Content uploaded by Suena H. Massey
Author content
All content in this area was uploaded by Suena H. Massey on Dec 06, 2023
Content may be subject to copyright.
Original Article
Journal of Drug Issues
2023, Vol. 0(0) 1–16
© The Author(s) 2023
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/00220426231159303
journals.sagepub.com/home/jod
Adolescent Substance Use
Outcomes in Response to Social
Consequences of Use: The Role
of Empathy
Drew E. Winters
1
, Suena H. Massey
2
, and Joseph T. Sakai
1
Abstract
Evidence suggests empathy deficits have a temporal relationship with substance use severity by
late adolescence theorized to decrease use via recognition of social consequences. However, this
has yet to be tested empirically along with differences in cognitive and affective empathy. Ad-
olescents admitted to substance use treatment (n= 3382) were followed through treatment and
12 months after treatment. Variable trajectories were fit using growth curve models; and cross-
lagged effects of cognitive and affective empathy (interpersonal reactivity index) on response to
social consequences of use were tested along with how response to social consequences affected
the mean trajectory of substance use. Results indicate higher cognitive empathy predicted greater
response to social consequences of use and response to these consequences at the end of
treatment predicted a steeper decrease in substance use. This evidence highlights the importance
of cognitive empathy for responding to social consequences of use for motivating less adolescent
substance use.
Keywords
adolescence, cognitive empathy, affective empathy, substance use, treatment
While it is well known that substance use has a typical age-related pattern that peaks in late
adolescence and drops rapidly as one moves into adulthood, a subset persist in substance use
(Dennis & Scott, 2007). Recent theories aiming to understand what may separate those that persist
in substance use suggest that social cognitive and affective functioning may play an important role
(Cousijn, Luijten, & Feldstein Ewing, 2018;Massey, Newmark, & Wakschlag, 2017). These
theories suggest empathy impairments underlie persistent use despite social consequences,
1
Department of Psychiatry, University of Colorado School of Medicine, Anschutz Medical Campus, Aurora, CO, USA
2
Department of Psychiatry and Behavioral Sciences, Northwestern University Feinberg School of Medicine, Chicago, IL,
USA
Corresponding Author:
Drew E. Winters, Department of Psychiatry, University of Colorado School of Medicine, Anschutz Medical Campus,
13001 e. 17th place, Aurora, CO 80045-2559, USA.
Email: Drew.winters@cuanschutz.edu
whereas normative levels of empathy underlie a recognition of social consequences and moti-
vation to change substance use –resulting in decreased substance use across varying stages of use
(Massey et al., 2017); and that these responses to the social environment explain normative
decreased use as adolescents transition into adult roles (Cousijn et al., 2018). Evidence dem-
onstrates socio -cognitive and -affective functioning (e.g., empathy) in youth has a temporal
association with substance use severity by late adolescence (For meta-analysis: Winters, Brandon-
Friedman, Yepes, & Hinckley, 2021). However, it has yet to be examined how empathy predicts
how the recognition of social consequences and related response to modify their use (i.e., response
to social consequences of use) and how this predicts substance use over time. This is a critical next
step for understanding empathic influences on substance use in adolescents. Thus, the present
study tests how empathy from a prior timepoint predicts response to social consequences of
substance use and how this impacts the trajectory of adolescent substance use.
The response to social consequences of substance use inspired by the work of Massey et al.
(2017) involves two distinct components that motivate change in substance use and are equally
important for defining the construct. The first factor is recognition of social consequences related
to their use which includes consequences for themselves as well as recognizing the impact their
use has on others. The second factor is engagement in doing something to reduce their substance
use. Recognition that consequences exist for themselves and for others is a potential motivating
factor, but, in and of itself, does not require a response to that recognition of a social consequence.
It is the response to consequence recognition, such as treatment motivation, that defines both the
recognition and motivation response of this construct. Massey et al. (2017) theorizes the response
to consequences of substance use is driven by the ability to understand other’s emotions (i.e.,
empathy; Decety, Bartal, Uzefovsky, & Knafo-Noam, 2016) or, in other words, response to social
consequences of use indirectly accounts for (i.e., mediates) empathy’s association with substance
use. Empathy not only aids recognition of social information but also an understanding what
others feel, which plausibly bridges the recognition and felt sense of loss of connection (whether
self-focused or other-focused) motivating a reduced use response.
Social consequences are different at different levels of use involving different responses to
consequences of substance use. For example, casual users with greater empathy would desist use
because of the concern expressed by those they care about or loss of social roles that incur
emotional costs to themselves; whereas those with problematic use or even full addiction that
develop high levels of empathy would be deterred because of a loss of relationship significant to
them or harm caused to their loved ones. On the other hand, those with low empathy would not be
deterred by any of these consequences, and would be more likely to persist in use (Massey et al.,
2017). Given that empathy is critical for bonding and responding to the social environment
(Anderson & Keltner, 2002;Cliffordson, 2002;Eisenberg et al., 1996), it is plausible that those
with lower empathy are less likely to respond to negative social signals related to their substance
use. Additionally, empathy is under development through adolescence and into adulthood (Decety
& Michalska, 2010;Eisenberg, 2005). Theory by Cousijn et al. (2018) suggests that neural
development underlying the capacity to respond to the social environment drives normative
decreased use as adolescents take on more adult roles –one may infer from this model that if this
development is impaired, young adults may persist in use as their peers decrease use. These
theories highlight the importance of social response and empathy relating to persistence in
substance use, especially during adolescence.
Empathy is supported by distinct processes with different developmental trajectories –cognitive
and affective empathy (Decety & Cowell, 2015;Singer, 2006;Smith, 2006;Walter, 2012). Cognitive
empathy involves taking the psychological point of view of others (i.e., perspective taking) whereas
affective empathy involves sharing other’s emotions, which can include a response of concern for
another’s wellbeing (i.e., empathic concern) (Decety, 2011;Decety & Cowell, 2015). Cognitive and
2Journal of Drug Issues 0(0)
affective empathy have distinct neural underpinnings in adolescence (Kral et al., 2017;Winters,
Pruitt, et al., 2021) that are still under development with subcortical regions associated with affective
empathy developing earlier whereas cortical regions involved in cognitive empathy do not fully
mature until early adulthood (Blakemore, 2012;Singer, 2006). These distinct empathy processes
likely have different roles when responding to social consequences. For example, cognitive empathy
can cue one into messages from the social environment; whereas affective empathy can generate the
affective resonance to understand the emotional concerns of others. These associations are likely
different between adolescents and adults. Thus, examining adolescent empathy is critical for un-
derstanding the interplay between responding to social consequences and substance use.
