The Longitudinal Association Between Exposure to Violence and Patterns of Health Risk Behaviors Among African American Youth

Abstract

Purpose:

We investigate whether exposure to violence (ETV) during adolescence and emerging adulthood predicts engagement in chronic disease-related health risk behaviors years later among African Americans.

Design:

A longitudinal study following youth from mid-adolescence (mean age = 14.8 years) to young adulthood (mean age = 32.0 years).

Setting:

Flint, Michigan.

Sample:

Four hundred forty-two African American (96.2%) and mixed African American and White (3.8%) participants.

Measures:

Outcomes were diet, smoking, drinking, and physical inactivity. Covariates were ETV, sex, mother’s educational attainment, and substance use by siblings, peers, and parents.

Analysis:

Latent profile analysis was conducted to identify distinct patterns of adult health risk behaviors and assess the association of youth ETV and identified patterns.

Results:

Four latent profiles were identified: high substance use (n = 46; 10.41%), high overall risk (n = 71; 16.06%), low overall risk (n = 140; 31.67%) and inactive (n = 185, 41.86%). Relative to the low overall risk profile, ETV was associated with being in the high overall risk profile (b = 0.37, p = 0.04), but not other profiles. Female gender and higher maternal education were associated with being in the inactive profile compared to the low overall risk profile. Peer alcohol and tobacco use were associated with being in the high substance use profile.

Conclusion:

ETV during adolescence and emerging adulthood increased the risk of engagement in multiple health risk behaviors later in life.

Keywords

violence risk behaviors health disparity latent class analysis adolescence emerging adulthood young adulthood

Purpose

Chronic noncommunicable diseases affect 6 in 10 adults in the US¹ and are often caused by behavioral risk factors such as alcohol use, tobacco use, physical inactivity, and poor diet quality.² These risk behaviors tend to occur in clusters and can have synergistic effects on health outcomes.³ Individuals engaging in all 4 risk behaviors had a 3-fold increased risk of cancer mortality than those engaging in single risk behavior.⁴ Researchers have found that the prevalence of having two or more risk factors for chronic disease is highest among African Americans,⁵ and consequently, have a higher risk for developing chronic diseases.⁶ The structural factors such as racism and segregation have uniquely influenced African Americans’ life experiences and health behaviors.^6,7 Yet, there is limited data available on how health risk behaviors cluster in African American populations and which social-environmental factors predict the patterns.⁸ Such gaps in the current literature represent a missed opportunity to more deeply understand the African American experience in order to effectively address disparities in chronic disease risk.

Understanding the clustering of health risk behaviors and their correlates is useful for informing prevention efforts. Researchers seeking to better understand the factors associated with clustered health risk behaviors have focused on demographic characteristics (e.g., SES, age, gender). They found being male and having lower SES were associated with the prevalence of unhealthy behavioral clusters.^3,9
-11 Nevertheless, one cross-sectional study focusing on multiple risk behaviors among African Americans found that social-environmental factors (e.g., perceived neighborhood cohesion), compared to demographic or individual-level psychological factors (e.g., perceived cancer risk), plays a much more critical role in differentiating the healthiest group from groups engaging in less healthy behaviors.⁸ In line with this finding, researchers have also found that exposure to violence (ETV) influences health outcomes. Boynton-Jarrett et al., for example, found that cumulative ETV was associated with increased poor self-rated health and reduced the predictive effects of household income.¹²

Further, Jackson and colleagues⁶ found that for African Americans, unhealthy behaviors were easily accessible coping strategies for the environmental stressors that are disproportionately experienced by African Americans. They found the number of unhealthy health behaviors was correlated with the number of chronic conditions for African Americans in their sample, but not for the Whites in the sample.⁶ These findings warrant the value of further understanding of whether ETV predicts the pattern of health risk behaviors among African Americans.

Exposure to Violence and Its Consequences

ETV includes both witnessing and experiencing violence, and is a social-environmental factor associated with negative behavior and health outcomes across the life span.¹³ Researchers have found that depression phenotypes commonly resulted from violence exposure and increased vulnerability to alcohol and substance use dependence.¹⁴ As a major source of psychological stress, ETV during childhood can activate disruptive psychological and physiological processes, give rise to unhealthy lifestyle choices, and eventually increase the susceptibility of chronic disease later in life.¹⁵ Such adversity experienced during adolescence and emerging adulthood carries similar negative effects later in life.^16,17 ETV may also be a significant barrier to physical activity, which may prevent or help manage chronic disease.^18,19 Okada et al., for example, reported an association between ETV and the adoption of unhealthy weight control among adolescents.¹⁹

African American youth endure a disproportionate burden of ETV and its consequences compared to White youth.²⁰ Cumulative exposure to multiple sources of violence may also be disproportionately experienced in urban areas²¹ and by African-American youth.²² Such disparity may be exacerbated by racism and inadequate response by multiple sectors (e.g. schools and law enforcement) to community violence experienced by African American youth.²³ Cumulative and chronic ETV among African American youth increases psychological and physiological stress,²⁴ which increases the risk of chronic conditions.^15,16

