Screening for Adolescent Risk Behaviors: Preliminary Evidence for a Family Functioning Tool

Abstract

Background

Family functioning is associated with adolescent drug use, alcohol use, cigarette use, and sexual risk behaviors. Assessing adolescents for family functioning, commonly associated with multiple risk behaviors, may help identify adolescents at risk for adverse health outcomes. This study examined whether a latent family functioning construct, encompassing multiple dimensions of family functioning, was associated with adolescents’ substance use and sexual risk behaviors.

Method

This study used data harmonization with three intervention trials, including data from 1451 adolescents (M = 13.6, SD = 1.0), to perform a full-information item bifactor analysis on 46 family functioning items from five pre-existing family functioning measures. Regression analysis was used to examine the association between the identified subset of items and the following outcomes: cigarette use, alcohol use, drug use, and condom use.

Results

Bifactor analysis identified a 26-item latent family functioning construct. Regression analysis indicated that a 26-item latent family functioning construct was associated negatively with lifetime and past 90-day cigarette use, alcohol use, and drug use.

Conclusion

In sum, the multi-dimensional latent family functioning construct may target specific barriers to risk screening in adolescent populations, including time constraint, hesitancy in discussing sensitive health topics, and use culturally appropriate and age-appropriate assessments.

Introduction

Adolescent drug use, cigarette use, alcohol use, and sexual risk behaviors are often associated with the leading causes of death for adolescents between 10 and 24 years old in the United States (U.S.) [1]. According to national reports, Hispanic adolescents in grades 9 through 12 reported higher than average levels of drug use, alcohol use, sexual risk behaviors, and higher levels of cigarette use when compared to their non-Hispanic Black adolescent counterparts [1]. Additionally, substance use behaviors have been shown to be associated with sexual behaviors, suggesting a comorbid association [2]. Substance use and sexual risk behaviors that develop during adolescence often predict substance use, drug dependence, and sexual risk behaviors in adulthood, along with other adverse health outcomes, such as major depressive disorders, violence, and STIs [3–6]. Although substance use and sexual risk behaviors are a serious public health concern among Hispanic adolescents, research suggests that positive family functioning may help prevent or lower levels of these health risk behaviors [7, 8].

The family systems theory posits that the better a family unit functions, the better the outcomes for family members [9, 10]. Thus, family functioning can be conceptualized as the positive or negative functioning of the family system or family environment surrounding an adolescent, and whether the family can function to adequately manage conflict and address adolescents’ needs for positive development [11]. Family functioning includes behaviors such as parental involvement, positive parenting, family communication, family cohesion, family support, parent-adolescent communication, and parental monitoring [12–14]. Hereafter, the term family functioning refers to any of the behaviors mentioned above or a combination of these behaviors. Positive family functioning is often inversely related with adolescent drug, alcohol, cigarette use, and sexual risk-related behaviors [7, 8]. A means of increasing positive family functioning behaviors is through the participation of adolescents and parents in family-based interventions.

Previous research shows that family-based interventions effectively improve family functioning behaviors and reduce adverse adolescent outcomes [15], particularly for adolescents at higher risk for adverse health outcomes [16]. However, the positive effects of family-based interventions may be limited if adolescents and their families, particularly those considered high risk, do not have the opportunity to participate in these interventions. The use of convenience sampling in existing studies examining the effects of family-based interventions may exclude Hispanic families who are not actively seeking out preventative interventions [17].

One way of increasing the reach of evidence-based family interventions is by recruiting adolescents and their families through a screening and referral process in primary care settings. In 2017, 74% of Hispanics/Latinos under 18 years of age reported visiting a doctor or other healthcare professionals in the last 6 months [18]. Primary care is an ideal setting for identifying at-risk adolescents. Parents often consult primary care providers regarding concerns with adolescent behaviors, and primary care providers are perceived to be a trusted source of information [19]. Primary care visits present an opportunity to identify and refer at-risk adolescents to evidence-based family interventions.

