University of Southern California Dissertations and Theses

Using integrative data analysis to evaluate gender differences in effects of multisystemic therapy for justice-involved youth

(USC Thesis Other)

Using integrative data analysis to evaluate gender differences in effects of multisystemic therapy for justice-involved youth

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Using Integrative Data Analysis to Evaluate Gender Differences in Effects of Multisystemic
Therapy for Justice-Involved Youth
by
Katharine Galbraith, M.A.
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
August 2024
Copyright 2024 Katharine Galbraith

GENDER DIFFERENCES IN MST EFFECTS
ii
Acknowledgments
Thank you, Dr. Stanley J. Huey, Jr., for being my mentor throughout grad school and
supporting me throughout my dissertation. You have taught me to think critically and write
thoughtfully, which I will take with me for the rest of my career. You have always prioritized my
well-being and encouraged me to remain positive even at the most difficult moments of graduate
school. Thank you to the rest of my dissertation committee, Drs. Christopher Beam, Jordan Davis,
and Richard John – your thoughtful feedback and willingness to connect me with helpful resources
made this study possible. I would like to extend an extra thanks to Chris, who spent tireless hours
walking me through the complexity of Integrative Data Analysis, and always went above and
beyond to keep me abreast of the latest research and resources in this area. Thanks to Gayla
Margolin, who provided me with invaluable support in my first funding application and has always
been so gracious in inviting me to and sharing her wisdom with me during her lab meetings. Thank
you to Dr. David Schwartz for supporting me throughout my clinical science career, from serving
on my master’s committee to writing a letter of recommendation on my behalf for grant
applications and postdocs. Thank you to Dr. Jonathan Tarbox, who played a huge role in my
master’s project, and has been such a positive force since. Thank you to Dr. Eraka Bath, who was
unbelievably willing to help support me in my endeavor to apply for an NIH grant despite hardly
knowing me. Thank you to Dr. Marina Tolou-Shams, who has been the most incredible mentor and
role model for me and is the reason I kept going in this field in the first place.
Thank you to my clinical supervisors who invested their time and energy into helping me
better understand existing interventions and become a better clinician: Drs. Latonya Wood, Gayla
Margolin, Shannon Couture, Vivian Credidio, and Chrystal Maher. My clinical knowledge and
practice have made me a better disseminator and evaluator of interventions in and outside of
research contexts.
Thank you to some of my dearest friends who have been right there with me along the way:
Ree Ree Li, Allie Souza, Catherine Wang, Natalie Abeysena, among many others – your love and
support have been my guiding light since day one of graduate school. Thank you to my mom and
dad – I quite literally could not have completed this degree without you. Most of all, thank you to
my husband, Ryan McGarry. You helped me get through this in more ways than you know, and you
are the best man I’ve ever known.

GENDER DIFFERENCES IN MST EFFECTS
iii
Table of Contents
Acknowledgments ...............................................................................................................................ii
List of Tables......................................................................................................................................iv
List of Figures......................................................................................................................................v
Abstract...............................................................................................................................................vi
Chapter 1: Introduction........................................................................................................................1
Overview of Multisystemic Therapy ...............................................................................................4
Effectiveness of Multisystemic Therapy .........................................................................................7
Overview of Integrative Data Analysis ...........................................................................................8
Study Aims and Hypotheses..........................................................................................................13
Chapter 2: Method .............................................................................................................................15
Overview of Studies.......................................................................................................................15
Study Outcome Measures..............................................................................................................18
Data Transformation Procedures...................................................................................................20
Analysis as a Collective Dataset....................................................................................................26
Power .............................................................................................................................................27
Chapter 3: Results..............................................................................................................................29
Measurement Invariance Testing – Moderated Nonlinear Factor Analysis..................................29
Pooled Data Analyses....................................................................................................................45
Chapter 4: Discussion ........................................................................................................................53
Limitations.....................................................................................................................................59
Implications ...................................................................................................................................61
Future Directions ...........................................................................................................................63
References..........................................................................................................................................65
Appendix A: Outcome Measures Used in the Current Study............................................................75
Personal Experiences Inventory.....................................................................................................75
Family Adaptability and Cohesion Scale (FACES) – Parent Report & Youth Report .................76
Self-Report Delinquency Scale......................................................................................................78
Appendix B: Overview of Missing Data ...........................................................................................81

GENDER DIFFERENCES IN MST EFFECTS
iv
List of Tables
Table 1: Nine Principles of MST.........................................................................................................4
Table 2: Overview of MST RCTs Included in the Current Study.....................................................15
Table 3: Data Timepoints Included Across Studies...........................................................................16
Table 4: Overview of Trial Outcomes...............................................................................................17
Table 5: Required Number of Females in Sample Needed to Achieve Specified Power..................28
Table 6: FACES Youth Report Parcel Level MNFLA Results.........................................................34
Table 7: FACES Youth Report Final MNLFA Model ......................................................................34
Table 8: FACES Parent Report Parcel Level MNLFA Results.........................................................36
Table 9: FACES Parent Report Final MNLFA Model......................................................................37
Table 10: SRDS Item Level MNLFA Results...................................................................................38
Table 11: SRDS Final MNLFA Model..............................................................................................42
Table 12: Self-Reported Alcohol Use................................................................................................45
Table 13: Self-Reported Marijuana Use ............................................................................................46
Table 14: Self-Reported Delinquency ...............................................................................................47
Table 15: Family Adaptability Ratings, Youth Report......................................................................49
Table 16: Family Cohesion Ratings, Youth Report...........................................................................50
Table 17: Family Adaptability Ratings, Parent Report......................................................................51
Table 18: Family Cohesion Ratings, Parent Report ..........................................................................52
Table A1: CFA Model Fit Statistics for Original Youth FACES......................................................77
Table A2: CFA Model Fit Statistics for Parceled Youth FACES .....................................................77
Table A3: CFA Model Fit Statistics for Original Parent FACES......................................................77
Table A4: CFA Model Fit Statistics for Parceled Parent FACES .....................................................77
Table A5: CFA Model Fit Statistics for Unidimensional SRDS.......................................................80
Table B1: Constructs/Measure Availability by Study and Timepoint...............................................82

GENDER DIFFERENCES IN MST EFFECTS
v
List of Figures
Figure 1: Intraclass Correlation Curves (ICCs) Uniform vs. Nonuniform DIF.................................11
Figure 2: Path Diagram for MNLFA Model......................................................................................24
Figure 3: Alcohol Use Over Time .....................................................................................................46
Figure 4: Self-Reported Delinquency Over Time..............................................................................48
Figure 5: Youth Reported Family Adaptability Ratings by Gender..................................................50

GENDER DIFFERENCES IN MST EFFECTS
vi
Abstract
Background: Justice-involved youth (JIY) have higher rates of substance use and substance use
disorders relative to their non-offending counterparts. Substance use co-occurs with delinquency
and exacerbates antisocial behavior. Thus, interventions for JIY often target co-occurring substance
use and delinquency. These interventions typically consider male-specific (e.g., higher rates of
violent offending) or gender-neutral (e.g., delinquent peers) risks for substance use and
delinquency, but neglect female-specific risks (e.g., elevated rates of PTSD), which has raised
concern about their appropriateness for female JIY. Females represent the fastest-growing subgroup
among JIY, but whether they benefit from existing interventions remains unclear due to poor
representation in randomized trials and low statistical power. Method: The current study used
Integrative Data Analysis (IDA) to pool raw data from three randomized controlled trials of
Multisystemic Therapy (MST), a widely used intervention for substance use and delinquency
among JIY. IDA allows for analysis as one “mega” dataset to generate the statistical power needed
to detect gender differences in treatment outcomes. The aims of the current study were three-fold.
Aim 1 was to establish measurement invariance across study, timepoint, and gender among study
outcome measures using moderated nonlinear factor analysis. Aim 2 was to determine whether
MST was effective in reducing substance use and delinquency, and improving family functioning
relative to the control group through one year follow-up when raw data from three trials was pooled
together as one dataset. Aim 3 was to determine whether gender moderated MST effects on
delinquency, substance use, and family functioning through one year follow-up. Results:
Moderated nonlinear factor analyses showed that measurement invariance could be established
across study, timepoint, and gender, although measures exhibited uniform and nonuniform
differential item functioning, which required constraining and releasing intercepts and loadings
before pooling data for analysis as a collective dataset. Analyses of the pooled dataset showed MST
was no more effective than the control group in reducing substance use and delinquency, nor in
improving family functioning among JIY through one year follow-up. Moreover, MST was equally
ineffective for male and female JIY. Conclusions: The current study is the first to use IDA to
analyze MST outcomes and whether effects vary by gender. Findings suggest that measurement
invariance can be established among outcome measures commonly used among JIY (e.g., selfreported delinquency scales), which is promising for continued use of IDA in future studies with
this population. Findings also suggest that, across these three pooled randomized trials, MST is no
more effective than usual services for male and female JIY; continued development and study of
effective psychosocial interventions for this population is sorely needed.

GENDER DIFFERENCES IN MST EFFECTS
1
Using Integrative Data Analysis to Evaluate Gender Differences in Effects of Multisystemic
Therapy for Justice-Involved Youth
Chapter 1: Introduction
Justice-involved youth (JIY) have higher rates of substance use and substance use
disorders relative to their non-offending counterparts (Chassin, 2008; Teplin et al., 2002). An
estimated 22-96% of JIY meet criteria for substance use disorders(Borschmann et al., 2020)
relative to 15-16% among all adolescents (Swendsen et al., 2012). Substance use is a mental
health issue and a form of delinquent behavior; it is also highly correlated with other types of
offending behavior (e.g., theft; Assink et al., 2015). As such, evidence-based interventions for
JIY often focus on treating co-occurring substance use and delinquency (Belenko et al., 2017;
Underwood et al., 2016).
Within the juvenile justice system, males represent the majority (71%) of JIY and engage
in more serious and frequent forms of delinquency relative to females (Ehrmann et al., 2019;
Farrington et al., 2009; Gorman-Smith & Loeber., 2005; Loeber et al., 2013; Tracy et al., 2009).
In contrast to delinquency, males and females in the juvenile justice system report similar rates
of substance use (Borschmann et al., 2020), but justice-involved females are more likely to
develop substance use disorders (Teplin et al., 2002; Teplin et al., 2006). Males and females
share risk factors (i.e., gender-neutral risks) for both delinquency and substance use, including
deviant peer involvement, high impulsivity, and low levels of parental monitoring (Pusch &
Holtfreder, 2018; Scott & Brown, 2018).
In addition to shared (i.e., gender-neutral) risks, females also possess gender-specific
developmental pathways and risks for substance use and delinquency leading to justice-system
involvement. For example, relative to males, females are more likely to become justice-involved

GENDER DIFFERENCES IN MST EFFECTS
2
for status offenses (i.e., offenses that are illegal due to being a minor), and their offenses are
more likely to be linked to experiences of victimization (e.g., retaliating against an abuser;
Belknap & Holsinger, 2006; Ehrmann et al., 2019; Kerig, 2018). Polyvictimization (i.e., having
experienced multiple types of traumatic events and/or chronic trauma exposure) and sexual abuse
are stronger predictors of delinquency and substance use in females compared to males (Baglivio
et al., 2014; Conrad et al., 2014; Kerig, 2018; Kerig & Ford, 2014; Modrowski et al., 2021).
Females are also more likely than males to use substances to cope with other psychiatric
symptoms, such as post-traumatic stress or depression (Smith & Saldana, 2013). Internalizing
problems are a stronger predictor of substance use and delinquency in female JIY, and rates of
psychiatric comorbidity are higher among females with substance use disorders (Abram et al.,
2003; Copeland et al., 2007; Docherty et al., 2016; McCabe et al., 2002; Wasserman et al.,
2005). Conflictual relationships, especially with family members or romantic partners, is a
stronger predictor of substance use and delinquency for females as well (Burgess-Proctor et al.,
2016; Cauffman et al., 2008; Kerig, 2014; Kerig & Ford, 2014; Kuhn, 2015; Liu & Miller, 2020;
Rusby et al., 2018; Skeer et al., 2011).
Taken together, these findings provide support for the gendered pathways theory
(Burgess-Proctor et al., 2016; Jones at al., 2014). In the criminology literature, gendered
pathways theory suggests males and females possess shared, gender-neutral risks and needs, but
that females also possess gender-specific risks and needs that account for their delinquent
behavior and co-occurring behaviors like substance use (Jones et al., 2014; Whaley et al., 2013).
The theory argues gender-neutral and gender-specific risks for reoffending need to be considered
to effectively treat co-occurring delinquency and substance use in females.

GENDER DIFFERENCES IN MST EFFECTS
3
Unfortunately, female-specific risks for substance use and delinquency are often
neglected in interventions disseminated in the juvenile justice system, largely because males
represent the majority of JIY (~79%; Leve et al., 2015; Lydia, 2020; Puzzancherra & Ehrmann,
2018; Sherman & Balck, 2015; Zahn et al., 2009). Feminist scholars hypothesize this lack of
attention to gender-specific needs suggests existing interventions delivered in justice system
contexts may be less effective for females (Bloom & Covington, 2001; Covington & Bloom,
2007) and thus advocate for the implementation of gender-responsive programming for JIY.
Although males represent the majority of JIY, the proportion of females has grown over the past
several decades (Puzzanchera & Ehrmann, 2018; Puzzanchera, 2022), so the study of whether
interventions for substance use and delinquency work equally well for male and female JIY is an
increasingly important topic of research.
There are several widely used interventions to treat substance use and delinquency among
JIY, including Multisystemic Therapy (MST), Multidimensional Family Therapy, and Treatment
Foster Care Oregon (Elliot et al., 2020). However, published research studies of these
interventions either have too few participants to detect gender effects in mixed-gender samples
(i.e., lacking sufficient statistical power) or have single-sex samples (Leve et al., 2015). As such,
there is a critical gap in the literature as to whether existing interventions for JIY work as well
for females compared to males. Increasing sample sizes through pooled data techniques, thereby
increasing statistical power to detect moderator effects, can address this gap in the literature. The
current study aims to assess the effectiveness of one of these widely used interventions, MST,
and whether these effects vary by gender using Integrative Data Analysis (IDA), a pooled data
technique that allows for increased statistical power to detect moderator effects. An overview of
MST and IDA are provided in the following sections.

GENDER DIFFERENCES IN MST EFFECTS
4
Overview of Multisystemic Therapy
MST is tailored to meet each youth’s unique needs; it is an intensive and flexible
intervention that targets diverse ecological factors, including family and peer factors, that
contribute to delinquency and substance use (Henggeler et al., 1998; Henggeler et al., 2009;
Huey et al., 2000). MST is informed by Bronfenbrenner’s ecological systems theory, which
posits that individuals are a product of the interactions within and between various systems in
which the youth is embedded (e.g., school, peer, neighborhood, family; Bronfenbrenner, 1979).
MST is guided by nine core principles, all of which are outlined in Table 1.
Table 1
Nine Principles of MST
Principle Description
1. Finding the fit Assessment is needed to understand the “fit”
between identified problems and their systemic
context.
2. Positive and strength focused Theraputic contacts should emphasize positives
and leverage existing systemic strengths to
promote change.
3. Increasing responsibility Intervention should focus on promoting
responsible behavior and decreasing
irresponsible behavior among youth and family
members.
4. Present focused, action oriented,
and well defined
Goals for treatment should be clear and
measurable, and focused on the present.
5. Targeting sequences Target sequences of behavior within and
between multiple systems (e.g., looking at how
miscommunication between youth and teachers
at school and between youth and caregiver at
home is causing difficulties across both
systems).

GENDER DIFFERENCES IN MST EFFECTS
5
6. Developmentaly appropriate Interventions should fit developmental needs of
youth.
7. Continuous effort Intervention should require daily or weekly
effort by family members.
8. Evaluation and accountability Effectiveness should be evaluted from multiple
perspectives (e.g., parent, youth, and therapist
report); providers assume accountability for any
barriers to improved outcomes.
9. Generalization Interventions should generalizable and facilitate
long-term maintenance of positive behavior
change.
MST emphasizes the need for clinicians to adapt interventions to reduce delinquency,
substance use, and other behavioral problems using strategies best suited to the youth and their
family’s specific risks, needs, and strengths. This personalized approach includes the individuals
who are key to sustaining youth behavior change. For example, within the family system, the
MST therapist typically involves caregivers in the implementation of evidence-based strategies
(e.g., tracking the youth’s behavior and implementing appropriate consequences and reinforcers
consistently) to reduce problem behaviors like delinquency and substance use and increase
desired behaviors like attending school and engaging with prosocial peers. Within the MST
framework, providers offer caregivers various services and resources as needed so that they are
best equipped to implement these strategies. For example, MST therapists might offer
caregiver(s) individual and/or couples counseling to help caregivers with their own emotion
regulation or conflict resolution skills so that they can be more effective in facilitating positive
behavior change in their child. Additionally, MST therapists have flexible work hours and low
caseloads (typically 4-6 families), and they provide services in settings most convenient for
youth and families (e.g., home, school, community). There is also a rotating on-call schedule

GENDER DIFFERENCES IN MST EFFECTS
6
such that an MST therapist is available 24/7 should any crisis occur. This model of delivery has
minimized dropout and increased treatment uptake among JIY (Henggeler et al., 1996).
The tailored nature of MST suggests it aligns with gendered pathways theory – that is, by
individualizing the treatment to a youth’s specific needs and strengths, it likely integrates genderneutral and female-specific risks and needs into treatment for female JIY. However, reviews of
MST manuals used for addressing delinquency among JIY reveal little explicit mention of
female-specific risks and needs (Henggeler et al., 1998; Henggeler et al., 2009), which suggests
MST may align more with gender-neutral or universal criminology theories, i.e., theories that
argue the same underlying factors explain both female and male risks for delinquency. For
example, although the manual instructs providers to personalize treatment to each youth’s needs,
instructions and case examples provided in the manual focus primarily on male youth with
externalizing behavior problems. There is limited case material presented on internalizing
problems, psychiatric comorbidity, and sexual abuse – all of which are more strongly associated
with female substance use and delinquency. Since females have historically and currently
represent the minority of JIY, providers in juvenile justice contexts may have a more limited
understanding of female-specific pathways into the juvenile justice system, as well as genderspecific treatment considerations for addressing delinquency, substance use, and other related
psychosocial concerns (Sherman & Balck, 2015). Thus, given the lack of explicit gender
considerations in MST protocols and manuals, MST might be less effective for females relative
to males, despite its individualized and tailored nature. Because MST is one of the most widely
used treatments for substance use and delinquency among JIY (MST Services, 2024),
establishing whether this intervention is equally effective for females relative to males has
substantial practical value in the field. Findings from the current study may directly inform

