University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Running head: CUMULATIVE RISK AS A MODERATOR OF MST

1

Cumulative Risk as a Moderator of Multisystemic Therapy Effects for Juvenile Offenders
Miriam P. Rubenson, B.A.
University of Southern California

Faculty Advisor: Stan Huey, Jr., Ph.D.
(CLINICAL PSYCHOLOGY)
Master’s Thesis
August 2017

CUMULATIVE RISK AS A MODERATOR OF MST

2

Abstract
The present study examined cumulative risk (CR) for delinquency, drug use, and arrests as a
moderator of treatment efficacy in a randomized trial of multisystemic therapy (MST) compared
to drug court (DC) and family court (FC) (Henggeler, Halliday-Boykins, Cunningham, Randall,
Shapiro, & Chapman, 2006). Participants were 161 juvenile drug offenders arrested in
Charleston, SC and their families. Although CR is a robust predictor of juvenile offending, no
studies to date have examined its role as a moderator of treatment efficacy for youth
psychosocial interventions. MST is an effective treatment for juvenile offenders that aims to
address contextual risk factors that contribute to criminal behavior. The MST literature has yet to
address two important questions: (1) is MST more effective than other interventions or usual
services across youth with different levels of cumulative risk, and (2) do potentially modifiable
risk factors and static risk factors moderate MST effects differently? The author hypothesized
that MST would be more effective than comparison conditions for youth with high CR, while
effects would not differ significantly across conditions for youth with low levels of risk. In
particular, participants with many risk factors in potentially modifiable domains were expected
to benefit the most from MST, compared to those in other conditions. Data did not support these
hypotheses; no significant moderator effects were detected. Results suggest that treatment
efficacy does not depend on level of cumulative risk, and MST may be equally effective for
youth with high and low levels of risk.

CUMULATIVE RISK AS A MODERATOR OF MST

3
Table of Contents
Background ..................................................................................................................................... 5
Method .......................................................................................................................................... 14
Participants ................................................................................................................................ 14
Treatment Conditions ................................................................................................................ 14
Measures ................................................................................................................................... 15
Youth characteristics. ............................................................................................................ 15
Family relations. ................................................................................................................... 16
Caregiver characteristics. ...................................................................................................... 16
Peer delinquency. .................................................................................................................. 17
Neighborhood poverty. ......................................................................................................... 17
Youth outcomes. ................................................................................................................... 17
Cumulative Risk........................................................................................................................ 18
Dichotomization .................................................................................................................... 20
Results ........................................................................................................................................... 21
Missing Data ............................................................................................................................. 21
Cumulative Risk........................................................................................................................ 22
Moderation Effects .................................................................................................................... 22
Form 90 drug use. ................................................................................................................. 22
PEI drug use. ......................................................................................................................... 24
Arrests ................................................................................................................................... 25
Delinquency .......................................................................................................................... 27
Discussion ..................................................................................................................................... 28
CUMULATIVE RISK AS A MODERATOR OF MST

4
Limitations ................................................................................................................................ 30
Conclusion .................................................................................................................................... 33
References ..................................................................................................................................... 34
Tables and Figures ........................................................................................................................ 41

CUMULATIVE RISK AS A MODERATOR OF MST

5
Cumulative Risk as a Moderator of Multisystemic Therapy Effects for Juvenile Offenders
Background
Juvenile offenders are frequently exposed to multiple risk factors for offending and drug
use in childhood. They are more likely to have experienced severe stress and trauma in childhood
than non-offending youth (Dierkhising, Ko, Woods-Jaeger, Briggs, Lee, & Pynoos, 2013;
Baglivio, Epps, Swartz, Huq, Sheer, & Hardt, 2014), and they tend to face multiple forms of
adversity. Frequently-cited risk factors for adolescent criminal behavior affect multiple social-
ecological systems (Bronfenbrenner, 1979), such family poverty, having drug-abusing parents, or
spending time with delinquent friends (Loeber & Farrington, 1998). As such, therapy for juvenile
offenders that can address problems across multiple domains simultaneously may be optimal.
Multisystemic therapy (MST) is a flexible intervention for juvenile offenders that treats
problems in multiple domains that may contribute to offending (Henggeler, Rodick, Borduin,
Hanson, Watson, & Urey, 1986; Henggeler, Schoenwald, Borduin, Rowland, & Cunningham,
1998). Such domains include family, peer groups, school, and any other relevant system that
affects the youth’s life (Bronfenbrenner, 1979; Henggeler, Schoenwald, Borduin, Rowland, &
Cunningham, 2009). MST is consistently effective, and is one of the best-validated interventions
for serious and chronic juvenile offenders (Henggeler, 2011). Youth treated with MST have
lower rates of delinquency, drug use, and rearrests compared to controls, as evidenced by 25
published randomized trials conducted mostly by independent researchers (Henggeler &
Schaeffer, 2016). Moreover, longitudinal studies of MST have shown remarkable effectiveness.
For example, a trial with chronic juvenile offenders and their caregivers yielded decreased
behavior problems and mental health symptoms, improved family relations, and a 63% decrease
in youth reoffending at post treatment for MST compared to individual treatment (Borduin et al.,
CUMULATIVE RISK AS A MODERATOR OF MST

6
1995). At the 14-year follow-up, MST rearrests decreased by 54% and days incarcerated by 57%
compared to controls (Schaeffer & Borduin, 2005), and similar outcomes were sustained 22
years later (Sawyer & Borduin, 2011).
In fact, MST seems to be effective regardless of youth demographics or symptom
severity at pretreatment. Many randomized trials of MST have tested SES, sex, ethnicity, and age
as moderators of MST outcomes, but in the vast majority of cases, these factors have had no
significant effects on outcomes (Henggeler, 2011). For example, Henggeler, Melton, and Smith
(1992) found that neither age, race, social class, gender, prior arrest, nor prior incarceration
moderated treatment effects in a trial of MST versus usual services for serious juvenile
offenders. Similarly, neither age, gender, ethnicity, single-parent family status, maternal
education nor maternal employment had any moderating effects on treatment effectiveness in a
randomized trial of MST in Sweden (Sundell, Hansson, Löfholm, Olsson, Gustle, & Kadesjö,
2008). These results suggest that MST works equally well for diverse families.
The demographic variables noted above (e.g., age, gender, race) have also been cited as
risk factors for adolescent delinquency and drug use. For example, living in poverty (Loeber,
Farrington, Stouthamer-Loeber, & Van Kammen, 1998), being an ethnic minority (Felson,
Deane & Armstrong, 2008), being male (Huizinga, Loeber, Thornberry & Cothern, 2000), living
with a single parent (Velez, Johnson, & Cohen, 1989), and having a mother with a low
educational background (Kandel, Simcha-Fagan, & Davies, 1986) have all been found to predict
delinquency and drug use in multiple longitudinal studies. Unsurprisingly, when youth face
multiple risk factors simultaneously, their odds of experiencing negative outcomes rise
precipitously (e.g., Loeber et al., 1998; Small & Luster, 1994; Thornberry, Krohn, Lizotte,
Smith, & Tobin, 2003).
CUMULATIVE RISK AS A MODERATOR OF MST

7
However, no studies to date have tested whether the accumulation of risks in a youth’s
life affects the likelihood of success in therapy. Prior research shows that MST is effective with
youth facing a variety of risk factors, but is it equally effective for clients with different levels of
cumulative adversity? Given its flexibility and focus on targeting a range of problems
(Henggeler, Schoenwald, Borduin, Rowland, & Cunningham, 2009), MST should be more
effective for juvenile offenders with risks across different social-ecological systems than
interventions that target a single problem domain, like behavioral parent training (Swenson,
Schaeffer, Henggeler, Faldowski, & Mayhew, 2010). However, clients with multiple
simultaneous risk exposures may present a greater challenge to MST therapists than clients with
risk in only a few domains. Hypothetically, an adolescent with clinically significant attention
problems who dropped out of school and lives with a single caregiver in a high-crime
neighborhood may require interventions that are more difficult to coordinate and implement than
an equally delinquent adolescent with only one risk factor.
One common method for measuring levels of adversity is to compute a cumulative risk
score. Cumulative risk (CR) is most often operationalized as the sum of dichotomized predictors
of a negative outcome present in a person’s life (Evans, Li, & Whipple, 2013). In general,
increases in CR are strongly associated with greater likelihood of problems including
delinquency, drug use, gang membership, physical illness, and early death (e.g., Loeber et al.,
1998; Thornberry et al., 2003; Felitti et al., 1998). CR is also a better predictor of negative
outcomes than any individual risk factor (Evans et al., 2013). In their review of cumulative risk
in the child development literature, Evans and colleagues (2013) point to 20 cross-sectional and
16 longitudinal studies to date finding significant main effects of CR as a predictor of conduct
problems.
CUMULATIVE RISK AS A MODERATOR OF MST

8
In the literature on juvenile offending, a CR score frequently represents a snapshot of the
youth’s current risk exposures.
1
Many large-scale longitudinal studies like the Pittsburgh Youth
Study (Loeber et al., 1998) and the Rochester Youth Development Study (Thornberry et al.,
2003) use this snapshot approach to predict negative outcomes like mental health problems,
delinquency, arrests, or gang involvement from large community samples. These studies build
CR scores from a number of sociodemographic, individual, peer, and family constructs, and
examine their additive predictive power for the negative outcome of interest. They generally find
CR is a stronger predictor of negative outcomes than any individual risk factor.
The combined presence of multiple risk factors may also have a significant impact on
treatment gains, but no studies to date have examined this question for youth interventions.
Indeed, cumulative risk is normally operationalized as a predictor of negative outcomes in the
risk literature (Evans et al., 2013), rather than a moderator of treatment efficacy. To my
knowledge, cumulative risk has not been tested as a moderator of treatment efficacy for any
youth psychosocial intervention, let alone for juvenile offenders. This is a significant gap in the
treatment literature, as juvenile offenders tend to have disproportionately high levels of risk
(Baglivio et al., 2014) that may interact with treatment. While most prediction studies do not
examine its role in treatment, studies of cumulative risk as a moderator of treatment efficacy may
provide important information about who benefits from treatment. High-risk youth may improve
less than low-risk youth, or interventions like MST may work equally well for youth with
different levels of cumulative risk.
Indeed, only one study to my knowledge examined CR as a moderator of behavior
problems, but not as a moderator of treatment efficacy. MacKenzie, Nicklas, Brooks-Gunn, and

