Psychometric Properties of the Family Assessment Task Parental Monitoring Scenario among Adolescents Receiving Substance Use Treatment

Abstract

Introduction:

The Family Assessment Task (FAsTask) is an observer-rated parent-child interaction task used in adolescent substance use intervention. The parental monitoring component of the FAsTask is thought to provide an objective assessment of parental monitoring that can guide treatment planning and circumvent the potential limitations of self-report measures. Yet, the factor structure, measurement invariance, and concurrent validity of the parental monitoring FAsTask has not been evaluated; doing so is essential to effectively guide clinical care. This study examined if the parental monitoring FAsTask can be reliably administered across adolescent age and sex, and to identify which components of the parental monitoring FAsTask are most consistently associated with adolescent substance use.

Methods:

The study pooled data from 388 adolescent-caregiver dyads across six separate clinical trials (adolescents [M_age= 15.7, 57.5% male, 61.9% White, 31.2% Latine]; caregivers [M_age=42.14, 88.7% female, 72.7% White, 24.2% Latine]). Dyads completed the FAsTask and the Timeline Followback at the baseline session, prior to randomization. Analyses proceeded in three steps. First, the study team conducted an exploratory factor analysis (EFA) in half of the sample, followed by a confirmatory factor analysis (CFA) in the second half of the sample. Second, the study tested measurement invariance as a function of adolescent age and biological sex. Third, the study used a series of structural equation models to assess the associations of each factor with alcohol use, binge drinking, and cannabis use.

Results:

EFA and CFA indicated the presence of four factors (labeled Supervised/Structured, Active Monitoring, Task Engagement, and Parental Rules/Strategies). The analysis found evidence of measurement invariance across adolescent age and sex. The Supervision/Structure was negatively associated with adolescent alcohol use, binge drinking, and cannabis use.

Conclusions:

The parental monitoring FAsTask demonstrates validity and retains its structure across adolescent age and sex. Items focused on parental supervision and structure are most strongly associated with adolescent substance use and may best inform clinical care decisions in adolescent substance use treatment.

1. Introduction

Parental monitoring, defined as actions parents take to supervise and maintain awareness of adolescents’ school functioning, interpersonal relationships, and whereabouts, is integral to preventing and reducing adolescent risk behaviors (Dishion & McMahon, 1998). Greater parental monitoring is associated with reductions in youth problematic behavior over time, indicating its utility as a prevention method (Fosco et al., 2012). Conversely, low parental monitoring is consistently and strongly associated with behaviors that put youth at risk of harm, even into emerging adulthood (Bendezú et al., 2018; DeVille et al., 2020; Fosco et al., 2012; Stone et al., 2012).

Parental monitoring has an especially strong association with adolescent substance use. A systematic review and meta-analysis of longitudinal studies of parental monitoring and adolescent alcohol use found that parental monitoring had the strongest protective effect on the probability and levels of adolescent alcohol use relative to other general parenting practices (Yap et al., 2017). In a longitudinal study of over 800 youth, parental monitoring at age 11 was associated with a reduced likelihood of initiating cannabis use through age 17 (Bohnert et al., 2012). A secondary analysis of a national sample of over 17,000 youth found that youth-reported parenting behaviors consistent with monitoring (e.g., checking if homework is done, limiting time spent outside on school nights) were protective against cannabis use, such that youth indicating that their parents seldom/never engaged in those behaviors were at higher risk of cannabis use in the past month and year compared to youth whose parents engaged in monitoring more frequently (King et al., 2015).

There is evidence to suggest that demographic factors moderate effects of parental monitoring on substance use outcomes. A longitudinal study of 400 8^th graders found that lower parental monitoring was associated with onset of cannabis and alcohol use and binge-drinking by 9th grade (Rusby et al., 2018). In this sample, adolescent-reported parental monitoring was more protective for females than males for both cannabis and alcohol onset. Relatedly, while these parenting behaviors were protective against cannabis use across all ages, the protective effect was highest among 12- to 13-year-olds (King et al., 2015).

Optimal measurement of parental monitoring has been of interest for decades. Observational procedures offer some advantages for assessing parental monitoring. For example, parent-adolescent interaction tasks in which dyads discuss household norms around parental monitoring are less likely to be subject to recall bias or social desirability and therefore may provide a more reliable, objective assessment (Girard & Cohn, 2016). Observational techniques also provide a snapshot of behaviors of interest (e.g., listening, instructions that facilitate monitoring) and how these behaviors are altered by other’s reactions. Importantly, observational tasks also provide a level of accessibility to measuring a construct when low levels of reading ability or comprehension negatively impact accuracy in responding to surveys (Dishion & Kavanagh, 2003). Despite their utility, studies utilizing observational methodologies to assess parenting, including monitoring, are very limited. Ratings of videotaped interaction tasks are key components of empirically supported interventions for children (e.g., Parent-Child Interaction Therapy [PCIT]), but have not been routinely incorporated into interventions for adolescents where real-time data could be extremely helpful in guiding interventions. Indeed, Dishion and Kavanaugh (2003) note that direct observation of families is likely more appropriate for families from a “wide range of socioeconomic and cultural backgrounds … than wordy questionnaires and lengthy interviews designed by researchers with advanced degrees” (p. 55-56).

One observational parent-adolescent interaction task that evaluates parental monitoring is the Family Assessment Task (FAsTask; (Dishion et al., 2003). Whereas most self-report tools define ‘parental monitoring’ as consisting of both ‘parental knowledge’ (i.e., what parents know about child behavior) and ‘sources of knowledge’ (i.e., how parents learn this information; Racz & McMahon, 2011; Kerr & Stattin, 2000; Stattin & Kerr, 2000), the FAsTask was designed to measure the youth’s avoidance of adult supervision and the adult’s involvement in the youth’s activities. The FAsTask is a component of the assessment battery used in an evidence-based treatment for substance use called the Family Check-Up (FCU; Dishion et al., 2003), a strengths-based, tailored familial intervention focused on improving parenting skills and family management strategies.

