Profiles of psychosocial risk and protective factors and their associations with alcohol use and regular smoking in Black adults

Abstract

Introduction:

This study aimed to identify the clustering of substance use related psychosocial risk and protective factors (subgroups) and the differential associations of those subgroups with current alcohol use and regular smoking among Black adults.

Methods:

Data were drawn from 4,462 participants (29% Afro Caribbean, 71% African American; median age=38; 63% female) in a nationally representative study of social, economic, and structural conditions and health in Black Americans. Latent classes, i.e., subgroups, were derived via latent profile analysis with 10 indicators representing social support and religious involvement (support); demands from family and religious community (demands); and socioeconomic and neighborhood factors and racial discrimination (adversity). Frequency of alcohol use and prevalence of regular smoking were compared across classes using regression analyses.

Results:

Four classes emerged: (1) high support, low demands and adversity, (2) high support and demands, low-moderate adversity, (3) low support and demands, low-moderate adversity, and (4) low support, high demands and adversity. Relative to Class 1, frequency of alcohol use and regular smoking prevalence were significantly higher only in Class 4.

Conclusions:

Results indicate substantive variations in the clustering of substance use related psychosocial risk and protective factors in Black adults. Furthermore, they suggest that neither the presence of high demands nor the absence of support alone differentiate likelihood of engaging in frequent alcohol use or regular smoking, but adverse experiences such as racial discrimination may be especially impactful.

Introduction

The prevalence of alcohol use is lower in Black compared to White populations in the U.S. [1], but Black drinkers appear to be more susceptible than their White counterparts to alcohol related physical health problems as well as negative social consequences associated with their alcohol use [2]. Similarly, although cigarette smoking is less prevalent in Black compared to White adults, Black smokers have lower rates of cessation and are at greater risk for smoking related health problems [3]. Identifying the psychosocial risk and protective factors associated with Black Americans’ drinking and cigarette smoking – two of the leading causes of preventable death in the U.S. [4] – is a critical public health goal.

Many well documented substance use related risk and protective factors are more prevalent in Black vs. White populations and thus merit particular attention in efforts to understand alcohol and smoking related health disparities. Black families are disproportionately represented in disadvantaged neighborhoods [5, 6] and low socioeconomic status households [7], both of which have been linked to problem drinking [8, 9] and smoking [10, 11]. Racial discrimination - which three in four Black American adults report experiencing [12] - has also been associated with elevated alcohol consumption [13] and smoking [14]. Furthermore, Black adults are more likely than White adults to experience multiple forms of social disadvantage [15] and interactive effects of different forms of disadvantage, e.g., racial discrimination and living below the poverty line, have been shown to elevate risk for problem drinking in Black adults [16]. Notably, one of the most consistently documented protective factors for both substances, religious involvement [15, 18], is greater in Black compared to White communities [19]. Social support, a broad acting protective factor against health problems [20] has likewise shown stronger effects in Black than White adults [21].

Although a wide range of psychosocial risk and protective factors for alcohol use and smoking have been identified, variation within Black populations in their co-occurrence and associated distinctions in outcomes have rarely been considered. That is, the extent to which protective factors such as religious involvement may exert their effects in the context of risk factors, e.g., financial difficulties, remains unknown. However, evidence that social support buffers risk for problem drinking and smoking conferred by racial discrimination [22, 23] suggests the importance of considering the joint contributions of psychosocial risk and protective factors to alcohol use and smoking to determine how they act in concert.

The present study aimed to identify prototypical patterns of endorsement, i.e., subgroups, of current substance use related psychosocial risk and protective factors among Black adults and the differential associations of those subgroups with current frequency of alcohol use and regular smoking. This is a critical step in tailoring intervention efforts, as the impact of a given risk factor on substance use outcomes may depend on the protective factors that co-occur with it. The present investigation focused on psychosocial risk and protective factors that are highly prevalent in Black Americans, including financial difficulties [7], neighborhood disadvantage [5, 6], racial discrimination [12], religious involvement [19], and support from friends, family, and religious community [24]. We included demands from family and religious community as well, a less commonly recognized but potentially stressful aspect of close social connections [21]. The use of an all-Black sample allowed for the assessment of these relationships without the confounding effects of known variations in the distributions of these factors by race, thus addressing how well they distinguish frequency of alcohol use and likelihood of regular smoking among Black adults. The inclusion of adults ranging in age from 18 to 65 allowed for the assessment of developmental stage differences in the relevance of psychosocial factors to drinking and smoking behaviors [25] and the inclusion of a large number of Afro Caribbeans provided a rare opportunity to examine these relationships in this understudied subpopulation.