1
Conversely, CR in the public health literature often measures historical exposures to traumatic
events or environmental contaminants. See Felitti et al. (1998) for a seminal example.
CUMULATIVE RISK AS A MODERATOR OF MST

9
Waldfogel (2014) examined how CR interacted with harsh parenting in the development of
externalizing problems in a high risk population of children. They found that CR was a
significant moderator of the relationship between high-frequency spanking and children’s
development of externalizing problems, such that young children with high CR who were
frequently spanked were more likely to develop externalizing problems later in childhood
compared to those with low CR. MacKenzie and colleagues (2014) provide precedent for the
study of CR as a moderator, but do not examine its role in a treatment context.
Cumulative risk may be an important moderator of treatment efficacy because it usually
includes certain modifiable risk factors that are targets of treatment, such as adverse parenting
practices. In a study using data from the Pittsburgh Youth Study, Ahonen, Loeber, and Pardini
(2015) describe risk factors as either static or potentially modifiable. Static risk factors, they
explain, include demographic variables and past traumatic exposures that correlate with negative
outcomes, such as low socioeconomic status or having parents with an arrest history (Lipsey &
Derzon, 1998; Hoffman & Maynard, 1983). Static risk factors cannot be changed through
intervention. On the other hand, potentially modifiable risk factors tend to be contextual, ongoing
issues that currently contribute to the youth’s offending. For example, lax parental supervision
and discipline both predict delinquency (Loeber et al., 1998; West & Farrington, 1998; McCord,
1979), but are potentially modifiable in parent- and family-based therapies. In the context of an
intervention, it is important to consider whether clinicians can reasonably address their clients’
risk factors. MST is one intervention that aims to treat a range of modifiable risk factors like
parental supervision and discipline that contribute to the youth’s criminal behavior (Henggeler et
al., 1986). CR may moderate treatment efficacy because it identifies participants with modifiable
problems who stand to gain optimally from certain forms of treatment.
CUMULATIVE RISK AS A MODERATOR OF MST

10
However, there is also evidence that CR may have a negative effect on treatment
response for youth externalizing problems. Bagner and Graziano (2013) examined the effect of
cumulative risk on Parent-Child Interaction Therapy (PCIT) outcomes and rates of premature
termination in a study of families with developmentally delayed children. They found that
families with higher CR saw diminished treatment gains and greater dropout compared to
families with low CR. However, all participants in the study received PCIT, so moderation
effects and efficacy were not assessed. The limited research leaves a gap in our understanding of
the role of CR in treatment.
Cumulative Risk Models in the Literature
While there is no standard CR measure, most studies that assess cumulative risk include
risk factors that are shown (or assumed) to be predictors of the negative outcome of interest
(Evans et al., 2013). Most models include risk factors from several domains, including
neighborhood, family, peer, school, and individual, and each factor is given equal weight. Often
some attempt is made to consolidate or remove redundant or highly correlated risk factors so as
to avoid over-counting risk factors that seem to measure the same construct (e.g., Loeber et al.,
1998; Thornberry et al., 2003).
However, the CR literature is characterized by heterogeneity of methodology, particularly
in regards to decisions about how to compute total scores. Some only count the number of
domains (e.g., individual, family, school, and neighborhood risk) for which a person is at risk, on
the assumption that risk factors within each domain have similar influences on the likelihood of
negative outcomes (e.g., Small & Luster, 1994; Thornberry et al., 2003). In such cases, all risk
factors from a given domain are collapsed into a single score, for example “academic risk,”
rather than separately computing risk attributed to poor grades, low school attendance, and being
CUMULATIVE RISK AS A MODERATOR OF MST

11
held back a year (e.g., Thornberry et al., 2003). Other CR models separately count all risk factors
regardless of domain (e.g., Felitti et al., 1998; Loeber et al., 1998; MacKenzie et al., 2014).
CR studies also differ in how they select risk factors to include in a CR score. While
some studies simply choose risk factors that are related to outcomes based on previous literature,
others determine the risk factors to be included in a CR score using an empirical method. Loeber
and colleagues (1998) analyzed correlations between dozens of variables and their outcomes
(i.e., mental health problems, delinquency, drug use, and arrests) in a large sample of Pittsburgh
school children. They included only variables with significant but low to moderate correlations
with the longitudinal outcomes into their CR model, so as to avoid adding risk factors that might
be proxies for the outcome measures. They dichotomized scale variables using sample estimates,
like third-quartile splits. Theirs was a rigorous method for selection and dichotomization of risk
factors, and it depended on a large heterogeneous sample and a longitudinal design. However,
using such a method with a small, clinical sample would result in invalid over- or underestimates
of risk that lack power and variability (Warner, 2013).
There is also considerable variability in how studies determine cutoffs to dichotomize
scale variables for any given construct. Dichotomization is not always a straightforward process.
For example, while researchers agree that being born to a young mother poses a risk to the youth,
what constitutes “young” ranges from under 18 to under 21 across studies and authors do not
always provide justification for these decisions (e.g. MacKenzie et al., 2014).
Yet despite the variability in methods, CR is consistently a robust predictor of negative
youth outcomes (Forehand, Biggar, & Kotchick, 1998; Loeber et al., 1998; Thornberry et al.,
2003; Mackenzie et al., 2014; Small & Luster, 1994). CR models tend to show that the number
of risk factors matters more than type when it comes to predicting negative outcomes (Evans et
CUMULATIVE RISK AS A MODERATOR OF MST

12
al., 2013). However, rigorous and consistent methods for operationalizing cumulative risk are
needed (Evans et al., 2013).
Present Study
The present study will test whether MST efficacy depends on the number of risk factors
present in the youth’s life at the start of treatment. Although MST is effective for diverse youth,
does it work as well as other interventions for youth with multiple simultaneous risk factors?
Given the tendency for juvenile offenders to face several risk factors at once, the interaction
between CR and treatment efficacy is important for a deeper understanding of MST.
To answer these questions, a secondary analysis was performed on data from a
randomized trial of MST conducted by Henggeler, Halliday-Boykins, Cunningham, Randal,
Shapiro, and Chapman (2006). MST provides an optimal platform for the study of cumulative
risk because its theoretical framework emphasizes that multiple systems, and therefore multiple
risk factors, influence youth behavior (Bronfenbrenner, 1979). The 2006 trial included drug-
abusing juvenile offenders and sought to test whether MST enhanced the effectiveness of
juvenile drug court on drug use, reoffending, rearrest, and out-of-home placement. Henggeler
and colleagues (2006) found that drug court and drug court enhanced with MST were
significantly more effective than family court, or treatment as usual, in reducing adolescent drug
use and criminal behavior, but not rates of rearrest or out-of-home placement. Contrary to
expectation, they also found that MST did not enhance the effects of drug court.
For the present study, a cumulative risk score was created for all participants in the
Henggeler et al. (2006) trial. Since there is no standard measure of cumulative risk, the method
for creating this score was based on the most common approach used in the CR literature (Evans
et al., 2013); that is, each risk factor exposure was dichotomized (0 = no risk; 1 = risk) and the
CUMULATIVE RISK AS A MODERATOR OF MST

13
dichotomous scores were summed. The present study drew from methods used by Loeber and
colleagues (1998) when appropriate for creating a cumulative risk score from available data and
includes justifications for all inclusion and dichotomization decisions.
Hypotheses
Overall, MST was expected to be more effective than drug and family court for youth
with high pretreatment levels of cumulative risk, whereas treatment effects were not expected to
differ significantly for youth with low risk across conditions. Specifically, CR was expected to
moderate MST treatment effects on rates of rearrests, drug use, and delinquency at 12-month
follow-ups, such that MST would be more effective for youth with high CR than drug and family
court. MST aims to make sustainable changes to multiple social-ecological systems (Henggeler
et al., 2009; Bronfenbrenner, 1979) in ways drug court does not uniformly aim to do (Belenko &
Dembo, 2003).
Modifiable risk was also expected to moderate efficacy, since modifiable risk factors are
explicit treatment targets for MST, but not for drug or family court. As a result, MST was
expected to be more effective for youth with high modifiable risk compared to DC and FC, while
treatment effects were not expected to differ significantly across conditions for low-risk youth.
No significant differences across conditions were expected for youth with high levels of static
risk, which do not change under any treatment condition. As a secondary hypothesis, it was also
expected that analyses would detect main effects for cumulative risk, such that higher total CR,
modifiable risk, and static risk scores would each predict less improvement at follow-up across
all conditions.
CUMULATIVE RISK AS A MODERATOR OF MST

14
Method
Participants
Data for the present study were derived from a randomized trial of drug court enhanced
with MST (Henggeler et al., 2006). One hundred sixty-one youths aged 12- to 17-years (M =
15.2, SD = 1.1) were recruited from the Department of Juvenile Justice (DJJ) in Charleston
County, South Carolina. All participants had been arrested and met DSM-IV criteria for alcohol
or drug abuse or dependence. The sample was 83% male, 68% African American
2
, and 32%
White. Median family income was in the $10,000 to $15,000 range annually, with 38% of
families receiving government financial assistance. Median educational attainment for primary
caregivers was 12
th
grade.
Treatment Conditions
Participants were randomized to one of four conditions in the Henggeler et al. (2006)
trial: family court (FC; usual services), drug court (DC), drug court enhanced with MST
(DC/MST), or drug court enhanced with MST and contingency management (DC/MST/CM).
The present study combined the two MST conditions and compared them to DC and FC for all
analyses. Drug court was a 12-month program that integrated drug treatment into the criminal
justice process. The program included regular attendance in court plus incentives and sanctions
for urine drug test results (Belenko & Dembo, 2013). Contingency management included
vouchers for clean substance screens and functional analysis of substance use behavior (Budney
& Higgins, 1998; Cunningham et al., 2003). Youth in the DC/MST condition and DC/MST/CM
condition received an average of 66 hours (SD = 32hr) and 57 hours (SD = 30hr), respectively, of
direct (i.e., face-to-face meetings with any family member present) or indirect (e.g., phone

