CI=confidence interval; ND=nicotine dependence; bold indicates significant at p<0.05; x indicates that separate hazard ratios were estimated for different levels of a variable in a significant interaction

^arisk period-specific hazard ratios generated to adjust for violation of PH assumption

^bonly one value reported because PH assumption was not violated in this model

^ccomparison group=middle household income

^dcomparison group=high school level education. Both models included gender, age at last interview, family risk group, and age at first cigarette.

Table 5.

Results of Cox proportional hazards (PH) regression analyses predicting transition from first cigarette use to nicotine dependence onset as a function of youth’s and mother’s experience of social class discrimination: White youth

	Baseline Model HR (95% CI)	Adjusted Model HR (95% CI)
Social Class Discrimination
Youth	1.14 (0.79-1.64)	0.81 (0.53-1.24)
Mother	1.27 (0.87-1.85)	1.46 (0.96-2.22)
Socioeconomic Status Indicators
Household income a
Low	1.03 (0.69-1.53)	1.14 (0.75-1.75)
High	1.10 (0.74-1.62)	1.15 (0.79-1.66)
Mother’s education level b
< High school	2.17 (1.37-3.44)	1.55 (0.92-2.59)
> High school
ND onset ≤ age 18 c	0.88 (0.64-1.22) d	1.08 (0.70-1.68)
ND onset ≥ age 19 c		0.57 (0.36-0.89)
Father’s education level b
< High school	0.99 (0.69-1.42)	0.75 (0.50-1.12)
> High school	0.70 (0.46-1.07)	0.54 (0.34-0.85)
Psychosocial/Psychiatric Risk Factors and Family History
Youth alcohol use	-	3.06 (1.62-5.75)
Youth cannabis use		3.40 (2.22-5.22)
Youth major depressive disorder	-
ND onset ≤ age 18 c		2.51 (1.59-3.96)
ND onset ≥ age 19 c		1.35 (0.75-2.40)
Youth conduct disorder	-	1.01 (0.61-1.67)
Youth childhood physical abuse/harsh punishment or neglect	-	1.38 (1.00-1.90)
Youth childhood sexual abuse	-	1.55 (0.98-2.45)
Mother regular smoker (lifetime)		1.17 (0.78-1.76)
Father regular smoker (lifetime)	-
ND onset ≤ age 16 c		0.71 (0.39-1.28)
ND onset ≥ age 17 c		1.04 (0.64-1.67)

HR=hazard ratio; CI=confidence interval; ND=nicotine dependence; bold indicates significant at p<0.05

^acomparison group=middle household income

^bcomparison group=high school level education

^crisk period-specific hazard ratios generated to adjust for violation of PH assumption

^donly one value reported because PH assumption was not violated in this model. Both models included gender, age at last interview, family risk group, and age at first cigarette.

Age at first cigarette

Black youth

Youth’s social class discrimination was not associated with first cigarette use (HR=1.04 [CI:0.84-1.28]) in Black youth. A significant interaction between mother’s social class discrimination and low household income was observed, so HRs were estimated separately for low and average to high household income. HRs were non-significant at both levels (HR=1.30 [CI:1.00-1.69] and HR=0.78 [CI:1.00-1.69], respectively). Paternal education level <high school (HR=1.29 [CI:1.00-1.66]) and mother’s history of regular smoking (HR=1.25 [CI:1.03-1.52]) in addition to youth’s alcohol use (HR=1.62 [CI:1.27-2.08]), cannabis use (HR=3.19 [CI:2.56-3.98]), and history of childhood sexual abuse (HR=1.44 [CI:1.11-1.87]) were associated with increased risk for first cigarette use.

White youth

Among White youth, in the base model, youth’s social class discrimination was associated with elevated risk for first cigarette use at age 13 or younger (HR=1.57 [CI:1.01-2.45]), but it was non-significant across ages after adjustment for covariates (Adjusted model: HR=0.85 [CI:0.65-1.13]). Mother’s social class discrimination was associated with increased risk for first cigarette use (HR=1.53 [CI:1.18-1.99]). Both maternal and paternal education levels >high school were associated with reduced risk for first cigarette use. Reduced risk was also observed across ages for maternal education >high school (HR=0.76 [CI:0.59-0.98]) and for first cigarette ≤age 15 (HR=0.54 [CI:0.37-0.77]) for paternal education >high school. Youth’s alcohol use (HR=2.75 [CI:2.09-3.63]) and cannabis use (HR=2.48 [CI:1.85-3.31]) were associated with elevated risk for first cigarette use.