The present study longitudinally examines how cognitive and affective empathy predicts
response to social consequences of substance use and how the trajectory of this response is related
to substance use over time amongst adolescents involved in substance use treatment. We hy-
pothesize that (1) higher levels of both cognitive and affective empathy will predict higher levels
of response to social consequences of substance use and (2) that greater levels of response to
consequences of substance use at the end of treatment will mediate (i.e., indirect effects) cognitive
and affective empathy prediction of decreased substance use trajectory over time. This study is
critical for understanding social cognitive and affective factors that may contribute to future
substance use patterns, which highlight novel factors to motivate changes in substance use.
Methods
Sample
We conducted a secondary data analysis of the Drug Abuse Treatment Outcome Studies for
Adolescents data set (DATOS-A; Kristiansen & Hubbard, 2001) that was collected between
November 1993 and November 1995. This data set recruited adolescents admitted into substance
use treatment from six large American cities (n= 3382). Participants were assigned to either
residential, short-term inpatient, or outpatient treatment. Those in residential treatment had the
highest drug dependence and those in outpatient had the lowest drug dependence and the least risk
on average (Kristiansen & Hubbard, 2001). All participants were followed through treatment at
baseline (T1), 1 month (T2), 3 months (T3), and 6 months (T4) with a post-treatment follow up at
12 months post-treatment (T5).
The DATOS-A collection focused on broad factors influencing treatment outcomes at varying
levels of substance use treatment. Therefore, we have no information on the types of interventions
used during treatment only the level of service (i.e.; residential, inpatient, outpatient) assigned by
level of immediate risk and severity of use (Kristiansen & Hubbard, 2001).
Measures
Measures were part of a structured interview adapted from the National Health Survey and
National Institute of Mental Health’s Epidemiological Catchment Area Studies for a measure
specific to the DATOS-A (Hubbard, Craddock, Flynn, Anderson, & Etheridge, 1997). This
structured interview is comprised of multiple other measures and some items were created for this
specific measure. The descriptive statistics and amount of missing data for the examined variables
are shown in Table 1.
Substance Use. Participants self-reported their use of different substances [i.e., alcohol, marijuana,
cocaine, or hallucinogen (Questionnaire defined as “LSD, PCP, or any other hallucinogen”)]. We used
a dichotomous measure from these self-reports indicating if they had used any of the highest-reported
Winters et al. 3
substances of use in the sample (alcohol, marijuana, cocaine, or a hallucinogen) over the previous
3 months (use = 1 no use = 0). Another option we could have chosen was to use a substance use variety
score indicating the number of different substances participants endorsed. However, we chose against
this as (1) it does not capture the severity of use of each substance endorsed and (2) we had no
hypotheses related to variety of substance use –our hypotheses directly relate to whether substance use
occurred or not. Given that there were a clear number of four substances that were endorsed by
participants and all measures of substance use only indicate whether they used, we decided using the
dichotomized use variable of the most used substances provided the variance necessary to answer the
current research question.
Empathy. The perspective taking and empathic concern subscales of the Interpersonal Reactivity
Index (Davis, 1980,1983) were used to assess cognitive empathy and affective empathy, respectively.
It is common practice to measure cognitive and affective empathy with these two subscales (Konrath,
2013), and to exclude other subscales (fantasy and personal distress) as they likely measure constructs
that are beyond empathy (Baron-Cohen & Wheelwright, 2004). The perspective taking subscale (our
measure of cognitive empathy; baseline α= .70) consists of seven items that measure the disposition to
adopt another’s point of view (e.g., “I try to look at everybody’s side of a disagreement before I make a
decision”). The affective empathy subscale (baseline α= .68) also consists of seven items that measure
the disposition to share another’s emotional experience and have concern for them (e.g., “When I see
someone being taken advantage of, I feel kind of protective towards them”). Participants rate each item
on this measure on how well each statement describes them on a five-point Likert scale from “does not
describe me well”to “describes me well”(0–4). Higher scores indicate higher levels of dispositional
cognitive or affective empathy. The IRI demonstrates convergent and divergent validity, shows a
consistent factor structure across samples and nation of origin, (Konrath, 2013), and the factor structure
of this measure has been confirmed in adolescents (e.g., Hawk et al., 2013).
Table 1. Descriptive Statistics of Examined Variables.
Continuous
variables M ± SD % missing Reliability
Categorical
variables n (proportion) % missing
Cognitive empathy Substance use
Wave 1 12.62 ± 5.33 2 .70 Wave 1 2904 (97% use) 14
Wave 2 13.37 ± 5.34 34 .73 Wave 2 2256 (13% use) 33
Wave 3 14.11 ± 5.46 67 .74 Wave 3 1116 (18% use) 67
Wave 4 14.70 ± 5.32 84 .73 Wave 4 530 (13% use) 84
Wave 5 13.76 ± 5.38 47 .75 Wave 5 1703 (77% use) 49
Affective empathy Treatment modality 0
Wave 1 17.07 ± 5.07 2 .69 Residential 1627 (48%)
Wave 2 17.36 ± 4.90 35 .70 Inpatient 929 (27%)
Wave 3 17.40 ± 4.75 67 .68 Outpatient 826 (25%)
Wave 4 17.74 ± 4.92 85 .68 Sex 0
Wave 5 17.56 ± 4.70 47 .70 Female 885 (26%)
Social response Male 2497 (74%)
Wave 1 2.24 ± .36 34 .70 Race 0
Wave 2 2.47 ± .30 67 .67 White 1758 (52%)
Wave 3 2.70 ± .35 84 .70 Black 807 (23%)
Age at wave 1 15.75 ± 1.36 0 –Hispanic 685 (20%)
Other 132 (3.9%)
4Journal of Drug Issues 0(0)
The IRI was created as a measure of dispositional empathy. Previous ideas on empathy were
that it was a trait-like function that remained static over time. However, our contemporary
knowledge of empathy is that it is malleable and can be cultivated by training it –like a skill
(Schumann, Zaki, & Dweck, 2014;Zaki, 2018). While the IRI is thought to measure dispositional
empathy, it has been proposed to use it to observe changes in medical students training over
schooling (Konrath, 2013) and this has been implemented by using the IRI to examine medical
students change in empathy over 3 years of medical training (Shin, Park, & Lee, 2022). Therefore,
we believe the IRI to be an adequate measure to measure both cognitive and affective empathy
while also allowing the sensitivity to observe variation throughout the course of treatment.