GENDER DIFFERENCES IN MST EFFECTS
7
existing clinical practice - that is, whether MST should be used to treat delinquency, substance
use, and other related psychosocial outcomes among males and females in the juvenile justice
system.
Effectiveness of Multisystemic Therapy
MST has been touted as one of the leading evidence-based treatments for reducing
delinquency and substance use among JIY (MST Services, 2024), with 25 randomized controlled
trials (RCTs) supporting the effectiveness of the intervention in reducing problem behavior
among JIY and other populations, including juvenile sex offenders and youth with serious
emotional disturbance (Henggeler et al., 2016). Some reviews and meta-analyses (e.g., Curtis et
al., 2004; Henggeler et al., 2016; Henggeler & Schaeffer, 2010; van der Stouwe et al., 2014)
show MST is significantly more effective than usual services in addressing self-reported
delinquency, substance use, and other related outcomes, including family functioning. However,
findings from more recent meta-analyses (e.g., Markham, 2017; Littell et al., 2021) and
individual trials of MST with large sample sizes are mixed. For example, a recent systematic
review and meta-analysis found that MST was more effective at reducing arrests and out-ofhome placements but was not more effective than control conditions at reducing self-reported
delinquency. Similarly, a large RCT conducted in the United Kingdom with a sample size of
N=684 found that MST was largely ineffective at reducing delinquent behavior compared to
services as usual (Fonagy et al., 2018). Fonagy et al. (2018) found MST was more effective than
services as usual at reducing the volume and variety of self-reported substance use at 6 months
follow-up, but effects dissipated by 12 and 18-month follow-up, and gender did not moderate
treatment effects. Notably, recent reviews and meta-analyses (e.g., Markham 2017; Littell et al.,

GENDER DIFFERENCES IN MST EFFECTS
8
2021) neglect examining gender as a moderator of treatment effects entirely, which further
highlights how little is known regarding whether female JIY benefit from MST.
Results from the three RCTs included in the current study on whether MST is more
effective than control treatments are also mixed. For example, MST was not more effective than
control conditions at reducing self-reported delinquency in any of the trials (Henggeler et al.,
1997; Henggeler et al., 1999; Henggeler et al., 2006). However, for some studies, MST was more
effective than the control treatment at reducing self-reported status offenses and crimes against
persons (Henggeler et al., 2006) and archival measures of delinquent behavior (total days in outof-home placements; Henggeler et al., 1999), as well as substance use (Henggeler et al., 1999).
More detailed findings from each of the individual trials are provided in the Method section.
Among the three RCTs included in the current study, moderating effects of gender on
delinquency, substance use, and other related outcomes were either not examined (i.e. Henggeler
et al., 1997; Henggeler et al., 2006) or not detected (Henggeler et al., 1999), although each
individual study is limited in power to detect gender differences in treatment effects. The use of
IDA in the current study will allow for a large enough sample size for appropriate moderator
testing by gender, which will further elucidate whether MST is indeed effective for female JIY.
Overview of Integrative Data Analysis
One way to address the issue of limited statistical power in evaluating gender differences
in treatment effects is to use Integrative Data Analysis (IDA), a novel statistical framework that
pools multiple independent samples into one “mega” dataset (Curran et al., 2008). IDA can
directly model influences on between-study differences, which allows for analysis of study
equivalence (Curran et al., 2008; Hussong et al., 2013). Study equivalence ensures that raw data
across studies can be combined to formulate one larger dataset (Curran et al., 2008; Hussong et

GENDER DIFFERENCES IN MST EFFECTS
9
al., 2013). IDA builds a cumulative knowledge base using raw, existing data to address
limitations often present in single studies (Curran & Hussong, 2009). For example, pooling
multiple datasets into one increases the overall sample size, thereby increasing statistical power
to detect moderator effects of group membership in underrepresented subgroups (e.g., females).
It also allows for more reliable analyses of low base-rate behaviors (e.g., hard substance use);
combining datasets increases the aggregate number of low base-rate behaviors than is typically
found in a single study. This enhances the stability of model estimation, reduces the influence of
outliers on analyses, and allows for fitting more complicated models to the data (Curran &
Hussong, 2009). In the current study, I use IDA to pool data from three RCTs to assess whether
MST is effective at reducing delinquency and substance use, and improving family functioning,
among JIY over time, and whether effects vary by gender. All three datasets in the current study
include both male and female JIY.
Measurement invariance, the statistical property that indicates the same construct is being
measured across specified groups (e.g., across studies) is critical in IDA (Hussong et al., 2013;
Van de Schoot et al., 2015). Establishing measurement invariance requires rigorous statistical
testing to assess for differential item functioning (DIF). DIF occurs when members from
different groups (e.g., participants from study 1 vs. study 2) have different probabilities of
endorsing a particular item after controlling for overall scale scores. Put more simply, DIF is “the
circumstance in which two individuals of similar ability do not have the same probability of
answering a question the same way” (APA, 2014). A frequently cited example of DIF is an item
asking whether someone “cries often” on a scale assessing overall levels of depression; boys
with the same levels of depression as girls are much less likely to endorse “crying easily” despite
being equally as depressed, which indicates this item is functioning differently by subgroup

GENDER DIFFERENCES IN MST EFFECTS
10
(boys vs. girls) in measuring the underlying trait of depression (Bauer, 2017; Steinberg &
Thissen, 2006).
There are two types of DIF: uniform and nonuniform DIF. Uniform DIF occurs when
there are significant differences in item intercepts by grouping variable, whereas nonuniform
DIF occurs when there are significant differences in item loadings by grouping variable.
Practically speaking, uniform DIF occurs when individuals from one group are consistently more
or less likely to endorse a response to a specific item across levels of the underlying latent trait
(Ackerman, 1992; Montoya & Jeon, 2020). For example, uniform DIF would be present if girls
with varying levels of the underlying latent trait of delinquency are consistently less likely to
endorse violent behavior relative to boys with the same level of underlying delinquency.
Nonuniform DIF occurs when an item’s ability to discriminate among people with varying levels
of an underlying latent trait differs based on grouping variable (Ackerman 1992; Montoya &
Jeon, 2020). For example, an item asking about a particular type of delinquent behavior may
provide more information about an individual’s underlying level of delinquency depending on
gender; if boys with high and low levels of underlying delinquency are equally likely to endorse
getting suspended from school, but girls with high levels of underlying delinquency are more
likely than girls with low levels of delinquency to endorse getting suspended, this item would
show nonuniform DIF. A pictorial example (adapted from Montoya & Jeon, 2020) of uniform
and nonuniform DIF by intraclass correlation coefficient is provided in Figure 1 (adapted from
Hubert, 2017).

GENDER DIFFERENCES IN MST EFFECTS
11
Figure 1
Intraclass Correlation Curves (ICCs) Uniform vs. Nonuniform DIF
If statistical tests indicate that DIF is present in any of the measures used in the three
datasets to be combined for the current study, appropriate constraints will be applied to control
for DIF (and establish measurement invariance) when analyzing the three RCTs as an aggregate
sample. In sum, IDA is a useful framework that allows researchers to work with larger sample
sizes than they would have with individual studies, but it does require intensive statistical testing
to ensure that data are invariant from one another, i.e., they are appropriate to combine for
analysis as a single collective dataset. In addition to assessing for measurement invariance across
study, establishing measurement invariance by gender and timepoint is also necessary given the
primary research questions of whether treatment effects differ by gender and whether they are
sustained over time. I can only accurately assess whether MST effects differ by gender if
measures are working the same way for males and females, and I can only accurately assess
whether treatment effects are changing over time if outcome measures are functioning the same
way across time (i.e., the measures are invariant).

GENDER DIFFERENCES IN MST EFFECTS
12
Configural invariance is the least “strict” type of invariance; it is required but not
sufficient to establish measurement invariance across studies. Configural invariance refers to
when factor pattern matrices may differ across studies but correlate with the latent variable in the
same direction. The next type of invariance is metric invariance, which refers to when factor
patterns are identical but intercepts and residual variances may vary. The “strictest” type of
invariance is scalar invariance, which refers to when factor patterns and item intercepts are
identical across studies, which ensures that the latent variable is measured the same way and on
the same scale (Gregorich, 2006; Van de Schoot et al., 2012). Testing for scalar and metric
invariance is required to determine whether a measure shows differential item functioning (DIF).
In cases where there is not full metric or scalar invariance, partial invariance tests must be used
to identify subsets of items that have identical factor patterns and intercepts across all studies.
To analyze the three datasets pooled together as one larger dataset, partial scalar
invariance must be achieved across the three studies at a minimum. Partial scalar invariance tests
involve constraining factor structure, loadings, and intercepts that are similar across studies while
releasing the intercepts that differ (i.e., are noninvariant) across studies (Putnick & Bornstein,
2016). Partial scalar invariance tests require at least two intercepts to be the same across three
comparison groups (e.g., study).
To date, no studies have assessed gender as a moderator of intervention effects on
delinquent behavior using IDA, and only one study has used IDA to assess if gender moderates
intervention effects on substance use (Greenbaum et al., 2015). Across five trials of
Multidimensional Family Therapy, Greenbaum et al. (2015) found that gender did not moderate
substance use outcomes. Notably, the aggregate sample did not include exclusively JIY, nor did
it assess other outcomes related to substance use, such as delinquency and family relations.

GENDER DIFFERENCES IN MST EFFECTS
13
While there are published meta-analyses (e.g., van der Stouwe, 2014) and RCTs with large
sample sizes (e.g., Fonagy et al., 2018) assessing the effectiveness of MST with JIY, the current
study is the first to pool raw data using IDA to look at treatment outcomes and gender
differences in MST across multiple trials conducted in the U.S.
Study Aims and Hypotheses
The current study includes three aims. The first is to establish measurement invariance
across three RCTs of MST to ensure I can combine the data for analysis as one, “mega” dataset. I
hypothesize that given the similarity between study populations, eligibility criteria, and setting,
(detailed in the Method section), I will be able to establish measurement invariance across latent
outcome measures used in the three RCTs. The second aim is to assess whether MST is indeed
more effective than control conditions through one year follow-up when data from three studies
are combined using IDA. I hypothesize that the pooled sample will indicate that youth
randomized to MST vs. control will show reductions in self-reported delinquent behavior and
substance use, as well as improved family functioning, consistent with some reviews of MST
trials delivered in U.S. contexts (e.g., Curtis et al., 2004; Henggeler et al., 2016; Henggeler &
Schaeffer, 2010). The third and primary aim is to analyze this larger, pooled dataset to assess for
gender differences in intervention effects of MST on substance use, delinquency, and family
functioning. I have two competing hypotheses regarding gender as a moderator of MST effects
for JIY. The “universal treatment hypothesis” argues that existing interventions like MST
target general criminogenic needs and are broadly effective across different risk profiles. Thus,
one prediction is that males and females will benefit equally from MST. In contrast, the
“gender-responsive treatment hypothesis” argues that conventional interventions delivered in
typical juvenile justice contexts often neglect salient, gender-specific criminogenic needs of

GENDER DIFFERENCES IN MST EFFECTS
14
female offenders (e.g., higher rates of comorbid substance use and PTSD). Thus, an alternative
prediction is that males will benefit more than females from MST, as the intervention was
delivered in typical (rather than gender-responsive) juvenile justice contexts across all three
trials, and MST manuals do not explicitly address female-specific risks and needs related to
delinquency, substance use, and family functioning.

GENDER DIFFERENCES IN MST EFFECTS
15
Chapter 2: Method
Overview of Studies
Data from three RCTs of MST conducted with JIY in the U.S. (Henggeler et al., 1997,
Henggeler et al., 1999; Henggeler et al., 2006) were aggregated into a single “mega” dataset to
analyze collectively using IDA. The combined sample size is N=434, with n=81 females. See
Table 2 for details of participant characteristics, eligibility criteria, and comparison conditions by
study. Given the focus on differences in effects of MST compared to non-MST interventions, I
collapsed conditions in Study 3 to create an MST vs. non-MST comparison.
Table 2
Overview of MST RCTs Included in the Current Study
Study Characteristic Study 1
(Henggeler et al., 1997)
Study 2
(Henggeler et al., 1999)
Study 3
(Henggeler et al., 2006)
Treatment Conditions MST vs. Usual Services
(US)
MST vs. US Family Court + US vs.
Drug Court (DC) + US
vs. MST + DC vs. MST +
DC+ Contingency
Management
Location South Carolina, U.S. South Carolina, U.S. South Carolina, U.S.
N 155 118 161
% Female (n) 18.1% (28) 21% (25) 17% (27)
Age (M; SD) 15.22 (SD = 1.40) 15.70 (SD = 1.0) 15.2 (SD = 1.1)
Race/Ethnicity 80.6% African American,
19.4% Caucasian
50% African American,
47% Caucasian, 1%
Asian, 1% Native
American, 1% Hispanic
American
67% African American,
31% White, 2% Biracial
Arrests at baseline
(M; SD)
3.07 (2.07) 2.9 (1.9) 3.6 (2.5)
Eligibility Criteria Aged 11-17; committed a
serious prior offense or at
least three prior criminal
offenses aside from status
offenses
Aged 12-17; Diagnosis of
substance abuse or
dependence according to
DSM-III-R criteria;
informal or formal
probationary status
Aged 12-17; Diagnosis of
substance abuse or
dependence according to
DSM-IV criteria; informal
or formal probationary
status

GENDER DIFFERENCES IN MST EFFECTS
16
Timepoints included Baseline, posttreatment, 6
month follow-up*
Baseline, posttreatment, 6
month follow-up, 12
month follow-up*
Baseline, posttreatment,
12 month follow-up
Retention Rates 91.3% retention at
posttreatment; 61% at 6
months follow-up*
97.0% posttreatment,
91.5% 6 months followup; 87.4% at 12 months
follow-up*
For all self and parent
report data: 96.3% at
posttreatment; 83.9% at
12 months follow up.
*Denotes follow-up data timepoint not published in preliminary efficacy study but data that is accessible to the lab.
As outlined in Table 2, the baseline characteristics of study participants and eligibility
criteria are relatively consistent apart from the substance use criteria. Studies 2 and 3 required
that youth have a substance use disorder diagnosis, whereas study 1 did not. Nonetheless, 40% of
study 1 participants endorsed substance use in the past three months at baseline. All studies
include data from baseline, posttreatment, and at least one follow-up timepoint. Data collection
timepoints across each of the three studies are provided in Table 3. Study 2 is the only study that
collected data across all four timepoints, however, each timepoint included data from at least two
of the three studies in the aggregate sample. Of note, some of the data included in the current
study was not reported in the published manuscripts of the primary outcome papers for each trial
(marked with an * in Tables 2 and 3, e.g., 6-month follow-up data for study 1).
Table 3
Data Timepoints Included Across Studies
Individual Study Results
An overview of treatment outcomes from each individual trial is provided in Table 4.
There were no differences in self-reported delinquency at post-treatment or follow-up across all
three trials. Substance use findings were only reported in two of the three studies and findings
Timepoint Study 1 Study 2 Study 3
Baseline X X X
Posttreatment X X X
6 Month Follow-up X* X
12 Month Follow-up X* X

GENDER DIFFERENCES IN MST EFFECTS
17
were mixed. Study 2 found MST reduced self-reported marijuana and alcohol use at
posttreatment but not six month follow-up (Henggeler et al., 1999) compared to the control
condition. Study 3 found that only one of the two MST conditions showed reduced alcohol use
compared to only one of the two control groups at posttreatment and twelve month follow-up.
There were no intervention effects for marijuana use at posttreatment, but both MST conditions
led to reduced marijuana use compared to one of the two control conditions at twelve month
follow-up (Henggeler et al., 2006). Family functioning outcomes were only reported in study 1
and no treatment effects were observed. Moderating effects of gender were investigated in only
one study (Henggeler et al., 1999) which found that MST was more effective at reducing
substance use among females compared to services as usual at posttreatment but not follow-up;
for males, no treatment effects were found for substance use at any time point.
Table 4
Overview of Trial Outcomes
Outcome
Study 1
(Henggeler et al., 1997)
Study 2
(Henggeler et al., 1999)
Study 3
(Henggeler et al., 2006)
Self-Reported
Alcohol Use
Alcohol use data not
reported.
MST condition showed
significantly lower selfreported alcohol at
posttreatment but not 6
months follow-up compared
US.
DC and DC/MST/CM
conditions, but not the
DC/MST condition reported
significantly less alcohol use
compared to the FC
condition at posttreatment; at
12 months follow-up, only
the DC/MST/CM condition
showed significant
reductions in alcohol use
compared to the FC
condition.
Self-Reported
Marijuana Use
Marijuana use data not
reported.
MST condition showed
significantly lower marijuana
use at posttreatment but not
6 months follow-up
compared to US.
No treatment effects on
marijuana use at
posttreatment; at 12 months
follow-up, the DC/MST and
DC/MST/CM conditions
reported significantly less
marijuana use compared to
the FC condition but not the
DC condition.