Cumulative Exposure to Violence During Adolescence and Emerging Adulthood

While most research focused on the relationship between childhood adversity and chronic disease risk factors in adulthood,^15,25,26 researchers have demonstrated that adolescents and emerging adults continue to be vulnerable to the negative consequences of ETV as adults.²⁷ ETV during the formative years of adolescence and emerging adulthood increases mental distress and disruption of social relationships,¹³ and increases maladaptive coping behaviors, including substance use.²⁷ Thus, cumulative ETV can be viewed as a significant chronic stressor that has long-lasting detrimental effects on individuals’ psychosocial, behavioral, and physical outcomes.

In sum, African Americans shoulder the burden of both higher risk for preventable chronic diseases⁵ and higher risk for exposure to violence.²² Given the lack of studies investigating clustering patterns of health risk behaviors that include racial/ethnic minorities,^8,11 our study further the understanding of such health disparities in this population. To the best of our knowledge, no other study has used longitudinal data spanning from adolescence to young adulthood to examine social-environmental factors (i.e., ETV) beyond demographic characteristics for their effects on the health risk behavior clustering among African Americans.

Current Study

Our study examines how violence exposure during adolescence and emerging adulthood may influence adult health risk behaviors associated with chronic disease among a sample of African Americans. We use data from a 19-year longitudinal study to identify early ETV and its association with adult health risk behavioral patterns using latent profile analysis.²⁸ We use a 2 stage multivariate approach to test our hypothesis that ETV increases adult patterns of risk behaviors while controlling for demographics and other covariates. The findings from this study will provide evidence of whether early-life exposure has long-lasting effects on certain health risk behaviors in young adulthood, which will inform preventive efforts for this population.

Methods

Design

We use the Flint Adolescent Study (FAS) data for our analysis. FAS is a 19-year longitudinal study of adolescents (n = 850) from 4 public high schools in Flint, Michigan. The original sample consists of 80.1% African American, 3.1% mixed African American and White, and 16.8% White/Caucasian. Because the current study is focused on African Americans, we examine a subset of the original sample (n = 707) composed of the self-identified African Americans or mixed-race (i.e., African American and White). We included mixed-race respondents because they share many of the challenges and societal pressures faced by African Americans. The goal of the original study was to understand better the promotive factors associated with substance use among adolescents at-risk of high school dropouts. The inclusion criteria were 1) an 8th-grade GPA of 3.0 or lower and 2) the absence of an emotional and developmental disability as diagnosed by the schools. The research team collaborated with the University of Michigan/Flint Research Office and Flint Community Schools for recruitment from the city’s 4 main public schools in 1994. Ninety-two percent of all eligible youth participated in the wave 1 survey. The high school and student participation rates suggest that the sample reflected the Flint high school students’ overall population in 1994.²⁹ The participants were assessed from mid-adolescence (mean age = 14.8 years) to young adulthood (mean age = 32.0 years), and data were collected annually in 3 time periods: waves 1 to 4; 1994 to 1997 (high school years), waves 5 to 8; 1999 to 2002 (2-5 years post high school), and waves 9 to 12; 2008 to 2012 (participants were in their late twenties and early thirties). The non-white sample, which this study focused on, generally remained steady for the first 8 waves (above 90% from waves 1 to 4 and 67% from waves 5 to 8), and only above 37% between waves 9 and 12. Information on detailed design, survey instruments, and data are available publicly on the Inter-university Consortium for Political and Social Research (ICPSR) website.²⁹

Procedure

For each wave, participants completed a 50-60 minute, structured face-to-face interview conducted by trained interviewers. We did not match participants with interviewers’ by race or gender as this was not feasible given practical issues, but we found no interviewer effects for any outcome.³⁰ The interview portion of the questionnaire included most of the psychosocial variables in the data, such as exposure to violence, family and friend influences, diet and exercise. Following each interview, participants completed a self-administered questionnaire about sensitive items such as substance use and sexual behavior. Participants were interviewed at school, home, or at a location in the community specified by the participant. Informed consent was obtained from each participant prior to data collection. This study was approved by the Institutional Review Board at the [Institute Name] for the protection of human subjects.

Analytic Sample

Our analysis focused on African American and mixed-race respondents with available data points from any of the waves 9 to 12 to measure health risk behaviors. Of the 707 participants in wave 1, 265 participants were excluded due to attrition or did not provide health risk behavior data in waves 9 to 12, resulting in n = 442 (62.5% of the wave 1 sample) for the latent profile analysis. The final sample was n = 406 (57.4%) in the multinomial regression model because an additional 36 participants were excluded due to missing data in the covariates. We conducted attrition analysis comparing the sample used in the multinomial regression model (n = 406) with the original 707 participants (See Attrition Analysis below).