Transition from first cigarette to ND

Black youth

Among Black youth, a significant interaction between youth’s social class discrimination and maternal education >high school was observed, so HRs were estimated separately for maternal education levels ≤high school and >high school. In the base model, the youth social class discrimination HR was significant for >high school (HR=1.53 [CI:1.03-2.26] but not ≤high school (HR=0.69 [CI:0.41-1.19]) in predicting transition from first cigarette to ND. In the adjusted model, HRs for youth social class discrimination were non-significant across levels of maternal education. Significant interactions between mother’s social class discrimination and both high and low household income variables were observed, so HRs were estimated separately for each income level. The HRs at low, middle, and high-income levels were 1.21 [CI:0.76-1.94], 0.29 [CI:0.13-0.63], and 0.90 [CI:0.38-2.10], respectively. Paternal education >high school was associated as well with reduced risk specifically for ND onset ≤age 17 (HR=0.46 [CI:0.23-0.90]). Youth’s alcohol use (HR=3.87 [CI:2.14-7.00]), cannabis use (HR=2.85 [CI:1.66-4.87]), and major depressive disorder (HR=1.52 [CI:1.02-2.26]) were associated with elevated risk for ND onset.

White youth

Neither youth’s nor mother’s social class discrimination predicted transition from first cigarette to ND in White youth (HRs=0.81 [CI:0.53-1.24] and 1.46 [CI:0.96-2.22]). Both maternal and paternal education >high school were associated with reduced risk for progressing to ND, with maternal education effects specific to ND onset ≥age 19 (HR=0.57 [CI:0.34-0.85]) and paternal education effects consistent across ages (HR=0.54 [CI:0.0.34-0.85]). Youth’s alcohol use (HR=3.06 [CI:1.62-5.75]), cannabis use (HR=3.40 [CI:2.22-5.22]), and major depressive disorder (specific to ND onset ≤age 18: HR=2.51 [CI:1.59-3.96]) were associated with elevated risk for ND onset.

Discussion

The current study investigated the association of social class discrimination with first cigarette use and ND onset in Black and White youth, revealing variations in the association by race, SES indicators, and stage of smoking. Our inclusion of both youth’s and mother’s discrimination experiences further allowed us to identify independent contributions of social class discrimination at the parental and individual level on youth smoking outcomes. Our findings paint a more complex picture of the link between social class discrimination and progression of youth smoking than we hypothesized.

The one known prior study to examine the link between social class discrimination and cigarette smoking, conducted with adults in the Netherlands, did not differentiate racial/ethnic groups or assess for parental experiences of discrimination, and examined only one smoking outcome: current regular smoking [14]. Findings from the present investigation are broadly consistent with that study, in that we observed significant associations between social class discrimination and certain smoking outcomes and some did not differ by SES indicators or were non-significant in the context of related risk factors. Most notably, with respect to our hypotheses and study design, associations were not universal in terms of race, outcome, the relevance of SES indicators, or parent vs. youth discrimination experiences. The association with social class discrimination was specific to White youth for first cigarette use and to Black youth for progression to ND (both only before accounting for psychiatric and psychosocial correlates); it varied by SES indicators for Black but not White youth for ND only; and links between mother’s but not youth’s experiences of discrimination and smoking outcomes were robust in the context of correlated psychiatric and psychosocial factors.