2
Two participants were described as “other:” one “Black and Iraqi” and one “mixed Black and
White.” These are reported as African American here.
CUMULATIVE RISK AS A MODERATOR OF MST

15
contacts, school visits without youth or family present) intervention from MST therapists over a
four-month period. MST is conducted predominantly in the youth’s home for an average of 2-15
hours of contact per week. Therapy is time-limited but intense, with therapists available 24 hours
a day and seven days a week (Henggeler et al., 2009).
Measures
Pretreatment and outcome measures were obtained using multiple methods and
informants. Demographic information, including caregivers’ level of education and arrest
history, estimated annual family income, family structure, and youth’s race and enrollment status
in school, came from an extensive interview with the primary caregiver at pretreatment.
The following measures were administered at three time points: when the youth entered
the study (T1), after four months (T2) and finally at a 12-month follow-up (T3). Only
pretreatment assessments are included in the cumulative risk scores in order to create a snapshot
of the youth’s global risk at the time he or she entered treatment.
Youth characteristics. Youths’ attention problems and social problems were measured
by both the adolescent’s and caregiver’s ratings on the 113-item Youth Self Report and Child
Behavior Checklist (CBCL and YSR; Achenbach, 1990, 1991). Caregivers and youth responded
to descriptions of the youth’s behavior problems (e.g., “Can’t concentrate, can’t pay attention for
long,” “I am not liked by other kids”) on a 3-point scale from 0 (“not true”) to 2 (“often or very
true”). Unlike the competence scales on the CBCL, high scores on the syndrome scales
(including social problems and attention problems) indicate clinically important deviance
(Achenbach & Rescorla, 2001). All other individual characteristics of the youth (i.e., ethnicity,
sex, and enrollment status in school) came from the demographic interview with the primary
caregiver.
CUMULATIVE RISK AS A MODERATOR OF MST

16
Family relations. The Family Adaptability and Cohesion Evaluation Scale-III (FACES-
III; Olson, Sprenkle, & Russell, 1979) measured emotional bonding within the family (cohesion)
and the degree to which families change in response to situational stressors (adaptability). Four
levels of cohesion (i.e., disengaged, separated, connected, and enmeshed) and four levels of
adaptability (i.e., rigid, structured, flexible, and chaotic) are represented on this scale. For the
present study, families with very low or very high levels of adaptability (i.e., rigid or chaotic) or
cohesion (i.e., disengaged and enmeshed) are considered to be at risk, while those with moderate
levels fall in the normal range (Olson, Sprenkle, & Russell, 1979).
Caregiver involvement, discipline and supervision of youths over the past 90 days were
measured using a 16-item questionnaire developed for the Pittsburgh Youth Study (Loeber et al.,
1998). Adolescents and caregivers each completed versions of the same questionnaire.
Supervision and involvement were measured with questions about whether and how often
caregivers discussed plans for the day with their child, knew where their child was, and how to
get in touch with him or her (e.g., “in the past 90 days, about how often have you talked to your
child about what he/she had actually done during the day?”). Questions about discipline
addressed caregivers’ follow-through, consistency, and effectiveness of punishments (e.g., “if
you have told your child that he/she is grounded for a period of time as a punishment, would
he/she be let off [released, forgiven] before the time is up?”). Poor supervision and lax discipline
were considered risk factors in the present study.
Caregiver characteristics. Caregivers’ psychiatric symptoms over the past 90 days were
measured with the Brief Symptom Inventory (BSI; Derogatis, 1993), a 53-item self-report
measure that assesses global psychological adjustment, where higher scores indicate more severe
CUMULATIVE RISK AS A MODERATOR OF MST

17
symptoms. Respondents rated on a 5-point scale (0 = not at all; 4 = extremely) the degree to
which they had been distressed by psychiatric symptoms in the past 90 days.
The 12-item self-report Personal Experience Inventory (PEI; Winters & Henly, 1989)
measured caregiver drug use over the past 90 days. Respondents recorded the number of days
they used any of 12 categories of substances: alcohol, marijuana, LSD, other psychedelics,
cocaine, amphetamines, Quaaludes, barbiturates, tranquilizers, heroin, other narcotics, and
inhalants (e.g., sniffing glue, aerosols). Respondents were instructed to exclude substances
prescribed by a doctor. Other caregiver characteristics (i.e., arrest history and age relative to the
youth) came from the demographic interview.
Peer delinquency. Youths’ affiliation with delinquent peers was measured using a 15-
item self-report scale developed for the Pittsburgh Youth Study (PYS-PD; Loeber et al., 1998).
Items asked youth to report how many of their friends committed delinquent acts over the past 90
days, like using illegal drugs, hitting someone with the idea of hurting that person, or skipping
school without an excuse. Response options for each item ranged from 0 (“none of them”) to 4
(“all of them”).
Neighborhood poverty. Data from the 2000 United States census provided measures of
neighborhood poverty levels for each census tract (US Census Bureau, 2000). Available data
included percentage of residents with 1999 incomes below the federal poverty line, percentage of
children living below the poverty line, and median family income. Considerable overlap was
observed between all three of these indices, so percentage of residents with incomes below the
poverty line was selected for inclusion in the present study as a single poverty measure.
Youth outcomes. Henggeler and colleagues (2006) assessed youth outcomes at three
time points, but results of the present study mainly focus on T3. Youths’ substance use was
CUMULATIVE RISK AS A MODERATOR OF MST

18
assessed with the Form 90 (Miller, 1996), an interview with the youth using a calendar of the
previous 90 days to record estimated quantities of alcohol and other drugs consumed. The
interviewer first helped the youth highlight important events on the calendar and then recorded
specific quantities and types of substances consumed each day. The PEI was also administered to
youth and provided a second measure of youth drug use over the past 90 days, and was identical
to the PEI administered to caregivers described above. Delinquent behavior was measured using
the 47-item Self-Report Delinquency Scale (SRDS; Elliott, Ageton, Huizinga, Knowles, &
Canter, 1983), which assessed a range of delinquent behaviors from the past 90 days. Arrest data
were retrieved from the South Carolina DJJ computerized records and the South Carolina Law
Enforcement Division. Data was gathered from both juvenile and adult divisions because youths
aged 16 years and older could be charged as adults.
Cumulative Risk
The CR score was comprised of risk factors that have been validated across multiple
studies as prospective predictors of delinquency, drug use, or arrest. Risk factors in the present
study were drawn largely from Murray and Farrington’s (2010) review of predictors of
delinquency (their definition of which includes drug use) and conduct disorder from prospective
longitudinal studies using large community samples. Individual-level risk factors in the present
study were youth social problems, attention problems, non-enrollment in school, sex, and ethnic
minority status. Affiliation with delinquent peers was the only measured peer group risk factor.
At the family level, family poverty, single-parent family, and risky levels of adaptability,
cohesion, supervision, and discipline were included. In addition, having a young caregiver, a
caregiver with low educational attainment, current psychiatric symptoms, prior arrests, and
CUMULATIVE RISK AS A MODERATOR OF MST

19
current drug use, were all included as caregiver risk factors. Neighborhoods considered to be
high risk had high proportions of families with incomes below the federal poverty line.
Risk factors included in the present study met two criteria. First, each was validated by at
least two longitudinal studies as a prospective predictor of delinquency, drug use, and/or arrest in
a large nonclinical sample. A nonclinical sample allows for enough variability to detect specific
relationships between risk factors and an outcome of interest. Conversely, using small,
delinquent, or clinic-referred samples makes conclusions about risk factors for delinquency and
drug use more difficult, when all or most members of a sample have the outcome of interest
(Warner, 2013).
Second, the present study included only risk factors that were not indicators of service
utilization or treatment. This approach differs from other CR prediction studies that include
receipt of certain services as proxies for risk; for example, welfare support is often considered a
proxy for family poverty (e.g., Loeber et al., 1998; Farrington, 1998). Welfare is a convenient,
dichotomous variable that may be less susceptible to error than estimates of annual income, since
only families with financial need are, in theory, eligible to receive welfare support. Also, no
arbitrary dichotomization criteria for a scale variable are needed; either a family receives welfare
support or not. However, such variables are at best ambiguous risk factors because they are
actually benefits, and they identify only those families already receiving assistance or services of
some kind. This makes them less sensitive measures of risk; for example, only about half of the
families in the present data set with estimated annual incomes below the federal poverty line
were receiving government financial assistance at pretreatment, but all of those receiving
assistance had incomes below the poverty line. Thus, several service-related variables in the
present data set that other studies have found to be positive predictors of drug use, delinquency,
CUMULATIVE RISK AS A MODERATOR OF MST

20
and/or arrests were not included in the cumulative risk indices. These include receiving
government financial assistance, prior psychiatric treatment, prior hospitalization, current
treatment with psychiatric medication, being adopted, and being in a remedial or alternative
education program.
3

Some risk factors in the present CR scores merit further explanation. Affiliation with
delinquent peers as a risk factor is somewhat controversial in the CR literature; although
numerous studies have found it to be a robust predictor of delinquency (Loeber et al., 1998;
Barnes & Welte, 1986; Thornberry et al., 1994; Huizinga et al., 2003; Laird et al., 2005), Loeber
and colleagues (1998) removed it from their CR model because it was highly correlated with
self-reported delinquency. The authors argued that most delinquent acts occur in the company of
peers, so while peer delinquency differs conceptually from individual delinquency, one may
frequently be a measure of the other. However, the aim of the present study is not to predict
outcomes but to capture a multidimensional snapshot of risk as a moderator of treatment
efficacy. And while peer and individual delinquency often co-occur, they are conceptually
distinct. As such, peer delinquency is included in the present CR model.
Dichotomization. Studies of cumulative risk overwhelmingly use a common method for
calculating a risk score; all variables are dichotomized to represent risk or no risk, then summed
for a total score (for a review, see Evans et al., 2013). There are significant advantages to
dichotomizing continuous variables. First, many demographic risk factors are dichotomous by
nature (e.g., gender or being an ethnic minority), so dichotomizing continuous variables allows
for creation of a coherent, additive CR index made of counts. Second, many continuous variables
are highly skewed and have nonlinear relationships with outcome variables. If, for example, the