Measures

Cumulative ETV included violence victimization, observing violence, and instances of violence among friends across waves 1 to 8. Violent victimization was measured with three 5-point Likert scale (1 = 0 times, 5 = 4 or more times) items. Participants were asked how often in the past 12 months they experienced: 1) someone threatening to hurt them, 2) someone physically assaulting or hurting them, and 3) someone taking something from them using physical force (α = 0.51–0.68). Observing violence was measured with two 5-point Likert scaled (1 = 0 times, 5 = 4 or more times) items: “During the last 12 months, how often have you 1) seen a violent crime where a person was hurt and 2) seen someone get shot, stabbed, or beaten up” (r = 0.52–0.74). Violence among friends was measured using 3 items 5-point Likert scale (1 = none, 5 = all): “How many of your friends 1) get into fights, 2) carry a knife, and 3) carry a gun” (α = 0.72–0.80). We standardized the average scores for violence victimization, observing violence, and violence among friends, and then summed the standardized scores for each wave. We then computed a mean score for the 8 waves to create a cumulative ETV score. Similar approaches of computing violence exposure were used in previous studies using the study data and have shown good criterion validity of ETV.^31
-33 The average inter-item correlation was 0.31, which demonstrates adequate internal consistency.³⁴ We do not report a traditional measure of internal reliability (e.g., Cronbach’s alpha) for the 3 measures because our ETV measure is not attitudinal in nature. It is entirely feasible, for example, for an individual to experience one form of ETV, but not another.

Health risk behaviors for chronic disease

Tobacco and alcohol use was assessed in waves 9 to 12 using questions adopted from the Monitoring the Future national survey.³⁵ Cigarette use was measured by one item: “How often have you smoked cigarettes in the past 30 days?” (1 = not at all to 7 = about 2 or more packs/day). Alcohol use was measured using 3 items (α = 0.71–0.77) in waves 9 to 12: “How many times have you had alcoholic beverages to drink during the last 30 days?” (1 = 0 times to 7 = 40+ times), “In the last 2 weeks, how many times have you had 5 or more drinks in a row?” (1 = none, 6 = 10+ times), “In the last 2 weeks, how often have you drunk enough to get high?” (1 = never, 5 = on nearly all occasions). Alcohol use items were standardized and summed for each wave. The average of the 4 waves during young adulthood was used to create the tobacco and alcohol use variables.

Diet was measured by 2 questions: “In a normal week, how often do you eat foods that are high in fat such as red meat, cheese, fried foods, and eggs?” and “In a normal week, how often do you eat fruits or vegetables?” Responses used a 5 point scale (1 = almost never to 5 = every day). Ratings for the fruits and vegetable consumption were reverse coded. We averaged these 2 items for each wave to create a mean score across the 4 waves, with higher scores reflecting poorer diets. Prior studies have used similar questions to measure dietary intake.^36,37

Exercise was assessed in waves 9 to 12 using the question, “In a normal week, how many times do you engage in vigorous exercise?” The item is similar to that used in the Global Physical Activity Questionnaire. A prior study demonstrated that single-item physical activity measure can perform as well as other short physical activity tools in reliability and validity.³⁸ We recoded the 5-point response scale so that higher scores reflect more weekly physical inactivity (1 = every day to 5 = almost never). We created an average score across the 4 waves.

Covariates

In Wave 1, participants self-reported their sex and mother’s educational attainment (1 = completed grade school or less; 7 = graduate/professional school after college.). During Wave 2 to 4, participants were asked the number of friends or older siblings who used each substance from 1 (none) to 5 (all) and how often the substance was used by adults raising them from 1 (never) to 5 (very often). We averaged cigarette and alcohol use across waves to create a combined siblings, peers, and parents measure. Descriptive statistics for all study variables are reported in Table 1.

Table 1.

Demographic Background and Primary Predicting and Outcome Variables Among African American Participants, by Sex, Flint Adolescent Study, 1994-2012.

	Female	Male
Variable	M (SD) or %	M (SD) or %
Age at baseline	14.74 (0.59)	14.84 (0.62)
Mother’s Educational Attainment _a	3.90 (1.41)	4.19 (1.40)
Sibling Tobacco and Alcohol Use (W2-4)	1.78 (0.99)	1.70 (0.95)
Peer Tobacco and Alcohol Use (W2-4)	1.74 (0.88)	1.62 (0.76)
Parent Tobacco and Alcohol Use (W2-4)	2.16 (0.79)	2.29 (0.85)
Adolescence to Emerging Adulthood: W1-8
Exposure to Violence (ETV)_b	−0.21 (1.20)	0.24 (1.24)
Young Adulthood: W9-12
Cigarette use _b	−0.16 (0.94)	0.03 (0.95)
Heavy alcohol use _b	−0.15 (0.96)	0.12 (1.05)
Poor Diet _c	−0.08 (0.79)	0.11 (0.81)
Physical inactivity _c	0.19 (1.00)	−0.27 (1.01)

_a 1 = completed grade school or less; 2 = some high school; 3 = completed high school; 4 = vocational/training school; 5 = some college; 6 = completed college; 7 = graduate/professional school after college.

