short literature review

Contents lists available at ScienceDirect

Addictive Behaviors

journal homepage: www.elsevier.com/locate/addictbeh

Substance use and suicidal ideation among child welfare involved adolescents: A longitudinal examination

Christina M. Sellersa,b,⁎, Ruth G. McRoyb, Kimberly H. McManama O'Briena,c,d

a Department of Psychiatry, Boston Children's Hospital, 300 Longwood Ave, Boston, MA 02115, USA b Boston College, School of Social Work, 140 Commonwealth Avenue, Chestnut Hill, MA 02467, USA c Department of Psychiatry, Harvard Medical School, 25 Shattuck St, Boston, MA 02115, USA d Department of Innovation in Practice and Technology, Education Development Center, 43 Foundry Ave, Waltham, MA 02453, USA

H I G H L I G H T S

• For child welfare involved youth, individual, family and peer factors predict alcohol and marijuana use, and suicide ideation • For child welfare involved youth, there is evidence of a reciprocal relationship between alcohol use and suicide ideation • Interventions for child welfare involved youth should focus on the relationship between alcohol use and suicide ideation

A R T I C L E I N F O

Keywords: Alcohol Marijuana Suicidal ideation Child-welfare involved adolescents

A B S T R A C T

Background: The purpose of this study was to investigate the longitudinal predictors of alcohol use, marijuana use, and suicidal ideation among maltreated adolescents. Methods: Longitudinal data from this study come from three waves of the National Survey of Child and Adolescent Wellbeing II (NSCAW II). Participants included 1050 adolescents (Mage = 14.13) who were subjects of child abuse or neglect investigations. Items from the Health Risk Behavior Questionnaire were used to measure alcohol and marijuana use. Suicidal ideation was measured using an item from the Childhood Depression Inventory. Data on deviant peer affiliation, caregiver health, maltreatment type, age, race, and gender were also collected. Results: Marijuana use, suicidal ideation, caregiver drug abuse, deviant peer affiliation, age, and race were predictive of alcohol use. Alcohol use, deviant peer affiliation, age, and time were predictive of marijuana use. Alcohol use, deviant peer affiliation, age, and gender predicted suicidal ideation. Conclusions: Longitudinal evidence indicated that individual, family, and peer factors played an important role in predicting alcohol use, marijuana use, and suicidal ideation among child welfare involved adolescents. In addition, this study provides evidence of a potentially reciprocal relationship between alcohol use and suicidal ideation among this population. Intervention efforts for reducing the public health problems of substance use and suicide among child welfare involved adolescents should focus on the importance of peers in influencing thoughts and behaviors, as well as the functional relationship between alcohol use and suicidal ideation.

1. Introduction

Approximately 3.6 million referrals alleging child maltreatment are received in the United States (US) each year (U.S. Department of Health & Human Services, Administration for Children and Families, Administration on Children, 2016). Nearly two-thirds (61%) of these referrals are screened for further investigation with a substantial pro- portion eventually defined as substantiated child abuse and neglect

cases. In 2014, for example, 702,000 children and youth were identified as child abuse and neglect victims (U.S. Department of Health & Human Services, Administration for Children and Families, Administration on Children, 2016). Since the circumstances and conditions in the child welfare system are stressful and often traumatic, youth involved in this system compared to other youth may be prone to engage in risky be- haviors that can have life-threatening consequences. Although not all maltreated youth are involved in risky behaviors, a large proportion use

https://doi.org/10.1016/j.addbeh.2019.01.021 Received 16 July 2018; Received in revised form 27 November 2018; Accepted 14 January 2019

⁎ Corresponding author at: Department of Psychiatry, Boston Children's Hospital, 300 Longwood Ave, Boston, MA 02115, USA. E-mail addresses: [email protected] (C.M. Sellers), [email protected] (R.G. McRoy),

[email protected] (K.H.M. O'Brien).

Addictive Behaviors 93 (2019) 39–45

Available online 17 January 2019 0306-4603/ © 2019 Elsevier Ltd. All rights reserved.

T

alcohol and other substances (Ireland et al., 2002) and endorse suicidal ideation (Brown et al., 1999). Research has demonstrated that a history of childhood abuse is a strong risk factor for suicidal ideation (Zapata et al., 2013) and alcohol misuse and related problems (Widom & Hiller- Sturmhöfel, 2001). Specifically, 29% of maltreated youth in the Na- tional Survey of Child and Adolescent Wellbeing I (NSCAW I) engaged in substance use, with 9% reporting moderate to high levels of use and 5% reporting risky suicidal behavior (Wall & Kohl, 2007). Another study compared adolescents in a child welfare involved sample with adolescents from public high schools and found that the child welfare involved adolescents were approximately 1.5 times more likely to ex- perience suicidal ideation, when compared to adolescents from the public high schools (Heneghan et al., 2013).

Alcohol use increases the risk for suicide attempts among adoles- cents with suicidal ideation (Schilling et al., 2009). Alcohol consump- tion results in disinhibition of behavior that can enhance the odds of acting on suicidal thoughts (Bagge et al., 2013; Bryan et al., 2016; McManama O'Brien et al., 2014; Sher, 2006). Research has demon- strated proximal and distal effects of alcohol use on suicide attempts, as well as proximal and distal effects of suicide attempts on alcohol use (Bagge & Sher, 2008). However, these relationships are complex and in need of further research (Bagge & Sher, 2008), particularly among a child welfare involved sample.

In addition, in a systematic review, Bridge, Goldstein, and Brent (2006) noted that common risk factors for both suicidal thoughts and behaviors and alcohol and marijuana use include parent/caregiver variables as well as peer variables. Peer relationships are often a source of influence for an adolescent's substance using behavior, as well for their suicidal thoughts and behaviors. One reason that peers are par- ticularly influential is because the peer group often defines the beha- vioral norms within adolescents' social context. In addition, teens begin to spend increasing time with their peers during adolescence14. Previous school based research suggests that adolescents often affiliate with peers who engage in similar behaviors as their own (Urberg et al., 2003). Research has also demonstrated that peers also influence each other's behavior (Hartup, 2005). Bridge and colleagues (2006) suggest that associating with a deviant peer group is a risk factor for suicidal thoughts and behaviors as well as alcohol and marijuana use. Moreover, research using structural equation modeling has found deviant peer affiliation is related to suicidal ideation such that having a deviant peer affiliation can increase substance use and depression, which ultimately increases risk for suicidal ideation (Prinstein et al., 2000).