3
To be clear, participants who received these services were not excluded from the analyses, only
receipt of services were not counted as risk factors
CUMULATIVE RISK AS A MODERATOR OF MST

21
risk for delinquency increases sharply only for youth above a certain threshold of exposure,
dichotomization allows the researcher to assign risk meaningfully (Loeber et al., 1998). For the
present study, dichotomization criteria for scale and ordinal variables were determined using
clinical cutoffs if applicable (e.g., clinical T-scores on the CBCL for social problems), or else
criteria were borrowed from Loeber and colleagues (1998), as their risk factors and
dichotomization criteria were determined empirically to maximize predictive power.
Table 1 shows the risk factors that comprise total cumulative risk, type of risk (i.e., static
or modifiable), their primary domain of influence, references to at least two large longitudinal
studies finding significant correlations with the outcomes, their corresponding measures from the
present data set, respondent(s), and dichotomization criteria. The maximum correlation between
these risk factors was != .582. Each youth was assigned three risk scores: a total CR score out of
19, a sum of modifiable risk factors, and a sum of static risk factors. Each of these three scores
were tested as moderators of the relationship between treatment condition and the outcome
measures: self-reported delinquency at T3, drug use at T3, and arrests T1 to T2 and T2 to T3.
Results
Missing Data
Missing data varied by measure. Of the original 161 families, six did not participate in T2
assessments and 24 did not participate in T3 assessments. One hundred thirty-five families
participated in all three assessments. Follow-up was 100% (N = 161) for all youth arrests.
Attrition for other outcome measures was evenly distributed across treatment conditions. For
example, 79% of FC, 82% of DC, 76% of DC/MST, and 86% of DC/MST/CM youths completed
the Form 90 at all assessment points. Reasons for attrition included a lack of family
responsiveness when the youth was incarcerated and frequent family moves (Henggeler et al.,
CUMULATIVE RISK AS A MODERATOR OF MST

22
2006). Listwise deletion was used for four subjects with no self-report outcome data.
Little’s MCAR chi-square test was conducted to explore how missing data related to
outcome variables. There was significant evidence that the data was not missing completely at
random, "
2
(5) = 11.979, p = .035. Youth with higher rates of pretreatment delinquency and drug
use were more likely to be missing by T3 compared to those with lower rates.
A full cumulative risk score could not be computed for seven participants where data on
at least one risk factor was missing. Three of these seven participants also failed to complete
outcome measures at T3 for general delinquency and drug use, and were excluded from the
present study.
Cumulative Risk
Total cumulative risk (M = 7.24, SD = 2.54, range = 1-14, kurtosis = -1.86), static risk (M
= 4.45, SD = 2.04, range 0-9, skew = -.113, kurtosis = -.875) and modifiable risk (M = 2.78, SD
= 1.702, range 0-8, skew = .444, kurtosis = -.359) were all approximately normally distributed
across the sample. Group means for each outcome variable by level and type of cumulative risk
are presented in Table 2.
Moderation Effects
Binomial logistic regressions were conducted separately for all outcome variables to test
whether the three indices of cumulative risk moderated the relationships between treatment
condition and outcomes. For all tests, cumulative risk scores were centered around the mean and
treatment condition was contrast coded (1 = MST conditions; -1 = DC/FC) to avoid problems
with multicollinearity with the outcome measures (Warner, 2013).
Form 90 drug use. Binomial logistic regressions were conducted with abstinent days on
the F90 at T3 as a dichotomous dependent variable. Number of abstinent days out of 90 were
CUMULATIVE RISK AS A MODERATOR OF MST

23
dichotomized to represent complete abstinence or any drug use (90 abstinent days = 0; < 90 = 1).
A first analysis was performed to ascertain the effects of the interaction between total CR and
treatment condition on the likelihood that participants would report any drug use at T3. Total CR
was not a significant moderator (odds ratio = 1.03, p = .844). The overall model was not
statistically significant, χ
2
(3) = 2.552, p = .466; the model explained 2.8% (Nagelkerke R
2
) of the
variance in drug use and correctly classified only 59.2% of cases. Sensitivity was 87%,
specificity was 21.6%, positive predictive value was 45% and negative predictive value was
60%. None of the three predictor variables—condition, total CR, or the interaction—were
statistically significant. Figure 1 shows the relationship between total CR, treatment condition,
and drug use. CR is separated into bars for low (0-5 risk factors), medium (6-7 risk factors), and
high (7 or more risk factors) risk for ease of visualization, although analyses were run with CR as
a continuous variable. These divisions are based on approximate thirds for the distribution of
total CR across the sample. Figure 1 demonstrates that while none of the observed differences
are significant, MST does not appear to be more effective than DC and FC at high levels of risk.
However, between youth receiving MST, the treatment appears to be equally effective regardless
of level of risk.
Next, a binomial logistic regression was performed to ascertain the effects of the
interaction between modifiable risk and treatment condition on the likelihood that participants
would report any drug use at T3. Modifiable risk was not a significant moderator (odds ratio =
.905, p = .654). The overall model was not statistically significant, χ
2
(3) = 0.221, p = .974. The
model explained 0.2% (Nagelkerke R
2
) of the variance in drug use and correctly classified 57.0%
of cases. Sensitivity was 100%, specificity was 0%, positive predictive value was uninterpretable
(the model predicted no cases with any drug use), and negative predictive value was 100%. This
CUMULATIVE RISK AS A MODERATOR OF MST

24
model was identical to the base rate of drug use, with no significant effects detected for any of
the predictor variables.
A third binomial logistic regression was performed with static risk as the moderator.
Static risk did not moderate effects (odds ratio = 1.239, p = .278), and the overall model was not
statistically significant, χ
2
(3) = 5.624, p = .131. The model explained 6% (Nagelkerke R
2
) of the
variance in drug use and correctly classified 58.5% of cases. Sensitivity was 83.1%, specificity
was 25%, positive predictive value was 62% and negative predictive value was 52%. Static risk
was a statistically significant predictor of T3 drug use, such that increases in static risk were
associated with decreased odds of using drugs at T3 (odds ratio = .725, p = .034). Figure 2 plots
the level of static risk by treatment condition for any drug use reported on the F90 at T3. Static
risk is separated into bars for low (0-3 factors), medium (4-5 factors), and high (6 or more
factors) static risk for illustrative purposes, although analyses were run with static risk as a
continuous variable. These divisions are based on approximate thirds for the distribution of static
risk across the sample. Figure 2 illustrates that across conditions, youth with low risk appear
more likely to report any drug use at T3 than youth with higher risk, but no pattern of moderation
is discernable.
PEI drug use. Binomial logistic regressions were run on the PEI for the same
combinations of predictors described above, but with dichotomized soft drug use (alcohol and
marijuana) at T3 as the dependent variable. None of the analyses with total CR, modifiable risk,
or static risk showed significant interaction effects with treatment condition. The overall model
with total CR as the moderator was not statistically significant, χ
2
(3) = 3.851, p = .278, and the
interaction between total CR and treatment condition was also not significant (odds ratio = 1.199,
p = .230). However, increases in total CR were significantly associated with decreased
CUMULATIVE RISK AS A MODERATOR OF MST

25
likelihood of drug use at the # = .10 level (odds ratio =.815, p = .062). The overall model
explained 4.2% (Nagelkerke R
2
) of the variance in drug use and correctly classified 60% of
cases. Sensitivity was 20.8%, specificity was 91%, positive predictive value was 64.7% and
negative predictive value was 59.2%.
The overall model with modifiable risk as the moderator was not significant, χ
2
(3) = .947,
p = .814, and the interaction between modifiable risk and treatment condition was also not
significant (odds ratio = 1.218, p = .376). This model explained 1% (Nagelkerke R
2
) of the
variance in drug use and correctly classified 56.2% of cases. Sensitivity was 9.3%, specificity
was 94%, positive predictive value was 55.6% and negative predictive value was 56.3%.
Finally, the overall model with static risk was also not significant, χ
2
(3) = 3.356, p = .340,
and static risk was not a significant moderator (odds ratio = 1.210, p = .321). However, increases
in static risk significantly predicted decreased likelihood of drug use at the # = .10 level (odds
ratio = .777, p = .087). The model explained 3.6% (Nagelkerke R
2
) of the variance in drug use
and correctly classified 56.9% of cases. Sensitivity was 20%, specificity was 86.8%, positive
predictive value was 55% and negative predictive value was 57.2%.
Arrests. Separate analyses were conducted for arrests between T1 and T2 and for arrests
between T2 and T3. Unlike the self-report outcomes, arrests before T2 directly affected the
likelihood of being arrested between T2 and T3, as some youth arrested in the first phase of the
study were incarcerated during the second phase and therefore could not be arrested again
(Henggeler et al., 2006).
A set of binomial logistic regressions were first run with arrests between T1 and T2 as the
outcome variable to determine whether cumulative risk moderated MST treatment efficacy
compared to DC and FC during the course of the intervention. No significant interactions
CUMULATIVE RISK AS A MODERATOR OF MST

26
between treatment condition and any of the three risk scores were detected. The overall model
was significant for total CR; omnibus tests of model coefficients yielded χ
2
(3) = 9.787, p = .020.
In addition, increases in total CR were associated with increased likelihood of arrests between T1
and T2 (odds ratio = 1.295, p = .011), and the $
%
estimate was significant ($
%
= -.491, p = .045).
The model explained 8% (Nagelkerke R
2
) of the variance in arrests and correctly classified
63.4% of cases. Sensitivity was 25%, specificity was 88.2%, positive predictive value was 57.7%
and negative predictive value was 64.6%.
The overall model was also not significant for modifiable risk; omnibus tests of model
coefficients yielded χ
2
(3) = 3.843, p = .279. The interaction between modifiable risk and
treatment condition was not significant (odds ratio = .970, p = .879), but the $
%
estimate was
significant ($
%
= -.491, p = .045). This model explained 3.3% (Nagelkerke R
2
) of the variance in
arrests and correctly classified 63% of cases. Sensitivity was 13.3%, specificity was 94.7%,
positive predictive value was 61.5% and negative predictive value was 63.1%.
The overall model was not significant for static risk; omnibus tests of model coefficients
yielded χ
2
(3) = 5.554, p = .135. The interaction between static risk and treatment condition was
not significant (odds ratio = .838, p = .301), but the $
%
estimate was significant ($
%
= -.528, p =
.027). Increases in static risk were associated with increased likelihood of arrests between T1 and
T2 (odds ratio = 1.313, p = .039). The model explained 4.7% (Nagelkerke R
2
) of the variance in
arrests and correctly classified 61.6% of cases. Sensitivity was 9.7%, specificity was 94.8%,
positive predictive value was 54.5% and negative predictive value was 62.2%.
Binomial logistic regressions were next run on arrests between T2 and T3 for the same
combinations of predictors described above. None of the overall models, interaction effects, or
main effects were significant. All models estimated the odds of arrest at less than .5 for all
CUMULATIVE RISK AS A MODERATOR OF MST

27
participants, resulting in 0% sensitivity. However, $
%
estimates for all three models (total CR:
$
%
= -.469, p = .044; static: $
%
= -.458, p = .048; and modifiable risk: $
%
= -.478, p = .041) were
significant.
Delinquency. Binomial logistic regressions were also run using dichotomized scores on
the SRDS General Delinquency scale at T3. Models aimed to predict the likelihood of youths
reporting any delinquency at T3 with treatment condition as the independent variable and the
three risk scores as moderators. The overall model testing with total CR was not statistically
significant, χ
2
(3) = 2.620, p = .454, and total CR was not a significant moderator (odds ratio =
1.265, p = .138). The overall model explained 3.1% (Nagelkerke R
2
) of the variance in
delinquency and correctly classified 66.9% of cases. Sensitivity was 98.7%, specificity was
2.6%, positive predictive value was 67.2%, and negative predictive value was 0.9%. While the
estimate for $
%
was significant ($
%
= .597, p = .030), none of the three predictor variables—
condition, total CR, or the interaction—were statistically significant.
The overall model with modifiable risk was not statistically significant, χ
2
(3) = 1.076, p =
.783, and the interaction between modifiable risk and treatment condition was not significant
(odds ratio = 1.147, p = .559). The overall model explained 1.3% (Nagelkerke R
2
) of the variance
in delinquency and correctly classified 67.2% of cases. Sensitivity was 100%, specificity was
0%, positive predictive value was 67.2% and negative predictive value was not interpretable, as
the model predicted no cases without delinquency. That is, this model estimated the odds of
reporting any delinquency as greater than .5 for all youth, resulting in 0% specificity, and was
identical to the base rate. While the estimate for $
%
was significant ($
%
= .578, p = .035), none of
the three predictor variables—condition, modifiable risk, or the interaction—were statistically
significant.
CUMULATIVE RISK AS A MODERATOR OF MST

28
The analysis for static risk yielded similar results. The overall model was not statistically
significant, χ
2
(3) = 2.460, p = .483, and the interaction between static risk and treatment
condition was not significant (odds ratio = 1.328, p = .163). This model explained 2.8%
(Nagelkerke R
2
) of the variance in delinquency and correctly classified 66.9% of cases.
Sensitivity was 100%, specificity was 0%, positive predictive value was 66.9% and negative
predictive value was not interpretable, as the model predicted no cases without delinquency. That
is, this model estimated the odds of reporting any delinquency as greater than .5 for all youth,
resulting in 0% specificity, and was identical to the base rate. While the $
%
estimate for static risk
was significant ($
%
= .639, p = .020), none of the three predictor variables—condition, static
risk, or the interaction—were statistically significant.
Discussion
The purpose of this paper was to study the role of cumulative risk as a moderator of MST
treatment effects. CR was expected to moderate efficacy such that MST would show differential
effects across levels of risk compared to DC and FC. For youth with high CR, MST was
expected to be more effective than comparison conditions, while for youth with lower levels of
risk, MST and comparison conditions were expected to show similar effects. Although the MST
literature has shown that many risk factors do not moderate treatment effects (Henggeler et al.,
2011; Henggeler et al., 1992), studies of individual moderators do not account for overall level of
risk. There was reason to suspect that high CR would moderate MST effects, given evidence
from the literature that CR is a strong predictor of negative youth outcomes (e.g., Loeber et al.,
1998; Thornberry et al., 2003), treatment attrition (Kazdin et al., 1993; Bagner & Graziano,
2013), and diminished treatment effects (Bagner & Graziano, 2013). If CR had moderated MST
effects in the expected direction, it would have suggested that while MST is flexible enough to
CUMULATIVE RISK AS A MODERATOR OF MST

29
treat youth any risk factors, there may be limits to the overall level of risk MST therapists can
effectively treat.
Moreover, modifiable risk was expected to have a significant moderating effect. Several
modifiable risk factors are explicit primary or secondary treatment targets in MST, including
parental discipline and supervision, and family adaptability and cohesion (Henggeler et al.,
2009), but not in drug or family court. As a result, MST was expected to be more effective for
youth with high modifiable risk than drug and family court. MST and comparison conditions
were expected to show roughly equal effects across levels of static risk, since these factors
cannot change in treatment. Secondarily, cumulative risk scores were expected to produce main
effects in all analyses, such that increases in risk scores would generally be associated with
greater likelihood of drug use, delinquency and arrests at T3 across conditions.
None of the three indices of cumulative risk in the present study were significant
moderators in any analysis. One interpretation of these null findings is that MST is equally
effective for youth with high and low levels of cumulative risk. These results are consistent with
dozens of trials showing that MST is an effective intervention for diverse youth and that efficacy
does not differ by race, gender, symptom severity or a host of other risk factors (Henggeler,
2011; Henggeler et al., 1992; Sundell et al., 2008). The present study adds cumulative risk to the
growing list of variables that do not moderate MST efficacy, and suggests that youth with
multiple simultaneous risk factors for offending benefit from this therapy as much as their lower
risk peers.
Also, contrary to expectation, modifiable risk was not a significant moderator of efficacy.
Indeed, modifiable risk had no significant moderating effect on any outcome measure. This was
surprising because changes in modifiable risk factors in particular were expected to highlight the
CUMULATIVE RISK AS A MODERATOR OF MST

30
differences between MST’s social-ecological approach (Bronfenbrenner, 1979) and the more
suppression-focused drug and family court conditions (Belenko & Dembo, 2003).
Results for main effects were also contrary to expectations. The three CR scores were not
consistently associated with poorer treatment outcomes; indeed, increases in static risk predicted
decreased likelihood of drug use on the PEI and the F90. The lack of consistent main effects for
CR has implications for its utility in the context of psychosocial interventions. While CR is
typically a strong predictor of negative outcomes like criminal behavior, the results of the present
study suggest it may not be a useful predictor of success or failure in treatment. This may be a
result of having what amounts to a truncated sample; since MST clients in this study are all
offenders, there may not be enough variability within the sample for CR to have significant
predictive power (Warner, 2013). The few significant main effects for static risk and total CR
may even be attributable to random noise, given the number of analyses conducted (Warner,
2013) and the unexpected direction of some effects.
Limitations
There are several limitations to the present study. These data do not include subjects’
comprehensive risk history. Child maltreatment and other trauma, important risk factors for
offending (Baglivio et al., 2014), were not available, so the risk factors in the CR indices may
thus be poor reflections of subjects’ true cumulative risk. This is a common problem in the CR
literature, as clients’ trauma history is not always available to researchers using secondary data
(e.g., MacKenzie et al., 2014) to construct CR scores. Additionally, the available data limited the
number of risk factors in certain domains. Neighborhood and peer risk were each represented by
only one available factor, while data on family risk were much richer. While presenting a
CUMULATIVE RISK AS A MODERATOR OF MST

31
somewhat imbalanced picture, the greater presence of family risk factors may be less
problematic in the context of a family based intervention like MST.
The heterogeneity of methods in the CR literature points to some inherent limitations to
CR as a measure. First, there is no standard CR measure. Risk factors are often drawn from the
available data rather than determined a priori, and often without explicit justifications for
including some risk factors and not others (e.g., MacKenzie et al., 2014). Dichotomization leads
to a loss of information concerning severity, dosage, and frequency of risk exposure (Loeber et
al., 1998; Warner, 2013). Dichotomization criteria in the literature are often described in vague
terms, such as at “conceptually meaningful cutoff points” (Small & Luster, 1994), or are based
on sample estimates like third-quartile splits (Loeber et al., 1998) that do not readily generalize
to other samples. Some of Loeber and colleagues (1998)’s cut points for dichotomizing scale
variables were used in the present study because no standardization or consensus exists in the
literature.
However, there are several advantages to using dichotomous variables that seem to
outweigh the limitations (Loeber et al., 1998). Dichotomization solves some of the problems
presented when working with highly skewed data. Many demographic risk factors are
dichotomous by nature, and dichotomization allows for creation of a uniform cumulative risk
score that risk factors equally. While an alternative option may be to use continuous variables
when possible in ordinary least squares multiple regression, Loeber and colleagues found similar
results when they compared this method to logistic regression using dichotomous data.
Cumulative risk scores constructed like those in the present study give equal weight to
each risk factor. This is almost certainly a poor reflection of the relative impact of each risk
exposure on the youth, and yet the majority of CR models use this method (Evans et al., 2013). A
CUMULATIVE RISK AS A MODERATOR OF MST

32
simple sum of risk factors in a CR score assumes an additive effect on outcomes, but the reality
may be more complex (Evans et al., 2013).
Missing data was also an important limitation of this study, with approximately 25%
attrition by T3. Participants were not missing at random; youth with greater pretreatment
delinquency and drug use were the most likely be lost to follow-up by T3, often as a result of
being incarcerated (Henggeler et al., 2006). These participants may otherwise have had
significant effects on group means if they were not missing. Listwise deletion was used here, but
one future direction for this research is to re-run analyses using an imputation method to
determine whether results may differ based on missing data (Warner, 2013).
Another limitation concerns internal validity. Treatment conditions in the present study
were collapsed from four to two in order to compare youth receiving MST to youth in other
conditions. This decision was made to increase statistical power by increasing sample size
(Lenth, 2006-9), but also created greater heterogeneity within each condition, possibly resulting
in a loss of statistical power (Shadish, Cook, & Campbell, 2002). Henggeler and colleagues
(2006) found that drug court was significantly more effective than family court, but effects were
not significantly different from MST conditions, suggesting that heterogeneity within the
combined DC and FC condition increased the likelihood of a type 2 error. An alternative to
combining groups would be to compare only the DC condition to the DC/MST condition in order
to reduce heterogeneity and isolate the effects of MST on any differences in outcomes. However,
with only 40 youth in each condition at T1 and 25% attrition, this comparison would have lacked
statistical power to detect effects (Lenth, 2006-9). Future research should explore cumulative risk
as a moderator within other randomized trials of MST with larger samples to determine whether
the null effects of the present study replicate across trials.
CUMULATIVE RISK AS A MODERATOR OF MST

33
Conclusion
This is the first paper to explore the moderating role of cumulative risk in the context of
any psychosocial intervention for juvenile offenders. The cumulative risk scores created in the
present study were not significant moderators of drug use, delinquency, or arrests in this
randomized trial of MST. While CR is typically a strong predictor of negative outcomes in the
normative population (Evans et al., 2013), it may be a poor indicator of who benefits from
therapy. This is encouraging from a policy perspective; while high CR can predict who needs
therapy, it does not appear to predict that the same youth will fail to benefit from therapy.
This paper adds to the literature on risk by establishing a set of transparent, rigorous
criteria for creating CR score. The methods presented here improve on those from many studies
that neither specify nor justify criteria for including risk factors in a CR score. The inclusion
criteria presented in this paper may help make the use of CR more meaningful in future clinical
research where sample-based metrics are not possible or valid.

CUMULATIVE RISK AS A MODERATOR OF MST

34
References
Achenbach, T. M. (1991). Manual for the child behavior checklist and 1991 profile. Burlington:
University of Vermont, Department of Psychiatry.
Achenbach, T. M. & Rescorla, L. A. (2001). Manual for the ASEBA school-age forms &
profiles: An integrated system of multi-informant assessment. Burlington, VT: ASEBA.
Ahonen, L., Loeber, R., & Pardini, D. (2016). The prediction of young homicide and violent
offenders. Justice Quarterly, 33(7) 1265–1291.
Bagner, D., & Graziano, P. (2013). Barriers to success in parent training for young children with
developmental delay: The role of cumulative risk. Behavior Modification, 37(3), 356–
377.
Baglivio, M., Epps, N., Swartz, K., Huq, M. S., Sheer, A., & Hardt, N. S. (2014). The prevalence
of adverse childhood experiences (ACE) in the lives of juvenile offenders. Journal of
Juvenile Justice, 3(2), 1–23.
Barnes, G. M., & Welte, J. W. (1986). Patterns and predictors of alcohol use among 7-12th grade
students in New York state. Journal of Studies on Alcohol, 47(1), 53-62.
Belenko, S., & Dembo, R. (2003). Treating adolescent substance abuse problems in the juvenile
drug court. International Journal of Law and Psychiatry, 26(1), 87-110.
Blum, H. M., Boyle, M. H., & Offord, D. R. (1988). Single-parent families: Child psychiatric
disorder and school performance. Journal of the American Academy of Child &
Adolescent Psychiatry, 27(2), 214-219.
CUMULATIVE RISK AS A MODERATOR OF MST

35
Borduin, C. M., Mann, B. J., Cone, L. T., Henggeler, S. W., Fucci, B. R., Blaske, D. M. et al.
(1995). Multisystemic treatment of serious juvenile offenders: Long-term prevention of
criminality and violence. Journal of Consulting and Clinical Psychology, 63, 569–578.
Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and
design. Cambridge, MA: Harvard University Press.
Brownfield, D., & Sorenson, A. (1994). Sibship size and sibling delinquency. Deviant
Behavior, 15(1), 45-61.
Budney, A. J., & Higgins, S. T. (1998). A community reinforcement plus vouchers approach:
Treating cocaine addiction. Rockville, Maryland: National Institute on Drug Abuse.
Cashwell, C., & Vacc, N. (1996). Family functioning and risk behaviors: Influences on
adolescent delinquency. The School Counselor, 44(2), 105-114.
Crum, R. M., Ensminger, M. E., Ro, M. J., & McCord, J. (1998). The association of educational
achievement and school dropout with risk of alcoholism: A twenty-five-year prospective
study of inner-city children. Journal of Studies on Alcohol, 59(3), 318-326.
Cunningham, P. B., Donohue, B., Randall, J., Swenson, C. C., Rowland, M. D., Henggeler, S.
W., & Schoenwald, S. K. (2003). Integrating contingency management into multisystemic
therapy. Charleston: Medical University of South Carolina, Family Services Research
Center.
Derogatis, L. R. (1993). Brief symptom inventory. Administration, scoring, and procedures
manual. Minneapolis, MN: Clinical Psychometric Research.
Dierkhising, C. B., Ko, S. J., Woods-Jaeger, B., Briggs, E. C., Lee, R., & Pynoos, R. (2013).
Trauma histories among justice-involved youth: Findings from the National Child
Traumatic Stress Network. European Journal of Psychotraumatology, 4, 1–12.
CUMULATIVE RISK AS A MODERATOR OF MST

36
Drapela, L. A. (2005). Does dropping out of high school cause deviant behavior? An analysis of
the national education longitudinal study. Deviant Behavior, 26(1), 47-62.
Elliott, D. S., Ageton, S. S., Huizinga, D., Knowles, B. A., & Canter, R. J. (1983). The
prevalence and incidence of delinquent behavior: 1976-80 (Report of the National Youth
Survey, Project Report No. 26). Boulder, CO.
Evans, G., Li, D., & Whipple, S. (2013). Cumulative risk and child development. Psychological
Bulletin, 139(6), 1342-1396.
Farrington, D. P. (1995). The development of offending and antisocial behaviour from
childhood: Key findings from the Cambridge study in delinquent development. Journal
of Child Psychology and Psychiatry and Allied Disciplines, 36(6), 929-964.
Farrington, D. P. (1998). Predictors, causes, and correlates of male youth violence. Crime and
Justice, 24, 421-475.
Farrington, D. P., Barnes, G. C., & Lambert, S. (1996). The concentration of offending in
families. Legal and Criminological Psychology, 1(1), 47-63.
Farrington, D. P., Jolliffe, D., Loeber, R., Stouthamer-Loeber, M., & Kalb, L. M. (2001). The
concentration of offenders in families, and family criminality in the prediction of boys'
delinquency. Journal of Adolescence, 24(5), 579-596.
Farrington, D. P., Ttofi, M. M., & Piquero, A. R. (2016). Risk, promotive, and protective factors
in youth offending: Results from the Cambridge study in delinquent development.
Journal of Criminal Justice, 45, 63-70.
Felitti, V. J., Anda, R. F., Nordenberg, D., Williamson, D. F., Spitz, A. M., Edwards, V., …
Marks, J. S. (1998). Relationship of childhood abuse and household dysfunction to many
CUMULATIVE RISK AS A MODERATOR OF MST

37
of the leading causes of death in adults: The adverse childhood experiences (ACE) study.
American Journal of Preventive Medicine, 14(4), 245–258.
Felson, R. B., Deane, G., & Armstrong, D. P. (2008). Do theories of crime or violence explain
race differences in delinquency? Social Science Research, 37(2), 624-641.
Forehand, R., Biggar, H., & Kotchick, B. (1998). Cumulative risk across family stressors: Short-
and long-term effects for adolescents. Journal of Abnormal Child Psychology, 26(2),
119–128.
Hawkins, D. F., Herrenkohl, T., Farrington, D., Brewer, D., Catalano, R., & Harachi, T. (1998).
A review of predictors of youth violence. In R. Loeber & D. Farrington (Eds.), Serious
and Violent Juvenile Offenders (pp. 106–146). Thousand Oaks, CA: Sage Publications.
Hawkins, D. F., Laub, J. H., & Lauritsen, J. L. (1998). Race, ethnicity, and serious juvenile
offending. In R. Loeber & D. P. Farrington (Eds.), Serious and Violent Juvenile
Offenders (pp. 30–46). Sage Publications.
Henggeler, S. W. (2011). Efficacy studies to large-scale transport: The development and
validation of multisystemic therapy programs. Annual Review of Clinical Psychology, 7,
351–381.
Henggeler, S. W., Halliday-Boykins, C. A., Cunningham, P. B., Randall, J., & Shapiro, S. B.
(2006). Juvenile drug court: Enhancing outcomes by integrating evidence-based
treatments. Journal of Consulting and Clinical Psychology, 74(1), 42–54.
Henggeler, S. W., Melton, G. B., & Smith, L. A. (1992). Family preservation using
multisystemic therapy: An effective alternative to incarcerating serious juvenile
offenders. Journal of Consulting and Clinical Psychology, 60(6), 953–961.
CUMULATIVE RISK AS A MODERATOR OF MST

38
Henggeler, S. W., Rodick, J., Borduin, C. M., Hanson, C. L., Watson, S. M., & Urey, J. R.
(1986). Multisystemic treatment of juvenile offenders: Effects on adolescent behavior
and family interaction. Developmental Psychology, 22, 132-141.
Henggeler, S. W., & Schaeffer, C. M. (2016). Multisystemic therapy ®: Clinical overview,
outcomes, and implementation research. Family Process, 55(3), 514-528.
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.). New York, NY: Guilford Press.
Herrenkohl, T. I., Maguin, E., Hill, K. G., Hawkins, J. D., Abbott, R. D., & Catalano, R. F.
(2000). Developmental risk factors for youth violence. Journal of Adolescent Health,
26(3), 176-186.
Hoffman, S. D., & Maynard, R. A. (2008). Kids having kids: Economic costs & social
consequences of teen pregnancy. Washington, DC: The Urban Institute Press.
Huizinga, D., Loeber, R., Thornberry, T. P., & Cothern, L. (2000). Co-occurrence of delinquency
and other problem behaviors. Washington, DC: U.S. Dept. of Justice, Office of Justice
Programs, Office of Juvenile Justice and Delinquency Prevention.
Jang, S. J., & Krohn, M. D. (1995). Developmental patterns of sex differences in delinquency
among African American adolescents: A test of the sex-invariance hypothesis. Journal of
Quantitative Criminology, 11(2), 195-222.
Johnson, G. M., Shontz, F. C., & Locke, T. P. (1984). Relationships between adolescent drug use
and parental drug behaviors. Adolescence, 19(74), 295-295.
Kandel, D., Simcha-Fagan, O., & Davies, M. (1986). Risk factors for delinquency and illicit drug
use from adolescence to young adulthood. Journal of Drug Issues, 16(1), 67.
CUMULATIVE RISK AS A MODERATOR OF MST

39
Kazdin, A. E., Mazurick, J., & Bass, D. (1993). Risk for attrition in treatment of antisocial
children and families. Journal of Clinical Child Psychology, 22(1), 2–16.
Laerd Statistics (2015). Binomial logistic regression using SPSS Statistics. Statistical tutorials
and software guides. Retrieved from https://statistics.laerd.com/
Laird, R. D., Pettit, G. S., Dodge, K. A., & Bates, J. E. (2005). Peer relationship antecedents of
delinquent behavior in late adolescence: Is there evidence of demographic group
differences in developmental processes? Development and Psychopathology, 17(1), 127-
144.
Lenth, R. V. (2006-9). Java applets for power and sample size [computer software]. Retrieved
April 5, 2016, from http://www.stat.uiowa.edu/~rlenth/Power.
Lipsey, M. W., & Derzon, J. H. (1998). Predictors of violent or serious delinquency in
adolescence and early adulthood. In R. Loeber & D. P. Farrington (Eds.), Serious and
Violent Juvenile Offenders (pp. 86–105). Thousand Oaks, CA: Sage Publications.
Loeber, R., Farrington, D., Stouthamer-Loeber, M., & Van Kammen, W. (1998). Antisocial
behavior and mental health problems: Explanatory factors in childhood and adolescence.
Mahwah, NJ: Lawrence Erlbaum Associates.
Lynne-Landsman, S., Bradshaw, C. P., & Ialongo, N. S. (2010). Testing a developmental
cascade model of adolescent substance use trajectories and young adult adjustment.
Development and Psychopathology, 22(4), 933-48.
MacKenzie, M., Nicklas, E., Brooks-Gunn, J., & Waldfogel, J. (2014). Repeated exposure to
high-frequency spanking and child externalizing behavior across the first decade: A
moderating role for cumulative risk. Child Abuse & Neglect, 38(12), 1895-1901.
CUMULATIVE RISK AS A MODERATOR OF MST

40
Maguin, E., Hawkins, J.D., Catalano, R.F., Hill, K., Abott, R., Herrenkohl, T. (1995, November).
Risk factors measured at three ages for violence at age 17-18. Paper presented at the
American Society of Criminology, Boston.
McCord, J. (1979). Some child-rearing antecedents of criminal behavior in adult men. Journal of
Personality and Social Psychology, 37(9), 1477-1486.
Miller, W. R. (1996). Form 90: A structured assessment interview for drinking and related
behaviors: Test manual. Washington, DC: National Institute on Alcohol Abuse and
Alcoholism.
Murray, J., & Farrington, D. (2010). Risk factors for conduct disorder and delinquency: Key
findings from longitudinal studies. Canadian Journal of Psychiatry, 55(10), 633-42.
Olson, D. H., Sprenkle, D. H., & Russell, C. S. (1979). Circumplex model of marital and family
system: Cohesion and adaptability dimensions, family types, and clinical applications.
Family Process, 18(1), 3-28.
Rhatigan, K. J. (2002). The relationships between family functioning, adolescent delinquent
behavior, and life satisfaction in single- and dual-career families (Doctoral dissertation).
Hofstra University. Retrieved from ProQuest Dissertations & Theses Global.
Robins, L. N., & Robins, L. N. (1978). Sturdy childhood predictors of adult antisocial behaviour:
Replications from longitudinal studies. Psychological Medicine, 8(4), 611-622.
Sakyi, K. S., Melchior, M., Chollet, A., & Surkan, P. J. (2012). The combined effects of parental
divorce and parental history of depression on cannabis use in young adults in France.
Drug and Alcohol Dependence, 126(1-2), 195-199.
CUMULATIVE RISK AS A MODERATOR OF MST

41
Sawyer, A. M., & Borduin, C. M. (2011). Effects of MST 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, 643–652.
Schaeffer, C. M. (2000). Moderators and mediators of therapeutic change in multisystemic
treatment of serious juvenile offenders (Doctoral dissertation). University of Missouri -
Columbia. Retrieved from ProQuest Dissertations & Theses Global.
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, 445–453.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental
designs for generalized causal inference. Boston: Houghton Mifflin.
Small, S. A., & Luster, T. (1994). Adolescent sexual activity: An ecological risk-factor approach.
Journal of Marriage and Family, 56(1), 181–192.
Sundell, K., Hansson, K., Löfholm, C. A., Olsson, T., Gustle, L., & Kadesjö, C. (2008). The
transportability of multisystemic therapy to Sweden: Short-term results from a
randomized trial of conduct-disordered youths. Journal of Family Psychology, 22(4),
550-560.
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed. ed.) Allyn &
Bacon/Pearson Education, Boston, MA.
Thornberry, T.P. & Krohn, M.D. (Eds.). (2002). Taking stock of delinquency: An overview of
findings from contemporary longitudinal studies. Secaucus, US: Kluwer Academic
Publishers.
CUMULATIVE RISK AS A MODERATOR OF MST

42
Thornberry, T. P., Krohn, M. D., Lizotte, A. J., Smith, C. A., & Tobin, K. (2003). Gangs and
delinquency in developmental perspective. Cambridge, U.K.: Cambridge University
Press.
Thornberry, T. P., Lizotte, A. J., Krohn, M. D., Farnworth, M., & Jang, S. J. (1994). Delinquent
peers, beliefs, and delinquent behavior: A longitudinal test of interactional theory.
Criminology, 32(1), 47.
United States Census Bureau. (2000). [Charleston, SC: 1999 Income Below Poverty Level].
Social Explorer. Retrieved from http://www.socialexplorer.com/6f4cdab7a0/explore
Velez, C. N., Johnson, J., & Cohen, P. (1989). A longitudinal analysis of selected risk factors for
childhood psychopathology. Journal of the American Academy of Child & Adolescent
Psychiatry, 28(6), 861-864.
Wadsworth, M. E. J. (1979). Roots of delinquency: Infancy, adolescence, and crime. New York:
Barnes & Noble Books.
Warner, R. M. (2013). Applied statistics: From bivariate through multivariate techniques.
Thousand Oaks, CA: SAGE Publications, Inc.
West, D. J. & Farrington, D. P. (1973). Who becomes delinquent? Second report of the
Cambridge study in delinquent development. London: Heinemann Educational.
Wikström, P. H., & Loeber, R. (2000). Do disadvantaged neighborhoods cause well-adjusted
children to become adolescent delinquents? A study of male juvenile serious offending,
individual risk and protective factors, and neighborhood context. Criminology, 38(4),
1109-1142.
Winters, K. C., & Henly, G. A. (1989). Personal Experience Inventory and manual. Los
Angeles: Western Psychological Services.
CUMULATIVE RISK AS A MODERATOR OF MST

43
Zhao, L., Tang, J., & Li, K. (2011). A study of the relationship of aggressive, social support,
family adaptability and cohesion of the left-behind delinquents. Chinese Journal of
Clinical Psychology, 19(6), 790-791.

CUMULATIVE RISK AS A MODERATOR OF MST 41
Tables and Figures
Table 1

Risk Factors Included in the Cumulative Risk (CR) Indices
Risk Factor Domain Type References Measure

Respondent Dichotomization Criteria
Ethnic minority Individual Static Loeber et al. (1998; PYS);
Felson, Deane & Armstrong
(2008)
Demographic
interview
CG Youth not White
Family poverty family Static Farrington (1995); MacKenzie
et al. (2014); Tremblay et al
(2003)
Demographic
interview
CG Estimated family income below
federal poverty guidelines for
family size based on the 2000
US Census
Large family family Static Loeber et al. (1998; PYS);
Wadsworth (1979);
Brownfield & Sorenson (1994)
Demographic
interview
CG Four or more children under 18
in the home
Lax discipline family Mod Loeber et al. (1988; PYS);
West & Farrington (1973)
PYS
discipline
CG, Y Mean score of PYS discipline
items > 2 (“sometimes”). Either
or both respondents at risk = 1
Low caregiver
education
caregiver Static Loeber et al. (1998; PYS);
Kandel et al. (1986)
Demographic
interview
CG Either parent completed less
than high school education
Neighborhood
poverty
neighborhood Static Loeber et al. (1998; PYS);
Maguin et al. (1995; Seattle
Social Development Project)
US census
data
Residents Greater than 25% of families
below the poverty line
Peer delinquency Peer Mod Loeber et al. (1998; PYS);
Barnes & Welte (1986);
Thornberry et al. (1994);
Huizinga et al. (2003); Laird et
al (2005; Child Development
Project)
PYS-PD Y PYS-D means score for all
items ≥ 2 (“half of them”)
CUMULATIVE RISK AS A MODERATOR OF MST 42
Table 1

Risk Factors Included in the Cumulative Risk (CR) Indices
Risk Factor Domain Type References Measure

Respondent Dichotomization Criteria
Poor family
adaptability
family Mod Rhatigan (2002); Cashwell &
Vaac (1996); Zhao et al.
(2011)
FACES-III CG FACES-III adaptability score <
19 or > 30. Either or both
respondents at risk = 1
Poor family
cohesion
family Mod Rhatigan (2002); Cashwell &
Vaac (1996); Zhao et al.
(2011)
FACES-III CG FACES-III cohesion score < 31
or > 44. Either or both
respondents at risk = 1
Poor supervision family Mod Loeber et al. (1998; PYS);
McCord (1979); Robins &
Robins (1978); Herrenkohl et
al. (2000)
PYS
supervision
CG, Y Mean score of PYS supervision
items ≤ 2 (“sometimes”). Either
or both respondents at risk = 1
Primary caregiver
current drug use
caregiver Mod Loeber et al. (1998; PYS);
Johnson et al. (1984)
PEI CG Any illegal drug use reported
on PEI
Primary caregiver
current psychiatric
symptoms
caregiver Mod Loeber et al. (1998; PYS);
Sakyi et al. (2012; TEMPO)
BSI CG BSI Global Severity Index T-
score ≥ 63
Primary/secondary
caregiver arrest
history
caregiver Static Farrington et al. (1996);
McCord (1979); Robins &
Robins (1978); Farrington et
al. (2001); Thornberry &
Krohn (2002)
Demographic
interview
CG Either or both parents ever
arrested
Sex Individual Static Huizinga et al. (2000); Jang &
Krohn (1995; RYDS)
Demographic
interview
CG Male
Single parent
family
Family Static Loeber et al. (1998); Velez et
al. (1989); Blum et al. (1988)
Demographic
interview
CG One caregiver
Young caregiver Caregiver Static Loeber et al. (1998);
Farrington, Ttofi, & Piquero
(2016; Cambridge Study)
Demographic
interview
CG Primary caregiver < 20 years
older than youth
CUMULATIVE RISK AS A MODERATOR OF MST 43
Table 1

Risk Factors Included in the Cumulative Risk (CR) Indices
Risk Factor Domain Type References Measure

Respondent Dichotomization Criteria
Youth attention
problems
Individual Static Loeber et al. (1998);
Farrington (1995); Maguin et
al. (1995; SSDP)
CBCL/YSR CG, Y T score for attention problems
scale >63
Youth not in
school
Individual Mod Loeber et al. (1998);
Farrington (1989); Crum et al.
(1998); Drapela (2005)
Demographic
interview
CG Youth not currently attending
school
Youth social
problems
Individual Static Loeber et al. (1998); Lynne-
Landsman, Bradshaw &
Ialongo (2010)
CBCL/YSR CG, Y T score on social problems
scale > 63
Note. Mod = modifiable; RYDS = Rochester Youth Development Study; PYS = Pittsburgh Youth Study; SSDP = Seattle Social Development Project; TEMPO =
Trajectoires Épidémiologiques en Population Study (France); BSI = Brief Symptom Inventory; CBCL = Child Behavior Checklist; FACES-III = Family Adaptability and
Cohesion Evaluation Scale-III; PEI = Personal Experience Inventory; PYS-PD, = Pittsburgh Youth Study Peer Delinquency Scale; YSR = Youth Self Report; CG =
caregiver; Y = youth.

CUMULATIVE RISK AS A MODERATOR OF MST 44

Table 2

Group Means for Outcome Variables by Level of Risk Across All Conditions
PEI Soft
Drug use T3
Arrested
yes/no T1-T2
Arrested
yes/no T2-T3
Number of
Abstinent Days
(F90) T3
SRDS
General
Delinquency
T3
Total Cumulative Risk
Low (0-5) Mean 17.6875 .2564 .3590 71.5313 36.3548
N 32 39 39 32 31
Std. Deviation 32.15029 .44236 .48597 30.17260 90.53417
Medium (6-
7)
Mean 6.8409 .3704 .4444 82.8605 14.7381
N 44 54 54 43 42
Std. Deviation 25.85194 .48744 .50157 19.81171 44.45555
High (>7) Mean 10.5000 .5000 .4167 77.4667 26.8444
N 44 60 60 45 45
Std. Deviation 23.35918 .50422 .49717 25.93926 55.78977
Static Risk
Low Mean 20.5814 .2727 .3455 75.1395 38.6429
N 43 55 55 43 42
Std. Deviation 37.89034 .44947 .47990 26.50748 88.90714
Medium Mean 5.3158 .3750 .3958 78.0526 19.2222
N 38 48 48 38 36
Std. Deviation 11.06273 .48925 .49420 24.14589 47.95064
High Mean 5.8571 .5179 .5179 81.1429 15.2093
N 42 56 56 42 43
Std. Deviation 19.51806 .50420 .50420 24.79139 36.52889
Modifiable Risk
CUMULATIVE RISK AS A MODERATOR OF MST 45

Low Mean 5.5625 .3684 .5263 83.4194 10.8387
N 32 38 38 31 31
Std. Deviation 16.82728 .48885 .50601 20.37609 18.11831
Medium Mean 13.3400 .3030 .3788 69.7647 28.4490
N 50 66 66 51 49
Std. Deviation 28.16541 .46309 .48880 32.00412 79.27407
High Mean 12.7949 .5200 .3800 83.8718 34.2051
N 39 50 50 39 39
Std. Deviation 31.40079 .50467 .49031 13.85865 64.49528
Total Mean 11.1074 .3896 .4156 77.8099 25.7479
N 121 154 154 121 119
Std. Deviation 26.84238 .48925 .49443 25.27789 63.75286

CUMULATIVE RISK AS A MODERATOR OF MST 46

Figure 1. F90 any drug use at T3 by total CR

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Low CR (0-5 risk factors) Medium CR (6-7 risk factors) High CR (8+ risk factors)
Proportion Reporting Any Drug Use on F90 at T3
by Total CR
DC/MST & DC/MST/CM DC & FC
CUMULATIVE RISK AS A MODERATOR OF MST 47

Figure 2. F90 any drug use at T3 by static risk

0
0.1
0.2
0.3
0.4
0.5
0.6
Low Static (0-3 risk factors) Medium Static (4-5 risk
factors)
High Static (6+ risk factors)
Proportion Reporting Any Drug Use on F90 at T3
by Static Risk
DC/MST & DC/MST/CM DC & FC

Abstract (if available)

Abstract The present study examined cumulative risk (CR) for delinquency, drug use, and arrests as a moderator of treatment efficacy in a randomized trial of multisystemic therapy (MST) compared to drug court (DC) and family court (FC) (Henggeler, Halliday-Boykins, Cunningham, Randall, Shapiro, & Chapman, 2006). Participants were 161 juvenile drug offenders arrested in Charleston, SC and their families. Although CR is a robust predictor of juvenile offending, no studies to date have examined its role as a moderator of treatment efficacy for youth psychosocial interventions. MST is an effective treatment for juvenile offenders that aims to address contextual risk factors that contribute to criminal behavior. The MST literature has yet to address two important questions: (1) is MST more effective than other interventions or usual services across youth with different levels of cumulative risk, and (2) do potentially modifiable risk factors and static risk factors moderate MST effects differently? The author hypothesized that MST would be more effective than comparison conditions for youth with high CR, while effects would not differ significantly across conditions for youth with low levels of risk. In particular, participants with many risk factors in potentially modifiable domains were expected to benefit the most from MST, compared to those in other conditions. Data did not support these hypotheses

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Within session therapist behavior in multisystemic therapy

PDF

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

PDF

Reducing gang involvement through employment: a pilot intervention

PDF

Revision of the multisystemic therapy (MST) adherence coding protocol: assessing the reliability and predictive validity of adherence to the nine MST principles

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

Links between discrimination and sleep difficulties: romantic relationships as risk and resilience factors

PDF

Evaluating a cultural process model of depression and suicidality in Latino adolescents

PDF

The link between maternal depression and adolescent daughters' risk behavior: the mediating and moderating role of family

PDF

The link between the expression of negative emotion and future well-being: the role of disclosure modality in an East Asian American population

PDF

The successful African American Parenting (SAAP) Program: an evaluation of a culturally sensitive parenting intervention

PDF

Opportunity beliefs and behavioral outcomes in Latinx youth: an exploration of Ogbu’s cultural-ecological model

PDF

Mental health consequences of structural racism on Latine youth: moderating role of racial-ethnic socialization

PDF

Marital conflict and parenting: moderating effects of ethnicity and parent sex

PDF

An investigation of the factors associated with electronic aggression among college students

PDF

Adolescent conduct problems and substance use: an examination of the risk pathway across the transition to high school

PDF

Family environment as a moderator of the association between theory of mind and social functioning in people with schizophrenia

PDF

Disaster near-miss appraisal: effects of attribution, individual differences, psychological distance, and cumulative sequences

PDF

Effectiveness of a heart plus brain health-focused nutrition intervention among Latinas: a randomized controlled trial of Buenos Hábitos Alimenticios Para Una Buena Salud (Good Eating Habits for ...

PDF

A dissonance-based intervention targeting cheating behavior in a university setting

PDF

Renin-angiotensin system antihypertensive medicines with and without blood-brain-barrier crossing potential: effects on cognitive outcomes in the elderly

Asset Metadata

Creator Rubenson, Miriam P. (author)

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

School College of Letters, Arts and Sciences

Degree Master of Arts

Degree Program Psychology

Publication Date 07/20/2017

Defense Date 03/28/2017

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

Tag Adolescents,cumulative risk,juvenile offenders,multisystemic therapy,OAI-PMH Harvest,substance use

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Huey, Stanley Jr. (committee chair), John, Richard (committee member), Margolin, Gayla (committee member)

Creator Email miriamprubenson@gmail.com,rubenson@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-406280

Unique identifier UC11265610

Identifier etd-RubensonMi-5570.pdf (filename),usctheses-c40-406280 (legacy record id)

Legacy Identifier etd-RubensonMi-5570.pdf

Dmrecord 406280

Document Type Thesis

Rights Rubenson, Miriam P.

Type texts

Source 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 a...

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