_bThe original scales were standardized across scales to facilitate interpretation and the mean across waves was used to summarize an individual’s score during a period.

_cHealthy behaviors were reverse coded so that the higher scores reflect poorer diet/lower physical activity.

Analysis

All statistical analyses were conducted on Mplus 8.2.³⁹ We examined distinct profiles of adult health risk behaviors using latent profile analysis (LPA).²⁸ To obtain a finite number of latent profiles across our 4 health risk behaviors, we assessed model fit, using Akaike information criterion (AIC)⁴⁰), Bayesian information criterion (BIC),⁴¹ and sample-adjusted Bayesian information criterion (aBIC).⁴² Vuong-Lo-Mendell-Rubin likelihood ratio test (VLMR-LRT)⁴³ and the bootstrap likelihood ratio test (B-LRT)⁴⁴ were conducted to ascertain significant improvement in model fit between the k and k+1 solution. We used latent class posterior probabilities to assess the association of youth ETV on adult health risk behaviors. We implemented an automated 3-step procedure (R3STEP)²⁸ to conduct a multinomial logistic regression examining whether the control variables and youth ETV was associated with latent profile classifications. Variance inflation factor (VIF) values were calculated to check for multicollinearity. Full information maximum likelihood was implemented to treat missing data for the LPAs, while list-wise deletion was implemented to treat missing data for the multinomial logistic regression.

Results

Profiles of Adult Health Risk Behaviors

Table 1 represents the characteristics of demographic and primary study variables in our sample. Compared to females, male youth reported higher ETV, cigarette use, alcohol use, and poorer diet while being more physically active. LPA Model fit indices suggested that models with 2 to 5 latent profiles fit the data better than a single latent profile and fit deteriorated beyond 4 latent profiles (i.e., higher BIC). We found no difference in VLMR-LRT model fit estimates between the 4 and 5 profile models. However, other fits statistics (B-LRT, AIC, and aBIC) suggested that a 5 profile model was a better fit than a 4 profile model. We found that the low risk profile (n = 140) in the 4 profile model was split into 2 profiles for the 5 profile solution. These 2 profiles, however, had nearly identical health risk behavior patterns indicating they are not conceptually distinct. Entropy was also higher in the 4 profile model (.85) compared to the 5 profile model (.82) indicating a better classification certainty for the 4 profile model. Thus, we chose the 4 profile model for analysis. The LPA results are reported in Table 2.

Table 2.

Model Fit Indicators for Latent Profile Analyses Among Study Participants.

k Profiles	AIC	BIC	aBIC	Entropy	VLMR-LRT	B-LRT	Smallest class size (%)
1	4666.24	4698.97	4673.58	--	--	--	--
2	4474.39	4527.58	4486.32	.84	< .01	< .01	184 (41.63%)
3	4427.47	4501.12	4443.99	.84	< .01	< .01	92 (20.81%)
4	4337.01	4431.11	4358.12	.85	< .01	< .01	46 (10.40%)
5	4320.69	4435.25	4346.39	.82	.12	< .01	45 (10.18%)

As depicted in Figure 1, the first and smallest profile (i.e., high substance use; n = 46; 10.41%) was characterized as having high levels of cigarette and alcohol use and physical inactivity, and an average level of unhealthy diet. The second profile (i.e., high overall risk; n = 71; 16.06%) included individuals who reported high levels of all 4 risk behaviors. The third profile (i.e., low overall risk; n = 140; 31.67%) characterized individuals who reported low levels of all 4 risk behaviors. The fourth and largest profile (i.e., inactive; n = 185, 41.86%) characterized individuals who reported a high level of physical inactivity, but low alcohol use, and moderate cigarette use and poor diet.

Figure 1.

Latent profiles of adult health risk behaviors among study participants.

Youth Violence as a Predictor of Adult Health Risk Behaviors

Table 3 contains multinomial regression results that test whether being female, mother’s educational attainment, peer, parent, and sibling tobacco and alcohol use, and youth violence exposure were associated with health risk profiles. Relative to the low overall risk profile, individuals reporting higher levels of peer alcohol and tobacco use were more likely to classify into the high substance use profile. Additionally, being female and having more maternal educational attainment increased the risk of being in the inactive profile, rather than the low overall risk profile. Lastly, as hypothesized, individuals reporting more cumulative ETV were more likely to be in the high overall risk profile, rather than the low overall risk profile. In other words, cumulative ETV was associated with the clustering of the 4 health risk behaviors during young adulthood: cigarette use, heavy alcohol use, unhealthy diet, and physical inactivity.

Table 3.

Multinomial Logistic Regression Results (Log Odds and Odds Ratios) Among Study Participants.