During the FAsTAsk parental monitoring task, youth are instructed to describe to their parent(s) a time that they spent time with peers for at least an hour without adults around. Parents are instructed to listen, and then comment or request additional information. Coders rate parental behaviors gleaned from the conversation, such as indications of avoiding adult supervision (child) and involvement in youth’s activities (parent). The original version of the FAsTask contained 6 items per informant and two subscales: “Child-specific Monitoring,” and “Parent-specific Monitoring.” This information was later combined into one parental monitoring composite score. The FCU session itself provides feedback tailored to the family regarding strengths and challenges, based on information derived from the FAsTAsk and other assessments.

A study of the 6-item version of the FCU (Dishion et al., 2003) in families with a high-risk adolescent indicated that observable changes in parental monitoring across the study (6^th to 9^th grade) mediated the link between the FCU intervention condition and adolescent reductions in substance use by 9^th grade. Since that study, other investigations have implemented a longer scoring of the FAsTask Monitoring and Listening task (hereafter called the parental monitoring FAsTask), expanded to 23 items in total. A clinical code book developed by Dishion and Kavanaugh (2003) is used to help guide the scoring of each item. “Macro” clinical scores, an average of the items within a domain (e.g., parental monitoring), are used as part of the feedback sessions to families. Macros scores are used as feedback in the FCU session and coded into one of the following categories to facilitate feedback: 1-3 is poor, 4-5 is average, and 6+ is strong, representing “challenge”, “needs improvement”, or “strength”, (Dishion & Kavanaugh, 2003). In a clinical trial testing integrated cognitive-behavioral therapy vs. treatment as usual in adolescents with co-occurring substance use and psychiatric disorders, greater FAsTask-rated parental monitoring was associated with reductions in depressive symptoms and suicidal ideation over 12 months regardless of treatment condition (MacPherson et al., 2021), whereas parent-reported parental monitoring predicted neither. Additional analyses revealed that low FAsTask-rated parental monitoring at baseline, but not self-report, predicted a greater percentage of cannabis use and heavy alcohol use days over the 6-month follow-up period (Spirito et al., 2021).

Despite over two decades of research showing that the FCU improves child outcomes (e.g., Connell et al., 2016), the long version of the parental monitoring FAsTask has not undergone rigorous psychometric assessment. The predictive validity of the FAsTask parental monitoring task in prior studies, paired with its recommended use to inform subsequent FCU intervention delivery, suggests that it is a clinically useful tool. However, to our knowledge, there has been no evaluation of the factor structure, reliability, and validity of the parental monitoring FAsTask. Factor analyses may suggest fewer items are needed to capture the monitoring construct, thereby reducing the burden of scoring multiple items, and making it easier for clinicians to derive feedback for families regarding monitoring. Furthermore, despite evidence to suggest that parental monitoring varies across sociodemographic factors (e.g., age, sex; King et al., 2015; Rusby et al., 2018), the task’s concurrent validity and measurement invariance have not been evaluated. Establishing measurement invariance, accomplished using structural equation modeling, is critical to determine if the FAsTask parental monitoring task can be reliably administered across diverse populations. Lastly, parent reports of parenting behaviors are positively, but only weakly, associated with observations of parenting behaviors (Hendriks, Van der Giessen, Stams, & Overbeek, 2018). This finding, in conjuction with the FasTask-rated parental monitoring associations with adolescent substance misuse, support how integral the use of observational methods is in assessment of adolescent risk factors.

The goal of this study was to establish the psychometric properties of the FasTask parental monitoring scale to provide confirmation that this task and its scoring perform as expected. An additional potential benefit of these analysis is the possibility of isolating a particular factor/subset of items that carry the most clinical utility in predicting substance use, as clinicians may opt to evaluate only these items. First, we examined the task’s factor structure via exploratory and confirmatory factor analysis. Exploratory factor analysis was utilized to uncover the underlying structure of the 23 items, while confirmatory factor analysis verified the factor structure. Second, we assessed if the task retained its factor structure across age and sex to establish measurement invariance. Finally, we examined the associations of the extracted factors with substance use (i.e., alcohol use, binge drinking, cannabis use) to examine concurrent validity. All analyses are considered exploratory to establish the psychometric properties of the task and guide its scoring, application, and interpretation in future studies.

2. Material and Methods

2.1. Participants

The study pooled data across six separate clinical research trials that enrolled adolescents who use substances and their parent and/or guardian and administered the FAsTask during the baseline visit (i.e., prior to intervention). All studies excluded participants if they: 1) had serious psychotic symptoms (e.g., hallucinations), a primary diagnosis of an eating disorder, or obsessive-compulsive disorder; or 2) were acutely suicidal or homicidal.

2.1 1. Study 1 (Spirito et al., 2011; NCT00247221)

Adolescents (n=128; 13-17 years old) who were treated at an urban level I trauma center in the Northeast United States were eligible for study if they had a positive blood alcohol concentration or self-reported alcohol use in the 6 hours before the emergency department visit. Adolescents and a legal guardian assigned to the FCU condition (47 dyads) completed the FAsTask.

2.1.2. Study 2 (Spirito et al., 2017; NCT00925340)

Study 2 included adolescents recruited from local high schools, family and truancy courts, advertisements, or referrals from emergency departments or mental health centers. Adolescents enrolled because their parents were concerned about their alcohol or cannabis use. Eligible adolescents were: 1) 12 to 19 years old; 2) lived at home with a parent/legal guardian who was willing to participate; 3) used alcohol or cannabis at least once in the past 90 days; and 4) had a sibling (between 11 and 21 years old) within 5 years of age of the target adolescent, who lived at home with the adolescent and participating parent(s). Dyads in the FCU condition (n=40) had FAsTask data.

