Special issue on alcohol biosensors: Development, use, and state of the field: Summary, conclusions, and future directions

Wearable alcohol monitors for alcohol use data collection among college students: Feasibility and acceptability

OBJECTIVE

We assessed the feasibility and acceptability of using BACtrack Skyn wearable alcohol monitors for alcohol research in a college student population.

METHODS

We enrolled n = 5 (Sample 1) and n = 84 (Sample 2) Indiana University undergraduate students to wear BACtrack Skyn devices continuously over a 5-day to 7-day study period. We assessed feasibility in both samples by calculating compliance with study procedures, and by analyzing amount and distributions of device output [e.g., transdermal alcohol content (TAC), temperature, motion]. In Sample 1, we assessed feasibility and acceptability with the Feasibility of Intervention Measure (FIM) scale and the Acceptability of Intervention Measure (AIM) scale.

RESULTS

All participants were able to successfully use the alcohol monitors, producing a total of 11,504 h of TAC data. TAC data were produced on 567 days of the 602 total possible days of data collection. The distribution of the TAC data showed between-person variation, as would be expected with between-person differences in drinking patterns. Temperature and motion data were also produced as expected. Sample 1 participants (n = 5) reported high feasibility and acceptability of the wearable alcohol monitors in survey responses with a mean FIM score of 4.3 (of 5.0 possible score) and mean AIM score of 4.3 (of 5.0 possible score).

CONCLUSIONS

The high feasibility and acceptability we observed underscore the promise of using BACtrack Skyn wearable alcohol monitors to improve our understanding of alcohol consumption among college students, a population at particularly high risk for alcohol-related harms.

Acceptability and validity of using the BACtrack skyn wrist-worn transdermal alcohol concentration sensor to capture alcohol use across 28 days under naturalistic conditions - A pilot study

Wrist-worn transdermal alcohol concentration (TAC) sensors have the potential to provide detailed information about day-level features of alcohol use but have rarely been used in field-based research or in early adulthood (i.e., 26-40 years) alcohol users. This pilot study assessed the acceptability, user burden, and validity of using the BACtrack Skyn across 28 days in individuals' natural settings. Adults aged 26-37 (N = 11, Mage = 31.2, 55% female, 73% non-Hispanic white) participated in a study including retrospective surveys, a 28-day field protocol wearing Skyn and SCRAM sensors and completing ecological momentary assessments (EMA) of alcohol use and duration (daily morning reports and participant-initiated start/stop drinking EMAs), and follow-up interviews. Day-level features of alcohol use extracted from self-reports and/or sensors included drinks consumed, estimated Blood Alcohol Concentration (eBAC), drinking duration, peak TAC, area under the curve (AUC), rise rate, and fall rate. Repeated-measures correlations (rrm) tested within-person associations between day-level features of alcohol use from the Skyn versus self-report or the SCRAM. Participants preferred wearing the Skyn over the SCRAM [t (10) = -6.79, p < .001, d = 2.74]. Skyn data were available for 5614 (74.2%) out of 7566 h, with 20.7% of data lost due to syncing/charging issues and 5.1% lost due to device removal. Skyn agreement for detecting drinking days was 55.5% and 70.3% when compared to self-report and the SCRAM, respectively. Correlations for drinking intensity between self-report and the Skyn were 0.35 for peak TAC, 0.52 for AUC, and 0.30 for eBAC, which were smaller than correlations between self-report and SCRAM, at 0.78 for peak TAC, 0.79 for AUC, and 0.61 for eBAC. Correlations for drinking duration were larger when comparing self-report to the Skyn (rrm = 0.36) versus comparing self-report to the SCRAM (rrm = 0.31). The Skyn showed moderate-to-large, significant correlations with the SCRAM for peak TAC (rrm = 0.54), AUC (rrm = 0.80), and drinking duration (rrm = 0.63). Our findings support the acceptability and validity of using the Skyn for assessing alcohol use across an extended time frame (i.e., 28 days) in individuals' natural settings, and for providing useful information about day-level features of alcohol use.

Leveraging Mobile Phone Sensors, Machine Learning and Explainable Artificial Intelligence to Predict Imminent Same-Day Binge Drinking Events to Support Just-In-Time Adaptive Interventions: A Feasibility Study

BACKGROUND

Digital Just-In-Time Adaptive Interventions (JITAIs) can reduce binge drinking events (BDEs: consuming 4+/5+ drinks per occasion for women/men) in young adults, but need to be optimized for timing and content. Delivering just-in-time support messages in the hours prior to BDEs could improve intervention impact.