Caregivers (i.e., parents) are also influential in an adolescent's substance using behavior and suicidal thoughts and behaviors. Caregiver health is one risk factor for substance use and suicidal thoughts and behaviors. Specifically, having a caregiver with depres- sion and/or a caregiver with alcohol or drug abuse has been identified as risks for poorer outcomes among adolescents and maltreated youth (Dubowitz & Bennett, 2007; Jaffee & Maikovich-Fong, 2011). More- over, research suggests associations between suicidal ideation and at- tempts and a poor family environment, parental psychiatric history, and low parental monitoring (King et al., 2000).

Although prior studies provide a substantive foundation in sub- stance use and suicidal ideation among maltreated youth, some sig- nificant gaps remain in the research literature. First, studies examining substance use and suicidal ideation among maltreated adolescents have often relied on data collected before 2007 (Ireland et al., 2002; Brown et al., 1999; Wall & Kohl, 2007). Older data limits our ability to gen- eralize findings to the present, especially when terminology and po- licies have changed (i.e., increased demands and decreased resources for child welfare agencies; state budget cuts; and federal assessment through Child and family Services Reviews aimed at higher account- ability on agencies (Dolan et al., 2011). Second, past research has es- tablished alcohol use as a risk factor for suicide ideation, but limited research has examined the effects of marijuana. Third, the combination of these problems has been understudied using a sample of maltreated

youth. This study intends to fill some gaps by using the NSCAW II, the most recent (2008–2012) longitudinal national data set on maltreated and other youth, focusing on both substance use and suicidal ideation. Specifically, this study aims to investigate the longitudinal predictors of alcohol and marijuana use and suicidal ideation among maltreated youth. This study adds to the empirical literature by adding marijuana as an additional variable, and examining an old research question within a new population, that of child welfare involved youth. Based on the empirical literature, the following research questions and hy- potheses were developed in order to address study aims:

Research Question 1. After controlling for time, what factors predict the odds of using substances among maltreated youth?

Hypothesis 1. Age, gender, suicidal ideation, deviant peer affiliation, caregiver health, maltreatment type, and placement type will predict substance use among child welfare involved youth over time.

Research Question 2. After controlling for time, what factors predict the odds of endorsing suicidal ideation among maltreated youth?

Hypothesis 2. Age, gender, alcohol use, marijuana use, deviant peer affiliation, caregiver health, maltreatment type, and placement type will predict suicidal ideation.

2. Method

This study used the restricted data from the NSCAW II. The NSCAW is a national, longitudinal survey of children and families who have had child protective service investigations. The NSCAW collects data from children, parents, and other caregivers. Reports from caseworkers, teachers, and administrative records are also collected. To date, there have been two rounds of NSCAW: NSCAW I (1996–2007, five waves) and NSCAW II (2008–2012, three waves). This study utilized NSCAW II data, as the landscape of the child welfare population and the policies impacting the child welfare agencies have evolved since NSCAW I. The study was approved under the exempt [Exempt 45 CFR 46. 101(b)] status by the Institutional Review Board (IRB) at the overseeing uni- versity.

2.1. Participants

Participants in this study included 1050 adolescents age 11–17.5 (Mage = 14.13) at W1, who were subjects of child abuse or neglect in- vestigations conducted by Child Protective Services. In this sample, 55.43% of participants identified as female, 52.85% identified as White, and 67.52% were in in-home care (i.e., living with their families).

3. Measures

Alcohol Use. Alcohol use was measured using the Health Risk Behaviors Questionnaire, which is an adolescent self-report measure developed from the Youth Risk Behavior Surveillance System (Kann et al., 2000). A single item was used to measure the frequency of any alcohol use over the past 30 days. Responses to this item were on a 7- point Likert scale, with options ranging from “0 = zero days” to “6=all 30 days.” In this sample, alcohol use was severely skewed and lepto- kurtic (skew = 3.01, kurtosis = 13.41; M = 0.36, SD = 0.86; range [0,6]). No data transformations showed significant improvement. Ad- ditional sensitivity analyses showed that using a dichotomous variable was the best approach. Therefore, for this study, alcohol use was di- chotomized into “1=past 30-day alcohol use”, or “0=no past 30-day alcohol use”.

Marijuana Use. Marijuana use was measured using a single item from the Health Risk Behaviors Questionnaire, measuring the frequency of marijuana use over the past 30 days. Responses to this item were on a 6-point Likert scale, with options ranging from “0 = zero times” to “5=50 or more times.” In this sample, marijuana use was severely

C.M. Sellers et al. Addictive Behaviors 93 (2019) 39–45

40

skewed and leptokurtic (skew = 3.37, kurtosis = 14.03; M = 0.33, SD = 1.00; range [0,5]). No data transformations showed significant improvement. Additional sensitivity analyses showed that using a di- chotomous variable was the best approach. Therefore, for this study marijuana use was dichotomized into “1=past 30-day marijuana use”, or “0=no past 30-day marijuana use”.

Suicidal Ideation. Suicidal ideation (SI) was measured from a single item from the Childhood Depression Inventory (CDI: Kovacs, 1992). The CDI measures symptom severity over the past 2 weeks. Adolescents were asked, “Which of these best says how you have felt [in the past 2 weeks]?” The first response (0 = I do not think about killing myself) indicates an absence of SI, whereas the second (1 = I think about killing myself but I wouldn't do it) and third (2 = I want to kill myself) represent SI and suicidal intent, respectively. For this study, adolescents who responded with 1 or 2 were identified as having sui- cidal ideation.