2.1.3. Study 3 (Spirito et al., 2018; NCT00247221)

Study 3 included adolescents (n= 69; 13-18 years old) who lived at home with parent or legal guardian, used cannabis at least three times in the prior 90 days, and had a history of past-year school truancy were recruited from family and truancy courts, school counselor referral, or from presentations in high school health classes. Sixty-seven dyads had FAsTask data.

2.1.4. Study 4 (Kemp et al., 2023; NCT03107117)

Study 4 recruited court-involved, non-incarcerated adolescents who reported past year cannabis at intake to a Family Court in the Northeast United States, were 18 years of age or younger, and lived at home with a parent or guardian (n= 83). Seventy-three dyads had FAsTask data.

2.1.5. Study 5 (Becker et al., 2021; NCT03592186)

Study 5 recruited adolescents (n=61; 13-17 years old) and their parent/guardian from two residential treatment facilities: one in the Northeast and one in the Midwest United States. Adolescents were admitted to residential treatment due to substance use-related problems and reported past 90-day substance use. Parents were eligible if they were the adolescent’s legal guardian, would remain the custodial guardian after discharge, and had reliable internet access to receive a technology-assisted parenting intervention. Sixty dyads had FAsTask data.

2.1.6. Study 6 (Wolff et al., 2020; NCT01667159)

Study 6 recruited adolescents (n=112) and their families from a community mental health clinic in the Northeast United States. Adolescents were eligible for the study if they were enrolled in the intensive outpatient, home-based program for co-occurring substance use and mental health problems, were 12–18 years old, reported past 3-month alcohol/substance use, and spoke English. One hundred one dyads had FAsTask data.

2.2. Measures

2.2.1. Demographics

Adolescents and their parent/guardian self-reported biological sex, age, race, and ethnicity. Adolescents reported relation to their caregiver. Caregivers reported household income.

2.2.2. Family Assessment Task Parental Monitoring Scenario

The FAsTask consists of multiple structured caregiver-child interaction tasks, including limit setting, parental norms about substance use, and parental monitoring. This study examined the parental monitoring component due to its well-documented utility in predicting adolescent outcomes (e.g., Spirito et al., 2021). Dyads are prompted to discuss a recent time when the adolescent was unsupervised. Adolescents are instructed to, “talk about a time in the last month when you spent at least an hour with friends without adults around. Go into as much detail as you’d like, describing where you were, who you were with and what you were doing.” Caregivers are instructed to “…first listen to your adolescent and then comment or gather any other information that you might be interested in.” Dyads privately discuss the prompt for 5 minutes. The study audio and/or video recorded interactions for subsequent coding.

The study rated twenty-three items (see Table 1) on a 9-point scale (1=not at all to 9=very much). Item 1 is open-ended (unscored, i.e., “what/where/with whom is the situation?”) and excluded from these analyses. In studies 1, 2, and 3, two raters coded the FAsTask separately and then met to arrive at a final set of consensus codes. For the remaining studies, raters double coded approximately 20% of recordings to calculate inter-rater reliability; agreement levels ranged from acceptable to very good (i.e., 80% to 94%).

Table 1.

Descriptive Information for the Original FAsTask Parental Monitoring Items

FAsTask Item		M	SD	Skew	Kurtosis	N
2.	“Does it seem that the child spends time away from adult supervision?”	4.27	2.16	0.29	−0.77	345
3.	“Does the child indicate being with friends in settings without adult supervision?”	3.81	2.18	0.55	−0.61	383
4.	“Does there seem to be a lack of adult involvement in this child’s daily life?”	5.50	2.15	−0.27	−0.78	344
5.	“Is there a lack of structure or lax rules with respect to this child’s daily routine?”	5.17	2.27	−0.18	−0.97	385
6.	“Is there any mention of the child’s peers planning or engaging in deviant behaviors?”	6.03	3.11	−0.52	−1.40	386
7.	“Does the child volunteer important information about activities and companions?”	4.13	2.47	0.39	−0.97	382
8.	“Does the child do or say anything to indicate avoidance of adult supervision?”	7.15	2.36	−1.36	0.81	342
9.	“Does this parent seem to be monitoring with whom the child spends time?”	5.27	2.26	−0.31	−0.96	385
10.	“Does this parent seem to be monitoring where the child spends time?”	5.15	2.37	−0.17	−1.08	384
11.	“Does this parent seem to be monitoring what the child is doing when outside of adult supervision?”	4.83	2.38	−0.08	−1.10	385
12.	“Does the parent listen to the child?”	7.62	1.56	−1.78	3.90	381
13.	“Does the parent effectively gather important information about the child’s activities?”	5.57	2.35	−0.36	−0.97	382
14P.	“Does parent indicate any rules or guidelines that facilitated parents’ monitoring?”	2.80	2.32	1.17	0.13	386
14A.	“Does adolescent indicate any rules or guidelines that facilitated parents’ monitoring?”	2.29	1.97	1.71	1.99	375
15.	“Does the parent indicate involvement in the child’s activities, such as planning, discussing, participating or providing transportation?”	2.57	2.11	1.35	0.73	345
16.	“Does the parent seem to know about the child’s friendships, knowing the friends by names and their family situations?”	5.40	2.51	−0.34	−1.11	385
17.	“Does the parent control his/her own reactions to allow the child to finish talking?”	7.62	1.60	−1.79	3.23	381
18P.	“Does the parent follow the directions for the task?”	6.93	1.83	−1.34	1.80	346
18A.	“Does the adolescent follow the directions for the task?”	7.15	1.71	−1.36	2.19	343
19P.	“Does the parent participate in the discussion?”	7.34	1.78	−1.70	2.91	343
19A.	“Does the adolescent participate in the discussion?”	7.25	1.74	−1.44	2.20	340
20.	“Does the parent provide positive reasons for monitoring the child’s activities?”	2.67	2.25	1.34	0.69	386
21.	“What percentage of the time does the family talk about the child’s activities?”	7.47	1.78	−1.61	2.37	347

Note. P=parent item; A=adolescent item.

