Transdermal alcohol concentration features predict alcohol‐induced blackouts in college students

Abstract

Background
Alcohol‐induced blackouts (AIBs) are common in college students. Individuals with AIBs also experience acute and chronic alcohol‐related consequences. Research suggests that how students drink is an important predictor of AIBs. We used transdermal alcohol concentration (TAC) sensors to measure biomarkers of increasing alcohol intoxication (rise rate, peak, and rise duration) in a sample of college students. We hypothesized that the TAC biomarkers would be positively associated with AIBs.

Methods
Students were eligible to participate if they were aged 18–22 years, in their second or third year of college, reported drinking 4+ drinks on a typical Friday or Saturday, experienced ≥1 AIB in the past semester, owned an iPhone, and were willing to wear a sensor for 3 days each weekend. Students (N = 79, 55.7% female, 86.1% White, Mage = 20.1) wore TAC sensors and completed daily diaries over four consecutive weekends (89.9% completion rate). AIBs were assessed using the Alcohol‐Induced Blackout Measure‐2. Logistic multilevel models were conducted to test for main effects.

Results
Days with faster TAC rise rates (OR = 2.69, 95% CI: 1.56, 5.90), higher peak TACs (OR = 2.93, 95% CI: 1.64, 7.11), and longer rise TAC durations (OR = 4.16, 95% CI: 2.08, 10.62) were associated with greater odds of experiencing an AIB.

Conclusions
In a sample of "risky" drinking college students, three TAC drinking features identified as being related to rising intoxication independently predicted the risk for daily AIBs. Our findings suggest that considering how an individual drinks (assessed using TAC biomarkers), rather than quantity alone, is important for assessing risk and has implications for efforts to reduce risk. Not only is speed of intoxication important for predicting AIBs, but the height of the peak intoxication and the time spent reaching the peak are important predictors, each with different implications for prevention.

Full text

880
|
Alcohol Clin Exp Res. 2024;48:880–888.wileyonlinelibrary.com/journal/acer
Received: 24 November 2023
|
Accepted: 18 February 2024
DOI: 10.1111/acer.15290
RESEARCH ARTICLE
Transdermal alcohol concentration features predict
alcohol- induced blackouts in college students
Veronica L. Richards1 | Shannon D. Glenn1,2 | Robert J. Turrisi1,2 |
Kimberly A. Mallett1 | Sarah Ackerman1 | Michael A. Russell1,2
This is an op en access article under the ter ms of the Creative Commons Attribution License, which pe rmits use, distribution and reproduction in any medium,
provide d the original wor k is properly cited.
© 2024 The Authors. Alcohol, Clinical and Experimental Research published by Wiley Periodicals LLC on behalf of Research Society on Alcohol.
1Edna Bennett Pierce Prevention Research
Center, The Pennsylvania State University,
University Park, Pennsylvania, USA
2Depar tment of Biobehavioral Health, The
Pennsylvania St ate Universit y, University
Park, Pennsyl vania, USA
Correspondence
Veronica L . Richards, Edna Bennett
Pierce Prevention Research Center,
The Pennsylvania State Institute, 320
Biobehavioral Health Building, Universit y
Park, PA , 16802, USA.
Email: vlr5157@psu.edu
Abstract
Background: Alcohol- induced blackouts (AIBs) are common in college students.
Individuals with AIBs also experience acute and chronic alcohol- related conse-
quences. Research suggests that how students drink is an important predictor
of AIBs. We used transdermal alcohol concentration (TAC) sensors to measure bio-
markers of increasing alcohol intoxication (rise rate, peak, and rise duration) in a
sample of college students. We hypothesized that the TAC biomarkers would be
positively associated with AIBs.
Methods: Students were eligible to participate if they were aged 18–22 years, in their
second or third year of college, reported drinking 4+ drinks on a typical Friday or
Saturday, experienced ≥1 AIB in the past semester, owned an iPhone, and were willing
to wear a sensor for 3 days each weekend. Students (N = 79, 55.7% female, 86.1% White,
Mage = 20.1) wore TAC sensors and completed daily diaries over four consecutive week-
ends (89.9% completion rate). AIBs were assessed using the Alcohol- Induced Blackout
Measure- 2. Logistic multilevel models were conducted to test for main effects.
Results: Days with faster TAC rise rates (OR = 2.69, 95% CI: 1.56, 5.90), higher peak
TACs (OR = 2.93, 95% CI: 1.64, 7.11), and longer rise TAC durations (OR = 4.16, 95% CI:
2.08, 10.62) were associated with greater odds of experiencing an AIB.
Conclusions: In a sample of "risky" drinking college students, three TAC drinking fea-
tures identified as being related to rising intoxication independently predicted the
risk for daily AIBs. Our findings suggest that considering how an individual drinks (as-
sessed using TAC biomarkers), rather than quantity alone, is important for assessing
risk and has implications for efforts to reduce risk. Not only is speed of intoxication
important for predicting AIBs, but the height of the peak intoxication and the time
spent reaching the peak are important predictors, each with different implications
for prevention.
KEYWORDS
alcohol- induced blackouts, alcohol- related consequences, college students, drinking,
transdermal alcohol concentration