At least three subgroups were expected to emerge. In addition to overall high liability (high on risk, low on protective factors) and overall low liability (low on risk, high on protective factors) subgroups, variation in the clustering of high vs. low levels of certain risk and protective factors, e.g., high religious involvement pairing alternatively with high and low racial discrimination, was anticipated. Furthermore, the evidence for lower alcohol use and smoking [3] despite the higher prevalence of many substance use related risk factors in Black compared to White adults suggests that psychosocial protective factors exert substantial buffering effects in Black adults. Thus, lower frequency of alcohol use and lower prevalence of regular smoking were expected in the subgroups high on protective factors, even in subgroups characterized by a substantial degree of exposure to risk factors.

Methods

Participants and Procedures

Data were drawn from The National Survey of American Life: Coping with Stress in the 21st Century (NSAL). NSAL is a large-scale study on the social, economic, and structural conditions, and mental and physical health of Black Americans that oversampled for Afro Caribbeans (who are underrepresented in psychosocial research). Data collection was conducted from 2001 to 2004 by the University of Michigan’s Institute for Social Research, using a national multi-stage probability design. The first stage used a stratification approach involving selection of primary sampling units and the second stage used a clustering approach of selecting area segments within each primary sampling unit. Details on design and procedures can be found in prior publications by the NSAL investigators [26]. Most interviews were conducted face-to-face (86%) in respondents’ homes, with the remaining 14% conducted over the telephone. Free and informed consent was obtained prior to starting the interview. The study was approved by University of Michigan’s Institutional Review Board. Respondents were compensated for their participation. The overall response rate was 72%, with 6,082 Black and White adults over age 18 completing interviews. Substance use data were only collected from Black (Afro Caribbean and African American) participants. The current investigation is based on the publicly available data set. Given the wide range and skewed distribution of age for participants over 65 and our interest in considering age group differences, we restricted analyses to adults aged 18–65. The final analytic sample consisted of 1312 Afro Caribbeans and 3150 African Americans (median age=38; 63% female).

Operationalization of Key Constructs

Latent Class (Subgroup) Indicators

Ten indicators, 2 single items and 8 derived via factor analyses, representing psychosocial risk and protective factors, were included in models. The indicators fell into three broad categories: support, demands, and adversity. Support (protective) consisted of three separate factors representing support from family (8 items), religious community (8 items), and friends (4 items). We also included religious involvement (13 items) in this category, as it can be a source of community level as well as individual level, e.g., spiritual, support. Demands (risk) consisted of separate factors representing demands from family (3 items) and religious community (3 items). (Demands from friends were not assessed.) Adversity consisted of three indicators representing socioeconomic and neighborhood factors: financial difficulties (7 items; risk), low neighborhood safety (2 items; high values reflecting high risk) and low neighborhood cohesion (1 item; high values reflecting high risk) and racial discrimination (risk), which was represented by score on the 10-item Everyday Discrimination Scale [27]. Detailed information on indicators, including wording of each item used to derive indicators and correlations of indicators with frequency of alcohol use and current regular smoking status, can be found in Supplemental Table 1.

Substance Use Outcomes

Past year frequency of alcohol use was assessed in participants who endorsed ever consuming at least 12 drinks a year with the question, ‘In the past 12 months, how often did you usually have at least one drink?’ The five response options were (1) less than once a month, (2) 1 to 3 days a month, (3) 1 to 2 days a week, (4) 3 to 4 days a week, and (5) nearly every day. Participants who did not endorse ever consuming at least 12 drinks a year were coded 0; thus, the possible range was 0–5. The alcohol variable was modeled as a continuous variable, as the scale suggests an underlying continuum and the method we used (referred to as the BCH method) is robust to non-normal outcome distributions [28].

Current regular smoking status (dichotomous) was assessed by first establishing lifetime regular smoking status, using the standard definition of having smoked at least 100 cigarettes over the lifetime, then asking those who met criteria for regular smoking if they currently smoke. Respondents who endorsed smoking at least 100 cigarettes and reported that they currently smoke were coded 1; all others were coded 0 for current regular smoking.