Item 1 is open-response (i.e., “what/where/with whom is the situation?) and is not included in these analyses.

2.2.3. Substance use

Adolescents reported their recent (past 30-day or 90-day, depending on the study) cannabis and alcohol use at baseline using the Timeline Followback (TLFB; Dennis et al., 2004; Sobell & Sobell, 1992). Three calculated variables measure the proportion of TLFB days in which the adolescent reported alcohol use, binge drinking (i.e., standard drinks ≥ 5 for males, ≥ 4 for females), and cannabis use, respectively. We calculated the proportion of days each substance was used because the substance use reporting periods differed across studies.

2.3. Data Analytic Plan

The study pre-registered the analytic plan for the current study at https://osf.io/wv2ck. Deviations from the pre-registered analytic plan can be found in the Supplemental Material. To address the three overarching aims, analyses proceeded in three steps. In Step 1, researchers divided the sample into two subsamples using the SOLOMON procedure (Lorenzo-Seva, 2022). Researchers conducted exploratory factor analysis (EFA) in Mplus Version 8.9 (Muthen & Muthen, 1998–2018) in the first subsample to assess the factor structure of the FAsTask. Observed eigenvalues greater than one and greater than corresponding randomly generated eigenvalues in parallel analysis determined the number of factors to extract (O’connor, 2000); Tabachnick & Fidell, 2019). The decision to remove individual items was guided by several criteria, including standardized factor loadings less than 0.32, significant cross-loadings, and theoretical models of parental monitoring (Tabachnick & Fidell, 2019). EFA models were initially estimated using an orthogonal rotation (varimax) then using oblique rotation (Geomin). Model rotation was determined by which model provided a closer approximation to simple structure. Next, we conducted a confirmatory factor analysis (CFA) based on the results of the EFA in the second subsample. We used McDonald’s omega to estimate internal consistency for each factor (ω < .70 is poor, .70 to .80 is acceptable, .80 to .95 is excellent, >.95 indicates redundant items (Boateng et al., 2018).

In Step 2, we conducted a series of multigroup CFA in the full sample to assess measurement invariance of the FAsTask across age (i.e., adolescents 15-years-old and younger compared to 16-years-old and older) and biological sex (i.e., female and male). The study evaluated Four distinct types of invariance: configural (i.e., equivalent factor structures), metric (i.e., equivalent factor loadings of FAsTask items), scalar (i.e., equivalent item intercepts), and residual (i.e., equivalent item residual variances). Model fit was assessed using Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error Approximation (RMSEA), and Standardized Root-Mean-Square Residual (SRMR). Ranges were used to determine acceptable model fit (for CFI and TLI, <0.90 is poor, 0.90 to 0.94 is acceptable, and >0.95 is excellent; for RMSEA, 0.08 is poor, 0.05 to 0.07 is acceptable, and <0.05 is excellent; and for SRMR, 0.09 is poor, 0.06 to 0.09 is acceptable, and <0.06 is excellent; Marsh et al., 2004). The study used guidelines proposed by Chen (2007) to assess for invariance, such that a change of ≤−.005 in CFI, a change of ≥.010 in RMSEA or a change of ≥.025 in SRMR for testing loading invariance would indicate noninvariance. For intercept and residual invariance testing, a change of ≤−.005 in CFI, or a change of ≥.010 in RMSEA or a change of ≥.005 in SRMR would indicate noninvariance.

Lastly, Step 3 estimated cross-sectional hybrid structural equation models consisting of the CFA measurement model for the FAsTask and structural paths assessing associations with the three substance use variables (i.e., proportion of alcohol use, binge drinking, and cannabis use days). Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values were compared for negative binomial and zero-inflated negative binomial models to select the optimal count model for each outcome (poisson regression was not considered because means and variances were unequal; Brandie Wagner et al., 2015). All models controlled for age, biological sex, ethnicity, race, and focal substance (defined as the primary substance of use outlined in each study’s eligibility criteria, i.e., alcohol vs. cannabis or unspecified). The study used robust maximum likelihood estimation to estimate count models.

3. Results

Table 2 presents descriptive statistics by study and pooled across studies. Adolescents were M_age=15.71 (SD = 1.17), 57.5% male, 61.9% identified as White, and 68% identified as non-Latine. Proportion of days substances used were 6.5% (SD = 11.72), 3.98% (SD = 9.26) and 33.61% (SD = 34.73) for alcohol, binge drinking, and cannabis, respectively. Observed ranges of proportions were 0-1 for all substance use variables. Caregivers were M_age=42.14 (SD = 7.29), 88.7% female, and 24.2% Latine. Parents identified as 72.7% White, 12.6% Black, 2.6% Asian, 0.8% American Indian/Alaska Native, 5.2% multiracial, 1.3% another race, and 4.9% did not disclose their race. Median household income ranged from $26,000 to $60,000, depending on the study. Over 90% of the adolescents reported that they were participating with a biological parent, with a small proportion reporting that they participated with an adoptive parent or other legal guardian.

Table 2.

Demographic characteristics of each study and the overall sample.