GENDER DIFFERENCES IN MST EFFECTS
18
Self-Reported
Delinquency
No treatment effects of MST
on self-reported delinquency
at post-treatment.
No treatment effects of MST
on self-reported delinquency
at post-treatment or 6 months
follow-up.
No treatment effects on selfreported delinquency at
posttreatment; the
DC/MST/CM, DC/MST, and
DC conditions showed
significantly lower overall
self-reported delinquency at
12 months follow-up
compared to the FC
condition.
Family
Functioning
No treatment effects on self
and parent report of family
functioning.
Family functioning outcomes
not reported.
Family functioning outcomes
not reported.
Moderating
effect of gender?
Moderator effects not
examined.
MST females showed
significantly lower selfreported marijuana and
alcohol use compared to US
females assigned at
posttreatment, but MST
females deteriorated by
follow-up; females assigned
to US showed reduced selfreported marijuana and
alcohol use at follow-up but
not posttreatment. No other
moderator effects were
found.
Moderator effects not
examined.
Study Outcome Measures
Substance Use
The Personal Experiences Inventory (PEI; Winters & Henley, 1989) is a self-report
measure that assesses substance use over the last 90 days. Youth are asked to report the number
of days they have used any of the following twelve categories of substances: alcohol, marijuana,
LSD, other psychedelics, cocaine, amphetamines, quaaludes, barbiturates, tranquilizers, heroin,
other narcotics, and inhalants. Only data from items asking about marijuana and alcohol use
were available for use in the current study. Individual scores from these two items were treated
as distinct outcome variables (i.e., marijuana use, alcohol use). The item scores were transformed

GENDER DIFFERENCES IN MST EFFECTS
19
to proportion of maximum scores (POMS; procedure described in Data Transformation
subsection of the Method). Items from the PEI are provided in Appendix A.
Delinquency
The Self-Report Delinquency Scale (SRDS; Elliot & Huizinga, 1989) was used to
evaluate delinquent behavior. The SRDS is a 40-item self-report measure of the frequency and
severity of delinquent acts (e.g., larceny, fighting, selling drugs, etc.). Respondents are asked to
report how many times they have engaged in each delinquent behavior over the last 90 days.
Like the PEI, all SRDS items were converted to POMS. All items from this scale are aggregated
into a “general delinquency” index; the latent factor score was used in the current study. The full
SRDS scale shows an internal consistency (Cronbach’s alpha) of 0.91 (Pechorro et al., 2019).
The full SRDS is provided in Appendix A.
Family Functioning
The Family Adaptability and Cohesion Scale (FACES; Evenson et al., 1980; Olson et al.,
1980) was used to assess family functioning from both youth and caregiver perspectives. The
FACES is a two-factor measure that assesses family emotional bonding (cohesion) and ability to
change in response to situational stressors (adaptability). Items are rated on a 5-item Likert scale
from 1 (almost never) to 5 (almost always). Internal consistency ratings for FACES youth and
parent versions of the measure vary; some studies have found Cronbach’s alpha of 0.77 and 0.62,
respectively, for adaptability and cohesion scales with mixed samples of adolescents and parents
(e.g., Olson et al., 1986) but more recent studies show improved internal consistency ratings of
0.89 and 0.70 (e.g., Ide et al., 2010).
A confirmatory factor analysis (CFA) with the pooled dataset was conducted to establish
whether the two-factor FACES model was appropriate for use in the current study. The two-

GENDER DIFFERENCES IN MST EFFECTS
20
factor model yielded poor fit for both youth and parent versions of the measure. As such, items
were parceled (procedure described in detail in the next section) based on modification indices
(i.e., the estimate of how much the chi-square statistic of model fit would be improved if a
specific parameter restriction were removed from the model; MacCallum et al., 1992) within the
two factor CFA, which yielded excellent model fit for both the youth and parent versions of the
measure. Analyses from the original CFA and CFA after parceling items based on modification
indices are provided in Appendix A for both youth and parent versions of FACES. Parceled
items from each factor were summed to generate respective latent factor scores for study
outcome analyses.
Data Transformation Procedures
Before testing for measurement invariance, data from each of the three studies had to be
transformed to optimize analysis as a collective dataset. Measures that were “count” items (e.g.,
the SRDS and the PEI substance use items) were converted into Proportion of Maximum Scores
(POMS). When possible, manifest variables from each of the latent variables (i.e., FACES,
SRDS) were parceled, i.e., individual item scores were averaged into parcels in place of the item
scores (Bandalos, 2002). Procedures and rationale for these data transformation procedures are
provided in detail in the subsequent sections.
Converting Data to Proportion of Maximum Score (POMS)
The first transformation procedure was converting data to proportion of maximum score
(POMS). Both substance use items on the PEI (i.e., number of days used marijuana, number of
days drank alcohol) and all items on the SRDS scale ask youth how many times they have either
used a substance (for the PEI) or engaged in a specific delinquent behavior (SRDS) over the past
3 months. Response options are numerical but open-ended (i.e., responses could range from 0-

GENDER DIFFERENCES IN MST EFFECTS
21
1000). To ensure that each item was assessed on the same scale, I transformed scores from the
SRDS and substance use items from the PEI into POMS. I adjusted the scales on items from both
measures from 0-1000 to 0-1, where 0 was the minimum value and 1 represented the maximum
score (Little, 2013). For example, for one of the SRDS items, “stole a vehicle,” the minimum
observed score was 0 and the highest observed value was 18. All scores on the SRDS were then
transformed into POMS by dividing the observed value by the highest observed value (i.e., 18).
Thus, an original score of 1 on the SRDS for the “stole a vehicle” item would be transformed
into POMS of 0.056, or 1/18.
Parceling
Given that the pooled sample included data missing not at random (MNAR) there were
several inherent challenges related to missing data, including bias and nonnormal distributions.
Parceling scale items of outcome measures can address some of the difficulties present in
analyses with large amounts of missing data. Parceling involves averaging scores of multiple
items to create an aggregate indicator (e.g., averaging scores from three items from a scale to
create one, aggregate level item), so instead of the individual items, each parcel becomes the
indicator of the latent factor (Bandalos, 2008; Little et al., 2002, Little et al., 2022). For example,
a respondent who provided scores of 1, 1, 2, and 1 on FACES items 6, 8, 10, and 20 would have
a new parcel score of 1.25, which would then become the single indicator rather than four
individual indicators. Parceling items can yield greater reliability and communality, as well as
higher ratios of common vs. unique factor variance relative to using individual items as
indicators (Little et al., 2013; Little et al., 2022). Generally, models based on parceled data
compared to data using individual items are more parsimonious as they involve fewer parameter
estimates. They also are less likely to have correlated residuals (i.e., violation of the local

GENDER DIFFERENCES IN MST EFFECTS
22
independence assumption) or cross-factor loadings since parcels yield fewer indicators and
therefore lower unique variance, which reduces sampling error (Little et al., 2022). As such,
combining items into parcels often yields improved model fit and stability (Little et al., 2002).
For the current study, parceling allowed me to retain the original two-factor structure for the
FACES for the pooled data analyses, as the original two-factor structure did not fit the pooled
data without combining the items into parcels.
Ideally, each latent variable is composed of three parcels to yield a measurement model
that is just identified. Parcels do not have to contain an equal number of items (Little et al.,
2022). The primary purposes of this study were to determine whether MST was more effective
than control treatments with a pooled sample and whether gender moderated intervention effects
on overall latent factors (e.g., overall delinquency) rather than on individual item scores. As
such, I parceled items to minimize problems related to violations of local independence and to
facilitate simple structure of the items used to estimate constructs of interest.
In the current study, I attempted to parcel both the FACES and SRDS measures using the
balancing technique, which involves averaging pairs of items that have the highest modification
indices with pairs of items that have the smallest modification indices (Little et al., 2002). Tests
of model fit were conducted (e.g., chi square test, Comparative Fit Index, Root Mean Square
Error of Approximation) to confirm whether measures yielded improved model fit when
parceled. Tests indicated that parceling the FACES measure from a 2 factor, 20 item scale into a
2 factor, 6 parcel scale yielded improved model fit for both youth and parent report versions of
the measure. Accordingly, both the adaptability and the cohesion factors on the FACES measure
contained the optimum number of parcels (i.e., just identified with three factors) but the overall
measure contained six parcels. A summary of which FACES items were allocated to parcels 1-3

GENDER DIFFERENCES IN MST EFFECTS
23
for the adaptability and cohesion factors for both youth and self-report versions is provided in
Appendix A. In contrast to the FACES measure, parceling items from the SRDS did not yield
improved fit; as such, I kept the SRDS as a single factor, 40-item scale. Results from the final
CFA of the unidimensional SRDS and the two-factor FACES measure with original items
compared to the parceled items for both youth and parents are provided in Appendix A.
Moderated Nonlinear Factor Analysis
I used moderated nonlinear factor analysis (MNLFA) to assess whether each latent
construct (i.e., SRDS, FACES) was measured in the same way across studies, measurement
occasion (i.e., timepoint), and gender. This step in my analysis ensured that these variables do
not influence heterogeneity in the measurement of delinquency and family functioning. As an
alternative to group-based tests of measurement invariance, MNLFA is ideal for handling a
mixture of variables that might affect heterogeneity of measurement across study, timepoint, and
gender (Bauer, 2017; Curran et al., 2014). MNFLA allows for all parameters (i.e., factor
loadings, variances) to differ as a function of known individual characteristics and for testing of
multiple cause/multiple indicators that may influence measurement invariance, as shown in
Figure 2 (adapted from Gottfredson et al., 2019). One of the advantages that MNFLA has over
traditional group-based measurement invariance approaches is that it allows for factor means,
factor variances, and factor correlations to vary as a function of external causes while testing
uniform and nonuniform DIF. For example, MNFLA can allow for parameter estimates to vary
by treatment assignment (e.g., MST vs. control) while testing for uniform and nonuniform DIF
by multiple grouping or predictor variables (e.g., study membership, timepoint, and gender). In
the present study, the MNFLA approach permitted simultaneous assessment of whether measures

GENDER DIFFERENCES IN MST EFFECTS
24
were invariant by study, in addition to gender and timepoint, which allowed for pooling data
from three trials into a single data set.
Figure 2
Path Diagram for MNFLA Model
Note. A common factor or factors is represented by η. Items that load onto the factor are
represented by y1-y4. Factor loadings are depicted by λ and intercepts are depicted by τ. The
predictor variable(s) are represented by x. Factor variance is denoted by ψ and the intercept is
denoted by α. Additionally, local dependence between items can be accommodated by allowing
for residual correlations between items; an example of a residual correlation is depicted between
y1 and y2 in the figure. Figure is adapted from Gottfredson et al., 2019.
All procedures assessing for measurement invariance were done in MPlus version 8
(Muthén & Muthén, 2017). Models were fit using maximum likelihood with robust standard
errors (MLR). To establish measurement invariance, hundreds of tests were conducted to assess
for metric and scalar invariance by gender, timepoint, and study across parameters for items,
parcels, and/or factors for each of the outcome variables included in the current study. As such,
controls for multiple comparisons were needed to avoid Type I error (i.e., false positives).
However, traditional controls for multiple comparisons, such as Bonferroni corrections, can be
too conservative and increase Type II error, i.e., false negatives. In the current study, missing

GENDER DIFFERENCES IN MST EFFECTS
25
false negatives could lead to under-identification of DIF, which could threaten the validity of
findings from measurement invariance procedures. To balance type I and type II error (i.e., to
identify as many true positives while maintaining a low likelihood of false positives), I
controlled for the false discovery rate, defined as the rate of significant results that are in fact
null. Using false discovery rates adjusts for the actual p-value distribution of the data rather than
using the standard alpha of 0.05 (Benjamini & Hochberg, 1995).
Fitting MNFLA
I followed the steps outlined in both Curran et al. (2014) and Bauer (2017) suggested for
fitting MNFLA models to establish measurement invariance and assess for DIF. The first step
suggested is to determine the factor structure of the measures to be tested. To complete this first
step, I consulted the literature on prior studies of the factor structures of the measures I used and
analyzed preliminary data. I was able to fit the 20-item FACES self-report and parent-report data
into the established two-factor structure documented in prior studies (i.e., one adaptability factor
and one cohesion factor; Henggeler et al., 1992; Henggeler et al., 1997) by parceling items based
on modification index; the adaptability factor consisted of three parcels, and the cohesion factor
also consisted of three parcels. Studies show items from the SRDS are typically combined to
create a unidimensional “general delinquency” index (Henggeler et al., 1997; Henggeler et al.,
1999; Henggeler et al., 2006; Huey et al., 2000; Piquero et al., 2002). As such, I analyzed the
SRDS as a unidimensional “general delinquency” scale but was not able to parcel any of the
items due to issues with model fit, as noted earlier in the Method section.
I then fit MNFLA models separately to each factor. For the SRDS measure, I fit MNFLA
models for each of the 40 items on the scale. For the youth and parent-reported FACES
measures, I fit the MNFLA models to the bi-dimensional FACES factor structure that consisted

GENDER DIFFERENCES IN MST EFFECTS
26
of three parcels within the Cohesion factor and three parcels within the Adaptability factor. I
used MNFLA to determine whether each parcel (i.e., on the FACES measures) or item (i.e.,
SRDS) showed DIF by gender, timepoint, or study. Baseline factor means and variances were
constricted to 1 and 0, respectively. When loadings were significant according to each of these
variables (i.e., gender, timepoint, and study), that indicated the item or parcel had nonuniform
DIF and violated scalar invariance. When intercepts were significant, that indicated the item or
parcel had uniform DIF and violated metric invariance. If MNFLA results showed items or
parcels with significantly different intercepts or loadings (i.e., presence of uniform and
nonuniform DIF, respectively) by grouping variable (gender, timepoint, study), that indicated
that the parcel or item needed to be released to ensure measurement invariance for analysis as a
collective dataset.
Analysis as a Collective Dataset
After measurement invariance procedures were complete, a new dataset was generated
combining the raw data from all three RCTs. This new dataset included adjusted outcome scores
after constraining intercepts and loadings of items and parcels on the SRDS and FACES
measures that were invariant across studies and releasing intercepts and loadings that showed
DIF based on MNFLA results, while constraining baseline latent mean factor scores and
variances to 1 and 0, respectively. To achieve Aims 2 and 3, linear mixed models (LMMs) with
maximum likelihood estimation for missing data were used to evaluate whether MST was more
effective than Usual Services at treating substance use, family functioning, and delinquency
through 1 year follow-up, and whether effects were moderated by gender.
LMMs are a statistical technique used for modeling continuous outcome variables in the
presence of clustered or longitudinal data (e.g., repeated measures within individuals, like in the

GENDER DIFFERENCES IN MST EFFECTS
27
current study). They extend the traditional linear regression model by incorporating both fixed
effects (predictors that are assumed to be constant across all levels of the clustering variable) and
random effects (random variability associated with different levels of the clustering variable).
LMMs are superior in handling missing data relative to other analyses, such as repeated
measures ANOVA (Cnaan et al., 2015; McCullagh & Nelder, 2019).
In LMMs, random intercepts and slopes are used to account forsubject-specific baseline
differences and changesin outcomes over time (McCullagh & Nelder, 2019). For the pooled
dataset in this proposal, assignment to treatment condition and gender (moderator) were
classified as fixed effects interaction variables at the individual level (i.e., between individuals,
level 2). Change within individuals (i.e., across timepoints, level 1) was classified as a random
effect.
Robust LMMs were used for analyzing outcome data that were not normally distributed,
i.e., POMS scored data, including the SRDS and self-reported alcohol and marijuana use. Robust
estimation techniques provide more reliable parameter estimates with non-normal data
distributions and missing data. All LMMs and robust LMMs were conducted in R using the lme4
(Bates, 2007) and rlmm (Koller, 2016) packages.
Power
Although providing power analyses post-hoc for secondary data analyses is not typical
practice, the primary purpose of the current study was to combine datasets to generate sufficient
power needed to test for gender differences in treatment effects relative to other studies; as such,
a power analysis is provided here. The following calculations are the estimated number of
females needed to detect gender differences in effects of MST assuming the following: 1) four
total assessment time points, 2) a conservative estimate of 80% retention across datapoints, and

GENDER DIFFERENCES IN MST EFFECTS
28
3) random assignment to one of two treatment conditions. For LMMs, power was determined
using the simR package in R, including 1000 replications with group membership, gender, and
time. Coefficients of the longitudinal indicators of outcome with the intercept and slope were set
at 0.50. Direct effects of treatment and gender were also set to 0.50 (a conservative estimate).
Estimates of the number of females needed in a mixed-gender sample (assuming at least as many
males) to detect gender differences in treatment effects with 80% power and alpha at 0.05 by
minimum detectable effect size are outlined in Table 5. The current study allows for detecting
small effect size differences in treatment effects by gender (d=0.22) with 81 females, whereas
moderator analyses in the original studies only had sufficient females to detect medium to large
effect size differences between males and females.
Table 5
Required Number of Females in Sample Needed to Achieve Specified Power
Females Minimum Detectable Effect Size (Cohen’s d)
100 0.2
46 0.3
26 0.4

GENDER DIFFERENCES IN MST EFFECTS
29
Chapter 3: Results
Measurement Invariance Testing – Moderated Nonlinear Factor Analysis
Results from MNFLA of each latent measure are presented in tables 6-11. The tables
show means, variances, intercepts, and loadings of factors, parcels or items on the FACES youth
self-report, FACES parent-report, and SRDS measures. A minimum of partial scalar invariance
was established (i.e., at least two intercepts were the same across studies) for each of these
measures. Items or parcels with intercepts significantly different by grouping variable showed
uniform DIF and items or parcels with loadings significantly different by grouping variable
showed nonuniform DIF. Italicized p values indicate parameters that were significant at the 0.05
level before controlling for false discovery rates; bolded p values indicate parameters that were
significantly different after controlling for false discovery rates for item and parcel level testing
(provided in tables 6, 8 and 10). A written summary of the items or parcels on the FACES and
SRDS measures that exhibited DIF is provided in the subsequent section. All results are reported
for item or parcel-level MNFLA testing; for final model fit only significant findings are reported.
FACES – Youth
The bifactor FACES model assessing youth reported family adaptability and cohesion
showed both uniform and nonuniform DIF. All three parcels (1, 2, and 3) of the adaptability
factor showed nonuniform DIF by study, indicating that studies varied in how well each parcel
discriminated between individuals with different levels of latent youth-reported family
adaptability. For later studies (studies 1, 2, and 3 are in chronological order), scores on parcels 1
and 3 were more discriminating of overall youth-reported family adaptability, whereas scores on
parcel 2 were more discriminating of overall youth-reported family adaptability for earlier
studies. Parcels 2 and 3 of the adaptability factor also showed nonuniform DIF by timepoint,

GENDER DIFFERENCES IN MST EFFECTS
30
which indicates that these parcels varied in how well they discriminated overall family
adaptability levels from the youth’s perspective over time. Generally, parcel 2 was better at
discriminating levels of underlying youth reported family adaptability in later timepoints (e.g.,
follow-up) than for earlier timepoints (e.g., baseline); the opposite was true for parcel 3.
For the cohesion factor, parcels 1 and 3 showed nonuniform DIF by study, indicating that
these parcels varied in their ability to discriminate between underlying levels of youth reported
family cohesion by study. Parcel 1 was more discriminant of underlying cohesion for earlier
studies whereas Parcel 3 showed the opposite pattern. Parcel 2 showed uniform DIF by study; in
Study 1, participants across all levels of underlying family cohesion had higher scores on parcel
2 relative to participants in other studies. Parcel 3 of the cohesion factor also showed uniform
DIF by timepoint; parcel 3 scores were generally higher across all levels of underlying family
cohesion at later timepoints. Across both youth FACES adaptability and cohesion parcels, there
was no DIF by gender, indicating that male and female JIY with similar levels of underlying
self-reported family adaptability and cohesion provided similar responses or parcel scores. All
results from the parcel-level analyses of the youth report of FACES are provided in Table 6.
When fitting the final MNFLA model based on parcel-level analyses controlling for the
various sources of DIF by grouping variable across both factors and all parcels, there were no
significant differences between the means of the Adaptability and Cohesion factors, although
there were significant differences in factor variance of Adaptability by study and Cohesion by
timepoint. Parcels retained for fitting the final MNFLA model (i.e., parcels that showed DIF) and
significant factor means, variances, and covariances are provided in Table 7.
FACES – Parent

GENDER DIFFERENCES IN MST EFFECTS
31
Fitting the bidimensional FACES measure to parent-report of family adaptability and
cohesion showed both uniform and nonuniform DIF. For the adaptability factor, there was
nonuniform DIF (i.e., loadings were significantly different) by study for parcels 1 and 2, and
uniform DIF by study for parcel 3. For earlier studies (studies 1, 2, and 3 are in chronological
order), scores on parcel 1 were more discriminating of overall parent-reported family
adaptability. The opposite pattern was true for parcel 2; results showed that parcel 2 was more
discriminant in detecting underlying parent reported adaptability in later studies. For parcel 3,
participants in later studies were generally more likely to have higher scores across varying
levels of underlying parent reported adaptability. There was also nonuniform DIF by timepoint
for parcels 1 and 3 of the adaptability factor; parcel 1 showed higher discriminability across
follow-up (i.e., later) timepoints compared to earlier (e.g., baseline) timepoints, whereas the
opposite was true for parcel 3.
The cohesion factor on the parent-reported FACES measure showed nonuniform DIF by
study on parcels 1 and 2; scores on parcels 1 and 2 were more discriminant of underlying family
cohesion based on parent perspectives in earlier (e.g., study 1) compared to later studies. Parcel 1
also showed uniform DIF by timepoint; across all levels of underlying parent-reported family
cohesion, scores at earlier timepoints were generally higher for parcel 1. Parcel 2 of the Cohesion
factor showed uniform DIF by gender; parents of females across all levels of underlying family
cohesion generally scored higher on this parcel compared to males with the same underlying
parent-reported family cohesion. Parcel 3 showed no DIF by grouping variable for parentreported cohesion. All results from the parcel-level analyses are provided in Table 8.
After controlling for the various sources of DIF by grouping variable across both factors
and all parcels, timepoint significantly influenced the variance of the Adaptability latent factor,