High Substance Use Profile	Odds ratio	b	s.e.	p
Female	0.865	−0.145	0.449	0.707
Mother’s Educational Attainment	0.973	−0.027	0.161	0.867
Peer Tobacco and Alcohol Use	2.286	0.827	0.350	0.018
Parent Tobacco and Alcohol Use	0.940	−0.062	0.272	0.820
Sibling Tobacco and Alcohol Use	1.212	0.192	0.237	0.416
Cumulative ETV (Waves 1 to 8)	1.093	0.089	0.243	0.714
High Overall Risk Profile	Odds Ratio	b	s.e.	p
Female	1.338	0.291	0.359	0.417
Mother’s Educational Attainment	1.154	0.143	0.136	0.291
Peer Tobacco and Alcohol Use	1.217	0.197	0.289	0.291
Parent Tobacco and Alcohol Use	1.285	0.251	0.228	0.271
Sibling Tobacco and Alcohol Use	1.036	0.036	0.175	0.838
Cumulative ETV (Waves 1 to 8)	1.446	0.369	0.181	0.041
Physical Inactivity Profile	Odds Ratio	b	s.e.	p
Female	3.121	1.138	0.318	<.001
Mother’s Educational Attainment	1.412	0.345	0.111	.002
Peer Tobacco and Alcohol Use	1.141	0.132	0.276	0.633
Parent Tobacco and Alcohol Use	1.249	0.222	0.214	0.300
Sibling Tobacco and Alcohol Use	1.020	0.020	0.171	0.907
Cumulative ETV (Waves 1 to 8)	1.028	0.027	0.175	0.876

Note. Low overall health risk profile is the reference group.

Attrition Analysis

We conducted the attrition analysis to compare the final study sample used in the multinomial regression model (n = 406) with the original 707 participants who identified themselves as African Americans or mixed African Americans in wave 1. We compare participants’ ETV in the first 8 waves and all the covariates in the study. The only difference in all study variables between those retained in the final sample and the 301 participants with missing values was for mother’s education. The retained sample had slightly lower levels of mother’s education (mean = 4.03) compared to excluded participants (mean = 4.95; t(704) = −6.54, p < 0.01).

Discussion

We found that cumulative ETV was associated with latent profiles of multiple risk behaviors implicated in developing chronic diseases, premature mortality, and health disparities.^4,9,10 We found 4 distinct latent profiles: low overall risk profile, physically inactive profile, high substance use profile, and high overall risk profile. Notably, individuals with high ETV were more likely to be in the high overall risk profile with high levels of alcohol and cigarette use, unhealthy diet, and increased physical inactivity.

Our findings extend existing research about the clustering of health risk behaviors^3,11 and indicate the importance of examining socio-environmental factors when addressing health risk behaviors among African Americans.²⁷ A prior cross-sectional study showed that neighborhood cohesion, instead of individual-level demographic correlates, is a crucial predictive factor for African Americans engaging in multiple health behaviors.⁸ Our findings of ETV adds to such evidence with longitudinal data that violence in the neighborhood and other social environments may play an important role in shaping various health risk behaviors later in life among African Americans.

Our findings also suggest that African Americans who experienced higher levels of ETV may be most susceptible to the burden of chronic disease. The risk of mortality from chronic diseases increases several-fold for individuals engaging in all 4 risk behaviors.⁴ Unfortunately, engaging in one of these risk behaviors also increases the probability of engaging in others.³ Researchers showed strong evidence for the long-term effects of ETV on mental health outcomes such as posttraumatic stress disorder (PTSD).⁴⁵ Risky health behaviors, such as substance use, poor diet, and physical inactivity, may be a part of coping mechanisms in response to the stress that ETV elicits. Because smoking and alcohol use are associated with enhanced serotonergic activities that are needed for emotion regulation and coping, these risk behaviors may be a form of self-regulation to cope with the psychological distress from ETV.¹²

In contrast to a prior study that reports the influence of stress on sedentary behaviors is stronger among men,⁴⁶ we found that women in our sample are more likely to be in the physically inactive group than men. Women likely tend to be physically inactive when ETV is the stressor, and the safety of the neighborhood is of concern.¹⁸ From this perspective, our finding is consistent with other research showing that stress is associated with higher body mass index and unhealthy eating among African American females.^47,48

Due to the increased risk of maladjusted health behaviors from ETV,²⁷ including those examined in the current study, health promotion professionals must be mindful of the presence and consequences of ETV when developing and implementing chronic disease prevention and control programs in areas where community violence is a problem. Individuals who have experienced high levels of ETV may have difficulties with coping, social support, and access to resources to be physically active, eat right, and avoid harmful substance use. These challenges may be particularly salient within communities that have social and economic disadvantages. For African Americans, frequent experiences with racism also interfered with their receiving support after violence exposure.²³ Chronic disease prevention and control programs in such communities could directly identify any issues of community violence and how, if at all, they may affect program implementation and outcomes. When necessary, such programs could explore how to address the potential influences of ETV on chronic disease-related risk behaviors. For example, they may increase their focus on coping strategies, social support, and community resources for individuals with histories of ETV. Similarly, local health departments may consider violence prevention and interventions to mitigate the consequences of violence as useful population-based approaches to address chronic disease disparities. It is important to note that community violence, specifically, is also a consequence of the structural inequalities (e.g., segregation, discrimination, and community disinvestment)⁴⁹ created by racism that unjustly burden African Americans’ lives.⁴⁹ Therefore it takes structural changes and multi-sector collaborations²³ between mental health services, schools, churches, and community to mitigate such health inequities as a result of violence exposure. Finally, these findings underscore the importance of addressing violence exposure among youth. Trauma-informed practices in family, clinical, school, and community settings, for example, are promising in improving psychosocial outcomes among youth exposed to violence,⁵⁰ and may eventually help the prevention of subsequent health risk behaviors.

Limitations

Several study limitations should be noted. First, the self-report measures may introduce social desirability bias. Nevertheless, self-report is the standard approach to assess behaviors such as substance use and violence exposure. Second, our sample may not be representative of all adolescents in the US. Because our sample included African-American youth in an urban, economically distressed community, the generalizability of our findings may be limited to similar populations in an economically challenged city with significant ETV exposure and high concentrations of non-white individuals. Similarly, our findings should not be generalized to other racial/ethnic groups because the current analysis focused exclusively on the experience of African Americans. Third, as in most longitudinal studies, the missing data and attrition in our study may compromise the generalizability of our findings. Our attrition analysis, however, indicated no differences in key study variables between the final sample and the participants who had missing data. One exception is that the retained sample had slightly lower levels of maternal education compared to the participants excluded due to attrition or missing data, but the difference was minimal. In addition, we used full information maximum likelihood for handling missing data, which helps to generate unbiased and consistent estimates relative to other approaches.⁵¹ Fourth, two of our risk measures were assessed with a single item. These items, however, assess behaviors and do not require the kind of judgment necessary for attitudinal measures, and the items are face valid and have been used in several published studies.

Conclusions and Implications

To the best of our knowledge, this is the first study that examined the longitudinal relationship between ETV and the patterns from adolescence to young adulthood in an urban sample of African Americans. Our findings indicated that risk behaviors clustered in our African American sample, but more notably, that ETV predicted such high-risk clustering after controlling for other covariates. Chronic disease prevention and control strategies may benefit from measuring and addressing violence exposure if violence is an issue in the community.

So What?

What is already known on this topic?

African Americans shoulder the burden of both higher risk for preventable chronic diseases and higher risk for exposure to violence. Chronic noncommunicable diseases are often caused by behavioral risk factors such as alcohol use, tobacco use, physical inactivity, and poor diet quality.

What does this article add?

Cumulative exposure to violence during adolescence and emerging adulthood predicted patterns of health risk behaviors during young adulthood. Individuals with higher exposure to violence were more likely to exhibit smoking, heavy alcohol use, poor diet and physical inactivity simultaneously.

Implications for health promotion practitioners and researchers?

Violence prevention may be a useful strategy to address chronic disease disparities among African Americans. Chronic disease prevention and control strategies may benefit from measuring and addressing violence exposure if violence is an issue in the community.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a grant from the National Institute of Drug Abuse (NIDA) (R01-DA07484 and R01-DA035811).

ORCID iDs

Hsing-Fang Hsieh, PhD

Briana A. Scott, BS

Andria B. Eisman, PhD, MPH

References

National Center for Chronic Disease Prevention and Health Promotion. Chronic diseases in America. Published 2019. Accessed January 15, 2021. https://www.cdc.gov/chronicdisease/resources/infographic/chronic-diseases.htm

World Health Organization. WHO | Noncommunicable Diseases Country Profiles 2014. World Health Organization; 2014.

Meader

King

Moe-Byrne

, et al. A systematic review on the clustering and co-occurrence of multiple risk behaviours. BMC Public Health. 2016;16(1):1–9. doi:10.1186/s12889-016-3373-6

Kvaavik

Influence of individual and combined health behaviors on total and cause-specific mortality in men and women. Arch Intern Med. 2010;170(8):711–718. doi:10.1001/archinternmed.2010.76

Mozaffarian

Roger

, et al. Heart disease and stroke statistics—2013 update: a report from the American Heart Association. Circulation. 2013;127(1):e6–e245. doi:10.1161/CIR.0b013e31828124ad

Jackson

Knight

Rafferty

. Race and unhealthy behaviors: chronic stress, the HPA axis, and physical and mental health disparities over the life course. Res Pract. 2010;100(5):933–939. doi:10.2105/AJPH.2008.143446

Williams

Williams-morris

. Racism and mental health: the African American experience racism and mental health: the African American. Ethn Health. 2010;7858(5):3–4. doi:10.1080/713667453

Cho

Nguyen

Strong

, et al. Multiple health behaviors engagement in an African American cohort: clustering patterns and correlates. Heal Educ Behav. 2019;46(3):506–516. doi:10.1177/1090198119826207

Conry

Morgan

Curry

, et al. The clustering of health behaviours in Ireland and their relationship with mental health, self-rated health and quality of life. BMC Public Health. 2011;11(1):692. doi:10.1186/1471-2458-11-692

10.