Covariates

We incorporated as covariates three demographic factors that have been linked to variation in alcohol and smoking outcomes: self-reported gender (female or male) [29, 30], educational attainment [31], and age [29,30] as well as Afro Caribbean or African American ethnicity, given the possible but understudied differences between Black ethnic groups. Dichotomous indicators of family history of problem drinking and family history of regular smoking were included in the alcohol and regular smoking models, respectively, to account for the well-documented contribution of familial liability to alcohol and smoking outcomes [32]. Family history was assessed by asking how many relatives the respondent knew to have had ‘problems from the use of alcohol’ (problem drinking) and ‘problems with the use of tobacco or smoking’ (regular smoking). Respondents who endorsed one or greater were asked, ‘How many had [this problem] bad enough to disturb and interfere with their life at times?’ Those who endorsed one or more relatives meeting this criterion were coded positive for family history of the corresponding condition.

Analytic Approach

Analyses were conducted in three stages, using Mplus [33], applying prescribed NSAL weighting procedures to adjust for stratification (selection of primary sampling units), clustering (selection of area segments within each primary sampling unit), and unequal probabilities of selection and nonresponse. In the first step, confirmatory factor analysis was performed to test the factor structures of 8 derived indicators, constraining indicators to load on a single factor and residuals to be uncorrelated. Results (reported in Supplemental Table 2) supported adequate model fit: Root Mean Square Error of Approximation (95% Confidence Intervals)=.020 (.019, .021), CFI=.894, TLI=.886, with all standardized factor loadings >.40 and significant at p<.001. In the second step, latent profile analysis (LPA) was conducted, with all indicators entered into these models, selecting 2-, 3-, 4-, 5-, and 6-class solutions. Auxiliary variables (the covariates listed above as well as variables relating to lifetime history and current alcohol use and regular smoking) were included in each model to allow classes to be correlated with but not indicated by the covariates and outcomes for the planned regression analyses (see Table 3). The final LPA model selection was based on Bayesian information criterion (which takes into account that data were not generated from a simple random sample), entropy, smallest class size, probability of class membership > .80, parsimony, and theoretical interpretability.

Table 3.

Regression of alcohol use and current regular smoking by class

	Past year frequency of alcohol use				Current regular smoking
	Class 1	Class 2	Class 3	Class 4	Class 1	Class 2	Class 3	Class 4
Intercept	1.374	1.665	1.945	2.875**	−0.570	0.156	-0.486	1.108*
Afro Caribbeana	−0.132	−0.413	−0.211	0.330	−0.589	−2.362	-1.152	−0.612
Female	-0.450	-0.793	-0.843	-0.813	−0.568	-0.479	−0.146	-0.865
≥12 years education	−0.362	−0.344	0.160	-0.812	-0.559	-0.847	−0.233	-0.735
Age 18–29b	−0.221	0.183	−0.227	−0.420	0.090	-1.230	-0.663	0.538
Age 30–44b	−0.051	0.357	0.014	−0.143	−0.093	0.209	−0.046	-0.650
Family history of problem drinking	0.192	−0.215	0.314	0.145	--------- -	--------- -	--------- -	---------
Family history of regular smoking	--------- -	--------- -	--------- -	--------- -	−0.145	0.212	0.483	0.385

Boldface=differs from zero at p<.05;

^*=differs from Class-1 estimate at p<.01;

^**differs from Class-1 estimate at p<.001;

^acomparison group=African American;

^bcomparison group=age 45–65

In the third step, regression analyses were used to compare the frequency of alcohol use and prevalence of current regular smoking across the latent classes. Multiple-group linear regression analysis was conducted for the alcohol outcome and multiple-group logistic regression analysis was conducted for the smoking outcome. Stratifying on latent class, differences between classes in each outcome were evaluated by comparing model intercepts, controlling for the association between class and covariates, and class-specific effects of each covariate on the outcome. Between class differences in covariate-adjusted intercept estimates were tested by Wald test, using Class 1 as the reference class. These primary outcomes were evaluated using Bonferroni-corrected alpha of .017 (i.e., .05/3 pairwise comparisons per model). Class weights were included in the model estimation to control for measurement error associated with latent class assignment, using the BCH method in Mplus [34], where classes are treated as known and the risk of class shift is therefore eliminated. In both regression models, covariates included gender (female relative to male), ethnicity (Afro Caribbean relative to African American), age groups 18–29 and 30–44 (relative to 45–65), and education level (≥12 years relative to <12 years). In the frequency of alcohol use model, family history of problem drinking was also included as a covariate. In the regular smoking model, family history of regular smoking was included. Differences in covariate effects were tested by Wald test in pairwise comparisons between Class 1 and each of the other classes. Missing data were handled by listwise deletion.