Several national health organizations recommend that primary care physicians screen for substance use starting in the pediatric age range [20]. For example, the American Academy of Pediatrics recommends that primary care providers screen adolescents using the CRAFFT (Car, Relax, Alone, Friends/Family, Forget, Trouble) to identify if adolescents are at risk for alcohol use and drug use [21]. Other examples of existing screening tools include the AUDIT (Alcohol Use Disorders Identification Test) used both in primary care settings and epidemiological studies to assess alcohol consumption, drinking behaviors, and alcohol-related problems [22]; the POSIT (Problem Oriented Screening Instrument for Teenagers) is a 17-item screening test for substance use [23]; and the National Institute on Alcohol Abuse and Alcoholism Youth Alcohol Screening Tool used to screen for adolescents own and their friends’ past-year alcohol use [22]. Additionally, the SHS (sexual health survey) is used to identify the risk for sexually transmitted infections [24]. These tools share similar strengths, such that they are relatively brief and have acceptable psychometric properties; however, time constraints in busy primary care settings may be a barrier to using multiple screening tools for multiple risk behavior.

The previously developed tools may be limited due to their specificity as they are generally focused on assessing adolescents for one or two risk behaviors. For example, using one screening tool may result in identifying adolescents at risk for alcohol but may not identify adolescents at risk for other behaviors, such as sexual risk behaviors. Considering adolescents often engage in more than one problem behavior [25], the use of multiple screening tools to identify at-risk adolescents who may benefit from evidence-based interventions can be cumbersome for clinicians who cite time constraints as one of the barriers to screening adolescents [20]. While individually each screening tool has the benefit of brevity and time efficiency, using multiple screening tools for individual outcomes may offset these benefits.

A way to circumvent using more than one screening tool may be to screen adolescents for a factor associated with multiple risk outcomes. Family functioning behaviors are not only shown to be associated with lower levels of risk outcomes, but they are also shown to act as an underlying mechanism in the association between intervention effects (for family-based interventions) and subsequent risk outcomes [15]. A screening tool based on family functioning may be advantageous over those that are disease/health-outcome specific because adolescents may not be forthright when asked about their substance use and sexual risk behaviors by a provider due to concerns with confidentiality [26]. A family functioning screening tool may also be beneficial in screening and recruiting adolescents into family-based interventions aimed at improving family functioning and health outcomes, reducing time constraints faced by providers [27], and perhaps limit prioritization of one risk outcome over another [28]. Although family functioning screeners exist, such as the Family Assessment Measure (FAM) [29] and the Brief Assessment of Family Functioning Scale (BAFFS) [30], these family functioning measures have not been assessed for their associations with adolescent risk behaviors. Further, existing family functioning screeners also have limitations related to their length and use with early adolescent populations [31].

The current study harmonized baseline data across three trials of a family-based intervention in the prevention of substance use and sexual risk behaviors in Hispanic adolescents to (1) investigate whether items from five separate family functioning scales are related to an underlying latent family functioning construct using a bifactor model, (2) determine construct validity of a multi-dimensional latent family functioning construct, and (3) examine the associations between the latent family functioning construct and adolescent reported lifetime and past 90-day substance use and sexual risk behaviors. The hypotheses are that a latent family functioning construct will emerge that represents multiple dimensions of family functioning (Hypothesis 1) and this latent family functioning construct will be negatively associated with lifetime and past 90-day substance use and sexual risk behaviors at baseline assessment (Hypothesis 2).

Method

Participants

This study used adolescent-reported baseline data harmonized across three longitudinal family-based intervention trials (n = 1451) evaluating the effects of a family-based intervention in the prevention of substance use and sexual risk behaviors in Hispanic adolescents. In each trial, adolescents and their families were randomly assigned to either the experimental or a control condition. All intervention trials used a universal recruitment strategy. Additionally, overlapping inclusion criteria for the harmonized trials included (1) early adolescent of Hispanic descent, (2) living with a primary caregiver willing to participate in intervention, and (3) living within the South Florida catchment area. Details for each intervention trial can be found elsewhere [15, 32, 33].

Raw data from the three trials was combined to increase power by combining relatively smaller datasets into a larger sample, such that behaviors that are less prevalent in early adolescent populations (e.g., substance use, sexual behaviors) have a larger prevalence in a larger sample and accommodating for statistical analyses. Data harmonization was used to increase the power to detect significant associations due to relatively lower prevalence of substance use behaviors reported among adolescent participants included in the three intervention trials. Additionally, commensurate measures were developed across combined datasets [34], which was facilitated by all study variables using identical questions and response items.