OBJECTIVE

We determined the feasibility of developing a machine learning model to accurately predict future, that is, same-day, 1 to 6-hours prior BDEs using smartphone sensor data. We aimed to identify the most informative phone sensor features associated with BDEs on weekend and weekdays, respectively, to determine the key features that explain prediction model performance.

METHODS

We collected phone sensor data from 75 young adults (ages 21-25; mean =22.4, SD=1.9) with risky drinking behavior who reported drinking behavior over 14 weeks. Participants in this secondary analysis were enrolled in a clinical trial. We developed machine learning models testing different algorithms (e.g., XGBoost, decision tree) to predict same-day BDEs (versus low-risk drinking events and non-drinking periods) using smartphone sensor data (e.g., accelerometer, GPS). We tested various "prediction distance" time windows (more proximal: 1-hour; to distant: 6-hour) from drinking onset. We also tested various analysis time windows (i.e., amount of data to be analyzed), ranging from 1 to 12 hours prior to drinking onset, because this determines the amount of data that needs to be stored on the phone to compute the model. Explainable AI (XAI) was used to explore interactions between the most informative phone sensor features contributing to BDEs.

RESULTS

The XGBoost model performed best in predicting imminent same-day BDE, with 95.0% accuracy on weekends and 94.3% accuracy on weekdays (F1 score = 0.95 and 0.94, respectively). This XGBoost model needed 12- and 9-hours of phone sensor data at 3- and 6- hours prediction distance from the onset of drinking, on weekends and weekdays, respectively, prior to predicting same-day BDEs. The most informative phone sensor features for BDE prediction were time (e.g., time of day) and GPS-derived, such as radius of gyration (an indicator of travel). Interactions among key features (e.g., time of day, GPS-derived features) contributed to prediction of same-day BDE.

CONCLUSIONS

We demonstrated the feasibility and potential use of smartphone sensor data and machine learning to accurately predict imminent (same-day) BDEs in young adults. The prediction model provides "windows of opportunity" and with the adoption of XAI, we identified "key contributing features" to trigger JITAI prior to the onset of BDEs, with the potential to reduce the likelihood of BDEs in young adults.

CLINICALTRIAL

Examining new-generation transdermal alcohol biosensor performance across laboratory and field contexts

BACKGROUND

Wrist-worn transdermal alcohol sensors have the potential to change how alcohol consumption is measured. However, hardware and data analytic challenges associated with transdermal sensor data have kept these devices from widespread use. Given recent technological and analytic advances, this study provides an updated account of the performance of a new-generation wrist-worn transdermal sensor in both laboratory and field settings.

METHODS

This work leverages machine learning models to convert transdermal alcohol concentration data into estimates of Breath Alcohol Concentration (BrAC) in a large-scale laboratory sample (N = 256, study 1) and a pilot field sample (N = 27, study 2). Specifically, in both studies, the accuracy of the translation is evaluated by comparing BAC estimates yielded by BACtrack Skyn to real-time breathalyzer measurements collected in the laboratory and in the field.

RESULTS

The newest version of the Skyn device demonstrates a substantially lower error rate than older hand-assembled prototypes (0% to 7% vs. 29% to 53%, respectively). On average, real-time estimates of BrAC yielded by these transdermal sensors are within 0.007 of true BAC readings in the laboratory context and within 0.019 of true BrAC readings in the field. In both contexts, the distance between true and estimated BrAC was larger when only alcohol episodes were examined (laboratory = 0.017; field = 0.041). Finally, results of power-law-curve projections indicate that, given their accuracy, transdermal BrAC estimates in real-world contexts have the potential to improve markedly (>25%) with adequately sized datasets for model training.

CONCLUSION

Findings from this study indicate that the latest version of the transdermal wrist sensor holds promise for the accurate assessment of alcohol consumption in field contexts. A great deal of additional work is needed to provide a full picture of the utility of these devices, including research with large participant samples in field contexts.

Validating transdermal alcohol biosensors: a meta-analysis of associations between blood/breath-based measures and transdermal alcohol sensor output

BACKGROUND AND AIMS

Transdermal alcohol sensors carry immense promise for the continuous assessment of drinking but are inconsistent in detecting more fine-grained indicators of alcohol consumption. Prior studies examining associations between transdermal alcohol concentration (TAC) and blood/breath alcohol concentration (BAC) have yielded highly variable correlations and lag times. The current review aimed to synthesize transdermal validation studies, aggregating results from more than three decades of research to characterize the validity of transdermal sensors for assessing alcohol consumption.

METHODS

Databases were searched for studies listed prior to 1 March 2022 that examined associations between transdermal alcohol sensor output and blood and breath-based alcohol measures, resulting in 31 primarily laboratory-derived participant samples (27 precise effect sizes) including both healthy and clinical populations. Correlation coefficients and lag times were pooled using three-level random-effects meta-regression. Independent raters coded study characteristics, including the body position of transdermal sensors (ankle- versus arm/hand/wrist-worn device) and methodological bias (e.g. missing data).

RESULTS

Analyses revealed that, in this primarily laboratory-derived sample of studies, the average correlation between TAC and BAC was large in magnitude [r = 0.87, 95% confidence interval (CI) = 0.80, 0.93], and TAC lagged behind BAC by an average of 95.90 minutes (95% CI = 55.50, 136.29). Device body position significantly moderated both TAC-BAC correlation (b = 0.11, P = 0.009) and lag time (b = -69.41, P < 0.001). Lag times for ankle-worn devices were approximately double those for arm/hand/wrist-worn devices, and TAC-BAC correlations also tended to be stronger for arm/hand/wrist-worn sensors.

CONCLUSIONS

This meta-analysis indicates that transdermal alcohol sensors perform strongly in assessing blood/breath alcohol concentration under controlled conditions, with particular promise for the newer generation of wrist-worn devices.

An Abstract Parabolic System-Based Physics-Informed Long Short-Term Memory Network for Estimating Breath Alcohol Concentration from Transdermal Alcohol Biosensor Data

The problem of estimating breath alcohol concentration based on transdermal alcohol biosensor data is considered. Transdermal alcohol concentration provides a promising alternative to classical methods such as breathalyzers or drinking diaries. A physics-informed long Short-term memory (LSTM) network with covariates for the solution of the estimation problem is developed. The data-driven nature of an LSTM is augmented with a first principles physics-based population model for the diffusion of ethanol through the epidermal layer of the skin. The population model in an abstract parabolic framework appears as part of a regularization term in the loss function of the LSTM. While learning, the model is encouraged to both fit the data and to produce physically meaningful outputs. To deal with the high variation observed in the data, a mechanism for the uncertainty quantification of the estimates based on a recently discovered relation between Monte-Carlo dropout and Bayesian learning is used. The physics-based population model and the LSTM are trained and tested using controlled laboratory collected breath and transdermal alcohol data collected in four sessions from 40 orally dosed participants (50% female, ages 21 - 33 years, 35% BMI above 25.0) resulting in 256 usable drinking episodes partitioned into training and testing sets. Body measurement (e.g. BMI, hip to waist ratio, etc.), personal (e.g. sex, age, race, etc.), drinking behavior (e.g. frequent, rarely, etc.), and environmental (e.g. temperature, humidity, etc.) covariates were also collected from participants. The importance of various covariates in the estimation is investigated using Shapley values. It is shown that the physics-informed LSTM network can be successfully applied to drinking episodes from both the training and test set, and that the physics-based information leads to better generalization ability on new drinking episodes with the uncertainty quantification yielding credible bands that effectively capture the true signal. Compared to two machine learning models from previous studies, the proposed model reduces relative L 2 error in estimated breath alcohol concentration by 58% and 72%, and relative peak error by 33% and 76%.

Blood and Breath Alcohol Concentration from Transdermal Alcohol Biosensor Data: Estimation and Uncertainty Quantification via Forward and Inverse Filtering for a Covariate-Dependent, Physics-Informed, Hidden Markov Model

Transdermal alcohol biosensors that do not require active participation of the subject and yield near continuous measurements have the potential to significantly enhance the data collection abilities of alcohol researchers and clinicians who currently rely exclusively on breathalyzers and drinking diaries. Making these devices accessible and practical requires that transdermal alcohol concentration (TAC) be accurately and consistently transformable into the well-accepted measures of intoxication, blood/breath alcohol concentration (BAC/BrAC). A novel approach to estimating BrAC from TAC based on covariate-dependent physics-informed hidden Markov models with two emissions is developed. The hidden Markov chain serves as a forward full-body alcohol model with BrAC and TAC, the two emissions, assumed to be described by a bivariate normal which depends on the hidden Markovian states and person-level and session-level covariates via built-in regression models. An innovative extension of hidden Markov modeling is developed wherein the hidden Markov model framework is regularized by a first-principles PDE model to yield a hybrid that combines prior knowledge of the physics of transdermal ethanol transport with data-based learning. Training, or inverse filtering, is effected via the Baum-Welch algorithm and 256 sets of BrAC and TAC signals and covariate measurements collected in the laboratory. Forward filtering of TAC to obtain estimated BrAC is achieved via a new physics-informed regularized Viterbi algorithm which determines the most likely path through the hidden Markov chain using TAC alone. The Markovian states are decoded and used to yield estimates of BrAC and to quantify the uncertainty in the estimates. Numerical studies are presented and discussed. Overall good agreement between BrAC data and estimates was observed with a median relative peak error of 22% and a median relative area under the curve error of 25% on the test set. We also demonstrate that the physics-informed Viterbi algorithm eliminates non-physical artifacts in the BrAC estimates.

Characterising patterns of alcohol use among heavy drinkers: A cluster analysis utilising alcohol biosensor data

INTRODUCTION

Previous research has predominately relied on person-level or single characteristics of drinking episodes to characterise patterns of drinking that may confer risk. This research often relies on self-report measures. Advancements in wearable alcohol biosensors provide a multi-faceted objective measure of drinking. The current study aimed to characterise drinking episodes using data derived from a wearable alcohol biosensor.

METHODS

Participants (n = 45) were adult heavy drinkers who wore the Secure Continuous Remote Alcohol Monitoring (SCRAM) bracelet and reported on their drinking behaviours. Cluster analysis was used to evaluate unique combinations of alcohol episode characteristics. Associations between clusters and self-reported person and event-level factors were also examined in univariable and multivariable models.

RESULTS

Results suggested three unique clusters: Cluster 1 (most common, slowest rate of rise to and decline from peak), Cluster 2 (highest peak transdermal alcohol concentration and area under the curve) and Cluster 3 (fastest rate of decline from peak). Univariable analyses distinguished Cluster 1 as having fewer self-reported drinks and fewer episodes that occurred on weekends relative to Cluster 2. The effect for number of drinks remained in multivariable analyses.

DISCUSSION AND CONCLUSIONS

This is the first study to characterise drinking patterns at the event-level using objective data. Results suggest that it is possible to distinguish drinking episodes based on several characteristics derived from wearable alcohol biosensors. This examination lays the groundwork for future studies to characterise patterns of drinking and their association with consequences of drinking behaviour.

A new generation of transdermal alcohol biosensing technology: practical applications, machine -learning analytics and questions for future research

The use of transdermal alcohol monitors has burgeoned in recent years, now encompassing hundreds of thousands of individuals globally. A new generation of sensors promises to expand the range of applications for transdermal technology exponentially, and advances in machine-learning modeling approaches offer new methods for translating the data produced by transdermal devices. This article provides (1) a review of transdermal sensor research conducted to date, including an analysis of methodological features of past studies potentially key in driving reported sensor performance; (2) updates on methodological developments likely to be transformative for the field of transdermal sensing, including the development of new-generation sensors featuring smartphone integration and rapid sampling capabilities as well as developments in machine-learning analytics suited to data produced by these novel sensors and; (3) an analysis of the expanded range of applications for this new generation of sensor, together with corresponding requirements for sensor accuracy and temporal specificity. We also note questions as yet unanswered and key directions for future research.

Accuracy of Consumer-marketed smartphone-paired alcohol breath testing devices: A laboratory validation study

BACKGROUND

Although alcohol breath testing devices that pair with smartphones are promoted for the prevention of alcohol-impaired driving, their accuracy has not been established.

METHODS

In a within-subjects laboratory study, we administered weight-based doses of ethanol to two groups of 10 healthy, moderate drinkers aiming to achieve a target peak blood alcohol concentration (BAC) of 0.10%. We obtained a peak phlebotomy BAC and measured breath alcohol concentration (BrAC) with a police-grade device (Intoxilyzer 240) and two randomly ordered series of 3 consumer smartphone-paired devices (6 total devices) with measurements every 20 min until the BrAC reached <0.02% on the police device. Ten participants tested the first 3 devices, and the other 10 participants tested the other 3 devices. We measured mean paired differences in BrAC with 95% confidence intervals between the police-grade device and consumer devices.

RESULTS

The enrolled sample (N = 20) included 11 females; 15 white, 3 Asian, and 2 Black participants; with a mean age of 27 and mean BMI of 24.6. Peak BACs ranged from 0.06-0.14%. All 7 devices underestimated BAC by >0.01%, though the BACtrack Mobile Pro and police-grade device were consistently more accurate than the Drinkmate and Evoc. Compared with the police-grade device measurements, the BACtrack Mobile Pro readings were consistently higher, the BACtrack Vio and Alcohoot measurements similar, and the Floome, Drinkmake, and Evoc consistently lower. The BACtrack Mobile Pro and Alcohoot were most sensitive in detecting BAC driving limit thresholds, while the Drinkmate and Evoc devices failed to detect BAC limit thresholds more than 50% of the time relative to the police-grade device.

CONCLUSIONS

The accuracy of smartphone-paired devices varied widely in this laboratory study of healthy participants. Although some devices are suitable for clinical and research purposes, others underestimated BAC, creating the potential to mislead intoxicated users into thinking that they are fit to drive.

Effects of stomach content on the breath alcohol concentration-transdermal alcohol concentration relationship

INTRODUCTION

Wearable devices that obtain transdermal alcohol concentration (TAC) could become valuable research tools for monitoring alcohol consumption levels in naturalistic environments if the TAC they produce could be converted into quantitatively-meaningful estimates of breath alcohol concentration (eBrAC). Our team has developed mathematical models to produce eBrAC from TAC, but it is not yet clear how a variety of factors affect the accuracy of the models. Stomach content is one factor that is known to affect breath alcohol concentration (BrAC), but its effect on the BrAC-TAC relationship has not yet been studied.

METHODS

We examine the BrAC-TAC relationship by having two investigators participate in four laboratory drinking sessions with varied stomach content conditions: (i) no meal, (ii) half and (iii) full meal before drinking, and (iv) full meal after drinking. BrAC and TAC were obtained every 10 min over the BrAC curve.

RESULTS

Eating before drinking lowered BrAC and TAC levels, with greater variability in TAC across person-device pairings, but the BrAC-TAC relationship was not consistently altered by stomach content. The mathematical model calibration parameters, fit indices, and eBrAC curves and summary score outputs did not consistently vary based on stomach content, indicating that our models were able to produce eBrAC from TAC with similar accuracy despite variations in the shape and magnitude of the BrAC curves under different conditions.

DISCUSSION AND CONCLUSIONS

This study represents the first examination of how stomach content affects our ability to model estimates of BrAC from TAC and indicates it is not a major factor.

Using machine learning for real-time BAC estimation from a new-generation transdermal biosensor in the laboratory

BACKGROUND

Transdermal biosensors offer a noninvasive, low-cost technology for the assessment of alcohol consumption with broad potential applications in addiction science. Older-generation transdermal devices feature bulky designs and sparse sampling intervals, limiting potential applications for transdermal technology. Recently a new-generation of transdermal device has become available, featuring smartphone connectivity, compact designs, and rapid sampling. Here we present initial laboratory research examining the validity of a new-generation transdermal sensor prototype.

METHODS

Participants were young drinkers administered alcohol (target BAC = .08 %) or no-alcohol in the laboratory. Participants wore transdermal sensors while providing repeated breathalyzer (BrAC) readings. We assessed the association between BrAC (measured BrAC for a specific time point) and eBrAC (BrAC estimated based only on transdermal readings collected in the immediately preceding time interval). Extra-Trees machine learning algorithms, incorporating transdermal time series features as predictors, were used to create eBrAC.

RESULTS

Failure rates for the new-generation prototype sensor were high (16 %-34 %). Among participants with useable new-generation sensor data, models demonstrated strong capabilities for separating drinking from non-drinking episodes, and significant (moderate) ability to differentiate BrAC levels within intoxicated participants. Differences between eBrAC and BrAC were 60 % higher for models based on data from old-generation vs new-generation devices. Model comparisons indicated that both time series analysis and machine learning contributed significantly to final model accuracy.

CONCLUSIONS

Results provide favorable preliminary evidence for the accuracy of real-time BAC estimates from a new-generation sensor. Future research featuring variable alcohol doses and real-world contexts will be required to further validate these devices.