GENDER DIFFERENCES IN MST EFFECTS
32
and gender influenced the variance of the Cohesion latent factor. There were no significant
differences in factor covariance. Parcels retained for the final MNFLA model (i.e., parcels that
showed DIF) and significant factor means, variances, and covariances are provided in Table 9.
SRDS
Across all 40 items of the SRDS, the following items showed nonuniform DIF by gender:
item 3 (purposely damaged or destroyed property belonging to school), item 13 (been paid for
having sexual relations with someone), item 17 (cheated on school tests), item 28 (used force
[strong-arm methods] to get money or things from other students), and item 39 (been suspended
from school). All of these items were less discriminating of overall underlying delinquency
among females compared to males.
Items 5 (stolen [or tried to steal] something worth more than $50), 8 (run away from
home), 21 (hit [or threatened to hit] one of your parents), and 35 (broken into a building or
vehicle [or tried to break in] to steal something or just look around) showed uniform DIF by
gender. Generally, males with similar levels of overall delinquency compared to females were
more likely to endorse trying to steal something more than $50, hitting or threatening to hit their
parents, and breaking into a building or vehicle relative to females. Females with similar levels
of overall delinquency compared to males were more likely to endorse running away.
None of the items on the SRDS showed nonuniform DIF by study. Only item 16 (sold
marijuana or hashish [“pot,” “grass,” “hash.”]) showed uniform DIF by study. Participants in
earlier studies were generally less likely to endorse selling marijuana across all levels of
underlying delinquency relative to later studies.
The following items showed nonuniform DIF by timepoint: Items 6 (knowingly bought,
sold, or held stolen goods [or tried to do any of these things]), 17 (cheated on school tests), 27

GENDER DIFFERENCES IN MST EFFECTS
33
(had [or tried to have] sexual relations with someone against their will, 29 (used force [strongarm methods] to get money or things from a teacher or other adult at school) and 35 (broken into
a building or vehicle [or tried to break in] to steal something or just look around). Items 6, 17,
and 35 were more discriminating of latent delinquency levels at earlier timepoints relative to
later timepoints. In contrast, follow-up scores on items 27 and 29 were more discriminant of
underlying delinquency relative to scores from earlier timepoints.
Items 16 (sold marijuana or hashish [“pot,” “grass,” “hash.”]), 22 (hit [or threatened to
hit] other students), 23 (been loud, rowdy, or unruly in a public place [disorderly conduct]), 28
(used force [strong-arm methods] to get money or things from other students), 30 (used force
[strong-arm methods] to get money or things from other people [not students or teachers]), 37
(skipped classes without an excuse), and 39 (been suspended from school) showed uniform DIF
by timepoint. Respondents with the same levels of underlying delinquency were generally more
likely to endorse items 16, 22, 37, and 39 at earlier timepoints than later timepoints, whereas the
opposite was true for items 23, item 28, and item 30. Overall, timepoint and study, but not
gender, influenced the means and variances of latent delinquency scores. All results from the
item-level analyses are provided in Table 10; items retained for the final MNFLA model (i.e.,
items that showed DIF) and significant factor mean, variance, and covariance are provided in
Table 11.

GENDER DIFFERENCES IN MST EFFECTS
34
Table 6
FACES Youth Report Parcel Level MNFLA Results
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Adaptability
Parcel 1
Intercept 3.152 0.088 <.001 0.071 0.072 .325 -0.021 0.022 .347 0.064 0.044 .150
Loading 0.401 0.104 <.001 -0.034 0.086 .693 -0.009 0.041 .822 0.222 0.057 <.001
Parcel 2
Intercept 2.957 0.115 <.001 -0.005 0.052 .930 0.034 0.019 .064 -0.025 0.041 .536
Loading 0.867 0.078 <.001 -0.024 0.052 .638 0.052 0.018 .005 -0.274 0.043 <.001
Parcel 3
Intercept 2.9 0.081 <.001 -0.023 0.06 .700 -0.033 0.02 .095 0.014 0.041 .731
Loading 0.766 0.07 <.001 0.061 0.072 .402 -0.076 0.023 <.001 0.197 0.05 <.001
Cohesion
Parcel 1
Intercept 1.826 0.141 <.001 -0.036 0.103 .725 -0.024 0.026 .363 0.946 0.078 <.001
Loading 0.751 0.143 <.001 -0.034 0.097 .722 0.061 0.027 .023 -0.481 0.071 <.001
Parcel 2
Intercept 2.524 0.075 <.001 -0.047 0.077 .537 -0.023 0.03 .434 -0.502 0.093 <.001
Loading 0.38 0.099 <.001 0.12 0.071 .090 0.025 0.022 .260 0.037 0.054 .492
Parcel 3
Intercept 2.478 0.083 <.001 -0.019 0.09 .836 0.073 0.029 .011 -0.534 0.425 .209
Loading 0.701 0.176 <.001 -0.006 0.083 .945 -0.073 0.077 .343 0.154 0.044 <.001
Table 7
FACES Youth Report Final MNFLA Model
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Adaptability Factor
Mean 0 - - - - - - - - - - -
Variance 1 - - - - - - - - 1.305 0.153 <.001

GENDER DIFFERENCES IN MST EFFECTS
35
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error P
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Cohesion Factor
Mean 0 - -
Variance 1 - - - - - 0.184 0.058 .001 - - -
Factor Covariance
Fisher’s z - - - - - - - - - - - -
Adaptability
Parcel 1
Intercept 3.152 0.088 <.001 - - - - - - - - -
Loading 0.401 0.104 <.001 - - - - - - 0.197 0.05 <.001
Parcel 2
Intercept 2.957 0.115 <.001 - - - - - - - - -
Loading 0.867 0.078 <.001 - - - 0.052 0.018 .005 -0.274 0.043 <.001
Parcel 3
Intercept 2.9 0.081 <.001 - - - - - - - - -
Loading 0.766 0.07 <.001 - - - -0.076 0.023 <.001 0.197 0.05 <.001
Cohesion
Parcel 1
Intercept 1.826 0.141 <.001 - - - - - - 0.946 0.078 <.001
Loading 0.751 0.143 <.001 - - - - - - -0.481 0.071 <.001
Parcel 2
Intercept 2.524 0.075 <.001 - - - - - - -0.502 0.093 <.001
Loading 0.38 0.099 <.001 - - - - - - - - -
Parcel 3 - - - - - - - - -
Intercept 2.478 0.083 <.001 - - - 0.073 0.029 .011 - - -
Loading 0.701 0.176 <.001 - - - - - - 0.154 0.044 <.001

GENDER DIFFERENCES IN MST EFFECTS
36
Table 8
FACES Parent Report Parcel Level MNFLA Results
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Adaptability
Parcel 1
Intercept 3.586 0.118 <.001 0.002 0.077 .976 -0.008 0.021 .688 -0.487 0.077 <.001
Loading 0.796 0.105 <.001 0.057 0.072 .430 0.048 0.019 .010 -0.641 0.062 <.001
Parcel 2
Intercept 3.618 0.07 <.001 0.013 0.061 .836 0.003 0.018 .854 0.093 0.06 .117
Loading 0.701 0.083 <.001 -0.08 0.078 .306 0.003 0.026 .907 0.161 0.046 <.001
Parcel 3
Intercept 3.333 0.089 <.001 -0.039 0.055 .474 -0.01 0.018 .571 0.166 0.047 <.001
Loading 0.741 0.069 <.001 0.011 0.077 .884 -0.073 0.018 <.001 0.057 0.042 .177
Cohesion
Parcel 1
Intercept 1.383 1.103 <.001 0.018 0.088 .836 -0.058 0.021 .006 1.353 0.049 <.001
Loading 0.508 0.097 <.001 -0.055 0.097 .571 -0.009 0.026 .724 -0.317 0.042 <.001
Parcel 2
Intercept 2.33 0.08 <.001 -0.039 0.069 <.001 -0.054 0.025 .029 0.255 0.053 <.001
Loading 0.257 0.103 .013 -0.013 0.035 .711 0.006 0.012 .641 -0.186 0.05 <.001
Parcel 3
Intercept 2.496 0.101 <.001 0.044 0.098 .657 0.038 0.034 .258 -0.658 0.429 .125
Loading 0.387 0.081 <.001 -0.014 0.025 .574 -0.01 0.009 .291 -0.052 0.047 .266

GENDER DIFFERENCES IN MST EFFECTS
37
Table 9
FACES Parent Report Final MNFLA Model
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Adaptability Factor
Mean 0 - - - - - - - - - - -
Variance 1 - - - - - 0.098 0.043 .024 - - -
Cohesion Factor
Mean 0 - - - - - - - - - - -
Variance 1 - - 0.426 0.212 .039 - - - - - -
Factor Covariance
Fisher’s z - - - - - - - - - - - -
Adaptability
Parcel 1
Intercept 3.586 0.118 <.001 - - - - - - -0.487 0.077 <.001
Loading 0.796 0.105 <.001 - - - 0.048 0.019 .010 -0.641 0.062 <.001
Parcel 2
Intercept 3.618 0.07 <.001 - - - - - - - - -
Loading 0.701 0.083 <.001 - - - - - - 0.161 0.046 <.001
Parcel 3
Intercept 3.333 0.089 <.001 - - - - - - 0.166 0.047 <.001
Loading 0.741 0.069 <.001 - - - -0.073 0.018 <.001 - - -
Cohesion
Parcel 1
Intercept 1.383 1.103 <.001 - - - -0.058 0.021 .006 1.353 0.049 <.001
Loading 0.508 0.097 <.001 - - - - - - -0.317 0.042 <.001
Parcel 2
Intercept 2.33 0.08 <.001 -0.039 0.069 <.001 0.255 0.053 <.001
Loading - - - - - - -0.186 0.05 <.001

GENDER DIFFERENCES IN MST EFFECTS
38
Table 10
SRDS Item Level MNFLA Results
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Item 1: Purposely damaged or destroyed property belonging to parents or other family
Intercept 0.03 0.008 <.001 0.009 0.006 .160 -0.004 0.003 .115 0.009 0.005 .072
Loading 0.03 0.015 .049 0.000 0.014 .983 0.005 0.007 .483 0.003 0.010 .733
Item 2: Purposely damaged or destroyed property belonging to a school
Intercept 0.023 0.007 .003 0.000 0.006 .958 0.001 0.002 .599 0 0.003 .070
Loading 0.03 0.013 .022 0.014 0.013 .282 -0.003 0.004 .448 -0.002 0.007 .811
Item 3: Purposely damaged or destroyed other property that did not belong to you (not counting family or school property)
Intercept 0.017 0.004 <.001 0.009 0.004 .051 -0.003 0.002 .085 0.003 0.003 .316
Loading 0.022 0.007 .014 0.026 0.01 .007 -0.007 0.003 .026 0.002 0.006 .722
Item 4: Stolen (or tried to steal) a motor vehicle, such as a car or motorcycle
Intercept 0.007 0.005 .216 0.008 0.003 .025 0.003 0.002 .116 0 0.003 .909
Loading 0.031 0.012 .012 0.005 0.007 .423 -0.001 0.004 .848 -6 0.006 .332
Item 5: Stolen (or tried to steal) something worth more than $50
Intercept 0.016 0.007 .026 0.016 0.006 .012 0.003 0.003 .381 0.001 0.005 .869
Loading 0.031 0.019 .105 0.012 0.019 .515 -0.005 0.006 .460 0.01 0.012 .389
Item 6: Knowingly bought, sold or held stolen goods (or tried to do any of these things)
Intercept 0.043 0.01 <.001 0.006 0.007 .414 -0.007 0.003 .007 0 0.004 .944
Loading 0.043 0.014 .003 0.009 0.017 .592 -0.011 0.004 .005 -0.001 0.007 .893
Item 7: Thrown objects (such as rocks, snowballs, or bottles) at cars or people
Intercept 0.015 0.008 .051 0.003 0.006 .578 -0.001 0.001 .424 0.003 0.004 .498
Loading 0.008 0.016 .613 -0.002 0.012 .844 0.001 0.003 .669 0.012 0.009 .153
Item 8: Run away from home
Intercept 0.026 0.006 <.001 -0.018 0.006 .003 0.004 0.003 .107 0.003 0.002 0.161
Loading 0.024 0.012 .036 -0.014 0.012 .228 0.009 0.006 .127 -0.001 0.004 0.795
Item 9: Lied about your age to gain entrance or to purchase something; for example, lying about your age to buy liquor or get into a movie
Intercept 0.009 0.005 .090 0.003 0.004 .530 -0.001 0.002 .558 0.006 0.003 .023
Loading 0.006 0.011 .604 0.003 0.007 .662 0.001 0.005 .757 0.006 0.006 .328

GENDER DIFFERENCES IN MST EFFECTS
39
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Item 10: Carried a hidden weapon other than a plain pocket knife
Intercept 0.029 0.009 .002 -0.006 0.008 .428 0.002 0.002 .449 -0.005 0.003 .133
Loading 0.069 0.025 .005 -0.035 0.019 .066 -0.002 0.003 .517 -0.007 0.005 .201
Item 11: Stolen (or tried to steal) things worth $5 or less
Intercept 0.022 0.005 <.001 -0.002 0.004 .645 -0.002 0.002 .295 -0.003 0.002 .170
Loading 0.034 0.013 .007 0.007 0.011 .520 0 0.004 .929 -0.012 0.006 ..047
Item 12: Attacked someone with the idea of seriously hurting or killing him/her
Intercept 0.024 0.007 <.001 -0.004 0.005 .371 0.003 0.002 .082 -0.003 0.003 .291
Loading 0.045 0.02 .022 -0.013 0.013 .329 -0.002 0.005 .635 0 0.007 .992
Item 13: Been paid for having sexual relations with someone
Intercept 0.003 0.006 .639 0.001 0.004 .868 0.002 0.002 .214 0.001 0.003 .817
Loading -0.007 0.013 .560 0.015 0.006 .014 0.003 0.005 .538 0.005 0.011 .633
Item 14: Had sexual intercourse with a person of the opposite sex other than your spouse or significant others
Intercept 0.016 0.01 .114 0.002 0.01 .820 0.001 0.002 .660 0.003 0.002 .152
Loading 0.008 0.011 .441 0.001 0.01 .908 0.006 0.003 .096 -0.002 0.004 .581
Item 15: Been involved in gang fights
Intercept 0.027 0.09 .003 0.006 0.006 .277 -0.001 0.002 .472 -0.004 0.003 .285
Loading 0.055 0.024 .022 -0.009 0.015 .557 -0.007 0.004 .085 -0.007 0.008 .378
Item 16: Sold marijuana or hashish ("pot," "grass," "hash")
Intercept 0.011 0.008 .142 0.01 0.005 .061 -0.005 0.002 .018 0.009 0.003 .004
Loading 0.061 0.024 .010 -0.02 0.016 .212 -0.01 0.004 .025 0.001 0.007 .884
Item 17: Cheated on school tests
Intercept 0.03 0.007 <.001 0.007 0.006 .189 -0.007 0.002 <.001 0.001 0.003 .771
Loading 0.009 0.007 .210 0.02 0.007 .008 -0.006 0.002 .015 0.004 0.006 .509
Item 18: Hitchhiked where it was illegal to do so
Intercept 0.013 0.007 .048 0.001 0.005 .806 0.003 0.002 .134 -0.003 0.003 .294
Loading 0.021 0.016 .206 0.012 0.012 .308 0.005 0.005 .330 -0.006 0.008 .471
Item 19: Stolen money or other things from your parents or other members of your family
Intercept 0.019 0.007 .007 -0.008 0.006 .204 -0.001 0.001 .416 0.002 0.002 .462
Loading 0.019 0.009 .044 0.011 0.011 .333 -0.002 0.004 .589 -0.004 0.006 .521

GENDER DIFFERENCES IN MST EFFECTS
40
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Item 20: Hit (or threatened to hit) a teacher or other adult at school
Intercept 0.031 0.007 <.001 -0.003 0.006 .658 -0.003 0.002 .105 -0.002 0.003 .522
Loading 0.03 0.015 .046 -0.003 0.012 .832 0 0.004 .997 -0.001 0.006 .925
Item 21: Hit (or threatened to hit) one of your parents
Intercept 0.034 0.008 <.001 -0.023 0.008 .004 -0.001 0.002 .648 0.002 0.003 .450
Loading 0.048 0.015 .001 -0.033 0.018 .062 -0.001 0.004 .883 0.002 0.007 .787
Item 22: Hit (or threatened to hit) other students
Intercept 0.064 0.009 <.001 -0.007 0.006 .218 -0.006 0.002 .002 -0.008 0.003 .024
Loading 0.052 0.011 <.001 0.003 0.009 .752 -0.004 0.003 .159 -0.008 0.005 .160
Item 23: Been loud, rowdy, or unruly in a public place (disorderly conduct)
Intercept 0.006 0.001 .001 -0.015 0.01 .115 0.006 0.002 .002 0.004 0.003 .152
Loading 0.077 0.029 .007 -0.041 0.022 .066 0.002 0.004 .573 0.001 0.005 .848
Item 24: Sold hard drugs, such as heroin, cocaine, and LSD
Intercept 0.017 0.009 .044 0.006 0.005 .213 0.001 0.002 .671 -0.006 0.004 .123
Loading 0.067 0.025 .007 -0.025 0.016 .123 -0.003 0.003 .314 -0.013 0.007 .067
Item 25: Taken a vehicle for a ride (drive) without the owner's permission
Intercept 0.014 0.006 .017 0.007 0.004 .065 0.001 0.002 .506 0 0.003 .884
Loading 0.029 0.012 .013 -0.003 0.007 .673 0.002 0.004 .705 0 0.006 .972
Item 26: Bought or provided liquor for a minor
Intercept 0.006 0.006 .319 0.006 0.005 .177 0.003 0.002 .205 0 0.003 .881
Loading 0.017 0.012 .170 0.019 0.01 .062 0 0.006 .981 -0.003 0.008 .726
Item 27: Had (or tried to have) sexual relations with someone against their will
Intercept 0.003 0.004 .397 0 0.003 .952 0.004 0.002 .109 -0.001 0.001 .603
Loading -0.011 0.009 .223 0.008 0.005 .103 0.015 0.006 .019 0.003 0.005 .502
Item 28: Used force (strong-arm methods) to get money or things from other students
Intercept 0.01 0.006 .108 0.002 0.005 .598 0.009 0.002 <.001 -0.005 0.003 .057
Loading 0.014 0.014 .308 0.026 0.008 .002 0.01 0.006 .071 -0.002 0.01 .885
Item 29: Used force (strong-arm methods) to get money or things from a teacher or other adult at school
Intercept 0.001 0.003 .582 0.002 0.002 .321 0.005 0.002 .003 -0.001 0.001 .398
Loading -0.007 0.01 .491 0.012 0.006 .040 0.016 0.006 .010 -0.001 0.006 .857