For this study, participants were included if they had data on study outcome variables. Adolescents ranged from 12 to 16 years old (M = 13.6, SD = 1.0) at baseline, 51.4% were male, and 51% reported being U.S. born. Participants indicating being born outside of the U.S. were born in one of 20 countries but the majority were from Cuba (14.5%), Venezuela (6.8%), and Colombia (5.0%). Less than half (34.3%) of adolescents’ parents reported a family income of more than $25,000. All studies were approved by the Institutional Review Board of the University of Miami. Informed consent was provided by parents/primary caregivers and assent was obtained from adolescent participants.

Measures

Lifetime and Past 90-Day Substance Use and Sexual Risk Behaviors

To assess lifetime and past 90-day drug (i.e., marijuana, cocaine, inhalants, and other drugs), alcohol, and cigarette use, participants responded to items from the Monitoring the Future Survey [35], including responding to whether they had used a particular substance during the 90 days before the assessment. To assess lifetime condom use and past 90-day condomless sex, participants responded to items from the Sexual Behavior Instrument [36], including whether they had oral, vaginal, or anal sex in the last 90 days and if they had used a condom during their last sexual encounter. All items were dichotomized, (0, no; 1, yes) with the exception of past 90-day condomless sex, which was on a Likert scale (0, never; 5, always).

Family Functioning Scales

The parental involvement (11 items; α = 0.84) and the positive parenting (6 items; α = 0.89) subscales from the Parenting Practice Scale were used [14] with adolescents responding to items such as “How often do you help your parent?” for parental involvement and “When you have done something that your parents like or approve of, how often does your parent give you a wink or a smile?” Family communication was assessed using the 3-item subscale from the Family Relations Scale with adolescents responding to items such as “My family knows what I mean when I say something” (α = 0.72) [13]. Parent-adolescent communication was assessed using 20 items from the Parent-Adolescent Communication Scale (α = 0.89) [12] with adolescents responding to items such as “My parent tries to understand my point of view.” Finally, parental monitoring of peers [37] was assessed using 5 items (α = 0.80) with adolescent responding to items such as “How often during a typical week do your parents supervise you and your best friends spending time together?” Cronbach’s alphas provided above represent the current study sample.

Data Analysis

First, considering the multi-dimensional nature of family functioning, a bifactor model [38] was used to identify a latent family function construct. The bifactor model is a multi-dimensional item response theory model. Items are first assigned to their respective subscales (i.e., subdomains), while also loading on a primary domain, i.e., the latent family functioning construct (see Fig. 1). Concurrently, correlational associations within subscales are being considered. Bifactor models are beneficial because they preserve the associations within family functioning subscales and the latent family functioning construct is identified independent from the correlations within subscales. Therefore, the bifactor model included the primary domain and the five family functioning subdomains (i.e., parental involvement, positive parenting, family communication, parent-adolescent communication, and parental monitoring of peers). Second, construct validity was assessed for the latent family functioning construct. Third, correlational associations and linear regression analysis were used to examine the association between the latent family functioning construct and substance use and sexual risk outcomes. This study used standardized coefficients to determine effect sizes as either small (r = 0.10), medium (r = 0.30), or large effects (r = 0.50) [39].

Fig. 1.

Bifactor model of latent family functioning construct and five family functioning dimensions

Results

Descriptive data of substance use and sexual risk behaviors can be found in Table 1. Results from the bifactors analysis, including the items and factor loadings on the primary family functioning dimensions are shown in Table 2. Of the total items from the five family functioning subscales, 26 items were retained with factor loadings ≥ 0.50 to compose the latent family functioning construct. The cut-off for factor loadings on the primary domain (i.e., ≥ 0.50) was chosen based on existing literature [40] and after consultation with one of the developers of the bifactor model [41]. Additionally, the association between the five family functioning scales and the latent family functioning construct (26 items) was examined, which showed that the linear association and bivariate association were strong (R² linear = 0.923; Pearson r = 0.961, p < 0.001). To determine whether a more parsimonious model indicated similar results, items with factor loadings ≥ 0.60, resulting in 13 items, were selected. The associations between the five family functioning scales and the 13-item latent family functioning construct were examined. Results indicated that the linear and bivariate associations between the five family functioning scales items and the 13-item latent family functioning construct slightly deteriorated (R² linear = 0.793; Pearson r = 0.890, p < 0.001); therefore, the 26-item latent family functioning construct was retained. The Cronbach’s α for the 26 items in the latent family functioning construct was 0.94.