We did not find support for the hypothesized overall higher magnitude of effect of social class discrimination on smoking outcomes in Black compared to White youth. In the absence of prior work comparing social class discrimination effects on substance use or other health behaviors in Black vs. White individuals, we could look to studies of racial discrimination that assessed race differences, but given the lack of equivalence of perceived racial discrimination and its impact in marginalized vs. non-marginalized groups [39, 40], we do not consider them to be valid reference points. The specificity of social class discrimination effects to first cigarette in White youth and progression to ND in Black youth is suggestive of distinctions by race in the manifestation of risk conferred by social discrimination early vs. later in the course of cigarette smoking. The findings are in fact largely consistent with our hypothesis that discrimination effects would be observed specifically for early age at first cigarette – given that early initiation is a marker of risk for later problems, whereas trying a cigarette at some point is normative - which we found for White youth’s own discrimination (in the base model). They are also aligned with our hypothesis that risk for ND – a problem behavior by definition - would manifest across the risk period, which we found for Black youth’s own discrimination experiences (in the base model) and mother’s discrimination experiences.

Evidence for variation by SES indicators in the magnitude of effect for social class discrimination on smoking outcomes emerged, namely for ND in Black youth, but it was largely inconsistent with our expectation that youth from lower SES backgrounds would be at elevated risk. Results from the base (but not the adjusted) model of ND for Black youth indicated that experiencing social class discrimination was associated with increased risk for progressing to ND among youth whose mothers had greater than a high school level education. Mother’s experience of discrimination was also associated with progression to ND; decreased risk was observed specifically among Black youth from middle vs. low or high-income families. These findings highlight the need to address the impact of social class discrimination on smoking – and health outcomes more broadly - in Black youth at the high end of the SES continuum, a frequently overlooked subpopulation

Investigations of associations between social class discrimination and health outcomes have rarely assessed for variation by SES in its impact. The couple of studies that have done so found no differences by SES, but notably, one did not include a substantial number of Black participants [14] and the other did not examine interactions separately by race [16]. The most comparable smoking outcome study, which investigated discrimination broadly (i.e., not attribute-specific) across racial groups, found that ND risk conferred by discrimination was elevated among Latinx adults with less than a high school level of education and among Asian adults with a lower than average income. Contrary to our findings, the association between discrimination and ND did not vary by SES indicators in Black adults [20]. This inconsistency may be attributable to specificity of social class discrimination effects or too few Black participants from high SES backgrounds in their study to detect associations – important considerations for future research.

One final component of this study that merits discussion is the informativeness of including mother’s experience of social class discrimination. Parental experiences of discrimination have only rarely been included in studies examining the impact of discrimination on youth’s substance use or mental health [18, 19, 28, 29] (three of the four examples being work by our group using the present sample), but our results here indicate that they are as influential and more robust than youth’s own discrimination experiences. Of note, where social class discrimination effects were found – first cigarette in White youth and ND in Black youth - they were present at both the individual and parental level, indicating that they may reflect differential pathways, e.g., via increased stress for youth’s own experiences [24] and via socialization processes [41] akin to racial socialization [42] for mother’s experiences.

Limitations

The current study has certain limitations. First, as age at first experiences of social class discrimination was not directly assessed, the ordering of discrimination experiences and smoking outcomes may not be accurate for all participants. Second, parental education level and household income are commonly used SES indices but do not capture all components of SES. Third, given our interest in race differences and the lack of comparability of racial discrimination endorsement between Black and White individuals, we did not incorporate racial discrimination into analyses but consider the compounded impact of social class and racial discrimination in Black populations critical to address. Fourth, as we were interested in any degree of exposure to social class discrimination and coded it accordingly, inferences cannot be made about possible variation in impact by frequency of discrimination experiences. Fifth, the subjective nature of perceived discrimination and its attribution to a given characteristic, as well as differences in recall and willingness to report discrimination should be kept in mind when interpreting study findings [43].

Future directions

Replication with populations in other geographic regions, given geographic variation in smoking prevalence [44], cigarette excise taxes [45] and other relevant factors, is needed. Expansion of this line of research to other racial and ethnic groups as well as other health behaviors is essential for grasping the scope of the impact of social class discrimination. More detailed assessments of specific settings where social class discrimination is commonly experienced in addition to the socioenvironmental context, for instance, the match between neighborhood and family level SES, would also be highly informative. Most importantly, identifying the mechanisms driving the link between social class discrimination (including parental level discrimination) and youth cigarette smoking is critical to guiding prevention efforts, e.g., that capitalize on buffering effects of social support against risk conferred by discrimination on substance use that has been found for racial discrimination [46, 47].