	Study 1	Study 2	Study 3	Study 4	Study 5	Study 6	Overall
N	47	40	67	73	60	101	388
Age (M/SD) Biological	15.38 (1.19)	15.63 (1.15)	15.81 (1.34)	15.93 (1.11)	15.68 (1.02)	15.67 (1.18)	15.71 (1.17)
Sex (N)
Male	19	25	41	53	26	59	223 (57.47%)
Female	28	15	26	20	30	41	160 (41.24%)
Missing	-	-	-	-	4	1	5 (1.29%)
Race (N)
American Indian/Alaskan	-	-	-	1	-	-	1 (0.26%)
Native Asian	2	-	-	1	1	-	4 (1.03%)
Black	4	1	15	19	7	11	57 (14.69%)
Native Hawaiian/Pacific Islander	-	-	-	2	-	-	2 (0.52%)
White	33	29	39	34	34	71	240 (61.86%)
Other	-	-	-	6	9	3	18 (4.64%)
Multiracial	3	10	13	6	9	12	53 (13.66%)
Missing	5	-	-	4	-	4	13 (3.35%)
Ethnicity (N)
Non-Latine	34	22	46	48	45	68	263 (67.78%)
Latine	12	18	21	25	15	20	121 (31.19%)
Missing	1	-	-	-	-	2	4 (1.03%)
Study Substance of Focus	Alcohol	Polysubstance Use	Cannabis	Cannabis	Polysubstance Use	Polysubstance Use	-
Proportion Use Days/Past 30 or 90 days (M/SD)
Alcohol Use	10.47 (15.13)	5.00 (5.94)	7.50 (8.95)	0.97 (2.04)	8.83 (15.41)	7.19 (18.99)	6.50 (11.72)
Binge Drinking	5.37 (11.64)	2.86 (5.06)	4.08 (6.22)	0.38 (1.05)	5.10 (10.26)	5.65 (18.93)	3.98 (9.26)
Cannabis Use	14.04 (23.22)	11.22 (21.10)	46.67 (35.96)	34.81 (32.77)	38.02 (33.33)	39.44 (37.72)	33.61 (34.73)

3.1. Step 1: EFA of FAsTask

Table 1 provides descriptive statistics for the FAsTask items (pooled across studies). Results of the scree plot suggested either a 4 or 5 factor solution and parallel analysis suggested a 4-factor solution. Items on the fifth factor had significant cross-loadings onto other factors and the fifth factor did not have a clear, unique theoretical interpretation. Hence, a 4-factor model was selected. Models estimated using an orthogonal (varimax) rotation appeared to provide a poor approximation of simple structure as marked by several ambiguous factor loadings (i.e., λ>0.32 for multiple factors). Comparatively, the oblique rotation (Geomin) provided a closer approximation of simple structure as indicated by stronger loadings on each item’s primary factor and smaller loadings on the other factors. Therefore, an oblique rotation was used to better approximate simple structure. Four items (7, 8, 13, and 15) still had significant cross-loadings that were nearly equivalent in magnitude on each factor and were removed. The final set of 19 items and factor loadings resulting from the EFA are presented in Table 3.

Table 3.

Four factor EFA solution using a GEOMIN rotation.

	Item	Supervised/Structured	Active Monitoring	Task Engagement	Parental Rules/Strategies
2.	“Does it seem that the child spends time away from adult supervision?”	0.93 *	0.01	−0.03	−0.05
3.	“Does the child indicate being with friends in settings without adult supervision?”	0.91 *	−0.10	0.03	−0.01
4.	“Does there seem to be a lack of adult involvement in this child’s daily life?”	0.68 *	0.14	0.09	0.13
5.	“Is there a lack of structure or lax rules with respect to this child’s daily routine?”	0.66 *	0.21	−0.01	0.12
6.	“Is there any mention of the child’s peers planning or engaging in deviant behaviors?”	0.53 *	0.02	−0.07	−0.18*
9.	“Does this parent seem to be monitoring with whom the child spends time?”	−0.03	0.95 *	0.02	−0.03
10.	“Does this parent seem to be monitoring where the child spends time?”	0.09	0.89 *	0.01	0.04
11.	“Does this parent seem to be monitoring what the child is doing when outside of adult supervision?”	0.13	0.82 *	−0.04	0.01
12.	“Does the parent listen to the child?”	0.06	0.05	0.60 *	−0.09
14P.	“Does parent indicate any rules or guidelines that facilitated parents’ monitoring?”	0.01	−0.02	−0.03	1.02 *
14A.	“Does adolescent indicate any rules or guidelines that facilitated parents’ monitoring?”	0.06	0.15	−0.01	0.59 *
16.	“Does the parent seem to know about the child’s friendships, knowing the friends by names and their family situations?”	−0.04	0.67 *	0.16*	−0.03
18P.	“Does the parent follow the directions for the task?”	0.04	−0.02	0.76 *	0.15*
18A.	“Does the adolescent follow the directions for the task?”	−0.05	0.01	0.76 *	−0.11
19P.	“Does the parent participate in the discussion?”	0.01	0.03	0.75 *	0.13
19A.	“Does the adolescent participate in the discussion?”	0.03	−0.08	0.83 *	−0.11
20.	“Does the parent provide positive reasons for monitoring the child’s activities?”	−0.06	0.02	0.07	0.61 *
21.	“What percentage of the time does the family talk about the child’s activities?”	−0.15	0.05	0.76 *	0.08

Notes. P=Parent item; A=Adolescent item.

^*p < .05. Items retained on each factor are bolded.

Factor 1 (“Supervised/Structured”) captured five items that assess unmonitored (e.g., “Does the child indicate being with friends in settings without adult supervision”) or unstructured (e.g., “Is there a lack of structure or lax rules with respect to this child’s daily routine?”) adolescent behavior. Factor 2 (“Active Monitoring”) reflected four items querying parent/guardian use of active monitoring strategies (e.g., “Does this parent seem to be monitoring where the child spends time?). Factor 3 (“Task Engagement”) included seven items probing task-specific behavior (e.g., “Does the parent listen to the child during the task?”). Factor 4 (“Parental Rules/Strategies”) consisted of three items which captured parent and adolescent report of the imposition and justification of parental rules (e.g., “Does the parent indicate any rules or guidelines that facilitated parental monitoring?”).