|
881
TRANSDERMAL ALCOHOL CONCENTRATION AND BLACKOUTS
INTRODUCTION
Alcohol- induced blackouts (AIBs) are a consequence of drinking
that act as an accelerant for other consequences (Richards, Glenn,
et al., 2023). AIBs are defined as experiencing periods of amnesia
for all or part of a drinking episode (Wetherill & Fromme, 2016).
Individuals experiencing an AIB remain conscious and continue to
engage with their environment. Several studies have shown that
AIBs are associated with experiencing additional alcohol- related
consequences, ranging in severity (e.g., feeling embarrassed to sex-
ual victimization; Hingson et al., 2016; Merrill et al., 2019; Voloshyna
et al., 2018). On nights when an AIB is experienced, college student s
experience an average of 3.5 additional consequences compared to
non- AIB nights (Richards, Glenn, et al., 2023). AIBs have also been
associated with longer term consequences such as experiencing
symptoms of alcohol use disorder (AUD; Glenn et al., 2023; Studer
et al., 2019; Yuen et al., 2021). What makes these statistics even
more concerning is the frequency in which student drinkers report
experiencing an AIB. In a longitudinal study of more than 1700 col-
lege students, approximately 80% reported at least one AIB during
college (Glenn et al., 2022). Upon closer examination, these same
students experienced an average of eight AIBs during college (Glenn
et al., 2023). The frequency of AIBs and associations with additional
alcohol- related consequences suggest that reducing AIBs could help
prevent alcohol- related harm in college students.
Associations between alcohol consumption and AIBs have been
routinely reported in published studies (Evans- Polce et al., 2022;
Mallett et al., 2011; Rose & Grant, 2010). It has been shown that
quantity of alcohol consumption is not the only contributing factor
for AIBs. Several studies have identified the speed of intoxication
(i.e., rapid rises of blood alcohol concentration) to be an import-
ant risk factor for AIBs (Carpenter & Merrill, 2021; White, 2003).
Behaviors that impact speed of intoxication (e.g., use of protective
behavioral strategies, playing drinking games) have also been asso-
ciated with AIBs at equivalent levels of alcohol consumption (Carey
et al., 2022; Ray et al., 2014; Richards, Turrisi, et al., 2023). Together,
these findings indicate that it is not just the amount someone drinks,
but the manner in which they drink that impacts AIBs.
Nearly all of the literature examining alcohol- related conse-
quences, including AIBs, rely on self- reported drinking data. While
self- report provides important information about how much some-
one drinks (and sometimes for how long they drink), it does not
provide complete information about the manner in which someone
drinks. Concerns have also been raised about using self- report on
heavier drinking nights (Northcote & Livingston, 2011). In the con-
text of studying AIBs, self- report is further limited by the notion that
individuals do not remember at least part of the drinking episode
they are reporting on. Transdermal alcohol concentration (TAC) sen-
sors offer a possible solution by passively measuring biomarkers of
alcohol intoxication in near real time that do not rely on self- reports.
The most commonly used TAC sensor has been the SCR AM-
Continuous Alcohol Monitor (SCRAM- CAM; Alcohol Monitoring
Systems, Littleton, CO). While the SCRAM- CAM has been well
validated and used extensively in intervention studies (e.g., Alessi
et al., 2017; Barnet t et al., 2017; Dougherty et al., 2014, 2015;
Mathias et al., 2018), it is limited due to its appearance/size, dis-
comfort, and the stigma associated with wearing an ankle monitor
(Courtney et al., 2022; Villalba et al., 2020). Studies using a newer
generation, wrist- worn TAC sensor, the BACtrack Skyn (BACtrack
Inc., San Francisco, CA) have emerged. These studies indicate the
potential to incorporate a valid and more user- friendly TAC sensor
into “real- world” alcohol studies (e.g., Ash et al., 2022; Courtney
et al., 2022; Richards et al., 2022; Rosenberg et al., 2023; Wang
et al., 2021). Evidence support s its validity compared to both the
SCRAM- CAM and self- reports (Courtney et al., 2022; Richards
et al., 2022). The Skyn uses fuel cell technology to measure TAC
every 20 s, resulting in the continuous measure of alcohol intoxica-
tion (Wang et al., 2018).
Previous studies have examined all parts of the TAC curve as pre-
dictors of outcomes, including those reflective of declining alcohol
intoxication (Russell et al., 2022; Simons et al., 2015). For example,
fall rate describes the speed of alcohol elimination during declining
portions of the curve. Duration (time spent alcohol positive) and area
under the cur ve (AUC; cumulative burden of alcohol exposure; the
sum of TAC level by time across the day) combine increasing and
decreasing segments of the curve. In this study, we focus on rising
portions of the TAC curve, including rise rate (the speed of alcohol
intoxication), rise duration (the length of time spent under rising in-
toxication), and peak TAC (maximum TAC observed). Focus on fea-
tures associated with rising portions of the TAC curve will translate
more effectively to preventive actions because people can change
how much and how quickly they drink, but they cannot change how
quickly it is eliminated once consumed. Table 1 shows these features,
describing their interpretation, operationalization, and calculation.
TAB LE 1 Day- level transdermal alcohol concentration (TAC) features defined.
Feature Interpretation Operationalization Calculation
Rise rate The speed of alcohol consumption/
exposure
The day's average rate of TAC increase
per hour
mean(
ΔTACidt
ΔHours
idt )
∣ΔTACidt >
0
Peak The level of intoxication “achieved” Maximum TAC observed for the day's
drinking event(s)
max(
TACidt
)
Rise duration The duration of time spent under rising
alcohol intoxication
The total number of hours in which the
curve was rising that day
sum(
ΔHoursidt
)
∣ΔTACidt >
0

882
|
RICHARDS et al .
This study
This study was an analysis of data from an intensive longitudinal
study incorporating the Skyn TAC sensor and daily diaries designed
to examine TAC features and AIBs in a “risky” drinki ng sample of co l-
lege students across 12 days (Thursday, Friday, and Saturday) over
four weekends. We include college students in their second and
third years because this is a developmentally important time for al-
cohol risk (e.g., students move off campus, experience less monitor-
ing/regulation, and approach their 21st birthdays). This study design
allows for the examination of TAC features that vary daily, to an-
swer the question: Does the way in which students drink predict AIBs?
Our hypotheses focused on the day- level since we were interested
in how daily differences (rather than average differences between
people) in drinking manner predicted AIBs. Previous work demon-
strates that days with faster rise rates and higher peaks were signifi-
cantly associated with reporting more alcohol- related consequences
(Russell et al., 2022). No study to the best of our knowledge has
examined the relationship between rise duration and any alcohol-
related consequences, but previous work showed a positive asso-
ciation between overall duration and alcohol- related consequences
(Russell et al., 2022). We aimed to extend this work by examining
rising TAC features as a predictor of AIBs. We hypothesized that TAC
features associated with the rising portion of the cur ve (rise rate,
peak, an d ri se dur ation) would be posi ti ve ly ass ociate d wi th AIBs be-
cause they index times in which alcohol intoxication was increasing.
MATERIALS AND METHODS
Procedures and participants
The study consisted of 79 par ticipants aged 18–22 years who reg-
ularly engaged in “risky” drinking. To recruit the sample, we ran-
domly selected 60 00 students (50% sophomores and 50% juniors
based on credits) from the university registrar's database of stu-
dents at a large, public university in the northeastern United States.
Participants received an email invitation describing the study and
inviting their participation. All recruitment materials (invitation email
and up to six daily reminder emails) included a personalized URL to
access a screening survey. Student s were eligible to participate if
they were aged 18–22 years, in their second or third years of college,
reported drinking four or more drinks on a typical Friday or Saturday
in the past semester, ex perience d at lea st one AIB in th e pa st semes-
ter, owned an iPhone, and were willing to wear a sensor for 3 days
each weekend ( Thursday night through Sunday morning over four
weekends). Eligible students were immediately redirected to a base-
line survey which took approximately 20 min to complete. At the end
of the baseline survey, students scheduled a 10- min appointment to
come into the laboratory to pick up the sensor and receive training
on how to use it. Training took place the Monday–Thursday prior to
the daily portion of the study beginning. Data collection was timed
to avoid a large philanthropic event that takes place on campus each
February, spring break, and final examination periods in order to re-
flect typical drinking patterns. All procedures were approved by the
University's Institutional Review Board.
At screening, 15.5% (n = 927) of invited students consented to
participate in the study, of which 73.1% (n = 678) completed the
screening. A total of 28.3% (n = 192) of screened participants were
eligible to continue, of which 100% completed the baseline survey.
Due to device availabilit y, 80 students were able to participate in the
field period. Prior to the field period, one student dropped from the
study, resulting in a final sample of 79 students.
Students were instructed to wear the sensor continuously start-
ing Thursday evening at 5 p.m. through Sunday morning when they
woke up for four weekends, with the exception of removing the de-
vice to shower. We used the BACtrack Skyn model T15. The sen-
sors can hold a charge for up to 10 days, so students were instructed
to charge them once a week prior to Thursday. They received an
email and text message at 4 p.m. each Thursday reminding them to
turn on the device and wear it by 5 p.m. Daily surveys were sent to
all students each weekend morning (Friday–Sunday) at 10 a.m. for
four weekends about the day/night before. Surveys were available
until 6 p.m. each day. To enhance compliance and retention, stu-
dents received an email and text reminder 3 days prior to the start of
each social weekend (i.e., students received a reminder on Monday
to begin on Thursday). An email reminder to complete the survey
at 1 p.m. and a text message reminder at 4 p.m. each weekend day
was also sent to students. There was an average completion rate of
89.9% across the 12 surveys (range: 1–12).
Participants were compensated $15 for completing the baseline
survey and an additional $5 for each survey completed, for up to $75
total. Once par ticipants completed nine surveys, they were entered
for a chance to win one of eight $100 gift cards. Completing nine
surveys resulted in nine chances per student, with this increasing
by two for each additional completed survey (up to 15 chances for
completing all 12 sur veys).
In the present sample of 79 students, the mean age at base-
line was 20.1 years (SD = 0.9) and the majority identified as female
(55.7%) and White (86.1%). Par ticipants also identified as Hispanic/
Latino (11.4%), Asian (5.1%), Black (1.3%), or multiracial (3.8%). The
sample was split evenly between sophomores (49.4%) and juniors
(50.6%). Table 2 sh ows part icipant dem ographics by AIB occurrence.
Measures
Alcohol- induced blackouts
The Alcohol- Induced Blackout Measure- 2 (ABOM- 2; Boness
et al., 2022) was used to assess whether students experienced a frag-
mentar y or en bloc AIB each night they reported drinking. Students
were asked, “As a result of drinking yesterday, did you/were you
______” Response options were dichotomous with “No” (0) and “Yes”
(1). Fragmentary AIBs were assessed with four items (e.g., have fuzzy
memories of events, unable to remember a few minutes). En bloc AIBs

|
883
TRANSDERMAL ALCOHOL CONCENTRATION AND BLACKOUTS
were assessed with four items (e.g., unable to remember what hap-
pen ed the night befo re, wake up wit h no id ea wher e you had been). The
eight items from the two types of AIBs were examined together due to
low frequen cy of en bl oc AIBs (n = 18 en bloc AIBs reported). Responses
from the eight items were summed and recoded to a dichotomous scale
with “No AIB reported” (0) and “Yes, AIB reported” (1).
Segmenting TAC data into “social days”
Because drinking of ten extends past midnight and is not encapsu-
lated neatly in midnight- to- midnight calendar days, we used “social
days” as the primary temporal unit of analysis. As in previous TAC
sensor studies of alcohol use, TAC data were segmented into “so-
cial days” using 10:00 a.m. as the boundary for the end of the social
day because it matched the time of the morning survey (Courtney
et al., 2022; Richards et al., 2022; Russell et al., 2022). If participants
had any valid TAC data on a social day, it was considered a drinking
day. Day- level features were calculated separately for each social
day if a TAC drinking episode spanned multiple social days.
Day- level TAC data
Initial processing
Drinking episodes were identified and coded using a combination
of previously published research guidelines (Courtney et al., 2022;
Didier et al., 2023; Richards et al., 2022). We applied algorithms
to filter out observations in which the sensor was turned on but
not worn. These algorithms are based on the sensor's tempera-
ture and movement sensors (similar to algorithms used by Didier
et al., 2023). The sensor assesses TAC every 20 s, resulting in
443,999 observations. First, if the temperature sensor indicated
a temperature greater than 28°C, it was considered worn (88% of
observations). Second, if the temperature was less than or equal
to 28°C but more than 5°C above the participants minimum tem-
perature, it was considered worn (an additional 7.4% of obser va-
tions). Third, if the temperature was more than 3°C above the
participant's minimum temperature and the motion sensor regis-
tered above 0.01 Gs, it was considered worn (an additional 0.4%
of observations).
Identifying drinking episodes
Once nonworn observations were filtered out of the dataset,
drinking episodes were identified. Smoothing is fitting a longitudi-
nal function to the data. It is commonly performed with TAC sen-
sor data to minimize the influence of random measurement noise
in th e time series and facilitate feature extraction (Courtney et al.,
2022; Richards et al., 2022; Russell et al., 2022). As a first step,
the TAC time series for each person was smoothed using a 30-
min centered moving average (e.g., Courtney et al., 2022; Richards
et al., 2022). Following recommendations from the manufacturer,
observat ions that were less than 0 after smoothi ng were recorded
to 0. Episodes were demarcated according to the following. The
start of an episode was indicated if (a) there were two consecu-
tive zero readings followed by at least one positive (TAC ≥ 5) read-
ing, and (b) there was a positive at the start of the data stream
and at least one positive on the next two measurement occasions.
The end of an episode was indicated if (a) an alcohol positive is
followed by two consecutive nonpositive readings, (b) the last
alcohol- positive observation is the last reading for the person, or
(c) an alcohol- positive observ ation is followe d by a neg ative that is
the person's last observation. Episodes were divided in two if con-
secutive observations were more than 30 min apart. False positive
episodes were filtered out using guidelines adapted from previ-
ously published studies (Courtney et al., 2022; Didier et al., 2023;
Richards et al., 2022). Our specific criteria focused on identify-
ing TAC curves with features that were biologically implausible
according to previous literature and manufacturer information.
Specifically, we removed episodes that (1) were less than 45 min
in duration, (2) were less than or equal to 60 min in duration and
with a peak greater than or equal to 400 μg/L air, and/or (3) had
rise and fall rates that do not fall between ±20 and 300. This re-
moved 38% of originally identified episodes, but these episodes
contained only 6.8% of the total sensor data, leaving 718 drinking
episodes with valid TAC data.
Calculating day- level TAC features
Three TAC features were extracted from each social day with TAC-
positive episode data. If a day contained multiple TAC episodes,
TAB LE 2 Participant characteristics.
Experienced an
AIB (N = 59)
Did not experience
an AIB (N = 20)
Frequency (%) Frequency (%)
Sex
Female 34 (77.3%) 10 (22.7%)
Male 25 (71.4%) 10 (28.6%)
Age
Mean (SD) 20.2 (0.9) 19.8 (1.0)
Race
Asian 2 (50.0%) 2 (50.0%)
Black 0 (0.0%) 1 (100.0%)
Multiracial 1 (33.3%) 2 (66.7%)
White 54 (79.4%) 14 (20.6%)
Other 2 (66.7%) 1 (33.3%)
Ethnicity
Hispanic/Latino 4 (44.4%) 5 (55.6%)
Not Hispanic/Latino 55 (78.6%) 15 (21.4%)
Year in school
Sophomore 28 (71.8%) 11 (28.2%)
Junior 31 (77.5%) 9 (22.5%)
Abbreviation: AIB, alcohol- induced blackout.

884
|
RICHARDS et al .
features were calculated using all data for the day. TAC features
included peak, rise rate, rise duration, and rise AUC. A description
of TAC features' interpretations and calculations are available in
Table 1. TAC features were set to zero if the sensor was worn for
80% or more of the hours of the social day but there were no epi-
sodes present (n = 232 days), as this suggested that no drinking had
occurred. TAC features were left missing if no episodes were pre-
sent but the sensor was worn for less than 80% of the hours of the
soc ial day (n = 152 days). Of the 152 missing days, daily diaries were
completed for 28 days, of which only 12 days were drinking days (M
drinks reported = 1.74, SD = 2.65).
Statistical analysis
All data analyses were conducted in R. Means, day- , week- , and
person- level SDs were generated from empty multilevel linear mod-
els. To test our hypothesis, multilevel logistic models conducted
to examine the associations between individual TAC features (rise
rate, peak, and rise duration) and AIBs. Each model had three lev-
els of variation (day, week, and person). We used a 3- level center-
ing strategy to partition the variance of each TAC feature into day,
week, and person levels. Raw TAC feature values were centered
on person- day- means (creating a day- level TAC feature variable),
person- week- means were centered on person- means (creating a
week- level TAC feature variable), and person- means were centered
on the grand mean (creating a person- level TAC feature variable).
All variables were z- scored to allow for meaningful comparison of
odds ratio (OR) effect sizes. Each individual TAC feature model in-
cluded random intercepts and slopes. A post hoc analysis was con-
ducted to examine one model containing all three TAC features. To
account for the strong correlations between rise rate and peak TAC
(rs between 0.78 and 0.83 at day, week, and person levels; Table S1),
we conducted three additional models including a combination of
two features per model (i.e., rise rate and rise duration; peak TAC
and rise duration; and rise rate and peak TAC). Models were esti-
mated in a Bayesian framework (with noninformative priors and
20,00 0 total iterations [50% warmup]) using the brms package in
R (Bürkner, 2 017). Model estimates are the means of the posterior
parameter distributions; significance is determined using 95% cred-
ible inter vals (CIs). Associations are presented as ORs; CIs that do
not contain 1.0 were considered significant.
RESULTS
Descriptive statistics
Table 3 shows descriptive statistics for study variables among
students with valid TAC data (n = 76 participants, 410 days with
reports on whether or not an AIB was experienced). A total of
486 TAC- positive days were recorded and 147 AIBs were re-
ported. More than two thirds (69.3%) of students experienced
at least one AIB. Of those who experienced at least one AIB, an
average of 2.2 AIBs (SD = 1.5) were experienced over the 12- day
study period.
Associations between transdermal alcohol
concentration features and blackouts
Results of the full multilevel logistic models are presented in
Table 4. All TAC drinking features examined predicted AIBs. Days
with fa ster TAC ris e rat es (Table 4, model 1: OR = 2.69, 95% CI: 1.56,
5.90), higher peak TACs (Table 4, model 2: OR = 2.93, 95% CI: 1.64,
7.11), and longer rise TAC durations (Table 4, model 3 OR = 4.16,
95% CI: 2.08, 10.62) were associated with greater odds of experi-
encing an AIB.
Results of the models containing combined TAC features are
presented in Table 5. When all three features were included, daily
rise rate duration remained significantly associated with AIBs,
while daily rise rate and peak TAC were not associated with AIBs
(Table 5, model 1). In the model containing rise rate and rise du-
ration, both features were associated with AIBs at the day level
(Table 5, model 2). In the model containing peak TAC and rise du-
ration, both features were associated with AIBs at the day level
(Table 5, model 3). In the model containing rise rate and peak TAC,
peak TAC was significantly associated with AIBs but rise rate was
not (Table 5, model 4).
DISCUSSION
More than 30% of all drinking days resulted in an AIB. Full suppor t
for our hypothesis was observed. We identified three biomarkers of
increasing intoxication (rise rate, peak, and rise duration) that predict
TABLE 3 Descriptive statistics for study variables.
N persons N weeks N days Mean
Person- level
SD Week- level SD Day- level SD
AIBsa76 224 410 −1 .15 0.94 0.39 –
TAC rise rate 78 247 717 53.11 14 .0 9 5.10 53.11
Peak TAC 78 247 718 120.87 66.55 23.48 148.0 4
Rise TAC duration 78 247 717 2.36 0.54 0.28 2.56
Abbreviations: AIB, alcohol- induced blackout; SD, standard deviation; TAC, transdermal alcohol concentration.
aThe AIB empt y model was logistic, thus there was no residual (i.e., no day- level SD).

|
885
TRANSDERMAL ALCOHOL CONCENTRATION AND BLACKOUTS
likelihood of experiencing an AIB in “risky” college student drinkers.
Our results support the notion that the manner in which students
drink is important for risk prediction of AIBs.
Days with faster rise rates predicted AIBs. This finding sup-
ports the literature which suggests that the speed in which alco-
hol intoxication increases is impor tant (Carpenter & Merrill, 2021;
Goodwin et al., 1969; White, 2003). Carpenter and Merrill (2021)
assessed drinking every hour during an episode using ecological
momentary assessment (EMA) to examine changes in estimated
blood alcohol concentration. While using these real- time methods
for measuring self- reported drinking may be more accurate than
relying on next morning, retrospective reports, EMA still relies on
the par ticipant to respo nd to pr ompts while activel y dr inking. TAC
sensors are less burdensome to participants since they measure
alcohol intoxication passively. Interventions that target speed
of drinking would likely reduce the risk of experiencing an AIB.
These might include encouraging students to use more protective
behavioral strategies (e.g., alternating alcoholic and nonalcoholic
drinks) or engaging in less risky drinking behaviors (e.g., avoiding
playing drinking games; Carey et al., 2022; Fairlie et al., 2015;
Pearson, 2013; Prince et al., 2013).
Days with higher peak TACs were associated with increased
likelihood of AIBs. Reducing the speed of intoxication (rise rate)
would also work to decrease peak TAC. Higher peak TACs cor-
respond with higher BACs which are indicative of acute alcohol
intoxication (Dougherty et al., 2012; Vonghia et al., 2008). Peak
BAC has often been a target for intervention in college stu-
dents using personalized normative feedback (PNF; e.g., Carey
et al., 2007; Collins et al., 2002; Miller et al., 2016). Peak TAC
coul d be emp ha sized in PNF to facili tate greater pr ecision in treat-
ment (Barnett, 2015).
Days with longer rise durations were associated with increased
likelihood of AIBs. Rise duration corresponds to the amount of time
spent increasing alcohol intoxication. In conjunction with rise rate
and peak, time spent becoming intoxicated can be a target for inter-
vention. Previous work in young adults showed that high intensity
drinking days (8+/10+ drinks for females/males) were character-
ized by longer times spent drinking and more rapid paces of drink-
ing (Patrick et al., 2023). It is possible that targeting reductions in
any one of the examined TAC features could result in a reduction
in others.
Our results extend previous findings that indicate TAC features
independently predict drinking and related consequences (Russell
et al., 2022; Simons et al., 2015). We identified three TAC features
that are strongly predictive of AIBs (ORs bet ween 2.69 and 4.16).
Estimates from models combining these features indicate that rise
rate duration contributes the largest amount of unique variance to
our prediction. It is unclear if this translates to the “most import-
ant” clinical TAC feature, as inter vention trials using TAC have not
yet been conducted and it is therefore unclear which features are
most amenable to change. Future studies are needed to identify the
degree to which each TAC feature is amenable to change in inter-
vention efforts.
An alternative approach could have been to examine all TAC fea-
tures available, such as fall rate and AUC . We declined to examine
fall rate in relation to AIBs because it is difficult to behaviorally alter
TAB LE 4 Results from multilevel logistic models predicting alcohol- induced blackouts from transdermal alcohol concentration (TAC)
features.
Model 1: TAC rise rate Model 2: Peak TAC Model 3: Rise TAC duration
OR 95% CI OR 95% CI OR 95% CI
Fixed effects
Intercept 0.16 0.07, 0.27 0.13 0.05, 0.25 0.08 0.02, 0.18
Daily TAC 2.69 1.56, 5.90 2.93 1.64, 7.11 4 .16 2.08 , 10.62
Weekly mean TAC 1.32 0.97, 1.90 1 .52 1.07, 2.23 1.14 0.75, 1.75
Person- mean TAC 2. 24 1.47, 3.84 1.86 1.24, 3.16 1.91 1.18, 3.34
Week- level random ef fect s
Intercept SD 0.82 0.03, 2.17 1.01 0.05, 2.51 1.64 0.51, 2.96
Daily TAC slope SD 1.05 0.08, 2.47 1.17 0.15, 2.66 0.65 0.03, 1.92
Correlation (intercept and daily TAC
slope)
0.03 −0.89, 0.93 −0.05 −0.90, 0.92 −0.36 −0.98, 0.84
Person- level random effects
Intercept SD 0.60 0.04, 1.32 0.77 0.07, 1.63 0.55 0.02, 1.46
Daily TAC slope SD 0.72 0.07, 1.59 0.78 0.06, 1.81 1.13 0.25, 2.11
Correlation (intercept and daily TAC
slope)
0.41 −0.79, 0.98 0.31 −0.76, 0.97 −0.07 −0.94, 0.93
Note: Significant fixed effects (95% CIs that do not contain 1) are in bold; all independent variables were z- scored to allow meaningful comparison of
OR effect sizes.
Abbreviations: CI, credible interval; OR, odds ratio; SD, standard deviation; TAC, transdermal alcohol concentration.

886
|
RICHARDS et al .
alcohol elimination, and may be limited in implications for prevention
and intervention efforts. TAC- AUC has been identified as a predic-
tor for alcohol- related consequences (Russell et al., 2022; Simons
et al., 2015). A large TAC- AUC may be indicative of several different
patterns of drinking. It may reflect a drinking episode with a high
peak and short duration or a drinking episode with a low peak but
long duration. Thus, it may be difficult to know which part of the
drinking curve to directly target for optimal harm reduction. We in-
stead focused on features with clearer implications for prevention
and intervention.
Limitations
The following limitations should be considered. First, our sample
consisted of primarily White, female college students who re-
ported recent heavy drinking and AIBs. Results may not general-
ize to more diverse populations or those with different drinking
patterns. Second, the study was conducted over a 1- month period
which included St. Patrick's Day and Easter weekend. St. Patrick's
Day may be associated with especially high risk drinking (Mallett
et al., 2013) and Easter may be associated with especially low risk
drinking because students tend to travel home for this holiday. The
study period is perhaps more representative of a students' “true”
drinking patterns because of this variability. Third, it is possible
that the sensors missed lower intensity drinking days (Barnett
et al., 2014). There is currently no “gold standard” algorithm for
detecting drink ing days for the Sk yn sensor. Our guidelines for de-
tec ting drinking days were informed by pre vi ou s studies (Cour tn ey
et al., 2022; Didier et al., 2023; Richards et al., 2022). Fourth, our
study had a analy tic sample of 76 student s which was chosen, in
part, based on costs for the sensors and providing incentives to
participants for four weekends of data collection. The sample size
did limit the examination of potential moderating effects (e.g.,
baseline AUD, contextual factors) which may have important im-
plications for risk reduction and prevention (Stevely et al., 2020).
Fifth, we were unable to examine the relationship between TAC
features and type of AIB because we observed only 18 en bloc
AIBs over the study period. Future research is needed to distin-
guish between types of AIBs. Sixth, a unique aspect of AIBs is that
stud ents ma y not rea li ze they had one th e nex t mor ni ng. It is possi-
ble that student s completed the morning surveys prior to speak ing
TABLE 5 Result s from multilevel logistic modes predicting alcohol- induced blackouts from combined transdermal alcohol concentration
(TAC) features.
Model 1: Rise rate, peak,
and rise duration
Model 2: Rise rate and rise
duration
Model 3: Peak and rise
duration
Model 4: Rise rate and
peak
OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Fixed effects
Intercept 0.09 0.04, 0.18 0.10 0.05, 0.19 0.11 0.05, 0.21 0.18 0.10, 0. 27
Daily TAC rise rate 1.35 0.80, 2.46 1.58 1.12, 2.23 ––1 .19 0.74, 1.99
Weekly mean TAC
rise rate
0.97 0.50, 1.81 1 .47 1.03, 2.15 ––0 .96 0.58, 1.53
Person- mean TAC
rise rate
2.62 1.25, 6.48 2.22 1.32, 4.03 ––2.03 1.12, 3.96
Daily Peak TAC 1.23 0.72, 2.16 ––1.53 1.10, 2.18 1.87 1.21, 3.96
Weekly mean Peak
TAC
2.02 0.94, 4.75 ––1.91 1. 23, 3.17 1.55 0.95, 2.74
Person- mean Peak
TAC
0.77 0.3 0, 1.73 ––1.56 0.82, 3.03 1.00 0.56, 1.70
Daily TAC rise
duration
2.65 1.61, 4.81 2.68 1.77, 4.36 2.30 1.45, 3.88 – –
Weekly mean TAC
rise duration
0.72 0.40, 1.24 0.99 0.67, 1.49 0.70 0.42, 1.14 – –
Person- mean TAC
rise duration
1.26 0.61, 2.66 1 .11 0.65, 1.89 1.18 0.62, 2.32 – –
Week- level random ef fect s
Intercept SD 1.15 0.10, 2.27 0 .98 0.06, 2.02 0.97 0.07, 1.98 0.66 0.03, 1.5 4
Person- level random effects
Intercept SD 0.79 0.08 , 1.57 0. 69 0.05, 1.41 0. 81 0.09, 1.48 0.71 0.08, 1.32
Note: Significant fixed effects (95% CIs that do not contain 1) are in bold; all independent variables were z- scored to allow meaningful comparison of
OR effect sizes.
Abbreviations: CI, credible interval; OR, odds ratio; SD, standard deviation; TAC, transdermal alcohol concentration.

|
887
TRANSDERMAL ALCOHOL CONCENTRATION AND BLACKOUTS
with friends who were present during the reported drinking epi-
sode, thus they may misreport the absence of an AIB.
CONCLUSION
We identified three TAC drinking features related to rising intoxi-
cation that independently predicted the risk for daily AIBs among
a sa mple of “risky” dr in kin g colle ge st ud ent s. Our find in gs su gge st
that considering manner in which someone drinks as assessed
using TAC biomarkers, rather than quantity alone, is important
for risk reduction. Not only is speed of intoxication important for
predicting AIBs, but peak intoxication and how long they spend
reaching their peak are also important predictors with prevention
implications.
ACKNOWLEDGEMENTS
This work was supported by the National Institutes of Health (NIDA
T32 DA017629; MPIs: J. Maggs and S. L anza) and by departmental
funds awarded to Rober t Turrisi. The content is solely the responsi-
bility of the authors and does not necessarily represent the official
views of the National Institutes of Health.
CONFLICT OF INTEREST STATEMENT
The authors have no conflicts of interest to disclose.
DATA AVAIL AB ILI T Y STAT EME N T
The data that support the findings of this study are available from
the corresponding author upon reasonable request.
ORCID
Veronica L. Richards https://orcid.org/0000-0002-1391-0607
Sarah Ackerman https://orcid.org/0000-0002-6833-2402
Michael A. Russell https://orcid.org/0000-0002-3956-604X
REFERENCES
Alessi, S.M., Barnet t, N. P. & Petry, N.M. (2017) Experiences with
SCRAMx alcohol monitoring technology in 100 alcohol treatment
outpatients. Drug and Alcohol Dependence, 178, 417–424.
Ash, G.I., Gueorguieva, R., Barnet t, N.P., Wang, W., Robledo, D.S.,
DeMart in i, K. S. et al. (202 2) Sen si ti vi ty, specificit y, an d to le ra bi lity
of the BACTrack Skyn compared to other alcohol monitoring ap-
proaches among young adults in a field- based setting. Alcoholism:
Clinical and Experimental Research, 46, 783–796.
Barnet t, N.P. (2015) Alcohol sensors and their potential for improving
clinical care: editorial. Addiction, 110, 1–3.
Barnett, N.P., Celio, M.A., Tidey, J.W., Murphy, J.G., Colby, S.M. & Swift,
R.M. (2017) A preliminary randomized controlled trial of contin-
gency management for alcohol use reduction using a transdermal
alcohol sensor. Addiction, 112, 1025–1035.
Barnet t, N.P., Meade, E.B. & Glynn, T.R. (2014) Predictors of detec-
tion of alcohol use episodes using a transdermal alcohol sensor.
Experimental and Clinical Psychopharmacology, 22, 86–96.
Boness, C.L., Gatten, N., Treece, M.K. & Miller, M.B. (2022) A mixed-
methods approach to improve the measurement of alcohol- induced
blackouts: ABOM- 2 . Alcoholism: Clinical and Experimental Research,
46, 1497–1514.
Bürkner, P.- C. (2017) Brms: an R package for Bayesian multilevel models
using Stan. Journal of Statistical Software, 80, 1–28.
Carey, K.B., Scott- Sheldon, L. A. J., Carey, M.P. & DeMartini, K .S. (20 07)
Individual- level interventions to reduce college student drinking: a
meta- analytic review. Addictive Behaviors, 32, 2469–2494.
Carey, K.B., Tempchin, J., DiBello, A.M. & Mastroleo, N.R. (2022) Use
of protective behavioral strategies and blackout experience among
mandated college student s. Addictive Behaviors, 132, 107340.
Carpenter, R .W. & Merrill, J.E. (2021) How much and how fast: alcohol
consumption patterns, drinking- episode affect, and next- day con-
sequences in the daily life of underage heavy drinkers. Drug and
Alcohol Dependence, 218, 108407.
Collins, S.E., Carey, K.B. & Sliwinski, M.J. (2002) Mailed personalized nor-
mative feedback as a brief intervention for at- risk college drinkers.
Journal of Studies on Alcohol, 63, 559–567.
Courtney, J.B., Russell, M.A . & Conroy, D.E. (2022) Acceptability and va-
lidity of using the BACtrack Skyn wrist- worn transdermal alcohol
concentration sensor to capture alcohol use across 28 days under
naturalistic conditions—a pilot study. Alcohol, 108, 30–43.
Di die r, N. A. , King , A.C., Pol ley, E.C . & Fri dbe rg, D. J . (20 23) Sign al pr oce s s-
ing and machine learning with transdermal alcohol concentration to
predict natural environment alcohol consumption. Experimental and
Clinical Psychopharmacology. Advance online publiication.
Dougherty, D.M., Charles, N.E., Acheson, A., John, S., Furr, R.M. & Hill-
Kapturczak, N. (2012) Comparing the detection of transdermal and
breath alcohol concentrations during periods of alcohol consump-
tion ranging from moderate drinking to binge drinking. Experimental
and Clinical Psychopharmacology, 20, 373–381.
Dough er ty, D.M ., Hil l- Kaptur czak, N ., Lia ng , Y., Kar ns , T.E., Ca tes, S.E .,
Lake, S.L. et al. (2014) Use of continuous transdermal alcohol
monitoring during a contingency management procedure to re-
duce excessive alcohol use. Drug and Alcohol Dependence, 142,
301–306.
Dougherty, D.M., Karns, T.E., Mullen, J., Liang, Y., L ake, S.L., Roache,
J.D. et al. (2015) Transdermal alcohol concentration data collected
during a contingency management program to reduce at- risk drink-
ing. Drug and Alcohol Dependence, 148, 77–84.
Evans- Polce, R.J., Stevenson, B.L. & Patrick, M.E. (2022) Daily- level
analysis of drinking intensity and acute physical consequences.
Addictive Behaviors, 128, 107246.
Fairlie, A.M., Mag gs, J.L. & Lanza, S.T. (2015) Prepartying, drinking
games, and extreme drinking among college students: a daily- level
investigation. Addictive Behaviors, 42, 91–95.
Glenn, S.D., Turrisi, R., Waldron, K.A ., Mallett, K. A., Russell, M. A. &
Reavy, R.R. (2022) Examining the impact of early college experi-
ences on the cumulative number of alcohol- related consequences.
Addictive Behaviors, 132, 107357.
Glenn, S.D., Turrisi, R.J., Richards, V.L., Russell, M. A. & Mallett , K.A .
(2023) A dual- process decision- making model examining the lon-
gitudinal associations between alcohol- induced blackouts and al-
cohol use disorder risk among college student drinkers. Journal of
Studies on Alcohol and Drugs, 85, 73–83.
Goodwin, D.W., Crane, J.B. & Guze, S.B. (1969) Phenomenological as-
pects of the alcoholic “Blackout”. The British Journal of Psychiatry,
115, 1033–1038.
Hingson, R., Zha, W., Simons- Mor ton, B. & White, A. (2016) Alcohol-
induced blackouts as predictors of other drinking related harms
among emerging young adults. Alcoholism, Clinical and Experimental
Research, 40, 776–784.
Mallett, K.A ., Varvil- Weld, L., Borsari, B., Read, J.P., Neighbors, C. &
White, H.R. (2013) An update of research examining college stu-
dent alcohol- related consequences: new perspectives and im-
plications for interventions. Alcoholism, Clinical and Experimental
Research, 37, 709–716.
Mallett, K.A., Var vil- Weld, L., Turrisi, R. & Read, A. (2011) An examina-
tion of college students' willingness to experience consequences

888
|
RICHARDS et al .
as a unique predictor of alcohol problems. Psychology of Addictive
Behaviors, 25, 41–47.
Mathias, C.W., Hill- Kapturc zak, N., Karns- Wright, T.E., Mullen, J.,
Roache, J.D., Fell, J.C. et al. (2018) Translating transdermal alcohol
monitoring procedures for contingency management among adults
recently arrested for DWI. Addictive Behaviors, 83, 56–63.
Merrill, J.E., Boyle, H.K., Jackson, K .M. & Carey, K.B. (2019) Event- level
correlates of drinking events charac terized by alcohol- induced
blackouts. Alcoholism, Clinical and Experimental Research, 43,
2599–26 0 6.
Miller, M.B., Leavens, E.L., Meier, E., Lombardi, N. & Leffingwell, T.R.
(2016) Enhancing the efficacy of computerized feedback interven-
tions for college alcohol misuse: an exploratory randomized trial.
Journal of Consulting and Clinical Psychology, 84, 122–133.
Northcote, J. & Livingston, M. (2011) Accuracy of self- reported drink-
ing: obser vational verification of ‘last occasion’ drink estimates of
young adults. Alcohol and Alcoholism, 46, 709–713.
Patrick, M.E., Parks, M.J. & Peterson, S.J. (2023) High- intensity drinking
and hours spent drinking. Alcohol, Clinical & Experimental Research,
47, 2 0 81–2 089.
Pearson, M.R. (2013) Use of alcohol protective behavioral strategies
among college students: a critical review. Clinical Psychology Review,
33, 1025–1040.
Prince, M.A., Carey, K.B. & Maisto, S.A. (2013) Protective behavioral
strategies for reducing alcohol involvement: a review of the meth-
odological issues. Addictive Behaviors, 38, 2343–2351.
Ray, A.E., Stapleton, J.L., Turrisi, R. & Mun, E.- Y. (2014) Drinking game
play among first- year college student drinkers: an event- specific
analysis of the risk for alcohol use and problems. The American
Journal of Drug and Alcohol Abuse, 40, 353–358.
Richards, V.L., Barnett, N.P., Cook, R.L ., Leeman, R.F., Souza, T., Case, S.
et al. (2022) Correspondence between alcohol use measured by a
wrist- worn alcohol biosensor and self- report via ecological momen-
tary assessment (EMA) over a two- week period. Alcoholism: Clinical
and Experimental Research, 47, 308–318.
Richards, V.L ., Glenn, S.D., Turrisi, R.J., Altstaedter, A., Mallett, K. A. &
Russell, M.A . (2023) Does it really matter that I do not remember
my night? Consequences related to blacking out among college
student drinkers. Alcoholism: Clinical and Experimental Research, 4 7,
1798–1805.
Richards, V.L ., Turrisi, R.J., Glenn, S.D., Waldron, K.A., Rodriguez, G.C.,
Mallett, K.A. et al. (2023) Alcohol- induced blackouts among col-
lege student drinkers: a multilevel analysis. Addictive Behaviors, 143 ,
107706.
Rose, M.E. & Grant, J.E. (2010) Alcohol- induced blackout: phenomenol-
ogy, biological basis, and gender differences. Journal of Addiction
Medicine, 4, 61–73.
Rosenberg, M., Kianersi, S., Luetke, M., Jozkowski, K., Guerra- Reyes, L .,
Shih, P.C., et al. (2023) Wearable alcohol monitors for alcohol use
data collection among college students: Feasibility and acceptabil-
it y. Alcohol (Fayetteville, N.Y.), 111, 75–83. https://doi.org/10.1016/j.
alcohol.2023.05.007
Russell, M.A., Turrisi, R .J. & Smy th, J.M. (2022) Transdermal sensor fea-
tures correlate with ecological momentary assessment drinking
reports and predict alcohol- related consequences in young adults'
natural settings. Alcoholism: Clinical and Experimental Research, 46,
10 0 –11 3 .
Simons, J.S., Wills, T.A ., Emery, N.N. & Marks, R.M. (2015) Quantifying
alcohol consumption: self- report, transdermal assessment, and pre-
diction of dependence symptoms. Addictive Behaviors, 50, 205–212.
Stevely, A.K., Holmes, J., McNamara, S. & Meier, P.S. (2020) Drinking
context s and their association with acute alcohol- related harm: a
systematic review of event- level studies on adults' drinking occa-
sions. Drug and Alcohol Review, 39, 309–320.
Studer, J., Gmel, G., Bertholet, N., Marmet, S. & Daeppen, J. (2019)
Alcohol- induced blackouts at age 20 predict the incidence, mainte-
nance and severity of alcohol dependence at age 25: a prospective
study in a sample of young Swiss men. Addiction, 114, 1556–1566.
Villalba, K., Cook, C., Dévieux, J.G., Ibanez, G.E., Oghogho, E., Neira,
C. et al. (2020) Facilitators and barriers to a contingency manage-
ment alcohol intervention involving a transdermal alcohol sensor.
Heliyon, 6, e03612.
Voloshyna, D.M., Bonar, E.E., Cunningham, R.M., Ilgen, M.A ., Blow, F.C. &
Walton , M. A. (2018) Blackout s among mal e an d female yo uth seek-
ing emergency department care. The American Journal of Drug and
Alcohol Abuse, 44, 129–139.
Vonghia, L., Leggio, L., Ferrulli, A., Bertini, M., Gasbarrini, G. & Addolorato,
G. (2008) Acute alcohol intoxication. European Journal of Internal
Medicine, 19, 561–567.
Wang, Y., Fridberg, D.J., Leeman, R.F., Cook, R.L. & Porges, E.C. (2018)
Wrist- worn alcohol biosensors: strengths, limitations, and future
directions. Alcohol, 81, 83–92.
Wang, Y., Fridberg, D.J., Shortell, D.D., Leeman, R.F., Barnett, N.P., Cook,
R.L. et al. (2021) Wrist- worn alcohol biosensors: applications and
usability in behavioral research. Alcohol, 92, 25–34.
Wetherill, R.R. & Fromme, K. (2016) Alcohol- induced blackouts: a review
of recent clinic al research with practical implications and recom-
mendations for future studies. Alcoholism, Clinical and Experimental
Research, 40, 922–935.
White, A. (2003) What happened? Alcohol, memor y blackout s, and the
brain. Alcohol Research & Health, 27, 186–196.
Yuen, W.S., Chan, G., Bruno, R ., Clare, P.J., Aiken, A ., Mattick, R. et al.
(2021) Trajectories of alcohol- induced blackouts in adolescence:
early risk factors and alcohol use disorder outcomes in early adult-
hood. Addiction, 116, 2039–2048.
SUPPORTING INFORMATION
Additional supporting information can be found online in the
Suppor ting Information section at the end of this article.
How to cite this article: Richards, V.L., Glenn, S.D., Turrisi,
R.J., Mallett, K. A., Ackerman, S. & Russell, M.A . (2024)
Transdermal alcohol concentration features predict
alcohol- induced blackouts in college students. Alcohol:
Clinical and Experimental Research, 48, 880–888. Available
from: ht tps://doi.org/10 .1111/a cer.1529 0