Results

Class Identification

Results of LPA model fit for 2 through 6 classes are shown in Supplemental Table 3. The 4-class solution was selected based on model fit, entropy, and clear interpretability. Means for class indicators are reported by class in Table 1 and Figure 1. Table 1 shows means for all standardized factors and raw means for the single-item neighborhood cohesion indicator and score on the Everyday Discrimination Scale. Figure 1 shows means for all LPA indicators, which were standardized to allow for presentation on a single scale. Class descriptions are based on the relative values of the support, demands, and adversity indicators across classes. Class 1 (high support, low demands and adversity; 17.8%) was high on all protective factors and low on risk factors. Support from family, religious community, and friends, and religious involvement, were high; demands from family and religious community were low, as were financial difficulties, neighborhood disadvantage indicators, and racial discrimination. Class 2 (high support and demands, low-moderate adversity; 27.3%) was high on religious involvement and demands as well as support from family, religious community, and friends, with a low to moderate level of adversity. Class 3 (low support and demands, low-moderate adversity; 38.7%) was low on religious involvement and demands as well as support from family, religious community, and friends, with a low to moderate level of adversity. Class 4 (low support, high demands and adversity; 16.3%) was low on all protective factors and high on all risk factors.

Table 1.

Means for indicators by class^a

	Class 1 (n=793)	Class 2 (n=1,217)	Class 3 (n=1726)	Class 4 (n=726)
	High support; low demands, adversity	High support, demands; low-moderate adversity	Low support, demands; low-moderate adversity	Low support; high demands, adversity	Omnibus test p-value*
Support
Religious involvement	0.78	0.81	−0.25	0.08	<.001
Support from family	0.23	−0.06	−0.29	−0.68	<.001
Support from friends	−0.12	−0.02	−0.46	−0.39	<.001
Support from religious community	0.71	0.81	−0.81	−0.40	<.001
Demands
Demands from family	−0.69	0.09	−0.23	0.58	<.001
Demands from religious community	−0.28	0.52	−0.26	0.59	<.001
Adversity
Socioeconomic/neighborhood
Financial difficulties	−0.14	0.01	0.00	0.26	<.001
Low neighborhood safety	−0.49	0.02	−0.12	0.37	<.001
Low neighborhood cohesionb	3.85	3.72	4.40	3.96	<.001
Racial discriminationc	5.79	9.23	8.73	15.81	<.001

^aMeans shown for standardized factor scores, raw means shown for single item indicators: low neighborhood cohesion and racial discrimination;

^bpossible range=1–4;

^cpossible range=0–50;

^*ANOVA

Fig 1. Standardized means for all indicators by class.

Note: The single item indicators low neighborhood cohesion and racial discrimination were converted to standardized means so they could be reported on the same scale as the factor scores.

Demographic characteristics, family history of problem drinking, family history of regular smoking, current regular smoking status, and past year frequency of alcohol use are shown by class in Table 2. Males and Afro Caribbeans were overrepresented in Class 3 (χ²₍₃₎=48.61, p<.001, standardized residual=6.7 and χ²₍₃₎=52.47, p<.001, standardized residual=6.6, respectively). Adults in Class 4 were disproportionately likely to have <12 years of education (χ²₍₃₎=21.91, p<.001, standardized residual=4.4). Family history of both alcohol problems and regular smoking were highest in Class 4 (χ²₍₃₎=27.49, p<.001, standardized residual=5.0 and χ²₍₃₎=27.17, p<.001, standardized residual=4.1, respectively). Significantly larger proportions of 18- to 29-year-olds were observed in Class 3 (χ²₍₆₎=106.23, p<.001; standardized residual 7.5.).

Table 2.