Deviant Peer Affiliation. Deviant peer affiliation was measured using the Deviant Peer Affiliation scale (Capaldi & Patterson, 1989). This 6 item scale measures involvement with peers who engage in risky or deviant behaviors with questions regarding how many friends cheated on school tests, how many friends suggested they broke the law, and how many stole. In this sample, deviant peer affiliation use skewed and leptokurtic (skew = 1.88, kurtosis = 6.89; M = 1.56, SD = 0.74; range (U.S. Department of Health & Human Services, Administration for Children and Families, Administration on Children, 2016; Widom & Hiller-Sturmhöfel, 2001). No data transformations showed significant improvement. Additional sensitivity analyses showed that using a dichotomous variable was the best approach. Therefore, for this study, this variable was recoded into a dichotomous variable. A score of 0 indicates the participants total score was below the median, and a score of 1 indicates the total score was above the median.

Caregiver Health. Four different measures were used to measure caregiver health. Caregivers physical health was measured using the standardized score from the Physical Health Summary, and caregivers mental health was measured using the standardized score from the Mental Health Summary, from the Short Form Health Survey (Ware et al., 1996). These two composite scales were calculated from 12 questions with the composite score ranging from 0 to 100 with higher scores indicating better health.

Caregiver alcohol dependence was measured using the total score from the Alcohol Use Disorders Identification Test (AUDIT); (Babor et al., 2001). The AUDIT consists of 10 questions with response options ranging from 0 to 4. In this sample, AUDIT was severely skewed and leptokurtic (skew = 4.43, kurtosis = 31.53; M = 1.77, SD = 3.32; range (U.S. Department of Health & Human Services, Administration for Children and Families, Administration on Children, 2016; 29 Legal Medical Marijuana States and DC - Medical Marijuana, 2017)). No data transformations showed significant improvement. Additional sensitivity analyses showed that using a dichotomous variable was the best ap- proach. Therefore, for this study, scores of 7 or below on the AUDIT were recoded with a 0=“non hazardous drinking” and scores > 7 were recoded to 1 = “hazardous drinking” mirroring clinical cut offs.

Caregiver drug abuse was measured using the Drug Abuse Screening Test (DAST; (Skinner, 1982). The DSAT consists of 28 self-response questions that measure the abuse of drugs other than alcohol. Each question has a yes or no answer and a score of “1” was given for each yes response, except for items 4,5, and 7, which are phrased in opposite directions and thus, the no response was given a score of “1”. In this sample, the DAST was severely skewed and leptokurtic (skew = 4.95, kurtosis = 36.89; M = 0.74, SD = 1.63; range [0,18]). No data trans- formations showed significant improvement. Additional sensitivity analyses showed that using a dichotomous variable was the best ap- proach. Therefore, for this study, the DSAT was recoded into a di- chotomous variable, using clinical cut off scores where 0 represents “No Drug Abuse” and 1 represents “Drug Abuse.”

Placement Type. Using administrative records, placement type was measured using W1 data, and recoded to include, 0 = “In-Home”, 1 = “Kinship Care”, 2 = “Foster Care”, and 3=“Other out of home arrangement, i.e., group home.”

Maltreatment Type. The most serious maltreatment type was measured by caseworker report, using W1 data, and was recoded to include: 0 = “Physical Maltreatment”, 1 = “Sexual Maltreatment”, 2 = “Emotional Maltreatment”, 3 = “Neglect”, and 4 = “Other.”

4. Data analysis

Data management and preliminary analyses were conducted using STATA 14 SE. Next, panel data analysis using logistic models for di- chotomous variables were run in order to investigate study hypotheses. This type of analysis was chosen as conventional logistic regressions do not account for dependency within each participant (Rabe-Hesketh & Skrondal, 2012). Specifically, random effect models were utilized to examine study aims, which invested the predictors for alcohol use, marijuana use, and suicidal ideation.

In the alcohol use model, the Hausman test indicated that for caregiver drug abuse, the between and within effects were different. Consequently, these effects were estimated separately in the model. For the marijuana use model, only within effects were estimated, as the Hausman test indicated no difference in within and between effects. Lastly, in the suicidal ideation model, the Hausman test indicated that the between and within effects for caregiver mental health were dif- ferent. Given these differences, the effects were estimated separately.

5. Results

Descriptive statistics of the sample are presented in Table 1, and Chi square analyses are presented in Table 2. These results indicate cross- sectional relationships between alcohol use and suicidal ideation at all three waves.

6. Random effect models

Model One: Alcohol with Covariates. The results for the model that tested the predictors of alcohol use among child welfare involved youth are presented in Table 3. In this model, there were 1402 ob- servations within 832 subjects. The model was statistically significant (Wald Chi2 = 117.13, p < .001), suggesting that the odds of drinking alcohol in the past 30 days can be predicted from the independent variables. In this model, marijuana use, suicidal ideation, the between effect for caregiver drug abuse, deviant peer affiliation, age, and race were statistically significantly predictive of alcohol use.

In regards to individual thoughts and behaviors, when all other variables were controlled for, the odds of drinking alcohol for those who use marijuana increased 3662% when compared to those who did not use marijuana. Similarly, when all other variables were controlled for, those who presented with suicidal ideation were 113% more likely to drink alcohol, than those who did not have suicidal ideation.

In regards to family and peer variables, the odds of past 30 day al- cohol use for an individual who had a caregiver with drug dependence was 67% less, when compared to an individual with a caregiver who did not have drug dependence (between effect) and the odds of drinking alcohol increased by 223% for adolescents whose deviant peer affilia- tion score was above the median, compared to those whose deviant peer affiliation score was below the median, when controlling for all other variables.

Lastly, demographic characteristics, namely age and race, also played a role in predicting the odds of drinking alcohol use. The odds of past 30 day alcohol use for Black youth was 52% lower than for White youth, when all other variables were controlled for. Additionally, for each one-month increase in age, there was a 4% increase in the odds of drinking alcohol when controlling for all other variables.