Table 1.

Adolescent demographics and substance use and sexual risk behaviors (n = 1451)

Variable	% (n)
Age
12	8.6 (124)
13	47.7 (691)
14	25.0 (362)
15	10.9 (158)
16	5.7 (83)
Gender
Female	47.1 (683)
Male	51.4 (745)
Nativity
Born in the U.S	51.0 (739)
Born outside the U.S	46.9 (680)
Lifetime cigarette use
No	89.7 (1302)
Yes	5.9 (86)
Past 90-day cigarette use
No	94.8 (1376)
Yes	1.7 (25)
Lifetime alcohol use
No	83.6 (1213)
Yes	13.1 (190)
Past 90-day alcohol use
No	91.0 (1321)
Yes	4.3 (63)
Lifetime drug use
No	90.4 (1311)
Yes	6.5 (95)
Past 90-day drug use
No	93.1 (1351)
Yes	3.0 (44)
Lifetime condom use
No	4.6 (67)
Yes	4.0 (58)
Past 90-day condomless sex use
Never	2.3 (34)
Less than half the time	0.8 (11)
About half the time	0.5 (7)
Not always, but more than half the time	0.5 (7)
Always	1.0 (14)

Missing cases account for values not equaling total sample

Table 2.

Factor loadings on the primary dimension from bifactor analysis

Scale and item number	Factor loading on primary dimension	Item descriptor
PI 1	0.216	When was the last time that you talked with your parent about what you were going to do for the coming day?
PI 2	0.444	How often did your parent talk to you about what you were going to do for the coming day?
PI 3	0.415	When was the last time that you talked with your parent about what you had actually done during the day?
PI 4	0.503	How often did your parent talk with you about what you had actually done during the day?
PI 5	0.531	How often did your parent talk with you about how things are going in school?
PI 6	0.672	How often does your parent find time to listen to you when you want to talk to her?
PI 7	0.609	How often do you and your parent do things together at home?
PI 8	0.441	How often do you help your parent?
PI 9	0.496	How often do you help to plan a family activity?
PI 10	0.545	How often do you like to do things together with the family?
PI 11	0.473	How often do you go out with members of the family to movies, sports events, or other outings?
PP 1	0.539	Does your parent wink or smile?
PP 2	0.593	Does your parent say something nice about ...?
PP 3	0.602	Does your parent give you a hug...?
PP 4	0.517	Does your parent give you a reward...?
PP 5	0.413	Does your parent give you special privilege ...?
PP 6	0.579	Does your parent do something special with ...?
PAC 1	0.714	I can discuss my beliefs with my parent without feeling restrained or embarrassed
PAC 2	0.374	Sometimes I have trouble believing everything my parent tells me
PAC 3	0.806	My parent is always a good listener
PAC 4	0.332	I am sometimes afraid to ask my parent for what I want
PAC 5	0.400	My parent has a tendency to say things to me which would be better left unsaid
PAC 6	0.658	My parent can tell how I’m feeling without asking
PAC 7	0.824	I am very satisfied with how my parent and I talk together
PAC 8	0.826	If I were in trouble, I could tell my parent
PAC 9	0.656	I openly show affection to my parent
PAC 10	0.314	When we are having a problem, I often give my parent the silent treatment
PAC 12	0.223	When talking to my parent, I have a tendency to say things that would be better left unsaid
PAC 13	0.772	When I ask questions, I get honest answers from my parent
PAC 14	0.796	My parent tries to understand my point of view
PAC 15	0.405	There are topics I avoid discussing with my parent
PAC 16	0.799	I find it easy to discuss problems with my parent
PAC 17	0.771	It is very easy for me to express all my true feelings to my parent
PAC 18	0.593	My parent nags/bothers me
PAC 19	0.591	My parent insults me when she is angry with me
PAC 20	0.585	I don’t think I can tell my parent how I really feel about some things
FC 1	0.570	My family and I have the same views about what is right and wrong
FC 2	0.573	My family knows what I mean when I say something
FC 3	0.535	My family and I have the same views about being successful
PM 1	0.352	How well do your parents know your best friends?
PM 2	0.362	How often during a typical week do your parents supervise you and your best friends spending time together?
PM 3	0.424	How often during a typical week do your parents spend time talking with you about your friends?
PM 4	0.298	How well do your parents know the PARENTS of your best friends?
PM 5	0.306	How often during a typical week do your parents see or talk to the PARENTS of your best friends?