As a measure of consistency of measurement across the different time periods, we calculated in
intraclass correlation using the psych package in r. We found Cognitive empathy had an intraclass
correlation of .79 and affective empathy had an intraclass correlation of .74.
Social Response to Substance Use Consequences. Inspired by theory by Massey et al. (2017),we
selected items from the University of Rhode Island change assessment scale (URICA) and drug
treatment questions created for the structured interview used in the DATOS. Items selected for
inclusion from each measure are outlined in Table 2.
The URICA (McConnaughy, Prochaska, & Velicer, 1983) is a 24-item scale that measures
one’s stage of change (precontemplation, contemplation, action, and maintenance). Responses on
this measure are rated on a 5-point Likert scale from strong disagreement to strong agreement (1–
5). We selected items in this measure that captured 1) perceptions of treatment being worthwhile
(suggesting a belief treatment will help) and 2) willingness to engage with treatment (suggesting
active participation in changing substance use). We used items from the action subscale and
reversed scored precontemplation scales to assess a participant’s willingness to work on identified
problems related to substance use. Higher scores indicate perceptions that treatment is worthwhile
and a greater participation in addressing substance use problems.
The selected scale items created for the DATOS-A reflected recognition of substance use
causing problems for themselves and others. These items were rated on a three-point scale from
“not important”or “don’t agree”to “very important”or “strongly agree”(0–2). Higher scores
indicate identifying their substance use has caused significant problems for themselves and others
and agreement that treatment will help.
Time-Invariant Control Variables. For control variables we used self-reported sex, age, site where
data was collected, and assigned treatment modality. We controlled for sex to account for sex
related differences observed in substance use (For review: McHugh, Votaw, Sugarman, &
Greenfield, 2018). We controlled for age to account for age related patterns of use (Dennis &
Scott, 2007). To control for differences in variation due to level of treatment, we controlled for
inpatient and outpatient with residential as the reference category. We had no reason to believe
ethnicity would drive differences in empathy or social response, thus we excluded ethnicity as a
covariate.
Analysis
All analyses were conducted using R statistical language (Version 4.02; R Core Team, 2021). The
“psych”package (Revelle, 2021) was used for factor analysis and “lavaan”was used for both
confirmatory factor analysis (CFA) and growth curve modeling (Rosseel, 2012). Prior to our
analysis we tested for distribution normality of residuals, autocorrelation, and multicollinearity,
which we found no violations. All models were evaluated using criteria for adequate fit suggested
by Hu and Bentler (1999) and Mulaik et al. (1989) (RMSEA ≤.06, CFI & TLI ≥.90).
Winters et al. 5
A Latent Construct for Social Response to Substance Use Consequences. First, inspired by theoretical
work of Massey et al. (2017), we selected items to define different components of a response to
social consequence that involved 1) recognition of substance use causing a problem and 2)
motivation to change their substance use (see items in Table 2). The items measuring this construct
were collected at three out of the five timepoints (timepoints two through four). Response fre-
quencies on the URICA revealed <5% endorsement on the lower end of all items; thus, we
collapsed these items leaving three responses for each item. Then, these items were used to
statistically define social response to substance use consequences.
Next, we randomly assigned a training a test set of participants (70% and 30% respectively) and
conducted a factor analysis on the training set as well as a CFA on the test dataset to verify the response
to social consequences of use construct. For the factor analysis we used a varimax rotation to examine
the factor structure. We determined the number of factors using a scree plot and that factor had an
eigenvalue for each factor were >1. These criteria revealed a 4-factor structure represented the
construct. This factor structure adequately fit the data (x
2
(17) = 24.55, TLI = .977, RMSEA = .033)
and accounted for 98% of the variance (see loadings in Tabl e 2).
Table 2. Items and Factor Analysis Loadings for Social Response to Substance Use Consequences.
Items
Factors
Not engaged.
(reverse scored)
Drug use cause
problems
Actively changing
substance use
Treatment
importance
URICA items
I am doing something about
my drug problem
aa
.671
a
Being here is a waste of time –
I am not the cause
b
.506
aaa
I am working on my problems
aa
.577
a
I am doing something about
changing myself
aa
.707
a
It is boring to talk about my
problems
b
.649
aaa
Does no good to think about
my problems
b
.639
aaa
I would rather live with my
faults than try to change
b
.472
aaa
DATOS-A measure items
How troubled are you with
your drug problem
a
.439
aa
How important is it to get
treatment for your drug
problem
aa a
.964
Your drug problem has
caused problems in your life
a
.826
aa
Your drug problem has
caused problems for others
a
.817
aa
a
= loadings <.40
b
= reverse scored.
6Journal of Drug Issues 0(0)
We then conducted an item level analysis to determine an optimal parceling scheme from this
factor structure. We examined correlated residuals (θmatrix) and parceled items together that
showed a tendency to correlate with each other after conditioning on the underlying construct
being measured (Little, Rhemtulla, Gibson, & Schoemann, 2013). We parceled items that had the
highest set of loadings >10 (Supplementary Table S1). This resulted in one item parcels for each
individual factor.