C.M. Sellers et al. Addictive Behaviors 93 (2019) 39–45

41

Model Two: Marijuana with Covariates. The results of testing the predictors for marijuana use among child welfare involved youth are presented in Table 4. There were 1360 observations within 821 parti- cipants and the random effect model was statistically significant (Wald Chi2 = 70.57, p < .001), suggesting that the odds of using marijuana in the past 30 days can be predicted from the independent variables. Alcohol use, deviant peer affiliation, age, and time were statistically significant predictors in this model. Furthermore, maltreatment type, and specifically sexual maltreatment approached significance.

Drinking alcohol in the past 30 days, compared with not drinking alcohol in the past 30 days, presented a 4861% increase in the odds of using marijuana, when controlling for all other variables. In addition, youth whose deviant peer affiliation score was above the median had a

Table 1 Descriptive statistics for key variables: NSCAW II panel of adolescents with child welfare involvement (Panel selected at W1, 2008–2009).

Variable Wave 1: Baseline (N)

Wave 2: 18- Months (N)

Wave 3: 36 Months (N)

Mean Age 14.13 (1050) 15.30 (854) 17.29 (768) Gender Male 44.57% (468) Female 55.42% (582) Race American Indian 12.30% (125) Asian/Hawiian/Pacific

Islander 4.72% (48)

Black 30.12% (306) White 52.85% (537) Substantiated maltreatment 52.94% (541) Placement In-Home 67.52% (709) Kinship Care 14.00% (147) Foster Care 12.38% (130) Other OOH Placement 6.10% (64) Lifetime history of Alc. Use

(Yes) 43.24%% (435)

47.35% (393) 55.77% (401)

P30 Day Alc days used (Yes)⁎ 16.22% (165) 20.22% (168) 28.47% (209) Lifetime history of MJ Use

(Yes) 23.01% (231) 30.53% (254) 38.69% (277)

P30 Day MJ used (Yes)⁎ 10.13% (103) 13.82% (115) 16.64% (122) P2 week SI (Yes)⁎ 19.56% (194) 17.00% (126) 12.80% (53) Deviant Peer Affiliation

(Above the Mean) 54.29% (570) 61.62% (647) 22.48% (236)

Caregiver Alcohol Use (Hazardous Drinking)

32.86% (345) 43.43% (456) 65.81% (691)

Caregiver Substance Dependence (Drug Abuse)

32.57% (342) 44.95% (472) 67.33% (707)

Caregiver Physical Healtha 46.44 (1008) 45.83 (789) 45.36 (462) Caregiver Mental Healtha 48.65 (1008) 50.10 (789) 49.18 (462)

Notes: unbalanced panel ⁎ P = Past a Caregiver Physical and Mental Health each ranged from 0 to 100 with

higher scores indicating better health, thus the sample was on average, of average health.

Table 2 Chi Square values for suicidal ideation related to alcohol use and marijuana use at all three waves.

Variable Chi Square p

Wave 1 Alcohol use 16.83 0.000 Marijuana use 6.81 0.009 Wave 2 Alcohol use 6.97 0.008 Marijuana use 0.20 0.652 Wave 3 Alcohol use 9.54 0.002 Marijuana use 3.54 0.060

Table 3 Random effect model of alcohol use on marijuana use; suicidal ideation; care- giver alcohol dependence, drug dependence, mental health, and physical health; age, race, gender, maltreatment type, placement type, and time.

Variable Coefficient Odds Ratio

Marijuana Use 3.63 (0.39)*** 37.62 Suicidal Ideation 0.76 (0.25)** 2.13 Caregiver Alcohol Dependence 0.15 (0.37) 1.16 Caregiver Drug Abuse (Between) −1.10 (0.54)*** 0.33 Caregiver Drug Abuse (Within) −0.19 (0.42) 0.83 Caregiver Mental Health −0.00 (0.01) 1.00 Caregiver Physical Health 0.00 (0.01) 1.00 Deviant Peer Affiliation 1.17 (0.23)*** 3.23 Age 0.04 (0.01)*** 1.04 Race American Indian 0.09 (0.34) 1.10 Asian/Hawaiian/Pacific Islander 0.08 (0.52) 1.08 Black −0.73 (0.23)** 0.48 Gender 0.15 (0.23) 1.16 Maltreatment Type Sexual Maltreatment −0.78 (0.56) 0.46 Emotional Maltreatment −0.89 (0.63) 0.41 Neglect Maltreatment −0.52 (0.42) 0.59 Other Maltreatment −0.26 (0.34) 0.77 Placement In-Home Care 0.09 (0.34) 1.10 Foster Care −0.05 (0.47) 0.95 Kinship Care −0.35 (0.60) 0.70 Time 0.29 (0.21) 1.34 Goodness of Fit Wald Chi2 117.13, p < .001 AIC 887.20 BIC 1007.854

*p < .05; **p < .01, ***p < .001 Note: Standard Errors in parenthesis

Table 4 Random effect model of marijuana use on alcohol use; suicidal ideation; care- giver alcohol dependence, drug dependence, mental health, and physical health; age, race, gender, maltreatment type, placement type, and time.

Variable Coefficient Odds ratio

Random Effect Alcohol Use 3.90 (0.49)*** 49.61 Suicidal Ideation 0.08 (0.35) 1.08 Caregiver Alcohol Dependence 0.42 (0.50) 1.51 Caregiver Drug Abuse −0.86 (0.56) 0.42 Caregiver Mental Health 0.00 (0.01) 1.00 Caregiver Physical Health −0.02 (0.01) 0.99 Deviant Peer Affiliation 0.53 (0.18)** 1.70 Age 0.03 (0.01)** 1.03 Race American Indian 0.33 (0.45) 1.39 Asian/Hawaiian/Pacific Islander −0.28 (0.80) 0.76 Black 0.17 (0.36) 1.19 Gender −0.13 (0.32) 0.88 Maltreatment Type Sexual Maltreatment 1.21 (0.63) 3.36 Emotional Maltreatment 0.78 (0.73) 2.19 Neglect Maltreatment 0.25 (0.53) 1.29 Other Maltreatment −0.39 (0.48) 0.68 Placement In-Home Care 0.32 (0.45) 1.38 Foster Care 0.67 (0.58) 1.96 Kinship Care 0.66 (0.74) 1.94 Time 0.58 (0.28)* 1.79 Goodness of Fit Wald Chi2 70.57, p < .001 AIC 627.13 BIC 741.86

*p < .05; **p < .01, ***p < .001 Note: Standard Errors in parenthesis

C.M. Sellers et al. Addictive Behaviors 93 (2019) 39–45

42

70% increase in the odds of using marijuana when compared to their peers who had lower deviant peer affiliation scores. Increases in age and time also increased the odds of using marijuana. Specifically, for each additional month in age, there was a 3% increase in the odds of using marijuana, and for each additional wave of data collection (18 months), there was a 79% increase in the odds of using marijuana, when all other variables were controlled for. Lastly, although mal- treatment type did not statistically significantly predict marijuana use with > 95% confidence, it did predict marijuana use at 94.5% con- fidence (p = .055), indicating that, we are 94.5% confident that for youth who have a history of sexual maltreatment, compared to other types of maltreatment, the odds for using marijuana increase by 235%.

Model Three: Suicidal Ideation with Covariates. The results testing the predictors for suicidal ideation among child welfare in- volved youth are presented in Table 5. There were 1360 observations within 821 participants. The random effect model for suicidal ideation was statistically significant (Wald Chi2 = 47.45, p < .001), suggesting that the odds of endorsing suicidal ideation in the past week can be predicted from the independent variables.

In this model, alcohol use, deviant peer affiliation, age, and gender were statistically significant predictors of suicidal ideation and the between effect for caregiver mental health was approaching sig- nificance. Child welfare involved youth who drank alcohol had a 134% increase in the odds for suicidal ideation, compared to their peers who did not drink alcohol, when controlling for all other variables. Similarly, when all other variables were controlled, youth whose de- viant peer affiliation score was above the median, had a 97% increase in the odds of having suicidal ideation when compared to their peers whose deviant peer affiliation score was below the median. For each month increase in age, there was a 1% decrease in the odds of having suicidal ideation. Lastly, compared to males, females had a 71% in- creased odds of having suicidal ideation.

7. Discussion

Bivariate results indicated significant differences in suicidal ideation based on alcohol use (at all three waves) and marijuana use (at W1 only). The random effect models demonstrated how individual, family, and peer factors affect alcohol use, marijuana use, and suicidal ideation among youth involved in the child welfare system. Specifically, mar- ijuana use, suicidal ideation, caregiver drug abuse, deviant peer af- filiation, age, and race were predictive of alcohol use over time; alcohol use, deviant peer affiliation, age, and time were predictive of marijuana use over time; and alcohol use, deviant peer affiliation, age, and gender predicted suicidal ideation over time.

Consistent with other studies of child welfare involved youth (Heneghan et al., 2013), our study also found higher rates of alcohol use, marijuana use, and suicide ideation when compare to youth not involved with the child welfare system. In this study, marijuana use, suicidal ideation, caregiver drug abuse, deviant peer affiliation, age, and race were significant predictors of alcohol use over time. As ex- pected, marijuana use, suicidal ideation, and deviant peer affiliation were particularly potent risk factors for alcohol use. When youth use marijuana, experience suicidal ideation, and/or spend time with de- viant peers, they are at an increased risk of using alcohol. In this sample, however, having a caregiver with drug abuse (other than al- cohol) served as a protective factor. Specifically, when comparing in- dividuals with caregivers with drug abuse to individuals with caregivers without drug abuse, those who had a caregiver with drug abuse were at a decreased risk of using alcohol. These results are contrary to other studies, which suggest caregiver drug abuse is a risk factor for alcohol use among adolescents (Hawkins et al., 1992; Kilpatrick et al., 2000). One potential explanation for this finding is that given the sample of participants, these families are closely monitored which may result in additional supports and services. Over the past two decades, re- searchers and clinicians have developed and identified effective stra- tegies and services to support child welfare involved parents and their children when a parent has a substance use problem1. Social Cognitive Theory31 may also help explain this finding. In order to identify the probabilistic nature of a behavior, social cognitive theory suggests that there are five basic cognitive capabilities common to individuals (symbolizing, forethought, vicarious, self-regulatory, and self-re- flective) (Bandura, 1986). The degree to which individuals utilize these capabilities can help to predict how probable it is that the individual will engage in any given behavior31. Given that more and more youth are being removed from their families and/or gaining child welfare involvement as a result of parental substance use, it is possible that youth in this sample utilized the cognitive capabilities of forethought and vicarious learning in order to predict major consequences of their own drinking from the consequences of their parents substance use. As a result, it is possible that through these predictions, youth are less likely to engage in underage drinking given the expectancies they de- veloped around consequences of substance use.

Alcohol use, deviant peer affiliation, age, and time predicted mar- ijuana use over time in this study. Surprisingly, caregiver health (i.e., physical health, mental health, alcohol dependence, or substance abuse) had no significant effect on the odds of using marijuana. These findings may be related to adolescents spending more time with peers and less time with caregivers, as they transition from childhood to adolescence (Steinberg, 2014). Similarly, recent research has demon- strated that peer influences may be stronger than parent influences, particularly when it comes to alcohol and other drug use (Sellers et al., 2018). Specifically, research has suggested that peer factors are one potential mechanism through which alcohol use occurs in adolescence, over and above parental factors such as parental monitoring (Sellers et al., 2018). This may be particularly important for child welfare in- volved youth who may be placed in out of home settings resulting in less family involvement than their same aged peers. At the same time, adolescence is a time of exploration and experimentation and many

Table 5 Random effect model of suicidal ideation on alcohol use; marijuana use; care- giver alcohol dependence, drug dependence, mental health, and physical health; age, race, gender, maltreatment type, placement type, and time.

Variable Coefficient Odds ratio

Alcohol Use 0.85 (0.33)* 2.34 Marijuana Use −0.04 (0.39) 0.96 Caregiver Alcohol Dependence 0.40 (0.38) 1.49 Caregiver Drug Abuse 0.14 (0.37) 1.15 Caregiver Mental Health (Between) −0.03 (0.01) 0.97 Caregiver Mental Health (Within) 0.01 (0.02) 1.01 Caregiver Physical Health −0.01 (0.01) 0.99 Deviant Peer Affiliation 0.68 (0.16)*** 1.97 Age −0.01 (0.01)* 0.99 Race American Indian 0.39 (0.39) 1.47 Asian/Hawaiian/Pacific Islander 0.41 (0.56) 1.51 Black −0.08 (0.29) 0.92 Gender 0.54 (0.25)* 1.71 Maltreatment Type Sexual Maltreatment 0.42 (0.56) 1.52 Emotional Maltreatment −0.50 (0.64) 0.61 Neglect Maltreatment −0.39 (0.44) 0.68 Other Maltreatment −0.26 (0.36) 0.77 Placement In-Home Care −0.41 (0.39) 0.66 Foster Care −0.19 (0.50) 0.83 Kinship Care 0.11 (0.62) 1.11 Time −0.30 (0.19) 0.74 Goodness of Fit Wald Chi2 47.45 p < .001 AIC 1232.69 BIC 1352.64

*p < .05; **p < .01, ***p < .001 Note: Standard Errors in parenthesis

C.M. Sellers et al. Addictive Behaviors 93 (2019) 39–45

43

youth begin to experiment with alcohol during their adolescence (Steinberg, 2014), which may be a gateway into the use of marijuana.

Results from this study indicated that alcohol use, deviant peer af- filiation, age, and gender predict suicidal ideation over time. Both al- cohol use and deviant peer affiliation are risk factors for suicidal ideation among child welfare involved youth. When youth use alcohol or affiliate with deviant peers, they are at an increased risk for suicidal ideation. Being surrounded by a deviant peer group can amplify suicide risk, including increasing suicidal thoughts (Winterrowd & Canetto, 2013). One mechanism through which this may occur is through the proliferation of low emotional and behavioral regulation skills that ultimately contribute to increased suicidal ideation (He et al., 2015). Similarly, Cognitive Behavioral Theory posits that cognition plays a large role in the development and maintenance of emotional and be- havioral responses to a variety of experiences, with thoughts, behaviors, and emotions being intricately tied together (Sowers et al., 2008). Consequently, it is possible that youth who are involved with deviant peers are more likely to engage in deviant behaviors that lead to a negative self-schema and poor self-esteem, contributing to increased suicidal ideation.

The longitudinal analyses indicated a relationship between alcohol use and suicidal ideation among child welfare involved youth, such that alcohol use predicted suicidal ideation and suicidal ideation predicted alcohol use. These results are consistent with previous literature sup- porting the relationship between alcohol use and suicidal ideation among clinical populations (Bagge & Sher, 2008; Nock et al., 2013). Youth in the child welfare system may use alcohol to cope with distress (Khantzian, 1997), and at the same time, alcohol use may exacerbate distress (Marschall-Lévesque et al., 2017), suggesting a potential bi- directional relationship. For youth in the child welfare system, the circumstances and conditions of their child welfare involvement are often stressful and/or traumatic, leading to increased distress. Without the emotion regulation skills to cope with this stress and trauma, youth may use alcohol as one way of coping. Despite this attempt to cope, alcohol use may unintentionally exacerbate that distress as the con- sequences of drinking may be heightened for child-welfare involved youth.

Although this study contributes to the knowledge base on substance use and suicidal ideation among child welfare involved youth, it has several limitations. First, although the NSCAW II is the most recent national level study with data on suicidal ideation and substance use specific to the child welfare population, the data are approximately 10 years old. This study did not find a relationship between marijuana use and suicidal ideation. However, marijuana policies and potency continue to change. In 2008, 12 states had legalized medical marijuana and no states had legalized recreational marijuana. In 2017, 29 states had legalized medical marijuana and 8 states had legalized recreational marijuana (29 Legal Medical Marijuana States and DC - Medical Marijuana, 2017). In addition, there has been, and continues to be, an increase in the strength of marijuana (ElSohly et al., 2016). Specifically, the potency of marijuana has increased from approximately 4% in 1995 to approximately 12% in 2014 (ElSohly et al., 2016). Given the change in marijuana policies and potency over the past ten years, it is possible that the effect marijuana has on suicidal ideation has changed.

Lastly, the NSCAW II collects limited data on suicidal thoughts and behaviors. Understanding all elements of suicidal thoughts and beha- vior are important in understanding the relationship between substance use and suicide, as well as in identifying implications for policy, prac- tice, and research. Only measuring suicidal ideation (as opposed to also including suicide attempts, suicide plans, and non-suicidal self-injury) limits our ability to further understand these nuanced relationships. In addition, measuring gender as a binary construct without allowing for other gender presentations limits our ability to understand the effects of gender on suicide, especially when research has demonstrated that transgender youth are at an increased risk for suicide (Veale et al., 2017).

Despite these limitations, the findings of this study demonstrate that the proportion of youth experiencing substance use and suicidal thoughts is substantially higher among youth involved with the child welfare system, indicating the need for programming and interventions in this area. In addition, results indicated the potency with which peers play a role in both substance use and suicidal thoughts among child welfare involved youth. Consequently, clinicians should be aware of the developmental trajectories of youth and the role peers play in these problems. In order to utilize peers in a positive way, adults should train peers to recognize warning signs of problematic substance use and suicidal thoughts. Lastly, results provide evidence of a bidirectional relationship between alcohol use and suicidal ideation among this sub- population of youth. Research and theory have posited a complex re- lationship between thoughts and behaviors, with changes in one area leading to changes in another (Hollon & Beck, 1994). In addition, the standard of care is that youth receiving treatment for alcohol use or suicidal ideation often do so separately, despite the strong connection. As such, clinicians should consider developing accessible interventions for youth in care by either integrating substance use and suicidal ideation treatment, or targeting a reduction in underage alcohol use as a way to indirectly target suicidal thoughts for this population.