Bolded values indicate retained items for the latent construct

PI Parental Involvement Subscale, PP Positive Parenting Subscale, PAC Parent-Adolescent Communication Subscale, FC Family Communication Subscale, PM Peer Monitoring Subscale

Finally, correlations between the 26-item latent family functioning construct and substance use and sexual risk behaviors indicated negative correlational associations, suggesting divergent validity (Table 3). Further, when substance use and sexual risk behaviors were regressed on the latent family functioning construct, findings indicated a negative association with baseline levels of lifetime cigarette use (β = − 0.14, p < 0.001), past 90-day cigarette use (β = − 0.09, p < 0.01), lifetime alcohol use (β = − 0.19, p < 0.001), past 90-day alcohol use (β = − 0.11, p < 0.001), lifetime drug use (β = − 0.20, p < 0.001), and past 90-day drug use (β = − 0.15, p < 0.001). The sample size for condom use behaviors was insufficient to calculate a standardized estimate. The absolute value of the standardized estimates for all substance use behaviors suggests that the latent family functioning construct had an approximate small effect size on most of the outcomes [39].

Table 3.

Pearson bivariate correlations (n = 1451)

	1	2	3	4	5	6	7	8	9
1. FFC	-	−0.14*	−0.10*	−0.19*	−0.11*	−0.20*	−0.15*	−0.13	0.02
2. Lifetime cigarette use		-	0.41*	0.27*	0.26*	0.29*	0.24*	0.12	−0.02
3. Past 90-day cigarette use			-	0.24*	0.30*	0.32*	0.36*	0.14	0.07
4. Lifetime alcohol use				-	0.58*	0.34*	0.29*	0.05	0.08
5. Past 90-day alcohol use					-	0.33*	0.39*	0.17	0.06
6. Lifetime drug use						-	0.71*	0.11	0.10
7. Past 90-day drug use							-	0.21*	0.08
8. Lifetime condom use								-	0.57*
9. Past 90-day condomless sex									-

FFC, sum of items included in the family functioning construct

^*p < 0.05

Discussion

Substance use and sexual risk behaviors are a significant problem among Hispanic adolescents in the U.S. [1, 35]. Screening tools have been developed to help researchers and health practitioners screen and identify adolescents who may be at-risk based on their previous or current involvement in risky behaviors [20]. This current study, however, is the first to our knowledge to examine if a latent family functioning construct may identify Hispanic adolescents at risk for drug use, alcohol use, cigarette use, and sexual risk behaviors based on reported family functioning behaviors, an underlying mechanism associated with each risk outcome.

This study showed that the 26 items from the latent family functioning construct may serve as an effective assessment due to its relative brevity, psychometric properties, and correlation with adverse outcomes. The bifactor model indicated that the 26-item family functioning construct was negatively associated with adolescents’ baseline levels of lifetime and past 90-day drug use, alcohol use, and cigarette use. Findings also indicated that the latent family functioning construct showed acceptable reliability. Therefore, the latent family functioning construct may best serve primary care practitioners who express time constraints as a barrier to screening [20]. Further, primary care practitioners who feel uncomfortable when discussing sensitive health outcomes with adolescents [42] may benefit from using the latent family functioning construct due to the focus on family functioning, while taking into consideration the small effect sizes detected in this study.

Of the 26 items in the family functioning construct, most (61.5%) were from the communication-related items, emphasizing the importance of positive communication in reducing adolescents’ substance use, and has been shown to both explain and offset adolescents’ risk of negative outcomes [8, 43, 44]. Additionally, the remaining items in the family functioning construct came from the parental involvement and positive parenting scales [14]. Individually, both scales have been shown to be negatively associated with adolescent substance use [45]. The latent family functioning construct taps into the important role of family functioning and its negative association with adolescent substance use.

Study participants for each trial included primarily adolescents of Hispanic descent, and due to cultural factors, the family functioning behaviors may be more influential to their positive development, while a lack thereof may be a clear indicator of Hispanic adolescents’ risk of substance use. Hispanic cultural values, such as familismo, emphasize the idea of family cohesion, respect, and interdependence, and have been identified as protective cultural factors [46, 47]. Therefore, the behaviors captured with the latent family functioning construct may support Hispanic adolescents’ existing cultural values and, in turn, these behaviors may be influential in predicting risk behaviors. Additionally, the latent family functioning construct may also be used with African American and Asian adolescents considering that both race/ethnic cultures similarly emphasize the important role of family in positive adolescent development [48, 49]. Additionally, research suggests that Hispanic adolescents may feel uncomfortable if a clinician discusses sensitive topics, such as substance use and sexual behaviors, because they fear the clinician may be stereotyping them due to their race/ethnicity [50]. The screening adolescents on their reported family functioning may be beneficial for primary care practitioners aiming to be culturally sensitive to the Hispanic families they service. Nevertheless, future research should examine the latent family functioning construct among a heterogeneous adolescent sample to determine its generalizability to other race/ethnic groups.

Although baseline reports of lifetime and past 90-day substance use and sexual risk behaviors were lower than national averages [1], we primarily believe this is due to the average age of adolescents in our sample being approximately 13 years. It is important to note that as adolescents move from early adolescence into mid and late adolescence, their risk-taking behaviors increase. Primary care practitioners may be hesitant to screen for substance use and sexual risk due to the perception of low prevalence of these behaviors among early adolescents; however, family functioning behaviors are also negatively associated with other risk outcomes, such as internalizing behaviors and suicidal ideation among early adolescents [43, 51, 52]. Perhaps, the latent family functioning construct may be beneficial in assessing adolescents for mental health outcomes, a critical public health concern among adolescents between the ages of 12 and 17 [53].

Limitations

First, the cross-sectional design of this study limits any possible causal inferences. However, the focus of this study is on examining the feasibility of assessing adolescents on their current reports of family functioning and the association with any previous reports of substance use and sexual risk. Future research, however, should examine whether screening for family functioning during adolescence may predict risk behaviors beyond adolescence. Second, this study used adolescent self-reported data which may lead to self-report bias, threatening the validity of measures used due to social desirability bias and/or recall bias [54]. Additionally, adolescents who self-report sexual behaviors may be reporting inconsistently and contradictory. Observed data and/or parent-reported data may be needed to corroborate results in future studies. Additionally, future studies should explore the association of the identified latent family functioning construct with a sample of adolescents who exclusively reported engaging in risk substance use behaviors and sexual risk behaviors. Third, study participants originated from one geographical location, limiting the generalizability of the results. Although study participants were primarily of Hispanic descent, several Hispanic subgroups were represented by these data, including individuals born in the U.S. and foreign born. Fourth, most of the family functioning items refer to the adolescent’s primary caregiver. The wording of the items may limit the capability of the construct in assessing family functioning and its association with risk behaviors for mother-adolescent or father-adolescent dyads. Adolescent reports of family functioning may differ when referring to their mother versus their father, which in turn may influence the associations with risk outcomes [55]. Further, certain aspects of family functioning, such as communication, may differ across parent-adolescent depending on the adolescent’s gender. For example, due to cultural gender norms (i.e., marianismo and machismo), mothers may communicate greater consequences related to risk behaviors with girls and, alternatively, be more permissive with boys [56].

Conclusion

The recommendation for preventative pediatric healthcare by the American Academy of Pediatrics is that pediatricians begin screening adolescents for substance use and HIV at age 11 [57]. This study provides preliminary evidence of the potential benefit of screening on family functioning as a means for healthcare professionals to achieve assessing adolescents on the multi-dimensionality of family functioning and multiple risk outcomes. Future research is needed to examine whether the family functioning construct identified in this study can be used to identify adolescents at risk for adverse mental health outcomes, also recommended for preventative pediatric healthcare [57]. The evidence provided here is preliminary and more research is needed in determining scale thresholds to increase sensitivity and comparability with other screening tools, such as the CRAFFT. Although adolescents are generally healthier than their adult counterparts regarding chronic diseases, this developmental stage is critical for primary care providers to identify and intervene with adolescents who indicate risk of negative health outcomes.

Data Availability

Data can be made available upon reasonable request from the Principal Investigator of the study.