Demographics, family history, frequency of current alcohol use, and current regular smoking by class

	Overall	Class 1	Class 2	Class 3	Class 4	Omnibus test p-value (Chi- square, ANOVA)
Demographic factors and family history of problem drinking and regular smoking (%)
Gender						<.001
Female	63.20	69.23	67.21	57.13	64.33
Male	36.80	30.77	32.79	42.87	35.67
Ethnicity						<.001
Afro Caribbean	29.40	21.94	26.95	35.11	28.10
African American	70.60	78.06	73.05	64.89	71.90
Age group						<.001
18–29	26.62	17.65	22.35	32.91	28.65
30–44	40.65	40.10	41.17	39.63	42.84
45–65	32.72	42.24	36.48	27.46	28.51
Education level
< 12 years	18.76	18.28	16.27	18.25	24.66	<.001
≥12 years	81.24	81.72	83.73	81.75	75.34
Family history
Problem drinking	24.56	20.82	23.71	23.72	32.14	<.001
Regular smoking	33.17	27.32	34.57	32.01	39.94	<.001
Substance use (%)
Past year alcohol use						<.001
None	50.41	65.32	58.67	42.23	39.72
< once/month	17.74	15.38	16.68	18.74	19.72
1–3 days/month	11.44	8.45	9.70	13.46	12.83
1–2 days/week	10.74	6.68	8.38	13.92	11.59
3–4 days/week	4.51	1.89	3.53	5.34	7.03
Nearly every day	5.16	2.27	3.04	6.32	9.10
Current regular smoking	24.02	17.56	19.12	25.00	36.79	<.001

Regression Analyses

Past Year Frequency of Alcohol Use

The covariate-adjusted mean frequency of past year alcohol use differed significantly between Class 1 (the reference class) and Class 4, but not Classes 2 or 3. Relative to Class 1, the mean frequency of alcohol use was significantly higher in Class 4 (Class 1=1.374 vs. Class 4=2.875, p<.001). Covariate effects were statistically similar to Class 1 effects across classes. (See Table 3.)

Current Regular Smoking

The conditional log-odds of being a current regular smoker differed significantly between Class 1 and Class 4; the Class 4 intercept estimate was significantly larger (Class 1 intercept=−0.570, Class 4 intercept=1.108; p<.01). Classes 2 and 3 did not differ from Class 1. As seen in Table 3, covariate effects did not differ significantly across classes.

Discussion

The current study applied a subgrouping approach to investigate the collective influence of substance use related psychosocial risk and protective factors on frequency of alcohol use and regular smoking status in Black adults in the U.S. Results of analyses from the all-Black sample revealed variation in the pairing of substance use related risk and protective factors beyond a simple dichotomy of high and low liability subgroups. They further suggested that neither the presence of high demands nor the absence of support alone differentiate likelihood of engaging in frequent alcohol use or regular smoking, but adverse experiences such as racial discrimination may be especially impactful.

Patterns of Endorsement of Psychosocial Risk and Protective Factors

Our hypothesis that at least three subgroups representing combinations of psychosocial risk and protective factors would emerge, with patterns of endorsement including variation in the pairing of certain protective factors with certain risk factors, was supported. Four subgroups emerged, with high support (religious involvement and support from family, friends, and religious community members) clustering alternatively with low and high demands from the social network in Classes 1 and 2, respectively. Low support clustered alternatively with low-to-moderate and high adversity (financial difficulties, low neighborhood safety, and racial discrimination) in Classes 3 and 4, respectively. These findings underscore the complexity of the co-occurrence of substance use related psychosocial factors.

Notably, although a simple dichotomy of high on risk, low on protective factors and low on risk, high on protective factors did not fully capture the heterogeneity in the sample, overall low and overall high liability subgroups (Classes 1 and 4, respectively) were observed. These subgroupings were anticipated, given the known clustering of substance use related risk factors such as racial discrimination with neighborhood disadvantage [35] and protective factors such as family support with religious involvement [36]. Prior evidence for the clustering of certain risk factors with certain protective factors is more limited, but associations for example, between low SES and high religious involvement [19] have been found. The observed co-occurrence in Class 2 of high support with high demands from the social network, which has been linked to psychological distress [37], is consistent with prior work demonstrating that extensive family networks can involve both support and demands from network members, particularly in Black families [38]. The pairing of low support with low to moderate adversity in Class 3 further highlights the fact that the absence of some protective factors does not consistently coincide with the presence of risk factors.