Declarations of interest

None.

Role of funding sources

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Contributions

Christina M. Sellers conducted the literature search and developed the research questions. She conducted the analyses and wrote the manuscript. Ruth G. McRoy and Kimberly H. McManama O'Brien as- sisted with research conception, interpretation of results, and revisions of the manuscript. All authors contributed to and have approved the final manuscript.

References

29 Legal Medical Marijuana States and DC - Medical Marijuana. ProCon.org. https:// medicalmarijuana.procon.org/view.resource.php?resourceID=000881 Published 2017. Accessed December 30, 2017.

Babor, T. F., Higgins-Biddle, J. C., Saunders, J. B., & Monteiro, M. G. (2001). The alcohol use disorders identification test guidelines for use in primary care.

Bagge, C. L., Lee, H.-J., Schumacher, J. A., Gratz, K. L., Krull, J. L., & Holloman, G. (2013). Alcohol as an acute risk factor for recent suicide attempts: A case-crossover analysis. Journal of Studies on Alcohol and Drugs, 74(4), 552–558 (http://www.ncbi.nlm.- nih.gov/pubmed/23739018. Accessed June 8, 2016).

Bagge, C. L., & Sher, K. J. (2008). Adolescent alcohol involvement and suicide attempts: Toward the development of a conceptual framework. Clinical Psychology Review, 28(8), 1283–1296. https://doi.org/10.1016/j.cpr.2008.06.002.

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory (Prentice-Hall, ed.). Engldwood Cliffs, N.J.

Brown, J., Cohen, P., Johnson, J. G., & Smailes, E. M. (1999). Childhood abuse and ne- glect: Specificity of effects on adolescent and young adult depression and suicidality. Journal of the American Academy of Child and Adolescent Psychiatry, 38(12), 1490–1496. https://doi.org/10.1097/00004583-199912000-00009.

Bryan, C. J., Garland, E. L., Rudd, M. D., et al. (2016). From impulse to action among military personnel hospitalized for suicide risk: Alcohol consumption and the re- ported transition from suicidal thought to behavior. General Hospital Psychiatry, 41, 13–19. https://doi.org/10.1016/j.genhosppsych.2016.05.001.

Capaldi, D. M., & Patterson, G. R. (1989). Psychometric properties of fourteen latent con- structs from the Oregon youth study. New York, NY: Springer New Yorkhttps://doi.org/ 10.1007/978-1-4612-3562-0.

Dolan, M., Smith, K., Casanueva, C., Ringeisen, H., & Webb, M. B. (2011). Introduction to NSCAW II NSCAW II BASELINE REPORT: INTRODUCTION TO NSCAW II FINAL REPORT. (http://www.acf.hhs.gov/programs/opre/index.html. Accessed November 24, 2018).

Dubowitz, H., & Bennett, S. (2007). Physical abuse and neglect of children. Lancet, 369(9576), 1891–1899. https://doi.org/10.1016/S0140-6736(07)60856-3.

C.M. Sellers et al. Addictive Behaviors 93 (2019) 39–45

44

ElSohly, M. A., Mehmedic, Z., Foster, S., Gon, C., Chandra, S., & Church, J. C. (2016). Changes in Cannabis potency over the last 2 decades (1995–2014): Analysis of cur- rent data in the United States. Biological Psychiatry, 79, 613–619. https://doi.org/10. 1016/j.biopsych.2016.01.004.

Hartup, W. W. (2005). Peer interaction: What causes what? Journal of Abnormal Child Psychology, 33(3), 387–394. https://doi.org/10.1007/s10802-005-3578-0.

Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention. Psychological Bulletin, 112(1), 64–105. https://doi.org/ 10.1037/0033-2909.112.1.64.

He, A. S., Fulginiti, A., & Finno-Velasquez, M. (2015). Connectedness and suicidal idea- tion among adolescents involved with child welfare: A national survey. Child Abuse & Neglect, 42, 54–62. https://doi.org/10.1016/J.CHIABU.2015.02.016.

Heneghan, A., Stein, R. E. K., Hurlburt, M. S., et al. (2013). Mental health problems in teens investigated by U.S. child welfare agencies. Journal of Adolescent Health, 52, 634–640. https://doi.org/10.1016/j.jadohealth.2012.10.269.

Hollon, S. D., & Beck, A. T. (1994). Cognitive and cognitive-behavioral therapies. In A. E. Bergin, & S. L. Garfield (Eds.). Handbook of psychotherapy and behavior change (pp. 428–466). Oxford, England: John Wiley http://psycnet.apa.org/record/1994-97069- 009Accessed December 30, 2017.

Ireland, T. O., Smith, C. A., & Thornberry, T. P. (2002). Developmental issues in the impact of child maltreatment on later delinquency and drug use. Criminology, 40(2), 359–400. https://doi.org/10.1111/j.1745-9125.2002.tb00960.x.

Jaffee, S. R., & Maikovich-Fong, A. K. (2011). Effects of chronic maltreatment and mal- treatment timing on children's behavior and cognitive abilities. Journal of Child Psychology and Psychiatry, 52(2), 184–194. https://doi.org/10.1111/j.1469-7610. 2010.02304.x.

Kann, L., Kinchen, S. A., Williams, B. I., et al. (2000). Youth risk behavior surveillance – United States, 1999. The Journal of School Health, 70(7), 271–285. https://doi.org/10. 1111/j.1746-1561.2000.tb07252.x.

Khantzian, E. J. (1997). The self-medication hypothesis of substance use disorders: A reconsideration and recent applications. Harvard Review of Psychiatry, 4(5), 231–244. https://doi.org/10.3109/10673229709030550.