GENDER DIFFERENCES IN MST EFFECTS
41
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Item 30 Used force (strong-arm methods) to get money or things from other people (not students or teachers)
Intercept 0.008 0.006 .165 0.006 0.005 .216 0.008 0.003 .004 -0.002 0.003 .615
Loading 0.018 0.016 .251 0.001 0.013 .943 0.014 0.006 .029 0.013 0.011 .224
Item 31: Avoided paying for such things as movies, bus or subway rides, and food
Intercept 0.016 0.007 .019 0 0.004 .929 0.001 0.002 .682 0.002 0.003 .461
Loading 0.035 0.018 .050 -0.003 0.009 .777 -0.002 0.005 .678 0.004 0.008 .620
Item 32: Been drunk in a public place
Intercept 0.025 0.009 .006 -0.002 0.007 .760 0.003 0.002 .113 0 0.003 .874
Loading 0.043 0.028 .124 -0.011 0.023 .625 0.004 0.005 .351 -0.004 0.007 .618
Item 33: Stolen (or tried to steal) things worth between $5 and $50
Intercept 0.042 0.009 <.001 -0.012 0.007 .097 -0.003 0.003 .197 0.001 0.004 .697
Loading 0.041 0.026 0.113 -0.006 0.024 .800 -0.001 0.006 .801 0.008 0.009 .364
Item 34: Stolen (or tried to steal) something at school, such as someone's coat from a classroom, locker, or cafeteria, or a book from the library
Intercept 0.013 0.006 .051 0 0.004 .986 0.001 0.002 .482 -0.003 0.002 .213
Loading 0.024 0.013 .072 0.007 0.011 .531 -0.001 0.004 .887 -0.006 0.007 .413
Item 35: Broken into a building or vehicle (or tried to break in) to steal something or just to look around
Intercept 0.014 0.006 .012 0.011 0.004 .008 -0.002 0.002 .210 -0.002 0.003 .526
Loading 0.029 0.013 .026 0.017 0.01 .091 -0.01 0.004 .018 0 0.01 .960
Item 36: Begged for money or things from strangers
Intercept 0.014 0.005 .004 -0.006 0.006 .336 0.002 0.002 .255 0 0.002 .844
Loading 0.03 0.013 .027 -0.027 0.022 .222 0.004 0.002 .120 0.001 0.007 .936
Item 37: Skipped classes without an excuse
Intercept 0.06 0.009 <.001 -0.011 0.008 .148 -0.007 0.002 .003 -0.004 0.004 .368
Loading 0.037 0.019 .045 -0.006 0.014 .679 -0.002 0.004 .558 -0.002 0.006 .671
Item 38: Failed to return extra change that a cashier gave you by mistake
Intercept 0.069 0.017 <.001 -0.018 0.014 .193 -0.004 0.003 .187 -0.001 0.007 .859
Loading 0.057 0.021 .006 -0.026 0.016 .094 0.002 0.006 .771 0.004 0.012 .733
Item 39: Been suspended from school
Intercept 0.076 0.009 <.001 0.010 0.008 .220 -0.018 0.003 <.001 -0.003 0.005 .606
Loading 0.022 0.009 .016 0.016 0.009 .016 -0.003 0.003 .397 -0.003 0.007 .652

GENDER DIFFERENCES IN MST EFFECTS
42
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Item 40: Made obscene telephone calls, such as calling someone and saying dirty things
Intercept 0.02 0.007 .004 -0.012 0.006 .035 0.002 0.002 .301 0.002 0.002 .359
Loading 0.044 0.02 .026 -0.031 0.017 .061 0.005 0.003 .106 0.003 0.005 .613
Table 11
SRDS Final MNFLA Model
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Delinquency Factor
Mean 0 - - - - - -0.196 0.057 0.001 0.397 0.131 0.002
Variance 1 - - - - - -0.444 0.152 0.004 0.936 0.194 <.001
Item 3: Purposely damaged or destroyed other property that did not belong to you (not counting family or school property)
Intercept 0.017 0.004 <.001 - - - - - - - - -
Loading 0.022 0.007 .014 0.026 0.01 .007 - - - - - -
Item 5: Stolen (or tried to steal) something worth more than $50
Intercept - - - 0.016 0.006 .012 - - - - - -
Loading - - - - - - - - - - - -
Item 6: Knowingly bought, sold or held stolen goods (or tried to do any of these things)
Intercept 0.043 0.01 <.001 - - - -0.007 0.003 .007 - - -
Loading 0.043 0.014 .003 - - - -0.011 0.004 .005 - - -
Item 8: Run away from home
Intercept 0.026 0.006 <.001 -0.018 0.006 .003 - - - - - -
Loading - - - - - - - - - - - -
Item 13: Been paid for having sexual relations with someone
Intercept - - - - - - - - - - - -
Loading - - - 0.015 0.006 .014 - - - - - -
Item 16: Sold marijuana or hashish ("pot," "grass," "hash")
Intercept - - - - - - -0.005 0.002 .018 0.009 0.003 0.004
Loading 0.061 0.024 .010 - - - - - - - - -

GENDER DIFFERENCES IN MST EFFECTS
43
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Item 17: Cheated on school tests
Intercept 0.03 0.007 <.001 - - - -0.007 0.002 <.001 - - -
Loading - - - 0.02 0.007 .008 -0.006 0.002 .015 - - -
Item 21: Hit (or threatened to hit) one of your parents
Intercept 0.034 0.008 <.001 -0.023 0.008 .004 - - - - - -
Loading 0.048 0.015 .001 - - - - - - - - -
Item 22: Hit (or threatened to hit) other students
Intercept 0.064 0.009 <.001 - - - -0.006 0.002 .002 - - -
Loading 0.052 0.011 <.001 - - - - - - - - -
Item 23: Been loud, rowdy, or unruly in a public place (disorderly conduct)
Intercept 0.006 0.001 .001 - - - 0.006 0.002 .002 - - -
Loading 0.077 0.029 .007 - - - - - - - - -
Item 27: Had (or tried to have) sexual relations with someone against their will
Intercept - - - - - - - - - - - -
Loading - - - - - - 0.015 0.006 .019 - - -
Item 28: Used force (strong-arm methods) to get money or things from other students
Intercept - - - - - - 0.009 0.002 <.001 - - -
Loading - - - 0.026 0.008 .002 - - - - - -
Item 29: Used force (strong-arm methods) to get money or things from a teacher or other adult at school
Intercept - - - - - - 0.005 0.002 .003 - - -
Loading - - - - - - 0.016 0.006 .010 - - -
Item 30 Used force (strong-arm methods) to get money or things from other people (not students or teachers)
Intercept - - - - - - 0.008 0.003 .004 - - -
Loading - - - - - - - - - - - -
Item 35: Broken into a building or vehicle (or tried to break in) to steal something or just to look around
Intercept 0.014 0.006 .012 0.011 0.004 .008 - - - - - -
Loading 0.029 0.013 .026 - - - -0.01 0.004 .018 - - -
Item 37: Skipped classes without an excuse
Intercept 0.06 0.009 <.001 - - - -0.007 0.002 .003 - - -
Loading - - - - - - - - - - - -

GENDER DIFFERENCES IN MST EFFECTS
44
Baseline Gender Timepoint Study
Reference
Parameter
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Point
Estimate
Standard
Error p
Item 39: Been suspended from school
Intercept 0.076 0.009 <.001 - - - -0.018 0.003 <.001 - - -
Loading 0.022 0.009 .016 0.016 0.009 .016 - - - - - -

GENDER DIFFERENCES IN MST EFFECTS
45
Pooled Data Analyses
After measurement invariance tests were concluded and a minimum of partial scalar
invariance was achieved, then a new dataset based on MNFLA model fits with factor scores from
the SRDS and FACES measures was generated for analysis as a collective dataset. Results from
RLMMs of each outcome variable are provided below. Means and standard deviations of all
outcome variables related to aims 2 and 3, i.e., outcomes for MST vs. Control over time and by
gender for the aggregate sample at each timepoint, are also provided.
Alcohol Use
Results showed that treatment assignment (MST vs. Control) did not impact self-reported
alcohol use. There was a significant effect by time (see Figure 3) but not by gender. Gender did
not moderate treatment effects and there were no other significant interaction effects. Detailed
results are provided in Table 12.
Table 12
Self-Reported Alcohol Use
Predictor Estimate Standard Error t p
Intercept 0.023 0.003 6.932 <.001
Condition -0.003 0.005 -0.535 .593
Time -0.006 0.002 -3.686 <.001
Gender -0.005 0.004 -1.366 .172
Condition*Time 0.003 0.002 1.330 .184
Condition*Gender 0.005 0.005 0.878 .380
Time*Gender 0.003 0.002 1.820 .069
Condition*Time*Gender -0.002 0.003 -0.836 .403
Timepoint MST Control
Mean SD Mean SD
Baseline Alcohol Use 0.063 0.148 0.045 0.093
Male 0.061 0.146 0.045 0.095
Female 0.073 0.159 0.044 0.086
Posttreatment Alcohol
Use
0.026 0.076 0.039 0.100

GENDER DIFFERENCES IN MST EFFECTS
46
Male 0.027 0.081 0.043 0.085
Female 0.023 0.058 0.062 0.151
6 months follow-up
Alcohol Use
0.034 0.105 0.038 0.081
Male 0.021 0.044 0.034 0.068
Female 0.086 0.208 0.062 0.133
1 year follow-up
Alcohol Use
0.033 0.085 0.057 0.133
Male 0.032 0.082 0.055 0.136
Female 0.036 0.095 0.063 0.122
Figure 3
Alcohol Use Over Time
Marijuana Use
Results showed that treatment assignment (MST vs. Control) did not impact self-reported
marijuana use. There were no significant effects by time nor gender. Gender did not moderate
treatment effects on marijuana use and there were no other significant interaction effects.
Detailed results are provided in Table 13.
Table 13
Self-Reported Marijuana Use
Predictor Estimate Standard Error t p
Intercept 0.030 0.006 5.295 <.001

GENDER DIFFERENCES IN MST EFFECTS
47
Condition 0.007 0.008 0.906 .365
Time -0.005 0.003 -1.616 .106
Gender 0.001 0.006 0.172 .863
Condition*Time -0.001 0.004 -0.280 .779
Condition*Gender -0.007 0.009 -0.747 .455
Time*Gender -0.003 0.003 -0.896 .370
Condition*Time*Gender 0.003 0.005 0.676 .499
MST Control
Timepoint/Gender Mean SD Mean SD
Baseline Marijuana Use 0.138 0.250 0.112 0.214
Male 0.137 0.247 0.107 0.200
Female 0.139 0.267 0.137 0.269
Posttreatment Marijuana Use 0.040 0.128 0.055 0.144
Male 0.042 0.137 0.045 0.119
Female 0.034 0.085 0.103 0.223
6 months follow-up
Marijuana Use
0.053 0.162 0.044 0.154
Male 0.035 0.126 0.043 0.158
Female 0.118 0.248 0.053 0.137
1 year follow-up Marijuana
Use
0.106 0.257 0.087 0.203
Male 0.114 0.273 0.078 0.191
Female 0.081 0.190 0.130 0.252
Self-Reported Delinquency
Results showed that treatment assignment (MST vs. Control) did not impact SRDS
scores. Time had a significant effect on self-reported delinquency across both treatment
conditions (see Figure 4). Gender had no significant effects overall. Gender did not moderate
treatment effects and there were no other significant interaction effects. Detailed results are
provided in Table 14.
Table 14
Self-Reported Delinquency
Predictor Estimate Standard Error t p
Intercept 0.042 0.006 6.758 <.001
Condition -0.005 0.009 -0.508 .611
Time -0.013 0.003 -4.279 <.001
Gender -0.008 0.007 -1.158 .247

GENDER DIFFERENCES IN MST EFFECTS
48
Condition*Time 0.003 0.004 0.737 .461
Condition*Gender 0.008 0.010 0.819 .413
Time*Gender 0.005 0.003 1.499 .134
Condition*Time*Gender -0.003 0.005 -0.540 .589
Timepoint/Gender MST Control
Mean SD Mean SD
Baseline SRDS
Male 0.038 0.056 0.043 0.056
Female 0.049 0.068 0.042 0.046
Posttreatment SRDS
Male 0.020 0.031 0.022 0.032
Female 0.018 0.027 0.021 0.023
6 months follow-up
SRDS
Male 0.021 0.043 0.018 0.028
Female 0.024 0.056 0.018 0.034
1 year follow-up SRDS
Male 0.018 0.033 0.027 0.052
Female 0.013 0.027 0.017 0.028
Figure 4
Self-Reported Delinquency Over Time
Family Adaptability and Cohesion – Youth Report
Adaptability. Results showed that treatment assignment (MST vs. Control) did not
impact youth-reported family adaptability. Gender had a significant impact on youth reported

GENDER DIFFERENCES IN MST EFFECTS
49
family adaptability; males reported significantly higher ratings than females (see Figure 5). Time
did not affect youth reported family adaptability scores. Gender did not moderate treatment
effects and there were no other significant interaction effects. Detailed results are provided in
Table 15.
Table 15
FACES – Family Adaptability Ratings, Youth Report
Predictor Estimate Standard Error t p
Intercept -0.391 0.278 -1.404 .161
Condition 0.033 0.396 0.082 .934
Time 0.182 0.119 1.527 .127
Gender 0.624 0.309 2.203 .043
Condition*Time -0.060 0.170 -0.353 .724
Condition*Gender -0.044 0.439 -0.101 .920
Time*Gender -0.175 0.102 -1.562 .118
Condition*Time*Gender -0.017 0.189 -0.088 .930
Timepoint/Gender MST Control
Mean SD Mean SD
Baseline Youth Adaptability 0.131 2.416 0.145 2.241
Male 0.282 2.332 0.244 2.181
Female -0.532 2.686 -0.290 2.466
Posttreatment Youth
Adaptability
0.104 2.047 0.002 1.921
Male 0.135 1.932 0.092 1.923
Female -0.030 2.500 -0.401 1.885
6 months follow-up Youth
Adaptability
-0.021 1.105 -0.185 1.126
Male -0.035 1.054 -0.223 1.181
Female 0.036 1.310 -0.020 0.855
1 year follow-up Youth
Adaptability
0.120 1.989 -0.091 1.893
Male 0.189 1.931 -0.082 1.854
Female -0.164 2.230 -0.128 2.095

GENDER DIFFERENCES IN MST EFFECTS
50
Figure 5
Youth Reported Family Adaptability Ratings by Gender
Cohesion. Results showed that treatment assignment (MST vs. Control) did not impact
youth report of family cohesion. There were no significant effects by time nor gender. Gender
did not moderate treatment effects and there were no other significant interaction effects.
Detailed results are provided in Table 16.
Table 16
FACES – Family Cohesion Ratings, Youth Report
Predictor Estimate Standard Error t p
Intercept -0.334 0.148 -2.249 .025
Condition 0.120 0.211 0.568 .570
Time 0.059 0.067 0.893 .372
Gender 0.213 0.165 1.295 .196
Condition*Time -0.062 0.095 -0.647 .517
Condition*Gender -0.091 0.234 -0.390 .697
Time*Gender -0.041 0.074 -0.560 .576

GENDER DIFFERENCES IN MST EFFECTS
51
Condition*Time*Gender -0.004 0.106 -0.042 .967
Timepoint/Gender MST Control
Mean SD Mean SD
Baseline Youth Cohesion -0.201 1.033 -0.137 0.983
Male -0.158 1.001 -0.113 0.923
Female -0.389 1.156 -0.242 1.217
Posttreatment Youth Cohesion -0.106 1.072 -0.147 0.997
Male -0.084 1.012 -0.132 1.009
Female -0.201 1.308 -0.212 0.957
6 months follow-up Youth
Cohesion
-0.032 0.911 -0.141 1.012
Male -0.001 0.906 -0.156 1.066
Female -0.156 0.934 -0.078 0.751
1 year follow-up Youth
Cohesion
-0.253 1.271 -0.399 1.194
Male -0.227 1.214 -0.396 1.185
Female -0.361 1.502 -0.414 1.258
Family Adaptability and Cohesion – Parent Report
Adaptability. Results showed that treatment assignment (MST vs. Control) did not
influence parent report of family adaptability. There was no significant effect by time nor gender
on parent report of family adaptability. Gender did not moderate treatment effects and there were
no other significant interaction effects. Detailed results are provided in Table 17.
Table 17
FACES – Family Adaptability Ratings, Parent Report
Predictor Estimate Standard Error t p
Intercept -0.170 0.154 -1.108 .268
Condition 0.256 0.219 1.173 .315
Time 0.060 0.060 1.005 .241
Gender 0.128 0.171 0.746 .456
Condition*Time -0.156 0.243 -1.751 .080
Condition*Gender -0.156 0.234 -0.641 .255
Time*Gender -0.078 0.068 -1.138 .552
Condition*Time*Gender 0.150 0.095 1.403 .161
Timepoint/Gender MST Control
Mean SD Mean SD

GENDER DIFFERENCES IN MST EFFECTS
52
Baseline Parent Adaptability -0.024 1.006 0.071 1.029
Male -0.006 1.012 0.040 1.042
Female -0.103 0.989 0.206 0.969
Posttreatment Parent Adaptability -0.110 1.080 0.002 0.973
Male -0.101 1.083 0.059 0.939
Female -0.145 1.081 -0.248 1.088
6 months follow-up Parent
Adaptability
-0.184 1.224 -0.018 1.078
Male -0.157 1.116 -0.018 1.111
Female -0.281 1.590 -0.021 0.909
1 year follow-up Parent Adaptability -0.051 1.042 -0.152 1.294
Male -0.125 1.049 -0.162 1.335
Female 0.181 1.005 -0.113 1.157
Cohesion. Results showed that treatment assignment (MST vs. Control) did not impact
parent report of family cohesion. There were no significant effects by time nor gender on parent
reported family cohesion. Gender did not moderate treatment effects and there were no other
significant interaction effects. Detailed results are provided in Table 18.
Table 18
FACES – Family Cohesion Ratings, Parent Report
Predictor Estimate Standard Error t p
Intercept -0.080 0.148 -0.436 .663
Condition -0.025 0.208 -0.119 .905
Time 0.045 0.062 0.727 .467
Gender 0.083 0.162 0.513 .608
Condition*Time -0.079 0.091 -0.868 .386
Condition*Gender 0.084 0.231 0.364 .716
Time*Gender -0.061 0.070 -0.879 .379
Condition*Time*Gender -0.022 0.102 -0.214 .830
Timepoint/Gender MST Control
Mean SD Mean SD
Baseline Parent Cohesion 0.010 1.032 0.014 1.041
Male 0.041 1.094 0.042 1.058
Female -0.125 0.693 -0.106 0.970
Posttreatment Parent Cohesion -0.026 1.034 0.024 0.994
Male -0.038 1.095 0.041 1.035
Female 0.022 0.756 -0.052 0.799
6 months follow-up Parent Cohesion 0.025 0.993 -0.213 1.048