Family history of problem drinking and family history of regular smoking were highest in the overall high liability subgroup (Class 4), as anticipated, given that family history is a robust predictor of alcohol misuse and smoking [32] and thus expected to co-occur with other substance use related risk factors (and less so with protective factors). Low educational attainment was also highest in Class 4, as would be expected, given the elevated likelihood of adults with less than a high school level of education experiencing financial difficulties [39] and living in disadvantaged neighborhoods (as resident education level is commonly used in neighborhood deprivation indices). Consistent with literature supporting a lesser degree of connectedness to social networks among men compared to women [40], males were overrepresented in the subgroup (Class 3) characterized by low social support and religious involvement along with low demands from family and the religious community. Class 3 also had the highest proportion of Afro Caribbeans. Although beyond the scope of the current investigation, potential sources of their overrepresentation in this subgroup, e.g., cultural differences in social network interactions or less extensive networks as a function of being a smaller ethnic group compared to African Americans, merit exploration.

The Role of Risk vs. Protective Factors in Frequency of Alcohol Use and Regular Smoking

We hypothesized that subgroups characterized by a high level of psychosocial protective factors, regardless of the corresponding level of risk factors, would evidence less severe substance use outcomes than subgroups low on protective factors. The proposal that substance use outcomes would be driven to a greater degree by protective factors than risk factors was based in part on the evidence for lower alcohol and cigarette smoking [1, 3], despite higher prevalence of many substance use related risk factors, such as discrimination [12] and neighborhood disadvantage [5, 6], in Black compared to White adults, suggesting that the risk they pose is buffered by protective factors. Our findings support a somewhat different model of the joint contributions of risk and protective factors to current alcohol and smoking behaviors in Black adults.

Only Class 4, the overall high liability subgroup, differed significantly from Class 1, the overall low liability subgroup, with respect to current frequency of alcohol use or odds of current regular smoking. The absence of distinctions between Class 1 and Class 2, which was characterized by high demands in addition to high religious involvement and social support, is consistent with the notion that protective factors are robust to the presence of risk factors such as stress associated with family and community ties. However, the fact that Class 3, which was low on both support and demands, also did not differ from Class 1 indicates that protective factors were not the sole drivers of alcohol and smoking outcomes. Class 4 was the only class characterized by high adversity, suggesting that risk factors, not protective factors, drove the class association with substance use outcomes. Both heavy alcohol use and regular smoking have been linked to the risk factors that are elevated in Class 4, including financial difficulties [10, 41], low neighborhood safety [9, 11], and racial discrimination [14, 42] in numerous prior studies.

Notably, interpretation of our findings should also include the absence of significant differences between Class 1 and the two classes characterized by a combination of risk and protective factors. While not consistent with our hypothesis that protective factors would drive the associations, the lack of distinctions from the overall high liability class indicate that neither the magnitude of demands nor the level of support from members of social network in and of themselves differentiate likelihood of engaging in heavy alcohol use or regular smoking, underscoring the importance of considering how psychosocial risk and protective factors act in concert to elevate or mitigate substance use related risk.

Limitations

Certain limitations should be kept in mind when interpreting study findings. First, as a cross-sectional study, conclusions regarding the directions of effect cannot be made. Second, NSAL data were collected from 2001–2004, so the possibility that the interrelationships among alcohol and smoking related psychosocial factors or the associations of those naturally occurring subgroups with frequency of alcohol consumption or regular smoking status has shifted since that time should be considered. Notably, although conducted well over a decade ago, NSAL remains the largest scale health survey of Black Americans to include a substantial number of Afro Caribbeans and detailed assessments of social, economic, cultural, and race related constructs, and thus, is a rare resource for addressing our research question. Third, findings may not generalize to other alcohol or smoking related phenotypes such as problem drinking or frequency of smoking, as the relevance of various psychosocial influences differ across levels of substance use involvement. Fourth, although only one adult per household was enrolled in NSAL, it is possible that biological relatives from different households participated and as such, there may be non-independence of family history for which we cannot account. Fifth, current smoking status of lifetime non-regular smokers could not be determined, as only respondents who endorsed smoking 100 or more cigarettes over the lifetime were asked if they currently smoke.

Future Directions

Our findings suggest a number of possible directions for additional work in this area, including the exploration of the natural clustering of psychosocial risk and protective factors in other understudied racial/ethnic groups (e.g., Latinx), and their links to other alcohol and smoking related outcomes. The inclusion of additional protective factors that may be especially salient in the Black population in the U.S. but were not assessed in the NSAL adult sample, such as ethnic identity, is another important next step. The application of the approach used in the present study to understanding the collective contributions of psychosocial risk and protective factors to alcohol and smoking outcomes in Black adolescents is currently underway.