A correlated factors CFA model with four factors (Figure 1) was specified in the second subsample based on the solution of the EFA model in the first subsample. The CFA model provided an adequate fit to the data (χ²(244) = 457.02, p<.0001, CFI = .906, TLI=.890, RMSEA= .069, SRMR=.092). Factors demonstrated good to excellent internal consistency: Supervised/Structured α=.86, Active Monitoring α=.92, Task Engagement α=.79, Parent Rules/Strategies α=.80.

Figure 1.

Correlated Factors Model.

Note. P=Parent item; A=Adolescent item.

3.2. Step 2: Measurement Invariance Testing

3.2.1. Age

Table 4 provides a summary of measurement invariance testing results for age and biological sex. Despite the initial configural invariance model for age providing an adequate fit to the data, the 16-years-old and older group had a larger chi-square contribution to model fit compared to the 15-year-olds and younger group. Including a residual covariance between items 2 and 3, 6 and 12, and 14A and 14P in the 16-years-old and older group significantly improved model fit. Invariance of all factor loadings (i.e., metric invariance) and item intercepts (i.e., scalar invariance) was supported across age groups. Partial residual invariance was supported after freeing the residual variances on items 2. These results indicate that the interpretation of FAsTask factors was not dependent on age.

Table 4.

Measurement invariance test results.

Model		Model Fit					Nested Model Test
	χ2	df	p-value	CFI	TLI	RMSEA (95% CI)	SRMR	χ2	df	p-value	Notes
Age (15 or younger, 16 or older)
Configural	494.106	262	0	0.939	0.929	0.068	0.095				Unequal χ contribution across ages (15 or younger=217.957, 16 or older=276.149). MI suggested including residual covariance between the following items in the older group: 2 with 3; 6 with 12; 14A with 19P.
Configural (Partial)	444.135	259	0	0.951	0.942	0.061	0.087	46.686	3	<.001	Significant improvement in model fit
Metric (Partial)	465.634	272	0	0.949	0.943	0.061	0.101	21.548	13	0.063	-
Scalar (Partial)	484.939	286	0	0.948	0.944	0.06	0.105	18.682	14	0.177	-
Residual (Partial)	520.019	304	0	0.943	0.943	0.061	0.106	33.615	18	0.014	MI suggested freeing residual variance for item 2
Residual (Partial)	510.127	303	0	0.945	0.945	0.059	0.105	26.419	17	0.067
Biological Sex (females, males)
Configural	534.04	262	0	0.929	0.917	0.074	0.104	-	-	-	Unequal χ2 contribution across sex: females=247.407, males=286.633. including residual covariance between the following items in the Male group: 14A with 19A; 6 with 12.
Configural (Partial)	498.926	256	0	0.937	0.925	0.07	0.104	40.023	6	<.001	Significant improvement in model fit
Metric (Partial)	508.559	271	0	0.938	0.93	0.068	0.107	11.660	15	0.705	-
Scalar (Partial)	543.298	285	0	0.932	0.927	0.069	0.109	35.386	14	0.001	Reject, CFI change=−.0005; MI suggested freeing intercept for item Fast 14A
Scalar (Partial)	532.418	284	0	0.935	0.93	0.068	0.109	23.823	13	0.033	-
Residual (Partial)	579.405	301	0	0.927	0.926	0.069	0.118	34.330	16	0.005	Reject, CFI change=−.0005 and SRMR change; MI suggested freeing the residual variance for item 19A
Residual (Partial)	548.501	300	0	0.935	0.934	0.066	0.114	13.827	15	0.539	-

Notes. MI=modification indices.

Item numbers listed in notes are based on the item numbers presented in Table 3.

3.2.2. Biological Sex

Despite the overall configural invariance model for biological sex providing an acceptable fit to the data (see Table 4), the male group had a larger chi-square contribution to model fit compared to the female group. Including residual covariances between items 14A and 19A significantly improved model fit. Invariance testing supported constraining all factor loadings (i.e., metric invariance) across sex. Constraining all item intercepts to be equal across conditions (i.e., scalar invariance) led to a significant decrement in model. Partial scalar invariance was supported after freeing the item intercept for item 14A. Partial residual invariance was supported after freeing the residual variance for item 19A. These results indicate that the interpretation of FAsTask factors was not dependent on biological sex.

3.3. Step 3: FAsTask associations with alcohol and cannabis use

The hybrid SEM prediction models for alcohol and cannabis use outcomes did not initially converge. Increasing the number of iterations from 500 to 1000 and lowering the convergence criteria from .01 to .05 in Mplus led to the models converging without estimation issues.

3.3.1. Alcohol Use

Alcohol use favored a zero inflated negative binomial distribution (AIC= 12311.08, BIC= 12567.84) over a negative binomial distribution (AIC= 12380.97, BIC= 12634.52). Full model results are presented in Table 5. The Supervised/Structured factor was negatively associated with the proportion of alcohol use days and the Active Monitoring factor was positively associated with alcohol use days. Task Engagement and Parental Rules/Strategies were not significantly associated with the proportion of alcohol use days.

Table 5.

FAsTask associations with alcohol and cannabis use.

	Alcohol Use			Binge Drinking			Cannabis Use
Predictors	β	SE	P-value	β	SE	P-value	β	SE	P-value
Demographic covariates
Age	0.163	0.118	0.165	0.157	0.129	0.221	0.212	0.056	0.000
Biological Sex (0=male, 1=female)	0.117	0.129	0.364	0.175	0.122	0.149	0.012	0.054	0.816
Ethnicity (0=not Latine, 1=Latine)	−0.082	0.131	0.531	0.136	0.128	0.289	−0.062	0.057	0.279
Race (0=White, 1=Non-White)	−0.220	0.139	0.111	−0.329	0.127	0.009	−0.086	0.059	0.147
Study (Alcohol Study [Study 1] is the reference group)
Cannabis Studies (Studies 3 and 4)	−0.080	0.199	0.689	−0.359	0.175	0.039	−0.345	0.122	0.004
Polysubstance Use Studies (Studies 2, 5, 6)	0.061	0.187	0.744	0.047	0.173	0.784	0.359	0.124	0.004
Monitoring Factors
Supervised/Structured	−0.814	0.242	0.001	−0.697	0.246	0.004	−0.176	0.086	0.041
Active Monitoring	0.694	0.268	0.005	0.271	0.266	0.307	−0.012	0.096	0.901
Task Engagement	−0.021	0.211	0.920	−0.106	0.170	0.541	−0.009	0.054	0.868
Parental Rules/Strategies	0.299	0.154	0.053	0.506	0.151	0.001	0.052	0.062	0.397

Note. Alcohol and cannabis use were modeled with a zero inflated negative binomial distribution, while binge drinking model was modeled with a negative binomial distribution.

3.3.2. Binge Drinking

The negative binomial model examining the proportion of binge drinking days provided a superior fit (AIC=24730.19, BIC=25035.23) relative to the zero-inflated negative binomial model (AIC=24732.19, BIC=25041.15). Not surprisingly, studies focused on cannabis use reported less binge drinking than studies that focused on alcohol. As with the proportion of alcohol use days analyses, the Supervised/Structured factor was negatively associated with the proportion of binge drinking days. Parental Rules/Strategies was positively associated with binge drinking. Other FAsTask factors were not significantly associated with the proportion of binge drinking days (Table 5).

3.3.3. Cannabis Use

The zero-inflated negative binomial model (AIC= 26428.56., BIC= 26737.51) provided a better fit to the data than the negative binomial model (AIC= 26448.49, BIC= 26753.54). As is indicated in Table 5, studies focused on cannabis and polysubstance use reported higher cannabis use rates than studies focused on alcohol. The Supervised/Structured factor was negatively associated with proportion of cannabis use days. Other FAsTask factors were not significantly associated with the proportion of cannabis use days.

4. Discussion

The parental monitoring FAsTask, an observer-rated interaction task designed for substance use interventions, is widely used in clinical and community settings, but has never been subjected to rigorous psychometric testing. This study assessed the factor structure of the task, the extent to which it can be reliably implemented across age and sex, and its concurrent validity. We pooled data across six clinical trials for a robust sample in which to test the psychometric properties of the parental monitoring FAsTask.

4.1. Factor Structure

Factor analysis of the parental monitoring FAsTask revealed a 4-factor structure, consisting of factors labeled as Supervised/Structured, Active Monitoring, Task Engagement, and Parental Rules/Strategies. Although no prior work has formally evaluated the factor structure of the longer FAsTask’s parental monitoring component (i.e., 23-item version), Dishion et al. (2003) proposed that the 12-item version contained two factors: “Child-specific Monitoring,” and “Parent-specific Monitoring.” These factors align with two of the factors identified in the current study, Supervised/Structured and Active Monitoring, respectively.

This study identified two additional factors: Parental Rules/Strategies and Task Engagement. Parental Rules/Strategies refers to the extent to which parents set house rules regarding monitoring their adolescent’s behavior and provided positive reasons for monitoring rules (e.g., to keep their adolescent out of legal trouble). This factor seems related to parent-adolescent communication, an important protective factor against adolescent risk behaviors (Ryan et al., 2015). Rule-setting skills are fundamentally distinct concepts than those Dishion first put forward, which instead relied upon a common definition of parental monitoring focused predominantly on maintaining knowledge and awareness of the adolescent’s whereabouts (Kerr & Stattin, 2000; Stattin & Kerr, 2000).

The determination that adherence to the FAsTask instructions was a separate factor (“Task Engagement”) is an interesting methodological finding. At face value, the Task Engagement factor measures the extent to which adolescents and their parents followed task instructions and engaged with task prompts in the FAsTask. However, the factor may actually capture deeper unmeasured aspects of family functioning, such as parent-child communication (i.e., the degree to which the dyad is engaged, present, and listen to each other, as is specified in the instructions) and family problem-solving, both of which have been found to be protective against adolescent substance use in other studies (Abar et al., 2014; Ryan et al., 2015).

A key question arising from these findings is whether we can/should map the current factor structure findings to self-report assessment of parental monitoring. Our view is that we should not. The goal of this study was not to provide a direct comparison of the structure of the FAsTask measure to the structure of widely used self-report measures, such as those developed by Stattin & Kerr (2000). First, the FAsTask is a different method for evaluating parental monitoring, and it was developed prior to many of the frequently used self-report tools. Second, the goal of mixed-method assessment is to provide a multifaceted and comprehensive assessment of the construct under study. Third, prior work indicates that the FAsTask is not redundant with self-report measures as it predicts unique variance in adolescent outcomes (MacPherson et al., 2021). Thus, we would not expect the FAsTask to yield the same factors as self-report measures.

4.2. Measurement Invariance

Exploratory analyses of measurement invariance across groups (i.e., age and sex) indicated that the interpretation of the FAsTask factors was not dependent on the group in which it was implemented. The factor structure held across age and sex; both of which have been associated with adolescent substance use (Fagan et al., 2013; Ritchwood et al., 2015; Talley et al., 2013). Additional analyses of other types of invariance (i.e., metric, scalar, and residual) similarly found evidence of partial measurement invariance after freeing a few parameters. Taken together, findings indicate that the parental monitoring FAsTask retains its factor structure across age and sex. This is an important finding given youth who present to substance use treatment are diverse with respect to demographic factors (Burlew et al., 2013).