GENDER DIFFERENCES IN MST EFFECTS
53
Male -0.039 1.035 -0.238 1.109
Female 0.249 0.793 -0.070 0.614
1 year follow-up Parent Cohesion -0.068 0.961 -0.302 0.940
Male -0.078 0.983 -0.323 0.977
Female -0.033 0.907 -0.222 0.803
Chapter 4: Discussion
The current study had three aims. The first aim was to successfully aggregate data from
three RCTs testing MST with JIY as a “mega” dataset using IDA. Aim 1 required measurement
invariance testing using MNFLA for multi-item latent outcome measures in the current study,
including the SRDS and youth and parent FACES measures. MNFLA results indicated these
measures achieved a minimum of partial scalar invariance. Achieving partial scalar invariance
allowed me to release loadings or intercepts of the items and parcels on the SRDS and FACES
measures that showed DIF as needed and thus permitted analysis as a collective dataset using
factor scores. The second aim of the current study was to assess whether MST yielded reduced
substance use and delinquency outcomes and improved family functioning through one year
follow-up relative to the control treatment when raw data from three RCTs were combined.
Results showed that youth randomized to MST did not show reductions in self-reported
delinquency and substance use (alcohol, marijuana) relative to youth randomized to control
conditions, nor did they show improved family functioning from youth and caregiver
perspectives through one-year follow-up. The third aim of the current study was to examine
whether gender moderated treatment effects. Results showed that treatment outcomes were not
moderated by gender through 1-year follow-up.
This paper was the first to examine whether commonly used measures in evaluating
family-focused treatments for substance use and delinquency among JIY showed measurement
invariance across studies. The SRDS and FACES measures appear to function consistently

GENDER DIFFERENCES IN MST EFFECTS
54
across studies; although MNFLA tests indicated some parcels from the FACES scales and items
from the SRDS scales showed DIF, there were enough items and parcels that did not show DIF
(i.e., were invariant by grouping variable, which in the current study were gender, timepoint, and
study) within each measure that could serve as anchors, which allowed for pooling raw data from
all three MST RCTs to create a larger sample size using IDA. These results show promise for
using IDA in future studies with JIY. IDA can allow for adequate sample sizes for appropriate
moderator testing among JIY or other populations that are similarly difficult to recruit and retain;
pooling data can provide a pathway to answer questions about historically neglected and/or
minoritized subgroups like female JIY. For example, IDA may allow for further study of the
trajectories of youth with intersectional minoritized identities (e.g., Black females) or genderexpansive youth, an overrepresented minority among JIY for whom even less is known regarding
the relative effectiveness of existing interventions (Irvine & Canfield, 2015).
As noted in the Method, parceling the SRDS measure before invariance testing yielded a
poorer model fit relative to the original 40-item unidimensional version of the measure. As a
result, I ended up testing measurement invariance at the item level for the SRDS. Although the
purpose of the current study was to assess whether the overall latent factor, rather than individual
items, were working consistently across grouping variables (study, timepoint, gender) per se,
results from measurement invariance testing of the SRDS did reveal interesting patterns of DIF
that may warrant further study, particularly concerning gender differences in how JIY endorse
various scale items. Across all 40 items of the SRDS, the following items showed nonuniform
DIF by gender: item 3 (purposely damaged or destroyed other property that did not belong to
you, outside of family or school property), item 13 (been paid for having sexual relations with
someone), item 17 (cheated on school tests), item 28 (used force [strong-arm methods] to get

GENDER DIFFERENCES IN MST EFFECTS
55
money or things from other students), and item 39 (been suspended from school). For females,
these items were less discriminant of underlying latent levels of delinquency relative to males.
Notably, although female JIY are far more likely to be paid for having sexual relations compared
to male JIY (Mitchell et al., 2010), this item was not as discriminant of overall underlying
delinquency for females relative to males.
The following items showed uniform DIF by gender: items 5 (stolen or tried to steal
something worth more than $50), 8 (run away from home) 21 (hit or threatened to hit one of your
parents), and 35 (broken into a building or vehicle [or tried to break in] to steal something or just
look around). Males were more likely to endorse all these items except item 8 (run away from
home) and item 21 (hit or threatened to hit one of your parents). These findings are generally
consistent with the literature that male JIY tend to endorse higher levels of violent behavior and
commit more serious offenses relative to females (Ehrmann et al., 2019). It is also consistent
with findings that females’ delinquent behavior more often occurs in the context of interpersonal
relationships (e.g., threatened or hit a parent), and that females are generally more likely to
engage in status offenses such as running away from home (Ehrmann et al., 2019; Henggeler et
al., 1987) compared to males.
Although MST is advertised as an effective treatment for substance use and delinquency
among JIY (MST Services, 2024), results of the current study are congruent with other work
showing mixed effects of MST (e.g., Fonagy et al., 2018; Littel et al., 2021). MST was not more
effective than control treatments at reducing substance use or delinquency nor improving family
functioning among JIY when data from three RCTs conducted by the developers of MST were
combined. Results from the current study are also generally consistent with findings reported in
the primary outcome papers for each trial. For example, there were no significant differences in

GENDER DIFFERENCES IN MST EFFECTS
56
self-reported delinquency by treatment condition across all three trials included in the current
study, and self-reported delinquency appeared to decline for youth assigned to both MST and
control groups over time. Consistent with findings reported in the primary outcome papers,
findings from the pooled dataset also showed MST was generally ineffective at reducing
substance use among JIY through one year follow-up. Finally, although only one of the three
RCTs included in the current study reported on family functioning, results from the pooled
dataset are consistent with null findings on family functioning in Study 1. Only Study 2
(Henggeler et al., 1999) investigated moderating effects of gender on MST outcomes. Contrary
to findings from the current study, Study 2 found that gender moderated effects of self-reported
marijuana and alcohol use.
The consistency of null findings across the three individual trials (all of which were
conducted by MST developers) and the pooled data from the current study may lead some to
question why MST has emerged as one of the preeminent treatments for addressing substance
use, delinquency, and other related behaviors (e.g., poor family functioning, psychological
distress) among JIY. An important point worth noting is that MST participants did yield
significant reductions in problem behaviors relative to control group participants for some of the
outcomes not included in the current study (due to missing data, e.g., archival delinquency data;
see Appendix B for details). For example, youth randomized to MST were incarcerated
significantly fewer days at follow-up and reported significantly lower rates of psychiatric
symptoms relative to youth randomized to usual services in Study 1 (Henggeler et al., 1997).
Youth randomized to MST in Study 2 (Henggeler et al., 1999) reported significantly fewer days
in out-of-home placements relative to youth receiving usual services. Additionally, youth

GENDER DIFFERENCES IN MST EFFECTS
57
randomized to both MST + Drug Court conditions in Study 3 (Henggeler et al., 2006) had
significantly fewer positive drug screens relative to youth in the non-MST Drug Court condition.
However, results from some measures unavailable for use in the current study (e.g., arrest
data, urine screen data) were also consistent with null findings from the pooled data. For
example, although youth randomized to MST reported lower rates of annualized rearrest relative
to youth randomized to usual services in Study 1, this difference was not statistically significant.
Additionally, urine screen results in Study 2 revealed no differences in substance use by
condition at follow-up. Similarly, while youth arrest data revealed lower rates of recidivism and
criminal arrests (controlling for time incarcerated) for youth randomized to MST vs. usual
services in Study 2, differences were not statistically significant. Archival delinquency data from
Study 3 also showed no significant differences in the percentage of youth arrested and number of
arrests by condition through one year follow-up.
Interestingly, results from the current study are consistent with other studies showing that
interventions for JIY are generally ineffective. A meta-analysis and systematic review (N=35
studies from 2000-2019) of interventions for justice-involved adolescents found that active
treatments were no more effective than control treatments at reducing recidivism (Olsson et al.,
2021), and effects were not moderated by gender, although the authors used a crude estimate
(i.e., they compared studies that included >70% males compared to <70% males in their sample)
of gender as a moderator of intervention effects. Similarly, a preliminary meta-analysis using
raw effect size data by gender investigating whether psychosocial interventions are equally
effective at reducing delinquent behavior for female vs. male JIY found no gender differences in
intervention effectiveness, and that interventions were broadly ineffective for both male and
female JIY (Galbraith & Huey, 2023).

GENDER DIFFERENCES IN MST EFFECTS
58
Systematic reviews and meta-analyses focusing specifically on the effectiveness of MST
with JIY also show mixed findings. A meta-analysis investigating MST effectiveness (n = 22
published RCTs or quasi-experimental studies; total N = 4,066) with samples of JIY and/or
youth with conduct disorder or antisocial behavior (van der Stouwe et al., 2014) showed that
unlike in the current study, MST yielded small but significant reductions in overall delinquency
and substance use, but consistent with my findings, MST did not significantly improve family
functioning. The authors found gender did not moderate MST outcomes, however, they used the
proportion of females included in a study sample to examine moderator effects (rather than a
more precise measure like raw effect size data by gender). More recent systematic reviews
evaluating MST vs. control show effects of MST are largely mixed in terms of addressing
delinquent behavior, substance use, and family functioning (Littell et al., 2021; Markham, 2017).
For example, a meta-analysis of MST for youth with social, emotional, and behavioral problems
conducted by Littell et al. (2021) showed that MST was more effective than control groups at
reducing arrests at one year follow-up, but not self-reported delinquency. They also found that
effects varied by country; studies conducted in the U.S. often reported outcomes that were
considerably more favorable compared to studies conducted outside of the U.S., likely because
services as usual (i.e., the typical control group) are far more intensive in Canada and Europe,
where non-U.S. MST trials have been conducted (Littell et al., 2021). Taken together, findings
indicate conflicting evidence as to whether MST is indeed effective for JIY.
The current study is among the first to investigate whether a widely used treatment like
MST is equally effective for males and females in U.S. juvenile justice contexts with pooled raw
data. Existing reviews of MST or other psychosocial interventions often neglect to investigate
gender as a moderator of outcomes entirely (e.g., Markham et al., 2017; Little et al., 2021) or use

GENDER DIFFERENCES IN MST EFFECTS
59
imprecise measures of whether gender moderates treatment effects for female JIY (e.g.,
percentage of males and females in study samples; Olsson et al., 2021; van der Stouwe et al.,
2014). Indeed, there is only one review study to date that investigates whether interventions for
JIY vary by gender using raw effect size data (Galbraith & Huey, 2023). Individual studies of
intervention effectiveness for JIY rarely investigate gender as a moderator of treatment effects
(Leve, 2015) and/or have limited power for appropriate moderator testing by gender due to low
numbers of females in study samples. To this end, only one of the three primary outcome papers
included in the current study examined gender as a moderator of MST outcomes. Moreover,
existing studies that do test for moderator effect of MST by gender and have sufficient sample
sizes to do so (e.g., Fonagy et al., 2018) have not been conducted in U.S. juvenile justice
contexts. The current study adds to the extremely sparse literature on whether psychosocial
interventions (MST or otherwise) work for females in the U.S. juvenile justice system.
Additionally, among studies synthesizing data from multiple trials of psychosocial
interventions for JIY, only one uses effect size data by gender to investigate whether treatment
effects vary for males and females (Galbraith & Huey, 2023).
Limitations
There are numerous limitations present in the current study. The data in this study are not
as current relative to other RCTs of MST and I cannot assume findings would be replicated with
youth who are currently justice-involved. However, the data are among the most current RCTs
that assess delinquency, substance use, and family functioning among JIY in the U.S; other more
recent RCTs have been with non JIY and investigate very different types of outcomes (e.g.,
diabetes management and glucose levels; Ellis et al., 2012) or only include very specific subsets
of JIY (e.g., sexual offenders, Letourneau et al., 2009). There are more recent RCTs that assess

GENDER DIFFERENCES IN MST EFFECTS
60
MST effectiveness on substance use and delinquency, in addition to other psychosocial outcomes
like family functioning (e.g., Butler et al., 2011; Fonagy et al., 2018), however, these were
conducted in Europe. European countries tend to have more rehabilitative, intensive services as
usual as control treatments, and the threshold for arrest is often higher than in the U.S. As such,
the context of the comparison of MST to services as usual and risk profiles of JIY in these
studies are likely different than those in the U.S.-based RCTs in the current study.
The most significant limitations of this study are related to missing data. Some data were
known to be missing ahead of time. For example, I knew that when the three RCTs included in
this study were conducted, there was very little discussion on the differences between biological
sex and gender. As such, the current study was only able to use cisgender identifiers, which
limits the application of findings to gender-expansive individuals. Moreover, only two of the
three studies collected urine screens; as such, I could only rely on self-report data to assess
substance use outcomes.
However, much of the data that was originally present was unexpectedly missing entirely
from certain studies or baseline timepoints (See Appendix B for a complete description and
detailed table of missing data) including parent and youth report measures of psychological
functioning, hard drug use items from the PEI, and measures assessing for peer delinquency.
Official arrest data was also unexpectedly missing at baseline from study 3 which prohibited
analysis of archival data across studies, so I was required to rely exclusively on self-report to
assess changes in delinquent behavior by time, gender, and treatment condition. Official arrest
data has traditionally been considered the most accurate indicator of reoffending behavior,
however, self-report data shows concurrent validity with official arrest data (e.g., Holloway et
al., 2024; Maxfield et al., 2000), is thought to be a more sensitive measure of delinquent

GENDER DIFFERENCES IN MST EFFECTS
61
behavior relative to arrest data, and is less subject to systemic racial bias (e.g., Black youth are
more likely to get arrested than white youth even if they engage in similar levels of delinquent
behavior, Rovner, 2016).
Implications
Despite limitations, this study has important implications for the dissemination of
interventions to JIY. Findings from the current study suggest that MST may not be effective at
addressing delinquency, substance use, nor family functioning among JIY. Although results are
consistent with findings from the individual trials included in the current study, as well as the
growing literature showing MST yields mixed effects (e.g., Little et al., 2021; Markham, 2017),
there is also data to suggest not all hope is lost for MST. The 2014 meta-analysis by van der
Stouwe et al., in addition to robust data on the cost-effectiveness of MST to taxpayers and crime
victims (e.g., Klietz et al., 2010; Vermuelen et al., 2017), suggests MST can benefit both
treatment recipients and the communities in which MST is implemented. Furthermore, most of
the MST studies included in recent systematic reviews and meta-analyses primarily evaluate
treatment effects immediately post-treatment through a few years follow-up. Studies evaluating
long-term outcomes of MST (e.g., over 20 years post-treatment) appear to have more robust
effects relative to studies assessing shorter-term delinquency outcomes (Borduin et al., 2021;
Sawyer et al., 2011; Schaeffer et al., 2005). One study even demonstrated that siblings of violent
offenders enrolled in MST were significantly less likely to be arrested compared to siblings of
offenders enrolled in usual services at 25 years post-treatment (Wagner, et al., 2014). As such,
there is evidence to suggest that MST can yield lasting effects on delinquent behavior well into
adulthood and at the family level, which are important considerations to make before writing off
the intervention as ineffective altogether.

GENDER DIFFERENCES IN MST EFFECTS
62
The current study also provides preliminary evidence that MST may not benefit female
JIY. As noted earlier in the Discussion, few adequately powered studies have investigated
whether existing interventions like MST are differentially effective for males and females. The
current paper supports the need for continued investigation on whether MST and other widely
used interventions are useful for female JIY, a growing and often neglected subgroup within the
juvenile justice system (Puzzanchera et al., 2022). Although MST is a flexible intervention
tailored to each individual youth’s needs, it may not be effective for females because it is
typically delivered in generic juvenile justice contexts that do not incorporate gender-specific
considerations into treatment (Sherman & Balck, 2015), and MST manuals reveal limited
explicit consideration of risks and needs related delinquent behavior, substance use, and other
related psychosocial outcomes more salient to treating for female JIY.
Although results from the current study suggest MST benefits neither male nor female
JIY, it is possible that incorporating gender-responsive principles into MST or implementing
entirely new gender-responsive interventions could yield stronger treatment effects for females.
In their review paper, Kerig and Schindler (2013) offer a comprehensive overview of empirically
based gender-responsive treatment principles for female JIY, most of which are derived from
recommendations from The Office of Juvenile Justice and Delinquency Prevention (OJJDP) and
leading feminist criminologists. Many of these recommendations directly address risk factors for
delinquency and related psychosocial problems that are more common among females compared
with males (e.g., sexual abuse, internalizing problems, polyvictimization, and interpersonal
victimization; Conrad et al., 2014; Docherty et al., 2016; Kerig, 2018). Examples of
recommendations outlined in Kerig and Schindler’s review (2013) include having treatment
providers “acknowledge that the lives of girl offenders may have been affected by sexism,

GENDER DIFFERENCES IN MST EFFECTS
63
victimization, poverty and racism” (p. 246) and explicitly include “gender-responsive curricula
focusing on issues such as abuse, trauma, violence, and healthy dating relationships” (p. 247).
Although MST is designed to be a personalized intervention, since it is often delivered in a
majority-male context (i.e., the juvenile justice system), MST adaptations that explicitly
incorporate gender-specific treatment considerations may be warranted to facilitate improved
treatment outcomes for female JIY. Alternatively, female JIY may benefit more from treatments
exclusively designed for females (e.g., Juvenile Justice Anger Management for Girls, Goldstein
et al., 2018); however, more research is needed on the efficacy of such interventions since very
few gender-responsive programs have been empirically tested in randomized controlled trials,
and none have been tested in comparison to existing, gender-neutral evidence-based treatments
(Parrish, 2020).
Future Directions
Findings from the current study indicate more research is needed to better understand the
contexts in which MST is effective for JIY. The null findings from the current study in addition
to mixed findings on short-term outcomes of MST compared to usual services from prior studies
suggest the development of new, and/or further refinement of existing interventions for JIY more
broadly, as well as female JIY more specifically, is needed. The current study was one of few
with a sample large enough for appropriate moderator testing by gender; future studies should
incorporate IDA or other pooled data techniques to expand the literature on whether
interventions for JIY are effective, and whether effectiveness varies by gender. Future studies
should also attempt to tease apart variables that may explain null effects for some of these
psychosocial interventions, and whether these explanatory variables differ by gender. Finally,
given that studies tend to show that MST has more robust outcomes when effects are assessed

GENDER DIFFERENCES IN MST EFFECTS
64
over longer-term relative to shorter-term follow-up periods, further study is needed to understand
how and why MST effects may improve relative to usual services over time, and whether
adjustments to the intervention can be made such that positive intervention effects can emerge
more immediately.

GENDER DIFFERENCES IN MST EFFECTS
65
References
Achenbach, T. M. (1991). Manual for the child behavior checklist. Burlington: Department of
Psychiatry, University of Vermont.
Achenbach, T. M. (1991). Manual for the youth self-report and 1991 profile. Burlington:
Department of Psychiatry, University of Vermont.
Ackerman, T. A. (1992). A didactic explanation of item bias, item impact, and item validity from
a multidimensional perspective. Journal of Educational Measurement, 29(1), 67-91.
Assink, M., van der Put, C. E., Hoeve, M., de Vries, S. L., Stams, G. J. J., & Oort, F. J. (2015).
Risk factors for persistent delinquent behavior among juveniles: A meta-analytic review.
Clinical Psychology Review, 42, 47-61.
Bandalos, D. L. (2008). Is parceling really necessary? A comparison of results from item
parceling and categorical variable methodology. Structural Equation Modeling: A
Multidisciplinary Journal, 15(2), 211-240.
Bates, D., Sarkar, D., Bates, M. D., & Matrix, L. (2007). The lme4 package. R package version
2:74.
Bauer, D. J. (2017). A more general model for testing measurement invariance and differential
item functioning. Psychological methods, 22(3), 507.
Belenko, S., Knight, D., Wasserman, G. A., Dennis, M. L., Wiley, T., Taxman, F. S., Oser, C.,
Dembo, R., Robertson, A. A., & Sales, J. (2017). The Juvenile Justice Behavioral Health
Services Cascade: A new framework for measuring unmet substance use treatment
services needs among adolescent offenders. Journal of Substance Abuse Treatment, 74,
80-91.
Belknap, J., & Holsinger, K. (2006). The gendered nature of risk factors for
delinquency. Feminist Criminology, 1(1), 48-71.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and
powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B
(Methodological), 57(1), 289-300. doi/10.1111/j.2517-6161.1995.tb02031.x
Bloom, B., & Covington, S. (2001, November). Effective gender-responsive interventions in
juvenile justice: Addressing the lives of delinquent girls. In annual meeting of the
American Society of Criminology, Atlanta, GA.
Bonynge, E. R. (1993). Unidimensionality of SCL‐90‐R scales in adult and adolescent crisis
samples. Journal of Clinical Psychology, 49(2), 212-215.