Rabel

Laxy

Thorand

Peters

Schwettmann

Mess

. Clustering of health-related behavior patterns and demographics. Results from the population-based KORA S4/F4 cohort study. Front Public Heal. 2019;6:1–9. doi:10.3389/fpubh.2018.00387

11.

Noble

Paul

Turon

Oldmeadow

. Which modifiable health risk behaviours are related? A systematic review of the clustering of smoking, nutrition, alcohol and physical activity (‘SNAP’) health risk factors. Prev Med (Baltim). 2015;81:16–41. doi:10.1016/j.ypmed.2015.07.003

12.

Boynton-Jarrett

Ryan

Berkman

Wright

. Cumulative violence exposure and self-rated health: longitudinal study of adolescents in the United States. Pediatrics. 2008;122(5):961–970. doi:10.1542/peds.2007-3063

13.

Macmillan

. Violence and the life course: the consequences of victimization for personal and social development. Annu Rev Sociol. 2001;27(1):1–22. doi:10.1146/annurev.soc.27.1.1

14.

Leventhal

Pettit

Lewinsohn

. Characterizing major depression phenotypes by presence and type of psychomotor disturbance in adolescents and young adults. Depress Anxiety. 2008;25(7):575–592. doi:10.1002/da.20328

15.

Miller

Chen

Parker

. Psychological stress in childhood and susceptibility to the chronic diseases of aging: moving towards a model of behavioral and biological mechanisms. Psychol Bull. 2012;137(6):959–997. doi:10.1037/a0024768.Psychological

16.

Hsieh

H-F

Heinze

Lang

Mistry

Buu

Zimmerman

. Violence victimization, social support, and papanicolaou smear outcomes: a longitudinal study from adolescence to young adulthood. J Womens Health (Larchmt). 2017;26(12):1340–1349. doi:10.1089/jwh.2016.5799

17.

Mcdougall

Vaillancourt

. Long-term adult outcomes of peer victimization in childhood and adolescence. Am Psychol. 2015;70(4):300–310. doi:10.1037/a0039174

18.

Gómez

Johnson

Selva

Sallis

. Violent crime and outdoor physical activity among inner-city youth. Prev Med (Baltim). 2004;39(5):876–881. doi:10.1016/j.ypmed.2004.03.019

19.

Okada

Miranda

Pena G das

Levy

Azeredo

. Association between exposure to interpersonal violence and social isolation, and the adoption of unhealthy weight control practices. Appetite. 2019;142:104384. doi:10.1016/j.appet.2019.104384

20.

Sharkey

The long reach of violence: a broader perspective on data, theory, and evidence on the prevalence and consequences of exposure to violence. Annu Rev Criminol. 2018;1(1):85–102. doi:10.1146/annurev-criminol-032317-092316

21.

Boynton-Jarrett

Hair

Zuckerman

. Turbulent times: effects of turbulence and violence exposure in adolescence on high school completion, health risk behavior, and mental health in young adulthood. Soc Sci Med. 2013;95:77–86. doi:10.1016/j.socscimed.2012.09.007

22.

Browning

Calder

Ford

Boettner

Smith

Haynie

. Understanding racial differences in exposure to violent areas. Ann Am Acad Pol Soc Sci. 2017;669(1):41–62. doi:10.1177/0002716216678167

23.

Woods-Jaeger

Berkley-Patton

Piper

O’Connor

Renfro

Christensen

. Mitigating negative consequences of community violence exposure: perspectives from African American youth. Health Aff (Millwood). 2019;38(10):1679–1686. doi:10.1377/hlthaff.2019.00607

24.

Heinze

Stoddard

Aiyer

Eisman

Zimmerman

. Exposure to violence during adolescence as a predictor of perceived stress trajectories in emerging adulthood. J Appl Dev Psychol. 2017;49:31–38. doi:10.1016/J.APPDEV.2017.01.005

25.

Brown

Anda

Felitti

, et al. Adverse childhood experiences are associated with the risk of lung cancer: a prospective cohort study. BMC Public Health. 2010;10(1):20. doi:10.1186/1471-2458-10-20

26.

Garner

Shonkoff

Siegel

, et al. Early childhood adversity, toxic stress, and the role of the pediatrician: translating developmental science into lifelong health. Pediatrics. 2012;129(1): e224–e231. doi:10.1542/peds.2011-2662

27.

Motley

Sewell

Chen

Y-C

. Community violence exposure and risk taking behaviors among black emerging adults: a systematic review. J Community Health. 2017;42(5):1069–1078. doi:10.1007/s10900-017-0353-4

28.

Vermunt

. Latent class modeling with covariates: two improved three-step approaches. Polit Anal. 2010;18(4):450–469. doi:10.1093/pan/mpq025

29.

Zimmerman

. Flint [Michigan] Adolescent Study (FAS): a longitudinal study of school dropout and substance use, 1994-1997. Interuniv Consort Polit Soc Res. (ICPSR 34598) https://www.icpsr.umich.edu/web/NAHDAP/studies/34598

30.

Zimmerman

Caldwell

Bernat

. Discrepancy between self-report and school-record grade point average: correlates with psychosocial outcomes among African American adolescents. J Appl Soc Psychol. 2002;32(1):86–109. doi:10.1111/j.1559-1816.2002.tb01421.x

31.

Heinze

Stoddard

Aiyer

Eisman

Zimmerman

. Exposure to violence in adolescence as a predictor of stress trajectories in emerging adulthood. J Appl Dev Psychol. 2017;49:31–38. doi:10.1016/j.appdev.2015.07.003

32.

Hope

Lee

Hsieh

Hurd

Sparks

Zimmerman

. Violence exposure and sexual risk behaviors for African American adolescent girls: the protective role of natural mentorship and organizational religious involvement. Am J Community Psychol. 2019;64(1-2):242–255. doi:10.1002/ajcp.12341

33.

Lee

Hsieh

Stoddard

, et al. Longitudinal pathway from violence exposure to firearm carriage among adolescents: the role of future expectation. J Adolesc. 2020;81:101–113. doi:10.1016/j.adolescence.2020.03.009

34.

Piedmont

. Inter-item correlations. In: Michalos

, ed. Encyclopedia of Quality of Life and Well-Being Research. Springer; 2014:3303–3304. doi:10.1007/978-94-007-0753-5_1493.

35.

Bachman

Johnston

O’Malley

Schulenberg

. Monitoring the Future National Survey Results on Drug Use, 1975–2008: Volume I, Secondary School Students. Ann Arbor, MI; 2009.

36.

Hagoel

Ore

Neter

Silman

Rennert

. Clustering women‘s health behaviors. Heal Educ Behav. 2002;29(2):170–182.

37.

Laska

Pasch

Lust

Story

Ehlinger

. Latent class analysis of lifestyle characteristics and health risk behaviors among college youth. Prev Sci. 2009;10(4):376–386. doi:10.1007/s11121-009-0140-2

38.

Milton

Bull

Bauman

Milton

Heart

. Reliability and validity testing of a single-item physical activity measure correspondence to. Br J Sport Med. 2010;45(3):203–208. doi:10.1136/bjsm.2009.068395

39.

Muthén

. Mplus User’s Guide. 8th ed. Muthén & Muthén; 2017.

40.

Akaike

. Factor analysis and AIC. Psychometrika. 1987;52(3):317–332. doi:10.1007/BF02294359

41.

Schwarz

. Estimating the dimension of a model. Ann Stat. 1978;6(2):461–464.

42.

Tofighi

Enders

. Identifying the correct number of classes in a growth mixture model. In: Hancock

Samuelsen

, eds. Advances in Latent Variable Mixture Models. Greenwich, CT: Information Age Publishing, Inc.; 2008:317–341.

43.

Mendell

Rubin

. Testing the number of components in a normal mixture. Biometrika. 2001;88(3):767–778. doi:10.1093/biomet/90.4.991

44.

Feng

McCulloch

. Using bootstrap likelihood ratios in finite mixture models. J R Stat Soc Ser B. 1996;58(3):609–617. doi:10.1111/j.2517-6161.1996.tb02104.x

45.

Fowler

Tompsett

Braciszewski

Jacques-Tiura

Baltes

. Community violence: a meta-analysis on the effect of exposure and mental health outcomes of children and adolescents. Dev Psychopathol. 2009;21(1):227–259. doi:10.1017/S0954579409000145

46.

Taylor

Makarem

Shimbo

Aggarwal

. Gender differences in associations between stress and cardiovascular risk factors and outcomes. Gend Genome. 2018;2(4):111–122. doi:10.1177/2470289718820845

47.

Sims

Gordon

Garcia

, et al. Perceived stress and eating behaviors in a community-based sample of African Americans. Eat Behav. 2008;9(2):137–142. doi:10.1016/j.eatbeh.2007.06.006

48.

Fowler-Brown

Bennett

Goodman

Wee

Corbie-Smith

James

. Psychosocial stress and 13-year BMI change among blacks: the Pitt county study. Obesity (Silver Spring). 2009;17(11):2106–2109. doi:10.1038/oby.2009.130

49.

Purtle

. Population mental health & community violence: advancing the role of local health departments. Am J Public Health. 2017;107(9):1358–1360. doi:10.2105/AJPH.2017.303979

50.

Lindstrom Johnson

Elam

Rogers

Hilley

. A meta-analysis of parenting practices and child psychosocial outcomes in trauma-informed parenting interventions after violence exposure. Prev Sci. 2018;19(7):927–938. doi:10.1007/s11121-018-0943-0

51.

Dong

Peng

CYJ

. Principled missing data methods for researchers. Springerplus. 2013;2(1):1–17. doi:10.1186/2193-1801-2-222