Using the parceled indicators, we then conducted a CFA with the test sample to confirm the
factor structure. The factor structure was confirmed using the item parcels (x
2
(2) = 2.12, CFI =
.998, TLI = .990, RMSEA = .040) with an adequate reliability (α= .69). This same parceling
scheme was used for all timepoints to examine mean growth trajectories of social response to
substance use consequences. To ensure the same construct was being measured at each timepoint
we established measurement invariance at the configural, loadings, and intercept level (Table 3).
Individual Models of Growth
We examined mean change for each variable of interest (i.e., substance use, social response,
cognitive empathy, and affective empathy) using growth curve modeling in the structural
equation modeling framework. Preliminary investigation of means for each timepoint sug-
gested changes in social response to use was linear, whereas empathy and substance use
variables were nonlinear. To test whether a linear trend adequately represents the data, we first
test a linear model against a non-linear latent basis model. If non-linear, we then attempted to
model the optimal functional form by adding a quadratic trend and compare the AIC to
determine which model more closely models the data (See Table 4). The latent basis curve
model estimates non-linear trajectories with level and shape latent growth factors. The level
factor is constrained to one for each timepoint and the shape factor constrains the initial
timepoint to 0 and either the second or last timepoint to 1 allowing the other timepoints to be
estimated freely.
For models with continuous variables (i.e., cognitive empathy, affective empathy, consequence
response) we used full information maximum likelihood estimation method (FIML). FIML
produces unbiased estimates of parameters by using all available data to maximize the likelihood
over each vector of observations to derive parameter estimates (Little & Rubin, 2019;Muth´
en,
Tam, Muth´
en, Stolzenberg, & Hollis, 1993). Models with binary variables, specifically substance
use, we used the weighted least squares mean and variance adjusted estimator (WSLMV) to model
the categorical outcomes. WLSMV assumes a continuous latent construct underlying substance
use at each timepoint, which allows us to measure the extent of variation over time (Asparouhov &
Muth´
en, 2010). Concerning missing data, FIML and WLSMV can produce unbiased estimates
and retain missing values without dropping cases (Asparouhov & Muth ´
en, 2010). In comparison
to traditional deletion of substitution methods which introduce significant bias into statistical
analyses, these modern missing data approaches are far superior for reducing bias (Little & Rubin,
2019).
Table 3. Longitudinal Invariance for Social Response.
Model AIC X
2
(df) CFI RMSEA ΔX
2
(df) pΔCFI ΔRMSEA Decision
Configural invariance 7847.1 111.51 (33) .970 .069 –––– –
Metric invariance 7842.3 118.7 (39) .968 .064 7.18 (6) .31 .000 .005 Pass
Scalar invariance 7863.3 151.71 (45) .959 .069 33.01 (6) <.001 .009 .005 Pass
Winters et al. 7
Cross-Lagged Effects of Empathy on Substance Use Trajectories Regressed on
Social Response
We then examined social response growth terms on substance use latent growth terms as well as
cross-lagged effects of cognitive and affective empathy as a time-varying predictor on social
response (Wu & Lang, 2016;Wu, Selig, & Little, 2012). We included time constant controls for
age, sex, and treatment modality for each regression path. For meaningful intercept interpretation,
we coded social response such that the intercept was at the final timepoint (timepoint four) and for
substance use at timepoint five. We examined how time-varying predictors influenced social
response after controlling for covariates and how social response influenced the latent growth
trajectory of substance use (See Supplementary Figure 1 for specification). And finally we tested
indirect effects for each cross lagged connection on substance use outcomes using the product of
coefficients (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002).
An important feature of our growth curves is placement of importance on the response to social
consequences of use at the end of treatment. Changing the loci of importance in a trajectory does
not impact the temporal ordering of a growth curve, it merely emphasizes the point at this is
important for the current study. Given treatment is expected to impact how one responds to social
consequences related to use –we place the importance of the intercept for at the end of treatment to
account for this expected effect. The importance of the intercept at the end of treatment would be
interpreted as the mean level of response to social consequences of use at the end of treatment.
Importantly, under no circumstance did we have any latent variables with less than three items
representing that latent factor. For the final model, response to social consequences of substance
use each have four items and the intercept and growth factors of response to social consequences
of substance use has three items each. Substance use growth and intercept factors have five items.
Finally, both cognitive and affective empathy are not estimated as latent factors but instead are
modeled as observed cross-lagged effects to improve causal evidence.
Missing Data Sensitivity Analysis. The data collection had a substantial amount of missing data (see
Table 1). To address this, we did a sensitivity analysis by running the analyses with all waves of
Table 4. Fitting Single Growth Curve Models.
AIC X
2
(df) CFI TLI RMSEA ΔX
2
(df) ΔX
2
p Best fit model
Affective empathy
Linear 51,673 74.29 (10) .961 .961 .044
Latent basis 51,658 53.61 (7) .972 .959 .045 20.67 (3) <.001 Latent basis
Quadratic 51,624 17.26 (6) .993 .989 .024 36.35 (1) <.001 Quadratic
Cognitive empathy
Linear 53,708 131.12 (10) .927 .927 .060
Latent basis 53,635 52.64 (7) .972 .961 .044 78.48 (9) <.001 Latent basis
Quadratic 53,612 27.98 (6) .987 .978 .033 24.65 (1) <.001 Quadratic
Social response
Linear 22,058 134.50 (52) .985 .974 .022
Latent basis 22,062 142.11 (54) .984 .973 .022 7.61 (2) .022 Linear
Substance use
Linear NA 1747.5 (10) .00 29.7 .827
Latent basis NA 7.51 (7) .991 .987 .017 2182.3 (3) <.001 Latent basis
Quadratic NA 144.49 (6) .00 3.08 .301 143.03 (1) <.001 Latent basis
8Journal of Drug Issues 0(0)
data and also dropping wave three and four (where missingness was at its highest). The results of
the models were the same; thus, we only report results with all five waves of data.
Additionally, we conducted a set of t-tests to try and identify if any bias in the data accounted
for patterns of missingness with demographic variables and variables of interest. These tests did
not find any reason for missing and therefore assume data to be missing at random for which we
can properly treat using FIML estimation.
Results
Sample. The sample consisted of 3382 adolescents in substance use treatment (2497 males [74%]
885 and females [26%]) between the ages of 12–18 (15.75 ± 1.36) that were predominantly White
(White = 1758 [52%], Black = 807 [24%], Hispanic = 685 [20%], other = 132 [4%]. Most of the
participants were assigned to residential treatment (residential 1627 [48%], inpatient = 929 [27%],
outpatient = 826 [25%]). When recording baseline use of cannabis, alcohol, cocaine, or hallu-
cinogens, 3% of participants reported using no substances, 15% reported using one substance,
48% reported using 2, 24% reported using three, and 10% reported using all four over the past
3 months. Marijuana was the most used with 92% of participants reported its use. Distributions of
the IRI for both cognitive (overall average 13.71 ± 5.37) and affective empathy (overall average
17.43 ± 4.87) were in the moderate range. Mean IRI scores for each timepoint in the present study
(Table 1) are within a SD of mean IRI scores of other large adolescent studies (e.g., Hawk et al.,
2013;Overgaauw, Rieffe, Broekhof, Crone, & Güro˘
glu, 2017).
Direct Association Between Empathy and Use. Cognitive and affective empathy had moderate to low
negative zero-order correlations with substance use (see Supplementary Table S2).
Individual Growth Trajectories. Individual growth curves suggest that the mean trajectories of social
response (mean growth = .535, p= .038) increase over time whereas substance use decreases over
time (mean growth = .225, p< .001). Cognitive empathy and affective empathy have a mean
increase over time with a quadratic decrease between timepoints 4 and 5 (Cognitive: mean
growth = 1.04, p< .001, quadratic .186, p< .001; Affective: mean growth = .311, p< .001,
quadratic .052, p= .017; see Figure 1).
Full Model. When testing the hypothesized model of social response regressed on cognitive and
affective empathy simultaneously as well as mean trajectory of substance use regressed on social
response, we found that a higher mean reported level of social response at the end of treatment
(timepoint 4) associated with a steeper decrease in mean substance use growth (β=.065, p=
.009); and the model accounted for 69% of the variance in substance use growth trajectory (R
2
=
.690). Additionally, cognitive empathy consistently positively predicted and accounted for over
75% of the variance of social response at each time point (T2: β= .017, p<.001R
2
= .786; T3: β=
.008, p= .006, R
2
= .785; T4: β= .015, p< .001, R
2
= .798) whereas affective empathy did not
predict social response at any timepoint (p> .05). Assessment of indirect effects demonstrated
that, although time three was insignificant, response to social consequences of use regressed on
empathy at time one and time three indirectly decreased substance use trajectory (β=.001, p=
.022; β=.001, p= .065; β=.001, p= .024). Age and sex did not significantly predict the
growth trajectory in substance use, but the outpatient sample had a less steep decrease in substance
use in reference to residential treatment participants (β= .085, p< .001). The final model ad-
equately fitted the data without any modifications (X
2
(266) = 668.79, CFI = .920, TLI = .903,
RMSEA = .058; see Figure 2 for depiction of significant paths). Within the model, cognitive and
Winters et al. 9
Affective empathy had high to moderate zero-order correlations within construct but cognitive had
moderate to low zero-order correlations with affective empathy (see Supplementary Table S3).
Discussion
The present study provides critical evidence regarding the temporal role empathy has with substance
use trajectories via response to social consequences of substance use. Specifically, increases in
cognitive empathy amongst adolescents in substance use treatment temporally predicts greater rec-
ognition of consequences and treatment motivation (i.e., response to social consequences of use),
which in turn predicts a stronger negative mean trajectory of substance use over time. Present findings
may indicate that recognizing social consequences of use is a motivating factor for reducing substance
use amongst adolescents, and this is driven by higher levels of cognitive empathy.
Confirmatory Factor Analysis. The finding that the construct of interest adequately fit the data brings
empirical evidence for a latent construct consisting of both recognition of consequences and
motivation for treatment. Because this latent construct consists of items involving recognition of
consequences to others and self-related to substance use and motivation for substance use
treatment, we assert this latent construct describes a response to social consequences related to use.
While it may be argued these two components are separate, the empirical evidence holds that these
items together explain our latent construct of interest, which is a significant finding in and of itself.
We hope this inspires future investigation into this latent construct and its association with
substance use trajectories in youth and adults.
Individual Growth Trajectories. Functional forms of growth on variables of interest suggest, on
average, cognitive and affective empathy as well as response to social consequences increased,
Figure 1. Depicts the functional form of mean trajectories over time for variables in full structural equation
model.
10 Journal of Drug Issues 0(0)
whereas substance use decreased (Figure 1). Cognitive and affective empathy had a steep increase
at the start of treatment and a slight decrease at 6-months post-treatment whereas substance use
had a sharp decrease at the start of treatment and an increase at 6-month follow up. While we do
not have specific information on what occurred during treatment, it appears that treatment had a
positive effect on all variables of interest with some drop off with follow up 6-months post-
treatment.
Empathy is expected to increase during adolescence, and as expected affective empathy started
higher than cognitive empathy. However, we are unable to determine why empathy dropped after
treatment. Because adolescents can have heightened sensitivity to the social environment
(Blakemore & Mills, 2014), this increase may be due to the support received during treatment that
when removed post-treatment saw a slight average drop. The drop in empathy may also be due to
an effect of substance use specifically as the average substance use also increased post treatment.
Because empathy is still under development, and therefore more malleable, this finding may be
specific to adolescents, and we may not see this same pattern in adults. This is an unexpected
finding important to investigate in future studies.
Response to social consequences of substance use had a steady and significant linear increase
during treatment. But, because this information was only available during treatment, we do not
know how this may have changed post-treatment. Future studies could build on these results by
capturing baseline and post-treatment response to social consequences related to use.
Full Model. Regression path results extend previous research by suggesting that cognitive empathy
has an indirect effect on substance use amongst adolescents (Figure 2). Nguyen et al. (2011) and
Laghi et al. (2019) demonstrate cross-sectionally that higher levels of cognitive empathy associate
with higher levels of drug refusal efficacy and, subsequently, lower levels of substance use. The
present methods extend these findings by demonstrating this relationship exists longitudinally and
higher responses to use related social consequences as a motivational factor for decreased use. It is
plausible that drug refusal may be the result of responding to social consequences related to
substance use, which motivates increases in drug refusal behavior. Future studies could parse apart
Figure 2. Depicts the full structural equation model with significant betas. All abbreviations are depicted in
the legend and indirect effects are reported in the upper left hand table.
Winters et al. 11
the mechanism of action by examining both drug refusal and response to social consequences
related to substance use amongst adolescents.
One potential concern by readers for interpretation is that participants learned responses for
empathy measures that are more socially desirable over time. However, we find this to be unlikely
because (1) respondents did not receive direct feedback on their responses as being “good”or “not
good”that would have allowed learning to occur (2) the timespan of multiple months is an
adequate time before retaking the measure and (3) empathy dropped, on average, after treatment
and if learning effects existed, we would have expected this to continue to increase.
While there is a growing literature both theoretically and empirically demonstrating temporal
precedence of empathy impairments prior to substance use (for review: Winters, Brandon-
Friedman, et al., 2021), it is important there plausibly some bidirectionality between empathy
and substance use and ceasing use in and of itself may have improved empathy. It is critical to
explore this bidirectionality in future studies. Moreover, given the developmental trajectory of
cognitive empathy during adolescence, the observed effects may have some relevance for natural
development of cognitive empathy over the course of treatment. However, the average rapid
decrease and slight decrease after treatment evidences some variance outside of development
alone and that something about the substance use treatment received had an impact on both
empathy and substance use.
Affective empathy did not predict how one responded to social consequences of substance use
as hypothesized. Although it plausibly has an impact on response to social consequences, affective
empathy consistently demonstrates a direct association with substance use in adolescents (Laghi,
Bianchi, Pompili, Lonigro, & Baiocco, 2019;Luengo, Otero, Carrillo-de-la-Peña, & Mirón, 1994;
Winters, Wu, & Fukui, 2020) whereas cognitive empathy associates with substance use indirectly
(For review: Winters, Brandon-Friedman, et al., 2021). The present results support previous
literature that cognitive empathy has an indirect effect on substance use where affective empathy
does not. This suggests that cognitive understanding of others’perspective may impact a sense of
connection to and response to consequences related to substance use –this may be necessary to
stimulate the affective resonance with others in this response. Further investigation is needed to
parse apart the direct and indirect effects of cognitive and affective empathy in relation to ad-
olescent substance use.
Limitations. The present findings need to be interpreted under the following limitations. First, the
sample was limited to adolescents in substance use treatment, which limits generalizability to the
general population. Second, the measure for social response is conceptual and empirically verified
but may not be generalizable to samples outside participants who are in substance use treatment.
Further verification of this measure with different samples is needed. Third, the data used were
collected for a different purpose and confounding variables relevant for the present analysis such
as socioeconomic status and demographic information were not available. Fourth, severity of
substance use was not captured in the outcome variable because severity ratings were not available
in this dataset –only weather or not a substance was used. Although the presence of substance use
is meaningful and substance use severity was partially accounted for by treatment groupings,
future studies could explicitly examine substance use severity. Fifth, there were a large number of
missing cases as the study continued. However, the missing data approaches we applied help to
minimize any bias the may be introduced by missing participants and maintain the integrity of the
analysis far better than leaving cases out (Little & Rubin, 2019). Moreover, we did a sensitivity
analysis of excluding T3 and T4 and found it did not change the results that further support
our analysis. Sixth, while the present study evidence empathy and response to social consequences
to substance use generally, it is entirely plausible that substance use type may be particularly
important for the relationships observed here; therefore, future studies could build on the present
12 Journal of Drug Issues 0(0)
results by including detailed analyses on severity of individual substances used. Seventh, the
substance use variable used the top substances endorsed by the present sample that may not reflect
current substance use trends in adolescents, thus future investigations could build on this result by
using current samples and examining those most endorsed by the population. Eighth, we did not
have specific information on use such as if substances were used individual or co-used. Although
we do not believe this could impact the important findings of the present study it this information
may be relevant for further nuance in future investigations. Ninth, comorbidity of symptoms such
as ADHD, depression, or psychosis can impact empathy. However, the present study did not have
adequate measures of symptom severity, therefore we did not include these symptoms as controls.
Future studies should consider the importance of comorbid symptoms and their impact on
empathy in relation to substance use. Also, the available data was over 30 years old. While
understandably raising questions of relevance for contemporary youth, this large longitudinal
dataset is appropriate for this question given the relevant variables available and population.
Additionally, this is an initial investigation with measures that are still published in the literature
with contemporary youth related to substance use; therefore, the present analysis evidences the
relevance of investigating this phenomenon in future data collections. Finally, there was no direct
manipulation on the variables of interest. Although we establish temporal precedence which may
lead one to consider potential causal associations, a clinical trial that directly tests manipulation of
these variables is necessary to confirm.
Conclusion
Despite these limitations, the present study partially substantiates that a latent construct inspired
by Massey et al. (2017) that both recognition of social consequences and motivating factors
related to treatment empirically form a latent construct the well call a response to social con-
sequences related to use. We demonstrate higher levels of response to social consequences of
substance use in adolescents at the end of their substance use treatment predicts steeper decreases
in substance use. The present analysis extends theory by Massey et al. (2017) revealing that greater
levels of cognitive empathy predicts a greater response to social consequences of substance use.
Clinically this suggests that processes of perspective taking rather than emotional sharing may be
relevant for understanding how one’s substance use impacts others and perceptions of treatment
being an important factor for them. Future studies could build on this finding by parsing apart drug
refusal and response to social consequences in relation to cognitive empathy. Moreover, lon-
gitudinal clinical trials may be warranted for establishing causality.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or
publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or
publication of this article: D. E. W. was supported by a training grant from National Institutes of Mental
Health, T32MH015442. S. M. was supported by grants K23DA037913 (PI Massey) and R01DA050700 (PI
Massey) from the National Institute on Drug Abuse.
ORCID iD
Drew E. Winters https://orcid.org/0000-0002-0701-9658
Winters et al. 13
Supplemental Material
Supplemental material for this article is available online.
References
Anderson, C., & Keltner, D. (2002). The role of empathy in the formation and maintenance of social bonds.
Behavioral and Brain Sciences,25(1), 21–22. https://doi.org/10.1017/s0140525x02230010
Asparouhov, T., & Muth´
en, B. (2010). Weight least squared estimation with missing data. statmodel. https://
www.statmodel.com/download/GstrucMissingRevision.pdf
Baron-Cohen, S., & Wheelwright, S. (2004). The empathy quotient: An investigation of adults with Asperger
syndrome or high functioning autism, and normal sex differences. Journal of Autism and Developmental
Disorders,34(2), 163–175. https://doi.org/10.1023/b:jadd.0000022607.19833.00
Blakemore, S. J. (2012). Development of the social brain in adolescence. Journal of the Royal Society of
Medicine,105(3), 111–116. https://doi.org/10.1258/jrsm.2011.110221
Blakemore, S. J., & Mills, K. L. (2014). Is adolescence a sensitive period for sociocultural processing?
Annual Review of Psychology,65(1), 187–207. https://doi.org/10.1146/annurev-psych-010213-115202
Cliffordson, C. (2002). The hierarchical structure of empathy: Dimensional organization and relations to social
functioning. Scandinavian Journal of Psychology,43(1), 49–59. https://doi.org/10.1111/1467-9450.00268
Cousijn, J., Luijten, M., & Feldstein Ewing, S. W. (2018). Adolescent resilience to addiction: A social
plasticity hypothesis. The Lancet Child & Adolescent Health,2(1), 69–78. https://doi.org/10.1016/
S2352-4642(17)30148-7
Davis, M. H. (1980). A multidimensional approach to individual differences in empathy. Journal of Per-
sonality and Social Psychology,10(85), 49.
Davis, M. H. (1983). Measuring individual differences in empathy: Evidence for a multidimensional approach.
Journal of Personality and Social Psychology,44(1), 113–126. https://doi.org/10.1037/0022-3514.44.1.113
Decety, J. (2011). Dissecting the neural mechanisms mediating empathy. Emotion Review,3(1), 92–108.
https://doi.org/10.1177/1754073910374662
Decety, J., Bartal, I. B.-A., Uzefovsky, F., & Knafo-Noam, A. (2016). Empathy as a driver of prosocial
behaviour: Highly conserved neurobehavioural mechanisms across species. Philosophical Transactions
of the Royal Society of London. Series B, Biological Sciences,371(1686), 20150077. https://doi.org/10.
1098/rstb.2015.0077
Decety, J., & Cowell, J. M. (2015). Empathy, justice, and moral behavior. AJOB Neuroscience,6(3), 3–14.
https://doi.org/10.1080/21507740.2015.1047055
Decety, J., & Michalska, K. J. (2010). Neurodevelopmental changes in the circuits underlying empathy and
sympathy from childhood to adulthood. Developmental Science,13(6), 886–899. https://doi.org/10.
1111/j.1467-7687.2009.00940.x
Dennis, M., & Scott, C. K. (2007). Managing addiction as a chronic condition. Addiction Science & Clinical
Practice,4(1), 45–55. https://doi.org/10.1151/ascp074145
Eisenberg, N. (2005). The development of empathy-related responding. Nebraska Symposium on Motivation,
51,73–117.
Eisenberg, N., Fabes, R. A., Murphy, B., Karbon, M., Smith, M., & Maszk, P. (1996). The relations of
children’s dispositional empathy-related responding to their emotionality, regulation, and social
functioning. Developmental Psychology,32(2), 195–209. https://doi.org/10.1037/0012-1649.32.2.195
Hawk, S. T., Keijsers, L., Branje, S. J. T., Graaff, J. V. d., Wied, M. d., & Meeus, W. (2013). Examining the
interpersonal reactivity index (IRI) among early and late adolescents and their mothers. Journal of
Personality Assessment,95(1), 96–106. https://doi.org/10.1080/00223891.2012.696080
Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional
criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal,6(1), 1–55.
https://doi.org/10.1080/10705519909540118
14 Journal of Drug Issues 0(0)
Hubbard, R. L., Craddock, S. G., Flynn, P. M., Anderson, J., & Etheridge, R. M. (1997). Overview of 1-year
follow-up outcomes in the drug Abuse treatment outcome study (DATOS). Psychology of Addictive
Behaviors,11(4), 261–278. https://doi.org/10.1037/0893-164x.11.4.261
Konrath, S. H. (2013). Critical synthesis package: Interpersonal reactivity index (IRI). Mededportal
Publications. https://doi.org/10.15766/mep_2374-8265.9596
Kral, T. R. A., Solis, E., Mumford, J. A., Schuyler, B. S., Flook, L., Rifken, K., Patsenko, E. G., & Davidson,
R. J. (2017). Neural correlates of empathic accuracy in adolescence. Social Cognitive and Affective
Neuroscience,12(11), 1701–1710. https://doi.org/10.1093/scan/nsx099
Kristiansen, P. L., & Hubbard, R. L. (2001). Methodological overview and research design for adolescents in
the drug abuse treatment outcome studies. Journal of Adolescent Research,16(6), 545–562. https://doi.
org/10.1177/0743558401166002
Laghi, F., Bianchi, D., Pompili, S., Lonigro, A., & Baiocco, R. (2019). Cognitive and affective empathy in
binge drinking adolescents: Does empathy moderate the effect of self-efficacy in resisting peer pressure
to drink? Addictive Behaviors,89, 229–235. https://doi.org/10.1016/j.addbeh.2018.10.015
Little, R. J. A., & Rubin, D. B. (2019). Statistical analysis with missing data. New York, NY: Wiley. https://
books.google.com/books?id=BemMDwAAQBAJ
Little, T. D., Rhemtulla, M., Gibson, K., & Schoemann, A. M. (2013). Why the items versus parcels
controversy needn’t be one. Psychological Methods,18(3), 285–300. https://doi.org/10.1037/a0033266
Luengo, M. A., Otero, J. M., Carrillo-de-la-Peña, M. T., & Mirón, L. (1994). Dimensions of antisocial
behaviour in juvenile delinquency: A study of personality variables. Psychology, Crime & Law,1(1),
27–37. https://doi.org/10.1080/10683169408411934
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of
methods to test mediation and other intervening variable effects. Psychological Methods,7(1), 83–104.
https://doi.org/10.1037/1082-989x.7.1.83
Massey, S. H., Newmark, R. L., & Wakschlag, L. S. (2018). Explicating the role of empathic processes in
substance use disorders: A conceptual framework and research agenda. Drug and Alcohol Review,
37(3), 316–332. https://doi.org/10.1111/dar.12548
McConnaughy, E. A., Prochaska, J. O., & Velicer, W. F. (1983). Stages of change in psychotherapy:
Measurement and sample profiles. Psychotherapy: Theory, Research & Practice,20(3), 368–375.
https://doi.org/10.1037/h0090198
McHugh, R. K., Votaw, V. R., Sugarman, D. E., & Greenfield, S. F. (2018). Sex and gender differences in
substance use disorders. Clinical Psychology Review,66,12–23. https://doi.org/10.1016/j.cpr.2017.10.012
Mulaik, S. A., James, L. R., Van Alstine, J., Bennett, N., Lind, S., & Stilwell, C. D. (1989). Evaluation of
goodness-of-fit indices for structural equation models. Psychological Bulletin,105(3), 430–445. https://
doi.org/10.1037/0033-2909.105.3.430
Muth´
en, B., Tam, T., Muth´
en, L., Stolzenberg, R. M., & Hollis, M. (1993). Latent variable modeling in the
LISCOMP framework: Measurement of attitudes toward career choice. In: New directions in attitude
measurement (pp. 277–290). Festschrift for Karl Schuessler.
Nguyen, A. B., Clark, T. T., & Belgrave, F. Z. (2011). Empathy and drug use behaviors among African-
American adolescents. Journal of Drug Education,41(3), 289–308. https://doi.org/10.2190/de.
41.3.d
Overgaauw, S., Rieffe, C., Broekhof, E., Crone, E. A., & Güro˘
glu, B. (2017). Assessing empathy across
childhood and adolescence: Validation of the empathy questionnaire for children and adolescents
(EmQue-CA). Frontiers in Psychology,8, 870–870. https://doi.org/10.3389/fpsyg.2017.00870
R Core Team (2021). R: A language and environment for statistical computing. R Core Team. https://www.R-
project.org/
Revelle, W. (2021). psych: Procedures for psychological, psychometric, and personality research. Evanston,
IL: The Comprehensive R Archive Network, Northwestern University. R package version 2.1.9 https://
CRAN.R-project.org/package=psych
Winters et al. 15
Rosseel, Y. (2012). Lavaan: An R package for structural equation modeling. Journal of Statistical Software,
48(2), 1–36. https://doi.org/10.18637/jss.v048.i02
Schumann, K., Zaki, J., & Dweck, C. S. (2014). Addressing the empathy deficit: Beliefs about the mal-
leability of empathy predict effortful responses when empathy is challenging. Journal of Personality
and Social Psychology,107(3), 475–493. https://doi.org/10.1037/a0036738
Shin, H. S., Park, H., & Lee, Y.-M. (2022). The relationship between medical students’empathy and burnout
levels by gender and study years. Patient Education and Counseling,105(2), 432–439. https://doi.org/
10.1016/j.pec.2021.05.036
Singer, T. (2006). The neuronal basis and ontogeny of empathy and mind reading: Review of literature and
implications for future research. Neuroscience and Biobehavioral Reviews,30(6), 855–863. https://doi.
org/10.1016/j.neubiorev.2006.06.011
Smith, A. (2006). Cognitive empathy and emotional empathy in human behavior and evolution. The
Psychological Record,56(1), 3–21. https://doi.org/10.1007/bf03395534
Walter, H. (2012). Social cognitive neuroscience of empathy: Concepts, circuits, and genes. Emotion Review,
4(1), 9–17. https://doi.org/10.1177/1754073911421379
Winters, D. E., Brandon-Friedman, R., Yepes, G., & Hinckley, J. D. (2021). Systematic review and meta-
analysis of socio-cognitive and socio-affective processes association with adolescent substance use.
Drug and Alcohol Dependence,219, 108479. https://doi.org/10.1016/j.drugalcdep.2020.108479
Winters, D. E., Pruitt, P. J., Fukui, S., Cyders, M. A., Pierce, B. J., Lay, K., & Damoiseaux, J. S. (2021).
Network functional connectivity underlying dissociable cognitive and affective components of empathy
in adolescence. Neuropsychologia,156, 107832. https://doi.org/10.1016/j.neuropsychologia.2021.
107832
Winters, D. E., Wu, W., & Fukui, S. (2020). Longitudinal effects of cognitive and affective empathy on
adolescent substance use. Substance Use & Misuse,55(6), 983–989. https://doi.org/10.1080/10826084.
2020.1717537
Wu, W., & Lang, K. M. (2016). Proportionality assumption in latent basis curve models: A cautionary note.
Structural Equation Modeling: A Multidisciplinary Journal,23(1), 140–154. https://doi.org/10.1080/
10705511.2014.938578
Wu, W., Selig, J. P., & Little, T. (2012). Longitudinal data analysis. Oxford Handbook of Quantitative
Methods,2, 387–410.
Zaki, J. (2018). Empathy is a moral force. In K. Gray, & J. Graham (Eds.), Atlas of moral psychology
(pp. 49–58). Guildford WA: The Guilford Press.
Author Biographies
Drew E. Winters, PhD. holds a T32 funded postdoctoral fellowship in developmental psy-
chobiology at the University of Colorado Anschutz Medical Campus in the department of
Psychiatry. His research focuses on the interaction between social cognition and executive
functioning in relation to adolescent mental health.
Joseph T. Sakai, M.D., is an associate professor of Psychiatry at the University of Colorado
Anschutz Medical Campus.
Suena Massey, M.D., is a Faculty of Medicine at Harvard Medical School in the department of
Psychiatry.
16 Journal of Drug Issues 0(0)