GENDER DIFFERENCES IN MST EFFECTS
66
Borduin, C. M., Blaske, D. M., Cone, L., Mann, B. J., & Hazelrigg, M. D. (1989). Development
and validation of a measure of peer relations: The Misouri Peer Relations Inventory. In C.
University of Missouri (Ed.).
Borduin, C. M., Quetsch, L. B., Johnides, B. D., & Dopp, A. R. (2021). Long-term effects of
multisystemic therapy for problem sexual behaviors: A 24.9-year follow-up to a
randomized clinical trial. Journal of Consulting and Clinical Psychology, 89(5), 393.
Borschmann, R., Janca, E., Carter, A., Willoughby, M., Hughes, N., Snow, K., Stockings, E.,
Hill, N. T., Hocking, J., & Love, A. (2020). The health of adolescents in detention: a
global scoping review. The Lancet Public Health, 5(2), e114-e126.
Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and
design. Cambridge MA: Harvard University Press.
Burgess-Proctor, A., Huebner, B. M., & Durso, J. M. (2016). Comparing the effects of maternal
and paternal incarceration on adult daughters’ and sons’ criminal justice system
involvement: A gendered pathways analysis. Criminal Justice and Behavior, 43(8), 1034-
1055.
Butler, S., Baruch, G., Hickey, N., & Fonagy, P. (2011). A randomized controlled trial of
multisystemic therapy and a statutory therapeutic intervention for young
offenders. Journal of the American Academy of Child & Adolescent Psychiatry, 50(12),
1220-1235. doi: 10.1016/j.jaac.2011.09.017.
Carpenter, J., & Kenward, M. (2012). Multiple imputation and its application. John Wiley &
Sons.
Cauffman, E., Farruggia, S. P., & Goldweber, A. (2008). Bad boys or poor parents: Relations to
female juvenile delinquency. Journal of Research on Adolescence, 18(4), 699-712.
Chassin, L. (2008). Juvenile justice and substance use. The Future of Children, 165-183.
Cnaan, A., Laird, N. M., & Slasor, P. (1997). Using the general linear mixed model to analyse
unbalanced repeated measures and longitudinal data. Statistics in Medicine, 16(20), 2349-
2380.
Conrad, S. M., Tolou-Shams, M., Rizzo, C. J., Placella, N., & Brown, L. K. (2014). Gender
differences in recidivism rates for juvenile justice youth: The impact of sexual
abuse. Law and Human Behavior, 38(4), 305.
Covington, S. S., & Bloom, B. E. (2007). Gender responsive treatment and services in
correctional settings. Women & Therapy, 29(3- 4), 9-33.
Curran, P. J., & Hussong, A. M. (2009). Integrative data analysis: the simultaneous analysis of
multiple data sets. Psychological Methods, 14(2), 81.

GENDER DIFFERENCES IN MST EFFECTS
67
Curran, P. J., Hussong, A. M., Cai, L., Huang, W., Chassin, L., Sher, K. J., & Zucker, R. A.
(2008). Pooling data from multiple longitudinal studies: the role of item response theory
in integrative data analysis. Developmental Psychology, 44(2), 365.
Curtis, N. M., Ronan, K. R., & Borduin, C. M. (2004). Multisystemic treatment: a meta-analysis
of outcome studies. Journal of Family Psychology, 18(3), 411.
Diggle, P., Diggle, P. J., Heagerty, P., Liang, K. Y., & Zeger, S. (2002). Analysis of longitudinal
data. Oxford University Press.
Docherty, M., Boxer, P., Veysey, B. M., & Ostermann, M. (2016). Gender differences in
prevalence of internalizing and externalizing symptoms in a justice-referred sample of
youth. Journal of Juvenile Justice, 5(2), 104.
Ehrmann, S., Hyland, N., & Puzzanchera, C. (2019). Girls in the juvenile justice system. Juvenile
Justice Statistics National Report Series Bulletin.
Elliott, D. S., Buckley, P. R., Gottfredson, D. C., Hawkins, J. D., & Tolan, P. H. (2020).
Evidence‐based juvenile justice programs and practices: A critical review. Criminology
& Public Policy, 19(4), 1305-1328.
Embretson, S. E., & Reise, S. P. (2013). Item response theory. Psychology Press.
Farrington, D. P., Jolliffe, D., Hawkins, J. D., Catalano, R. F., Hill, K. G., & Kosterman, R.
(2009). Why are boys more likely to be referred to juvenile court? Gender differences in
official and self-reported delinquency. Victims & Offenders, 5(1), 25-44.
Fortier, I., Raina, P., Van den Heuvel, E. R., Griffith, L. E., Craig, C., Saliba, M., Doiron, D.,
Stolk, R.P., Knoppers, B. M., Ferretti, V., Granda, V., & Burton, P. (2017). Maelstrom
Research guidelines for rigorous retrospective data harmonization. International Journal
of Epidemiology, 46(1), 103-105.
Goldstein, N. E. S., Giallella, C. L., Haney-Caron, E., Peterson, L., Serico, J., Kemp, K.,
Romaine, C. R., Zelechoski, A. D., Holliday, S. B., Kalbeitzer, R., Kelley, S. M., Hinz,
H., Sallee, M., Pennacchia, D., Prelic, A., Burkard, C., Grisso, T., Heilbrun, K., Núñez,
A., Leff, S., Lochman, J. (2018). Juvenile Justice Anger Management (JJAM) Treatment
for Girls: Results of a randomized controlled trial. Psychological Services, 15(4), 386–
397. https://doi.org/10.1037/ser0000184.
Gorman-Smith, D., & Loeber, R. (2005). Are developmental pathways in disruptive behaviors
the same for girls and boys?. Journal of Child and Family Studies, 14(1), 15-27.
Gottfredson, N. C., Cole, V. T., Giordano, M. L., Bauer, D. J., Hussong, A. M., & Ennett, S. T.
(2019). Simplifying the implementation of modern scale scoring methods with an
automated R package: Automated moderated nonlinear factor analysis (aMNLFA).
Addictive Behaviors, 94, 65-73. https://doi.org/10.1016/j.addbeh.2018.10.031

GENDER DIFFERENCES IN MST EFFECTS
68
Greenbaum, P. E., Wang, W., Henderson, C. E., Kan, L., Hall, K., Dakof, G. A., & Liddle, H. A.
(2015). Gender and ethnicity as moderators: Integrative data analysis of multidimensional
family therapy randomized clinical trials. Journal of Family Psychology, 29(6), 919.
Henggeler, S. W., Edwards, J., & Borduin, C. M. (1987). The family relations of female juvenile
delinquents. Journal of Abnormal Child Psychology, 15, 199-209.
Henggeler, S. W., Halliday-Boykins, C. A., Cunningham, P. B., Randall, J., Shapiro, S. B., &
Chapman, J. E. (2006). Juvenile drug court: enhancing outcomes by integrating evidencebased treatments. Journal of Consulting and Clinical Psychology, 74(1), 42.
Henggeler, S. W., Pickrel, S. G., & Brondino, M. J. (1999). Multisystemic treatment of
substance-abusing and- dependent delinquents: Outcomes, treatment fidelity, and
transportability. Mental Health Services Research, 1(3), 171-184.
Henggeler, S. W., Pickrel, S. G., Brondino, M., & Crouch, J. L. (1996). Eliminating (almost)
treatment dropout of substance abusing or dependent delinquents through home-based
multisystemic therapy. The American Journal of Psychiatry, 153(3), 427-428.
Henggeler, S. W., & Schaeffer, C. M. (2016). Multisystemic Therapy((R)): Clinical Overview,
Outcomes, and Implementation Research. Family Process, 55(3), 514-528.
https://doi.org/10.1111/famp.12232.
Henggeler, S. W., Schoenwald, S. K., Borduin, C. M., Rowland, M. D., & Cunningham, P. B.
(2009). Multisystemic therapy for antisocial behavior in children and adolescents (2nd
ed.). Guilford Press.
Henggeler, S. W., Melton, G. B., Brondino, M. J., Scherer, D. G., & Hanley, J. H. (1997).
Multisystemic therapy with violent and chronic juvenile offenders and their families: the
role of treatment fidelity in successful dissemination. Journal of Consulting and Clinical
Psychology, 65(5), 821.
Holloway, E. D., Irgens, M., McPhee, J., Folk, J. B., & Tolou-Shams, M. (2024). Youth
recidivism: youth self-report matters. Frontiers in Psychology, 14, 1208317.
doi.org/10.3389/fpsyg.2023.1208317.
Hubert, S.K. (2017). A comparison of multivariate methods for the detection of Differential Item
Functioning. [Master’s thesis, Ludwig-Maximilians-University Munich]. Doi:
10.5282/ubm/epub.38407.
Huey Jr, S. J., Henggeler, S. W., Brondino, M. J., & Pickrel, S. G. (2000). Mechanisms of
change in multisystemic therapy: Reducing delinquent behavior through therapist
adherence and improved family and peer functioning. Journal of Consulting and Clinical
Psychology, 68(3), 451.

GENDER DIFFERENCES IN MST EFFECTS
69
Hussong, A. M., Bauer, D. J., Giordano, M. L., & Curran, P. J. (2021). Harmonizing altered
measures in integrative data analysis: A methods analogue study. Behavior Research
Methods, 53(3), 1031-1045.
Hussong, A. M., Curran, P. J., & Bauer, D. J. (2013). Integrative data analysis in clinical
psychology research. Annual Review of Clinical Psychology, 9, 61-89.
Ide, B., Dingmann, C., Cuevas, E., & Meehan, M. (2010). Psychometric testing of the FACES III
with rural adolescents. Journal of Family Social Wselfork, 13(5), 410-419.
Irvine, A., & Canfield, A. (2016). The overrepresentation of lesbian, gay, bisexual,
questioning, gender nonconforming and transgender youth within the child welfare to
juvenile justice crossover population. American University Journal of Gender, Social
Policy & the Law, 24(2), 2.
Jiang, J., & Nguyen, T. (2007). Linear and generalized linear mixed models and their
applications (Vol. 1). Springer.
Jones, N. J., Brown, S. L., Wanamaker, K. A., & Greiner, L. E. (2014). A quantitative
exploration of gendered pathways to crime in a sample of male and female juvenile
offenders. Feminist Criminology, 9(2), 113-136.
Keenan, K., Loeber, R., Zhang, Q., Stouthamer-Loeber, M., & Van Kammen, W. B. (1995). The
influence of deviant peers on the development of boys' disruptive and delinquent
behavior: A temporal analysis. Development and Psychopathology, 7(4), 715-726.
Kerig, P. K. (2014). Introduction: for better or worse: intimate relationships as sources of risk or
resilience for girls' delinquency. Journal of Research on Adolescence, 24(1), 1-11.
Kerig, P. K. (2018). Polyvictimization and girls’ involvement in the juvenile justice system:
Investigating gender-differentiated patterns of risk, recidivism, and resilience. Journal of
Interpersonal Violence, 33(5), 789-809.
Kerig, P. K., & Schindler, S. R. (2013). Engendering the evidence base: A critical review of the
conceptual and empirical foundations of gender-responsive interventions for girls’
delinquency. Laws, 2(3), 244-282.
Kuhn, C. (2015). Emergence of sex differences in the development of substance use and abuse
during adolescence. Pharmacology & Therapeutics, 153, 55-78.
Leve, L. D., Chamberlain, P., & Kim, H. K. (2015). Risks, outcomes, and evidence-based
interventions for girls in the US juvenile justice system. Clinical Child and Family
Psychology Review, 18(3), 252-279. MST Services. (2022). mstservices.com.
Littell, J. H., Pigott, T. D., Nilsen, K. H., Green, S. J., & Montgomery, O. L. (2021).
Multisystemic Therapy® for social, emotional, and behavioural problems in youth age 10

GENDER DIFFERENCES IN MST EFFECTS
70
to 17: An updated systematic review and meta‐analysis. Campbell Systematic
Reviews, 17(4), e1158.
Little, T. D. (2013). Longitudinal structural equation modeling. Guilford press.
Little, T. D., Cunningham, W. A., Shahar, G., & Widaman, K. F. (2002). To parcel or not to
parcel: Exploring the question, weighing the merits. Structural Equation Modeling, 9(2),
151-173.
Little, T. D., Rioux, C., Odejimi, O. A., & Stickley, Z. L. (2022). Parceling in Structural
Equation Modeling: A Comprehensive Introduction for Developmental Scientists.
Cambridge: Cambridge University Press.
Little, R. J., & Rubin, D. B. (2019). Statistical analysis with missing data (Vol. 793). John Wiley
& Sons.
Liu, L., & Miller, S. L. (2020). Protective factors against juvenile delinquency: Exploring gender
with a nationally representative sample of youth. Social Science Research, 86, 102376.
Loeber, R., Capaldi, D. M., & Costello, E. (2013). Gender and the development of aggression,
disruptive behavior, and delinquency from childhood to early adulthood. In Disruptive
behavior disorders (pp. 137-160). Springer, New York, NY.
Lydia, V. P. (2020). The Gendered Context of System Experience and its Impact on Girls in the
Juvenile Justice System (Doctoral dissertation, Nova Southeastern University).
Maxfield, M. G., Weiler, B. L., & Widom, C. S. (2000). Comparing self-reports and official
records of arrests. Journal of Quantitative Criminology, 16, 87-110.
MacCallum, R. C., Roznowski, M., & Necowitz, L. B. (1992). Model modifications in
covariance structure analysis: The problem of capitalization on chance. Psychological
Bulletin, 111, 490-504.
McArdle, J. J. (2009). Latent variable modeling of differences and changes with longitudinal
data. Annual Review of Psychology, 60, 577-605.
https://doi.org/10.1146/annurev.psych.60.110707.163612
McCullagh, P., & Nelder, J. A. (2019). Generalized linear models. Routledge.
Mellenbergh, G. J. (1982). Contingency table models for assessing item bias. Journal of
Educational Statistics, 7(2), 105-118. https://doi.org/10.3102/1076998600700210
Millsap, R. E., & Everson, H. T. (1993). Methodology review: Statistical approaches for
assessing Measurement bias. Applied Psychological Measurement, 17(4), 297-334.
https://doi.org/10.1177/014662169301700401

GENDER DIFFERENCES IN MST EFFECTS
71
Mitchell, K. J., Finkelhor, D., & Wolak, J. (2010). Conceptualizing juvenile prostitution as child
maltreatment: Findings from the National Juvenile Prostitution Study. Child
Maltreatment, 15(1), 18-36.
Modrowski, C. A., Rizzo, C. J., Collibee, C., Houck, C. D., & Schneider, K. (2021).
Victimization profiles in girls involved in the juvenile justice system: a latent class
analysis. Child Abuse & Neglect, 111, 104774.
Montoya, A. K., & Jeon, M. (2020). MIMIC Models for Uniform and Nonuniform DIF as
Moderated Mediation Models. Applied Psychological Measurement, 44(2), 118–136.
https://doi.org/10.1177/0146621619835496
Muthén, B., Kaplan, D., & Hollis, M. (1987). On structural equation modeling with data that are
not missing completely at random. Psychometrika, 52(3), 431-462.
Muthén, L. K., & Muthén, B. O. (2017). Mplus user’s guide (Version 8). Los Angeles, CA.
Newsom, J. T. (2015). Longitudinal structural equation modeling: A comprehensive
introduction. Routledge.
Olson, D. H., Portner, J., & Lavee, Y. (1985). Family adaptability and cohesion evaluation scales
(FACES III). St. Paul: University of Minnesota, Family Social Science.
Parrish, D. E. (2020). Achieving justice for girls in the juvenile justice system. Social Work,
65(2), 149-158.
Pechorro, P., Moreira, K., Basto-Pereira, M., Oliveira, J. P., & Ray, J. V. (2019). The self-report
delinquency scale from the national longitudinal study of adolescent to adult health
among at-risk for delinquency youths. Violence and Victims, 34(1), 120-135.
Piquero, A. R., Macintosh, R., & Hickman, M. (2002). The validity of a self-reported
delinquency scale: Comparisons across gender, age, race, and place of
residence. Sociological Methods & Research, 30(4), 492-529.
https://doi.org/10.1177/0049124102030004002
Pusch, N., & Holtfreter, K. (2018). Gender and risk assessment in juvenile offenders: A metaanalysis. Criminal Justice and Behavior, 45(1), 56-81.
Putnick, D. L., & Bornstein, M. H. (2016). Measurement invariance conventions and reporting:
The state of the art and future directions for psychological research. Developmental
Review, 41, 71-90.
Puzzanchera, C., & Ehrmann, S. (2018). Spotlight on girls in the juvenile justice system. OJJDP
Statistical Briefing Book.

GENDER DIFFERENCES IN MST EFFECTS
72
Puzzanchera, C., Hockenberry, S., & Sickmund, M. (2022). Youth and the juvenile justice
system: 2022 national report. National Center for Juvenile Justice, Pittsburgh, PA.
Ratitch, B., O'Kelly, M., & Tosiello, R. (2013). Missing data in clinical trials: from clinical
assumptions to statistical analysis using pattern mixture models. Pharmaceutical
Statistics, 12(6), 337-347.
Rovner, J. (2016). Racial disparities in youth commitments and arrests. The Sentencing Project,
1-16.
Rosseel, Y. (2012). lavaan: An R package for structural equation modeling. Journal of Statistical
Software, 48, 1-36.
Rubin, D. B. (1996). Multiple imputation after 18+ years. Journal of the American Statistical
Association, 91(434), 473-489.
Rusby, J. C., Light, J. M., Crowley, R., & Westling, E. (2018). Influence of parent–youth
relationship, parental monitoring, and parent substance use on adolescent substance use
onset. Journal of Family Psychology, 32(3), 310.
Sawyer, A. M., & Borduin, C. M. (2011). Effects of multisystemic therapy through midlife: A
21.9-year follow-up to a randomized clinical trial with serious and violent juvenile
offenders. Journal of Consulting and Clinical psychology, 79(5), 643.
Schaeffer, C. M., & Borduin, C. M. (2005). Long-term follow-up to a randomized clinical trial of
multisystemic therapy with serious and violent juvenile offenders. Journal of Consulting
and Clinical Psychology, 73(3), 445.
Scott, T., & Brown, S. L. (2018). Risks, strengths, gender, and recidivism among justiceinvolved youth: A meta-analysis. Journal of Consulting and Clinical Psychology, 86(11),
931.
Sherman, F. & Balck, A. Gender Injustice: System-Level Juvenile Justice Reforms for Girls.
Portland, OR: The National Crittenton Foundation, 2015.
Skeer, M. R., McCormick, M. C., Normand, S. L. T., Mimiaga, M. J., Buka, S. L., & Gilman, S.
E. (2011). Gender differences in the association between family conflict and adolescent
substance use disorders. Journal of Adolescent Health, 49(2), 187-192.
Smith, D. K., & Saldana, L. (2013). Trauma, delinquency, and substance use: Co-occurring
problems for adolescent girls in the juvenile justice system. Journal of Child &
Adolescent Substance Abuse, 22(5), 450-465.
Steinberg, L., & Thissen, D. (2006). Using effect sizes for research reporting: Examples using
item response theory to analyze differential item functioning. Psychological Methods,
11(4), 402–415. https://doi.org/10.1037/1082-989X.11.4.402.