4.3. Parental Monitoring FAsTask Associations with Substance Use

The results provided evidence that three of the identified factors were associated with adolescent substance use, though the effects were not consistently in the anticipated direction. The Supervised/Structured factor was negatively associated with the proportion of alcohol use days, binge drinking and cannabis use days, indicating that adolescents who reported more parental supervision and structure engaged in less alcohol use and cannabis use. These findings were consistent with a wealth of literature demonstrating that parenting supervision and structure are protective against adolescent substance use (Costello et al., 2010; Richardson et al., 1993; Snyder & Merritt, 2015). By contrast, Parental Rules/Strategies factor was positively associated with binge drinking. While perhaps counterintuitive, two of the items that load on this factor assess parental use of rules/guidelines that facilitated monitoring, but they do not necessarily represent effective use of monitoring rules/strategies, nor do they specify the types of strategies being employed. For example, parents might set specific rules (e.g., not to drink and drive, to be home by midnight), none of which would necessarily curb drinking even if the child complies. In other words, parents may set rules, but they still may not address the behavior in question. Moreover, the third item captured parental positive reasons for monitoring; it may be the case that parents of children who use alcohol more frequently provided more positive reasons for monitoring due to their child’s substance use. Active Monitoring was positively associated with alcohol use, which is also counter to the direction that would be expected. The items that load on this factor tap the degree to which parents know where/with whom the child spends time. Considering the cross-sectional nature of this data, one possibility is that caregivers engage in greater levels of monitoring after their children begin to engage in alcohol use. Prospective studies using the FAsTask would help tease apart the temporal relations between Active Monitoring and alcohol use. Task Engagement was not significantly associated with any substance use outcomes. Future work should seek to replicate these findings.

In sum, of the components of the parental monitoring task, those components focused on actively supervising the adolescent and ensuring structure in their routine were the most strongly associated with substance use. The 5 W’s (Who, What, Where, When, Why) is a common mnemonic used by therapists working with parents of adolescents to gather information about their adolescent’s whereabouts (Guilamo-Ramos, Jaccard, & Dittus, 2010); however, our findings suggest that direct adult supervision, and not simply knowledge of the adolescent’s activities, is most associated with lower risk for substance use. If replicated, our work could also point towards assessing only the Supervised/Structured items, which could substantially reduce measurement burden associated with using and scoring the parental monitoring component in clinical practice.

4.4. Limitations

These findings should be interpreted considering the following limitations. First, though this study benefitted from the inclusion of high-risk adolescents recruited across locations (e.g., emergency departments, family/truancy courts, psychiatric hospitals), the heterogeneity may have introduced a range of unmeasured constructs that could contribute to the observed effects (e.g., legal involvement, psychiatric diagnoses). Second, the sample size was not sufficient to test for measurement invariance across race and ethnicity, thus we were not able to capture culturally specific parenting practices that may be relevant to the FAsTask. Third, the analyses of the associations between FAsTask factors and the adolescent substance use variables were cross-sectional; therefore, it is not possible to conclude that the parental monitoring factors (and particularly the supervised/structured variable) longitudinally predict adolescent substance use. It is equally plausible that adolescent substance use led to increased parental supervision and structure, or that a third variable (e.g., psychiatric diagnoses) contributed to both heightened parental supervision and heightened substance use. Fourth, while the focus on substance use as the primary outcome was a strength, we did not have sufficient data to test how the four factors predict other high-risk behaviors associated with monitoring (e.g., truancy). Future longitudinal studies should seek to confirm this factor structure in other diverse samples and test a broader range of high-risk behaviors.

5. Clinical Implications

This study is the first to establish the structure, measurement invariance, and concurrent validity of the FAsTask parental monitoring component, a parent-adolescent interaction task that is used as both an assessment tool and as part of a preventative intervention. The FAsTask parental monitoring component appears to consist of four factors that are stable across age and sex: Supervised/Structured, Active Monitoring, Task Engagement, Parental Rules/Strategies.

We view the structure (EFAs/CFAs) and concurrent criterion validity analyses (SEM hybrid models) as each contributing important clinical utility. Clinicians use the FAsTask to guide clinical care in the FCU and our findings (i.e., identifying separate factors) can inform accurate and efficient scoring practices that can effectively guide clinical care. The SEM hybrid models, on the other hand, inform how each factor is related to correlates of treatment outcomes for youth who use substances. Knowledge of the correlates of a FAsTask factor may alert clinicians to youth who are particularly at risk for engaging in a specific substance of use. Developing a broader nomological network of correlates of the FAsTask (e.g., co-occurring psychopathology, self- and parent-report measures of family functioning, peer risk taking) may further help clinicians understand the correlates of the factors on the FAsTask monitoring task.

The FAsTask was recorded and coded by two raters in each of the studies reported in this paper because they were research studies. Regarding the practicality of administering the FAsTask in clinical practice, it is ideal for the task to be recorded and later rated using the codebook to generate clinical feedback for the FCU. When that is not possible, clinicians can observe the family interaction in real time and complete the coding sheet. Results of this study have three additional key clinical implications. First, clinicians and researchers seeking to use the FAsTask parental monitoring component can consider using the four subscales identified herein to score the task which also allows for more fine-grained recommendations regarding these four different aspects of monitoring. Second, clinicians and researchers should feel confident using the FAsTask parental monitoring component across age and sex. Third, if the FAsTask parental monitoring component is being used in a research study with the goal of understanding risk of substance use, then the items on the Supervised/Structured factor are likely the most valuable to score. By using the factor structure identified in this study, clinicians may more efficiently use the FAsTask parental monitoring component to inform clinical care.

Highlights.

• The parental monitoring FAsTask consists of four factors.

• The Supervision/Structure factor was negatively associated with substance use.

• Parental monitoring FAsTask can be reliably administered across age and sex.