Kilpatrick, D. G., Acierno, R., Saunders, B., Resnick, H. S., Best, C. L., & Schnurr, P. P. (2000). Risk factors for adolescent substance abuse and dependence: Data from a national sample. Journal of Consulting and Clinical Psychology, 68(1), 19–30. https:// doi.org/10.1037/0022-006X.68.1.19.

King, R. D., Gaines, L. S., Lambert, E. W., Summerfelt, W. T., & Bickman, L. (2000). The co-occurrence of psychiatric and substance use diagnoses in adolescents in different service systems: Frequency, recognition, cost, and outcomes. The Journal of Behavioral Health Services & Research, 27(4), 417–430. https://doi.org/10.1007/BF02287823.

Kovacs, M. (1992). Childrens depression inventory manual. Marschall-Lévesque, S., Castellanos-Ryan, N., Parent, S., et al. (2017). Victimization, sui-

cidal ideation, and alcohol use from age 13 to 15 years: Support for the self-medication model. https://doi.org/10.1016/j.jadohealth.2016.09.019.

McManama O'Brien, K. H., Becker, S. J., Spirito, A., Simon, V., & Prinstein, M. J. (2014). Differentiating adolescent suicide attempters from ideators: Examining the interac- tion between depression severity and alcohol use. Suicide & Life-Threatening Behavior, 44(1), 23–33. https://doi.org/10.1111/sltb.12050.

Nock, M. K., Green, J. G., Hwang, I., et al. (2013). Prevalence, correlates, and treatment of lifetime suicidal behavior among adolescents: Results from the National Comorbidity

Survey Replication Adolescent Supplement. JAMA Psychiatry. 70(3), 300–310. https://doi.org/10.1001/2013.jamapsychiatry.55.

Prinstein, M. J., Boergers, J., Spirito, A., Little, T. D., & Grapentine, W. L. (2000). Peer functioning, family dysfunction, and psychological symptoms in a risk factor Model for adolescent inpatients' suicidal ideation severity. Journal of Clinical Child Psychology, 29(3), 392–405. https://doi.org/10.1207/S15374424JCCP2903_10.

Rabe-Hesketh, S., & Skrondal, A. (2012). Multilevel and longirudinal modeling using stata, Volume II: Categorical responses, counts, and survival (3). College Station, TX: Stata Press.

Schilling, E. A., Aseltine, R. H., Glanovsky, J. L., James, A., & Jacobs, D. (2009). Adolescent alcohol use, suicidal ideation, and suicide attempts. Journal of Adolescent Health, 44(4), 335–341. https://doi.org/10.1016/j.jadohealth.2008.08.006.

Sellers, C. M., O'Brien, K. H. M., Hernandez, L., & Spirito, A. (2018). Adolescent alcohol use: The effects of parental knowledge, peer substance use, and peer tolerance of use. Journal Soc Social Work Res. 9(1), https://doi.org/10.1086/695809.

Sher, L. (2006). Alcohol consumption and suicide. QJM, 99(1), 57–61. https://doi.org/10. 1093/qjmed/hci146.

Skinner, H. A. (1982). The drug abuse screening test. Addictive Behaviors, 7(4), 363–371. https://doi.org/10.1016/0306-4603(82)90005-3.

Sowers, K. M., Karen, M., & Dulmus, C. N. (2008). Comprehensive handbook of social work and social welfare. John Wiley & Sons (https://books.google.com/books?hl=en&lr= &id=Z_o3LkQgk2YC&oi=fnd&pg=PA101&dq=cognitive+behavioral+theory& ots=BCGDNDSSKv&sig=0sYvkwm0NCOqGAkpP8EWy1g5K94#v=onepage& q=cognitive behavioral theory&f=false. Accessed November 27, 2018).

Steinberg, L. (2014). Adolescence (10). Boston, MA: McGraw Hill. U.S. Department of Health & Human Services, Administration for Children and Families,

Administration on Children (2016). Youth and families CB. Child Maltreatment 2014 (http://www.acf.hhs.gov/programs/cb/research-data-technology/statistics-re- search/child-maltreatment).

Urberg, K. A., Luo, Q., Pilgrim, C., & Degirmencioglu, S. M. (2003). A two-stage model of peer influence in adolescent substance use: Individual and relationship-specific dif- ferences in susceptibility to influence. Addictive Behaviors, 28, 1243–1256. https:// doi.org/10.1016/S0306-4603(02)00256-3.

Veale, J. F., Watson, R. J., Peter, T., & Saewyc, E. M. (2017). Mental health disparities among Canadian transgender youth. The Journal of Adolescent Health, 60(1), 44–49. https://doi.org/10.1016/j.jadohealth.2016.09.014.

Wall, A. E., & Kohl, P. L. (2007). Substance use in maltreated youth: Findings from the national survey of child and adolescent well-being. Child Maltreatment, 12(1), 20–30. https://doi.org/10.1177/1077559506296316.

Ware, J., Kosinski, M., & Keller, S. D. (1996). A 12-item short-form health survey: Construction of scales and preliminary tests of reliability and validity. Medical Care, 34(3), 220–233 (http://www.ncbi.nlm.nih.gov/pubmed/8628042 Accessed November 14, 2017).

Widom, C. S., & Hiller-Sturmhöfel, S. (2001). Alcohol abuse as a risk factor for and consequence of child abuse. Alcohol Res Heal. 25(1), 52–57.

Winterrowd, E., & Canetto, S. S. (2013). The long-lasting impact of adolescents' deviant friends on suicidality: A 3-year follow-up perspective. Social Psychiatry and Psychiatric Epidemiology, 48(2), 245–255. https://doi.org/10.1007/s00127-012-0529-2.

Zapata, L. B., Kissin, D. M., Bogoliubova, O., et al. (2013). Orphaned and abused youth are vulnerable to pregnancy and suicide risk. Child Abuse & Neglect, 37(5), 310–319. https://doi.org/10.1016/j.chiabu.2012.10.005.

C.M. Sellers et al. Addictive Behaviors 93 (2019) 39–45

45