GENDER DIFFERENCES IN MST EFFECTS
73
Swendsen, J., Burstein, M., Case, B., Conway, K. P., Dierker, L., He, J., & Merikangas, K. R.
(2012). Use and abuse of alcohol and illicit drugs in US adolescents: Results of the
National Comorbidity Survey–Adolescent Supplement. Archives of General
Psychiatry, 69(4), 390-398.
Teplin, L. A., Abram, K. M., McClelland, G. M., Dulcan, M. K., & Mericle, A. A. (2002).
Psychiatric disorders in youth in juvenile detention. Archives of General
Psychiatry, 59(12), 1133-1143.
Teplin, L. A., Abram, K. M., McClelland, G. M., Mericle, A. A., Dulcan, M. K., & Washburn, J.
J. (2006). Psychiatric Disorders of Youth in Detention. Juvenile Justice Bulletin. Office of
Juvenile Justice and Delinquency Prevention.
Tracy, P. E., Kempf-Leonard, K., & Abramoske-James, S. (2009). Gender differences in
delinquency and juvenile justice processing: Evidence from national data. Crime &
Delinquency, 55(2), 171-215.
Van De Schoot, R., Kluytmans, A., Tummers, L., Lugtig, P., Hox, J., & Muthén, B. (2013).
Facing off with Scylla and Charybdis: a comparison of scalar, partial, and the novel
possibility of approximate measurement invariance. Frontiers in Psychology, 4, 770.
Van de Schoot, R., Lugtig, P., & Hox, J. (2012). A checklist for testing measurement
invariance. European Journal of Developmental Psychology, 9(4), 486-492.
Van De Schoot, R., Schmidt, P., De Beuckelaer, A., Lek, K., & Zondervan-Zwijnenburg, M.
(2015). Measurement invariance. Frontiers in Psychology, 6, 1064.
Van der Stouwe, T., Asscher, J. J., Stams, G. J. J., Deković, M., & van der Laan, P. H. (2014).
The effectiveness of multisystemic therapy (MST): A meta-analysis. Clinical Psychology
Review, 34(6), 468-481.
Whaley, R. B., Hayes-Smith, J., & Hayes-Smith, R. (2013). Gendered pathways? Gender,
mediating factors, and the gap in boys’ and girls’ substance use. Crime &
Delinquency, 59(5), 651-669.
Whittaker, T. A. (2012). Using the modification index and standardized expected parameter
change for model modification. The Journal of Experimental Education, 80(1), 26-44.
Widaman, K. F., Grimm, K. J., Early, D. R., Robins, R. W., & Conger, R. D. (2013).
Investigating factorial invariance of latent variables across populations when manifest
variables are missing completely. Structural Equation Modeling: A Multidisciplinary
Journal, 20(3), 384-408.
Winters, K. C., & Henly, G. A. (1989). Personal Experience Inventory (PEI): A Measure of
Substance Abuse in Adolescents.

GENDER DIFFERENCES IN MST EFFECTS
74
Zahn, M. A., Day, J. C., Mihalic, S. F., & Tichavsky, L. (2009). Determining what works for
girls in the juvenile justice system: A summary of evaluation evidence. Crime &
Delinquency, 55(2), 266-293.

GENDER DIFFERENCES IN MST EFFECTS
75
Appendix A: Outcome Measures Used in the Current Study
Personal Experiences Inventory
(Response Range: 0-100) across studies. Note only items 1 and 2 are included in the current
study
1. How many days have you used alcohol in the last 3 months?
2. How many days have you used marijuana in the last 3 months?
3. How many days have you used LSD in the last 3 months?
4. How many days have you used other psychedelics in the last 3 months?
5. How many days have you used cocaine in the last 3 months?
6. How many days have you used amphetamines in the last 3 months?
7. How many days have you used quaaludes in the last 3 months?
8. How many days have you used barbiturates in the last 3 months?
9. How many days have you used tranquilizers in the last 3 months?
10. How many days have you used heroin in the last 3 months?
11. How many days have you used other narcotics in the last 3 months?
12. How many days have you used inhalants in the last 3 months?

GENDER DIFFERENCES IN MST EFFECTS
76
Family Adaptability and Cohesion Scale (FACES) – Parent Report & Youth Report
Scale Response Options are the following for each item: 1) Almost never, 2) Once in a while, 3)
Sometimes, 4) Frequently, 5) Almost always.
Even numbered items are part of the adaptability subscale, and odd items are part of the cohesion
subscale.
1. Family members ask each other for help
2. In solving problems kids’ suggestions are followed
3. We approve of each other’s friends
4. Children have a say in their discipline
5. We like doing things with just our immediate family
6. Different persons act as leaders in our family
7. Family members feel closer to each other than outsiders
8. Our family changes its way of handling tasks
9. Family members like to spend free time together
10. Parents and children discuss punishment together
11. Family members feel very close to each other
12. The children make the decisions in the family
13. When our family gets together for activities, everyone is present
14. Rules change in our family
15. We can easily think of things to do together as a family
16. We shift household responsibilities from person to person
17. Family members consult other family members on their decisions
18. It is hard to identify the leader(s) in our family
19. Family togetherness is very important
20. It is hard to tell who does household chores

GENDER DIFFERENCES IN MST EFFECTS
77
Table A1
CFA Model Fit Statistics for Original Youth FACES
RMSEA (CI) TLI SRMR p value (chi square)
0.065 (0.061, 0.068) 0.794 0.054 <.001
Parcels for Youth FACES based on modification indices:
Youth Adaptability Factor:
Parcel 1: 12, 18
Parcel 2: 4, 8, 10, 20
Parcel 3: 2, 6, 14, 18
Youth Cohesion Factor:
Parcel 1: 7, 19
Parcel 2: 1, 3, 9, 15
Parcel 3: 5, 11, 13, 17
Table A2
CFA Model Fit Statistics for Parceled Youth FACES
RMSEA (CI) TLI SRMR p value (chi square)
0.027 (0.00, 0.048) 0.992 0.012 0.052
Table A3
CFA Model Fit Statistics for Original Parent FACES
RMSEA (CI) TLI SRMR p value (chi square)
0.091 (0.087, 0.095) 0.603 0.085 <.001
Parcels for Parent FACES based on modification indices:
Parent Adaptability:
Parcel 1:12, 18
Parcel 2: 6, 8, 10, 20
Parcel 3: 2, 4, 14, 16
Parent Cohesion:
Parcel 1: 7, 15
Parcel 2: 3, 11, 17, 19
Parcel 3: 1, 5, 9, 13
Table A4
CFA Model Fit Statistics for Parceled Parent FACES
RMSEA (CI) TLI SRMR p value (chi square)
0.049 (0.031, 0.069) 0.972 0.025 <.001

GENDER DIFFERENCES IN MST EFFECTS
78
Self-Report Delinquency Scale
Youth are asked to report the number of times they engaged in each of the following behaviors
over the last 3 months:
How many times in the last three months have you:
1. Purposely damaged or destroyed property belonging to your parents or other family members.
2. Purposely damaged or destroyed property belonging to a school.
3. Purposely damaged or destroyed other property that did not belong to you (not counting
family or school property).
4. Stolen (or tried to steal) a motor vehicle, such as a car or motorcycle.
5. Stolen (or tried to steal) something worth more than $50.
6. Knowingly bought, sold or held stolen goods (or tried to do any of these things).
7. Thrown objects (such as rocks, snowballs, or bottles) at cars or people.
8. Run away from home.
9. Lied about your age to gain entrance or to purchase something; for example, lying about your
age to buy liquor or get into a movie.
10. Carried a hidden weapon other than a plain pocket knife.
11. Stolen (or tried to steal) things worth $5 or less.
12. Attacked someone with the idea of seriously hurting or killing him/her.
13. Been paid for having sexual relations with someone.
14. Had sexual intercourse with a person of the opposite sex other than your spouse or significant
others.
15. Been involved in gang fights.
16. Sold marijuana or hashish ("pot," "grass," "hash").
17. Cheated on school tests.

GENDER DIFFERENCES IN MST EFFECTS
79
18. Hitchhiked where it was illegal to do so.
19. Stolen money or other things from your parents or other members of your family.
20. Hit (or threatened to hit) a teacher or other adult at school.
21. Hit (or threatened to hit) one of your parents.
22. Hit (or threatened to hit) other students.
23. Been loud, rowdy, or unruly in a public place (disorderly conduct).
24. Sold hard drugs, such as heroin, cocaine, and LSD.
25. Taken a vehicle for a ride (drive) without the owner's permission.
26. Bought or provided liquor for a minor.
27. Had (or tried to have) sexual relations with someone against their will.
28. Used force (strong-arm methods) to get money or things from other students.
29. Used force (strong-arm methods) to get money or things from a teacher or other adult at
school.
30. Used force (strong-arm methods) to get money or things from other people (not students or
teachers).
31. Avoided paying for such things as movies, bus or subway rides, and food.
32. Been drunk in a public place.
33. Stolen (or tried to steal) things worth between $5 and $50.
34. Stolen (or tried to steal) something at school, such as someone's coat from a classroom,
locker, or cafeteria, or a book from the library.
35. Broken into a building or vehicle (or tried to break in) to steal something or just to look
around.
36. Begged for money or things from strangers.

GENDER DIFFERENCES IN MST EFFECTS
80
37. Skipped classes without an excuse.
38. Failed to return extra change that a cashier gave you by mistake.
39. Been suspended from school.
40. Made obscene telephone calls, such as calling someone and saying dirty things.
Table A5
CFA Model Fit Statistics for Unidimensional SRDS
RMSEA (CI) TLI SRMR p value (chi square)
0.048 (0.045, 0.050) 0.963 0.044 <.001

GENDER DIFFERENCES IN MST EFFECTS
81
Appendix B: Overview of Missing Data
The initial proposal sought to assess all study measures outlined in the table below.
Outcomes with measures that differed by study (e.g., the PYS and MPRI for peer delinquency)
were supposed to be harmonized, such that overlapping items from disparate measures were to
be combined to create one aggregate measure to assess the construct across all three studies (for
example, overlapping items both in the MPRI, used in studies 1 and 2, and the PYS measure,
used in study 3, were going to be combined to create one overall measure of peer delinquency to
be used across all three studies). I confirmed that data from all measures listed in Table 3 were
indeed collected (an “X” denotes whether a measure was used in Study 1, 2, or 3). I did so by a)
referencing the published outcomes from each RCT, b) consulting with my graduate school
advisor, Dr. Stan Huey, and c) reviewing the study codebooks made available to Dr. Huey, and
d) contacting team members (e.g., Dr. Phillipe Cunningham, Dr. Colleen Halliday-Boykins)
from all three studies to confirm the existence of data from each of the listed measures.
Unfortunately, I, Dr. Huey, and other members of the research team involved in the MST trials
used in the current study were only able to retrieve the data needed to analyze the three datasets
in aggregate (i.e., as one collective dataset) using IDA for measures listed in bold typeface in
Table 3. I also contacted the principal investigator (Dr. Scott Henggeler; now retired) of all three
studies, and he was also unable to provide the data needed to use IDA for any measure not listed
in bold in Table 3. As such, only data from the following measures were available for use in the
current study: the Self-Report Delinquency Scale (SRDS, Elliot & Huizinga, 1989), parent and
self-report data from Family Adaptability and Cohesion Scale (FACES, Olson et al., 1980), and
the Personal Experiences Inventory (PEI; substance use self-report measure, Winters & Henly
1989). In addition, data from the items on the PEI assessing hard drug use were often missing at

GENDER DIFFERENCES IN MST EFFECTS
82
various timepoints which prohibited analysis of the PEI as a collective drug use measure. As
such, only two items with complete data from the PEI, number of days of marijuana use and
number of days of alcohol use, in the past 90 days were used as substance use outcome variables
in the present study.
Only data from four measures (FACES self and FACES parent report are analyzed
separately) across three constructs (family functioning, delinquency, substance use) were
available because the other measures/constructs of interest were either missing completely across
a study/studies or missing from any timepoint in study 2. Since study 2 was the only study that
contained overlapping timepoints with both studies 1 and 3, if any measure or construct was
missing from study 2, it meant that IDA could not be conducted with that specific
measure/construct at all. As such, I was precluded from psychological functioning, peer
relations, and peer delinquency constructs all together. I was also unable to use official arrest
data as a measure of delinquency as it was missing from study 2, timepoint 4, and study 3,
timepoints 1 and 2. A table that shows what data was missing by timepoint/study is provided
below. Cells that are shaded grey indicate that data for that measure was never collected at the
study/timepoint (i.e., the data was supposed to be missing), blank cells indicate data that was
collected but unretrievable (i.e., data that was not supposed to be missing), and cells marked with
a “ü” indicate that the data was available for use (i.e., data that was collected and retrievable).
Table B1
Constructs/Measure Availability by Study and Timepoint
Construct/Measure Study 1 Study 2 Study 3
Substance Use T1 T2 T3 T4 T1 T2 T3 T4 T1 T2 T3 T4
PEI ü ü ü ü ü ü ü ü ü ü
CBCL

GENDER DIFFERENCES IN MST EFFECTS
83
Delinquency T1 T2 T3 T4 T1 T2 T3 T4 T1 T2 T3 T4
SRDS ü ü ü ü ü ü ü ü ü ü
Official Arrest Records ü ü ü ü ü ü ü ü
Truancy
Psychological
Functioning
T1 T2 T3 T4 T1 T2 T3 T4 T1 T2 T3 T4
CBCL
YSR
SCL-90 ü ü ü
Family Functioning T1 T2 T3 T4 T1 T2 T3 T4 T1 T2 T3 T4
FACES – youth report ü ü ü ü ü ü ü ü ü ü
FACES – parent report ü ü ü ü ü ü ü ü ü ü
Peer Relations T1 T2 T3 T4 T1 T2 T3 T4 T1 T2 T3 T4
MPRI ü ü ü ü ü ü ü
PYS
CBCL
Peer Delinquency T1 T2 T3 T4 T1 T2 T3 T4 T1 T2 T3 T4
MPRI ü ü ü ü ü ü ü
PYS
*T1 = Baseline, T2 = Posttreatment, T3 = 6 months follow-up, T4 = 1 year follow-up

Abstract (if available)

Abstract Background: Justice-involved youth (JIY) have higher rates of substance use and substance use disorders relative to their non-offending counterparts. Substance use co-occurs with delinquency and exacerbates antisocial behavior. Thus, interventions for JIY often target co-occurring substance use and delinquency. These interventions typically consider male-specific (e.g., higher rates of violent offending) or gender-neutral (e.g., delinquent peers) risks for substance use and delinquency, but neglect female-specific risks (e.g., elevated rates of PTSD), which has raised concern about their appropriateness for female JIY. Females represent the fastest-growing subgroup among JIY, but whether they benefit from existing interventions remains unclear due to poor representation in randomized trials and low statistical power. Method: The current study used Integrative Data Analysis (IDA) to pool raw data from three randomized controlled trials of Multisystemic Therapy (MST), a widely used intervention for substance use and delinquency among JIY. IDA allows for analysis as one “mega” dataset to generate the statistical power needed to detect gender differences in treatment outcomes. The aims of the current study were three-fold. Aim 1 was to establish measurement invariance across study, timepoint, and gender among study outcome measures using moderated nonlinear factor analysis. Aim 2 was to determine whether MST was effective in reducing substance use and delinquency, and improving family functioning relative to the control group through one year follow-up when raw data from three trials was pooled together as one dataset. Aim 3 was to determine whether gender moderated MST effects on delinquency, substance use, and family functioning through one year follow-up. Results: Moderated nonlinear factor analyses showed that measurement invariance could be established across study, timepoint, and gender, although measures exhibited uniform and nonuniform differential item functioning, which required constraining and releasing intercepts and loadings before pooling data for analysis as a collective dataset. Analyses of the pooled dataset showed MST was no more effective than the control group in reducing substance use and delinquency, nor in improving family functioning among JIY through one year follow-up. Moreover, MST was equally ineffective for male and female JIY. Conclusions: The current study is the first to use IDA to analyze MST outcomes and whether effects vary by gender. Findings suggest that measurement invariance can be established among outcome measures commonly used among JIY (e.g., self-reported delinquency scales), which is promising for continued use of IDA in future studies with this population. Findings also suggest that, across these three pooled randomized trials, MST is no more effective than usual services for male and female JIY; continued development and study of effective psychosocial interventions for this population is sorely needed.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Cumulative risk as a moderator of multisystemic therapy effects for juvenile offenders

PDF

A penta-dimensional longitudinal analysis of the predictors of compulsive internet use among adolescents using linear mixed model (LMM)

PDF

Substance use disparities in adolescents of lower socioeconomic status: emerging trends in electronic cigarettes, alternative tobacco products, marijuana, and prescription drug use

PDF

Association of carotid artery stiffness with measures of cognition in older adults

PDF

A "second chance" program for vulnerable young adults: a program evaluation and a randomized controlled trial of adjunctive motivational interviewing to improve retention

PDF

A network analysis of online and offline social influence processes in relation to adolescent smoking and alcohol use

PDF

Evaluating social-cognitive measures of motivation in a longitudinal study of people completing New Year's resolutions to exercise

PDF

Pregnancy in the time of COVID-19: effects on perinatal mental health, birth, and infant development

PDF

Chronic eye disease epidemiology in the multiethnic ophthalmology cohorts of California study

Asset Metadata

Creator Galbraith, Katharine (author)

Core Title Using integrative data analysis to evaluate gender differences in effects of multisystemic therapy for justice-involved youth

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Psychology

Degree Conferral Date 2024-08

Publication Date 07/12/2024

Defense Date 05/02/2024

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag delinquency,integrative data analysis,juvenile justice,multisystemic therapy,OAI-PMH Harvest,substance use

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Huey, Stanley Jr. (committee chair), Beam, Christopher (committee member), Davis, Jordan (committee member), John, Richard (committee member)

Creator Email katharig@usc.edu,katharinegalbraith1@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113997OHE

Unique identifier UC113997OHE

Identifier etd-GalbraithK-13228.pdf (filename)

Legacy Identifier etd-GalbraithK-13228

Document Type Dissertation

Format theses (aat)

Rights Galbraith, Katharine

Internet Media Type application/pdf

Type texts

Source 20240715-usctheses-batch-1182 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu