Clinical applications of wearable devices for individuals with at‐risk substance use and substance use disorders: A systematic review

Abstract

Background and Objectives

Managing substance use disorders (SUDs) is limited by the lack of longitudinal clinical data during active use and recovery. Wearable devices may help address this gap by capturing physiological and behavioral data in real‐world settings. This review evaluated the feasibility and acceptability of wearable devices among individuals with SUDs.

Methods

A systematic search of PubMed, PsycINFO, and MEDLINE identified peer‐reviewed studies published between 2000 and 2023 examining wearable devices that passively measured physiological data among adults with SUDs or at‐risk substance use.

Results

Thirty‐three studies met inclusion criteria and were grouped into feasibility and acceptability outcomes, sleep‐related outcomes, and other clinical outcomes including craving, stress, and detection of substance use or overdose events. Most studies used wrist‐worn devices and focused on opioid and alcohol use. The literature was heterogeneous and predominantly observational, with few experimental designs, and several studies included individuals with at‐risk use rather than diagnosed SUDs. Short‐term wearable use was generally feasible and acceptable when privacy and safety concerns were addressed. Wearables captured indicators such as sleep disruption, stress, and overdose risk. However, none of the included studies evaluated whether wearable use was associated with improved substance use outcomes over time.

Discussion and Conclusions

Wearable devices show promise as adjunct tools for monitoring clinically relevant indicators in SUDs, but evidence remains limited. Future research should prioritize longitudinal outcomes, validate clinically meaningful thresholds, and support equitable, patient‐centered implementation.

Scientific Significance

This review provides a structured synthesis of an emerging and fragmented literature, offering a foundation to guide future hypothesis‐driven, longitudinal, and implementation‐focused research.

INTRODUCTION

Substance use disorders (SUDs) are major contributors to global morbidity and mortality, accounting for a substantial proportion of preventable disease burden worldwide. ¹ Alcohol and drug use together contribute millions of disability‐adjusted life years (DALYs) and deaths annually, highlighting the longstanding global impact of SUDs. ¹ Even with evidence‐based treatment, relapse rates are high, ranging from 40% to 60% one year after discharge from pharmacological and behavioral interventions. ²

Contemporary models conceptualize relapse as a temporally evolving phenomenon in which risk of relapse fluctuates over time as motivation, craving, and self‐regulatory capacity change. ³ Longitudinal studies using ecological momentary assessments (EMAs) reveal that moment‐to‐moment variations in affect, stress, and craving are critical determinants of substance use and relapse and are not adequately captured by cross‐sectional or retrospective assessments. ³

Wearable devices offer a means of capturing objective movement‐based and physiological processes over several days or longer, which may improve our understanding of how these variables mediate substance use. ⁴ , ⁵ , ⁶ In the context of this review, movement‐based wearable data included actigraphy/accelerometry signals used to estimate sleep and rest‐activity patterns, identify substance‐use‐related movements or gestures, or detect movement changes associated with substance use or overdose risk. Wearable recording targets included movement, heart rate, respiratory measures, oxygen saturation, electrodermal activity, skin temperature, and sleep estimates. ⁴ , ⁵ , ⁶ , ⁷ Wearable devices often measure proxy signals, sometimes alongside EMAs, to infer clinical outcomes. For example, wrist‐worn actigraphs directly measure movement to estimate sleep outcomes such as sleep efficiency, total sleep time, and sleep onset latency. ³ , ⁵ , ⁶ , ⁷ , ⁸ Sleep disruptions are important to measure and are common during substance use and withdrawal and may also predispose individuals to relapse. ⁷ These devices have been increasingly explored in SUD management and provide opportunities for personalized treatment based on both objective data and clinical visits. ⁹ Remote monitoring may improve patient confidence by empowering patients to self‐manage their disorder and identify clinical symptoms in real time, leading to timely and accessible care. ¹⁰ Moreover, the passive nature of data collection enabled by wearables may balance some EMA burden for patients and complement their data, which may promote future adoption in SUD management. ¹¹

Prior reviews exploring wearable and remote monitoring technologies in substance use disorder research have been limited, have not been inclusive of all substance use types, have focused on biosensors for drug detection and do not comprehensively synthesize evidence on feasibility and acceptability, which is the specific focus of our paper. ⁴ , ¹² The objectives of the exploratory review are to synthesize the evidence for (1) the feasibility and acceptability of using wearable devices to detect clinical outcomes related to SUDs among individuals at risk of developing or currently diagnosed with SUDs, and (2) the ability of wearable devices to detect clinical outcomes related to SUDs, including measures of sleep, substance use, abstinence, craving, and overdose.

METHODS

This systematic review was registered on Prospero CRD42023477839 and conducted according to the Preferred Reporting of Systematic Reviews and Meta Analyses (PRISMA) Guidelines. ¹³

Inclusion criteria

English language studies published in peer‐reviewed journals were included. We restricted the years of publication from January 1, 2000, to the search date on July 24, 2023. We decided to exclude studies published before 2000, as they are likely to use outdated technology. All original research study types were included. Inclusion criteria: (1) adult participants (≥18 years old); (2) participants with SUDs or pre‐determined criteria for defining problematic or at‐risk substance use; (3) use of a wearable device or wireless technology, defined as an accessory or clothing item that passively measures physiologic data or geolocation, is portable, wireless, is not only a mobile device such as a smartphone or tablet, does not require medical assistance to use, and is intended for use over multiple days or longer; (4) inclusion of predefined outcomes of interest, including: clinical substance use outcomes (e.g., quantitative measures of substance use, abstinence, overdose, craving/desire to use), qualitative or quantitative measures of feasibility (e.g., facilitators or barriers to engagement by either providers or participants). Exclusion criteria included: (1) non‐peer‐reviewed studies, preprint articles, or conference abstracts, and (2) involved devices that only act as biosensors to detect the presence of the drug or substance, in the absence of any other physiologic or geolocation data. We excluded direct biochemical/analyte‐detection biosensors because they were outside this review's focus on physiological, behavioral, and geolocation‐based wearable monitoring and have been reviewed recently elsewhere. ¹⁴ , ¹⁵ , ¹⁶ Standard polysomnography was outside the scope of this review because it is primarily a diagnostic sleep‐recording method that typically requires specialized equipment and technical support, and is not generally intended for passive multi‐day wearable monitoring in naturalistic settings.

Information sources and search strategy

Three databases were systematically searched for relevant studies: MEDLINE, PubMed and PsycINFO. The search strategy used database‐specific subject headings, natural‐language keywords, and advanced search operators to encompass two main concepts: wearable devices and substance use disorders. A comprehensive search strategy is available in the Supporting Information S1. All studies were imported into Covidence for title and abstract screening, full‐text review, and data extraction. Abstract and full‐text screening were independently conducted by two authors, with disagreements resolved by consensus or consultation with a third author weekly. In addition to database searching, we screened the reference lists of relevant reviews and included eligible studies that were not captured in the original search. All additionally identified studies were assessed using the same inclusion and exclusion criteria as database‐identified records.

Risk of bias tool

A National Institutes of Health (NIH) Quality Assessment Tool for Observational Cohort and Cross‐Sectional Studies, which includes 14 methodological criteria, was adapted into a shorter scoring system tailored to the needs of this review. ¹⁷ Key items assessing clarity of research objectives, population definition, recruitment consistency, sample size justification, and validity of exposure measurements were preserved and translated into simple binary scoring (1 = yes, 0 = no) with added guidance to ensure consistent interpretation. Certain general NIH domains were adapted to better align with wearable‐device studies in substance use research. A certainty assessment for each study was not conducted due to the heterogeneous study designs and outcomes of interest.

RESULTS

Study characteristics

Following the removal of duplicates, 2743 titles and abstracts were screened; 91 were selected for full‐text review. A total of 33 studies met the inclusion criteria for extraction, seven of which were identified in the reference lists of relevant reviews. See Figure 1 for the PRISMA flowchart of included studies and reasons for excluded studies.

Figure 1.

PRISMA flowchart.

A wide variety of wearable devices have been reported in the literature. Of the 33 studies, 25 used wrist‐worn devices, four used chest‐worn bands, and one used a clip‐on wearable device. We included three studies that investigated perspectives about hypothetical devices that met the inclusion criteria of our study but were not actually used by the participants; two studies included wearable phases with actual wearable devices that can detect physiologic metrics associated with overdoses but do not have the ability to alert others about an overdose event. The most studied substances were opioids (nine studies), followed by alcohol and stimulants (five studies each), cannabis and multiple‐substance use (four studies each), and tobacco and at‐risk substance use (three studies each). Most studies were observational (23 studies), followed by randomized controlled trials (nine studies) and one case study. The most studied objectives for wearable devices included acceptability studies, sleep studies, and physiological monitoring associated with substance use, overdose, and craving in sample sizes that ranged from one participant to up to 119 participants. Details of the included studies are summarized in Tables 1, 2, 3. In these, we delineate between the target physiological or behavioral process that the wearable device aims to measure (i.e., sleep quality) and the technological methodology by which it aims to do so (i.e., actigraphy to measure movement during sleep).

Table 1.

Feasibility, acceptability and compliance findings for wearable devices.

Population	Reference	Device	Study objective	Study design and setting	Sample size	Comorbid psychiatric and medical diagnoses	Other measures of substance use and clinical symptoms	Therapeutic intervention	What the device measures	What the device aims to measure	Duration of wear for device	Acceptability (%), Perspectives, Factors Associated with Acceptability, Compliance Rates
Opioid use disorder.	[18]	Withings Scan Watch (wrist worn device).	Investigate perspectives about feasibility and acceptability of a wearable device to detect and reverse a potential overdose.	Observational study (Setting: Homeless shelter or prison)	Participant Focus groups: 16 (7 male: 9 female); Group that wore device: 6 (4 male, 2 female); Street Injectors Support Service workers: 4	Not listed	Not Applicable	Not Applicable	Heart rate and SPo2.	Measures physiologic measurements associated with an overdose.	1.5–2 h in evenings	Perspectives that facilitate engagement: the wearable device should have a proper wristband size for opioid users, no resale value, discreet appearance, and automatic monitoring of heart rate and blood oxygen levels. It should include educational initiatives, overdose alarms with GPS activation, and a cancel button for recovered users.
People who have used illegal drugs at least once in the past month. (Focus on opioids).	[19]	Hypothetical device would be the size of a phone and would be strapped to the chest or wrist.	To evaluate willingness to wear a hypothetical device that would alert others in the event of an overdose.	Observational study (Hypothetical Naturalistic setting)	N = 1061 participants (male 63.1%, female 36.9%)	HIV (304%), chronic pain (52.4%)	Not Applicable	Not Applicable	Not Reported ‐ Biometric data that would indicate an overdose.	Measures physiologic measurements associated with an overdose and alert bystanders and/or emergency medical services.	Not Applicable	Acceptability: 54.3% were willing to wear the device. Individuals who had overdosed in the past, were currently being treated with methadone, were female, or had a history of chronic pain were more likely to be willing to wear the device. Conversely, being homeless was not associated with willingness to wear the device.
At risk for opioid use disorder: Opioid use history of ≥3 months or prescribed opioids for surgery or chronic pain.	[20]	Hypothetical device is strapped to the thigh or shoulder, measures overdose, and injects naloxone.	To understand the preferences for wearing a device capable of sensing and reversing opioid overdose	Observational study (Hypothetical Naturalistic setting)	Survey N = 97 (59% male, 41% female) Interview: N = 95	Not Reported	Not Applicable	Not Applicable	Not Reported ‐ Biometric data that would indicate an overdose.	Measures physiologic measurements associated with an overdose and subsequently administers overdose reversal medication.	Not Applicable	Acceptability: 76% were willing to use the device. A wrist bracelet and a device that monitors vital signs without alerting others or administering naloxone were the most popular choices, closely followed by devices that alert medical first responders and/or indicate that the wearer is at risk of overdose and/or administers naloxone. Reliance on alerting bystanders to administer naloxone was the least popular design choice. Factors that increased willingness to use the device were discreetness and comfort. Respondents identified that homeless people and those with low tolerance for opioids would be in greatest need of the device.
Opioid use disorder receiving treatment	[5]	Garmin vívosmart 4	To investigate the combined use of wearable technologies and EMA as a potential mechanism for identifying physiological and behavioral biomarkers of drug use recurrence.	Observational study (Outpatient Setting)	77 (31 male, 46 female)	None listed	Urine toxicology, electronic momentary assessments about stress, fatigue, mood, pain, substance craving.	A structured group‐based, program in the Comprehensive Opioid Addiction Treatment Program (COAT)	Monitors HR and HRV via photoplethysmography, blood oxygen saturation, sleep, activity, and the Garmin Stress score (the root mean square of successive HRV)	Physiological measures of substance use recurrence.	Daily wear for at least 14 days.	Compliance with the EMA app and the wearable device was lowest among participants during the week of a drug use recurrence and was similarly significantly low during the week before a drug use recurrence. Additionally, 14 out of the 96 patients who consented to the trial were excluded from analyses because they had less than 14 days of data, and five were excluded due to insufficient information to determine drug use recurrence status.
Prescription opioid use disorder.	[21]	Actigraphy (wrist worn device) Philips Respironics®	To examine subjective and objective sleep disturbances among prescription opioid (PO) dependent individuals.	Observational study (Inpatient Hospital Setting)	68 non‐treatment seeking individuals (33 prescription opioid dependent, 35 healthy controls).	Pain condition	Cigarette smokers were provided with a nicotine.	None	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	one night	~50% of the eligible participants withdrew from the study but it is not noted at which point in the study they withdrew and why.
Opioid use disorder.	[22]	Actigraphy (wrist worn device) Philips Respironics	To measure sleep duration and timing in opioid dependent participants treated with methadone or buprenorphine.	Observational study (Setting: Outpatient treatment/research clinic)	48 enrolled; 37 completed (28 men, 8 women)		Urinalysis, EMA self‐report drug use, daily sleep diaries, Pittsburgh Sleep Quality Index (PSQI)	Methadone (100 mg/day) or Buprenorphine (16–24 gm/day)	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	six sets of three nights	two individuals withdrew because of the device wearing requirements.
Individuals with opioid use disorder	[23]	Spire® Health Tag, attached to underwear waistband or wing of bra.	To investigate respiratory data related to overdose events.	Observational study (Naturalistic setting)	N = 15	Their average drug injection career was 19 years (SD = 7.5). The majority (87%) had experienced an opioid‐related overdose in their lifetime and 27% believed they were likely to overdose again within the next year.	Not Applicable	None	Tag collects respiratory rate (RR), physical movement, and heart rate.	Collect physiologic measures to indicate overdose.	5‐day period, removing for bathing.	Participants wore biosensors for an average of 21.7 h daily (SD = 3.1). Those with shorter injection histories and those who used fentanyl‐only collected more RR data (p = .04, p = .008).
Opioid use disorder.	[24]	Actigraphy (wrist worn device) Micro Motionlogger Sleep Watch	To assess the changes in sleep in methadone maintained individuals with opioid use disorder who are treated with mirtazapine, zolpidem, a combination of mirtazapine and zolpidem, and placebo.	Randomized, double blind crossover with repeated latin square design (Outpatient setting)	N = 10 (4 male, 6 female)	No other SCID diagnoses	Pittsburg Sleep Quality Index; daily sleep diaries	3 active medications (zolpidem CR 12.5 mg, mirtazapine 30 mg, and zolpidem 10 mg plus mirtazapine 30 mg) and placebo	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	8 weeks	No Report of compliance or feasibility related to the wearable device.
Alcohol use disorder and post‐traumatic stress disorder.	[25]	Actigraphy (wrist worn device) Respironics Actiwatch 2	To assess sleep in participants who receive treatment for alcohol use and for PTSD.	Randomized Controlled Trial (Naturalistic setting)	119 (90% male, 10% female)	PTSD and Insomnia	Urine samples, Clinician Administered PTSD Scale for DSM‐5 (CAPS‐5); Insomnia Severity Index (ISI)	Participants were randomized to either 1) COPE (integrated treatment using prolonged exposure) 2) Seeking Safety (integrated therapy using cognitive behavioral, interpersonal techniques and case management).	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	seven nights pretreatment, seven nights post treatment	49% wore the device throughout the study.
Alcohol use disorder.	[26]	Actigraphy (wrist worn device) Philips Respironics Actiwatch 2.0	To assess the effect of oxytocin versus placebo on actigraphy‐measured motor activity and sleep during detoxification and compare actigraphy‐recorded motor function between patients and healthy controls.	Randomized Controlled Trial (Setting: Inpatient alcohol dependence treatment program)	74 (40 in RCT (20 in treatment, 20 in control), and 34 controls)	None reported	Phosphatidylethanol blood concentration	Oxazepam (benzodiazepine) for alcohol withdrawal symptoms, 24 IU intranasal oxytocin twice daily	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	3 days	All patients completed 24‐h recording; 38/40 completed the full 3‐day recording.
Alcohol use disorder	Lahti et al., 2012	Actiwatch Plus®, Cambridge Neurotechnology Ltd, UK	To assess of the effect of nitrous oxide on sleep in inpatient individuals during alcohol withdrawal.	Randomized Controlled Study (Inpatient setting)	N = 105 voluntary inpatients (80 men and 25 women)	NA	Severity of Alcohol Dependence Data (SADD) questionnaire; Obsessive‐Compulsive Drinking Scale (OCDS); Clinical Institute Withdrawal Assessment of Alcohol Scale (CIWA‐Ar)	(1) nitrous oxide/O2 (O2; starting concentration 30%, maximum 70%, aiming at a 30% EtN2O concentration); (2) normal air/O2 (flow 50%/50%, final O2 concentration 70%); (3) normal air	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	1–3 days (depending on individual inpatient detox stay)	No Report
Individuals with cocaine use disorder.	[27]	Wristwatch biosensor Q sensor, Affectiva	To identify biophysiometric patterns associated with cocaine use in natural environments.	Observational study (Naturalistic setting)	N = 15 (53% male, 47% female)	None Reported	Urinalysis, time line follow back	None	EDA, skin temperature, and movement in the x, y, z directions (minute by minute)	Physiologic measures of the sympathetic nervouse system associated with substance use.	daytime for 30 days, removed at night	Study participants found mobile biosensors to be acceptable, and compliance with the protocol was high. About half of participants (7/15) indicated that the sensor made them less likely to use drugs as it made them more mindful of their actions.
Cocaine use disorder.	[28]	Actigraphy (wrist worn device) Phillips Respironics	To assess the effect of 8 weeks of citicoline treatment on sleep in individuals with cocaine use disorder.	Randomized double blind controlled trial (Naturalist setting)	N = 12 (7 male; five female)	Adults with cocaine use disorder, some of which have either cannabis use disorder, alcohol use disorder or both. No Axis I disorders (depression, bipolar, anxiety or eating disorders).	Trail Making B Test; Block Design Task; Digit Symbol Substitution Test; Wisconsin Card Sorting Task; Daily Sleep Diaries	Citicoline (1000 mg/day)	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	Worn for 10 weeks (except when showering or charging): 2 weeks baseline, 8 weeks treatment	There were 43 original participants and majority (17) dropped out before the medication phase began. Another eight dropped out during the medication phase, four had no actigraphy data, and two were removed because of non‐compliance with sleep diaries. It is unclear why these participants dropped out, and if this is reflective of the acceptability of the wearable device.
Cocaine use disorder.	[29]	Actigraphy (wrist worn device) Misfit Flash, Fossil Group	To assess the changes in sleep, stress, and cue‐reactivity in individuals treated with suvorexant, an orexin receptor antagonist, in individuals with cocaine use disorder.	Placebo controlled randomized trial (Naturalist setting)	N = 23 (placebo = 10; suvorexant = 10. Three dropped within first 3 days)	No other SCID diagnoses	Pittsburg Sleep Quality Index; daily sleep diaries	Suvorexant (10 mg/day during week 1, 20 mg/day during week 2) versus placebo	Movement in x, y, z directions (minute by minute)	Sleep duration, awake bands, restful sleep, and sleep/wake switching	2 weeks	Three participants withdrew and were lost to contact within the first 3 days of the study with no reports about the reason for dropout.
Cannabis use disorder.	[30]	Actigraphy (wrist worn device) Respironics Actiwatch	To assess the effect of bupropion on cannabis withdrawal symptoms in individuals with cannabis use disorder.	Randomized double blind controlled trial (Naturalist setting with daily lab visits)	N = 22 who were randomized; nine completers (five men; four women)	No other SUDs or SCID diagnoses	Marijuana Withdrawal Checklist; Beck Depression Inventory; Beck Anxiety Inventory; Urine test for THC	2 weeks of placebo or bupropion (days 1–3:150 mg/day; days 4–14: 300 mg/day)	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	14 days	The bupropion group had 50% completion of the study compared to the placebo group (33%). No reports of non‐compliance related to the wearable device.
Cannabis use disorder.	[31]	Actigraphy (wrist worn device) Actiwatch Activity Monitoring System	To assess the effectiveness of quetiapine for reducing cannabis withdrawal symptoms in individuals with cannabis use disorder.	Within subject, double blind, placebo controlled randomized trial (Part naturalist setting part inpatient laboratory setting)	N = 20 enrolled, 14 completers (12 male, two female)	No other SUDs or SCID diagnoses	7 Item sleep quality visual analog scale	200 mg quetiapine per day	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	30 days	No reports of participants dropping out of the study due to the wearable device.
Cannabis use disorder.	[32]	Actigraphy (wrist worn device) Actiwatch Activity Monitoring System	To assess the effectivness of lithium carbonate for reducing cannabis withdrawal symptoms in individuals with cannabis use disorder.	Double blind placebo controlled randomized trial (Inpatient Setting)	N = 38 (Placebo = 20; Lithium = 13)	No other SCID diagnoses	Cannabis withdrawal scale; Athens Insomnia Scale	Lithium carbonate (500 mg twice daily) versus placebo	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	8 days	Retention in treatment increased by 14.6% for every 1% improvement in sleep efficiency. No reports related to the compliance or feasibility of the wearable device.
One or more SUDs	[7]	E4 wrist worn biosensor	To determine the accuracy of a wearable sensor to detect and differentiate episodes of self‐reported craving and stress in individuals with substance use disorders, and to assess the feasibility of sensor based monitoring in this population.	Observational study (Outpatient setting)	30 (57% male, 43% female)	50% multiple primary substances of use. Alcohol use in 73% of individuals with a single substance use disorder, heroin use disorder 27%. Psychiatric disorder in 70%.	Self‐reported stress and craving in their wearable device.	Medication assisted treatment (33% of sample).	Electrodermal activity (EDA), skin temperature, three‐dimensional acceleration, and heart rate.	Physiologic measures of sympathetic nervous system arousal associated with stress and craving.	Daily wear for 4 days.	Majority reported willingness to wear more than one sensor, but they expressed preference for a less bulky deign that includes features like telling time. Over 60% s reported increased mindfulness while wearing the device, which was perceived as therapeutic. Participants also reported positive user experiences.

Abbreviations: HIV, human immunodeficiency virus; SPO2, peripheral oxygen saturation; PTSD, post traumatic stress disorder.

Table 2.

Devices used to measure sleep in individuals with substance use disorders.

Population	Reference	Manufacturer of device/brand name	Research question	Study design and setting	Sample size	Comorbid psychiatric and medical diagnoses	Other measures of substance use and clinical symptoms	Therapeutic intervention	What the device measures	What the device aims to measure	Duration of wear for device	Clinical outcomes
Tobacco use disorder.	[33]	Actigraphy (wrist worn device) Mini Motionlogger, Ambulatory Monitoring Inc., New York	To investigate differences in sleep between young adult smokers and nonsmokers.	Observational study (Naturalistic Setting)	Smokers N = 40 (12 male, 28 female) Non‐smokers N = 46 (14 male, 32 female)	Not Reported	piCO Breathalyzer, QSU‐brief (craving), FTND (nicotine dependence), STAI (anxiety), BDI‐II (depression), Pittsburg Sleep Quality Index (PSQI)	NA	Movement in x, y, z directions (minute by minute)	Sleep measurements that may be associated with cigarette smoking including Total sleep time (TST); Sleep onset latency (SOL); Sleep efficiency (SE); Wake time after sleep onset (WASO)	1 week	Among smokers and nonsmokers, there was a negative correlation between PSQI sleep duration scores and actigraphy‐measured total sleep time, significant among nonsmokers. Smokers showed lower overall sleep quality according to PSQI, with greater nicotine dependence linked to lower sleep efficiency as measured by actigraphy. Actigraphy indicated lower sleep continuity (reflected by lower sleep efficiency and heightened WASO) in smokers, but PSQI sleep measures did not differ between groups. Smokers’ self‐reported sleep onset times correlated with actigraphy measurements, unlike nonsmokers.
Alcohol use disorder.	[34]	Actigraphy (wrist worn device) Philips Actiwatch‐2 model	To investigate changes in sleep regularity index (SRI) during treatment for alcohol use disorder.	Observational study (Setting: Inpatient alcohol dependence treatment program)	N = 124 (66.1% male, 33.9% female)	Anxiety (50.8%); Mood Disorder (56.5%)	SCID IV SCID = Structured Clinical Diagnostic Interview	Inpatient treatment for alcohol withdrawal as well as psychosocial management.	Movement in x, y, z directions (minute by minute)	Sleep efficiency, wake after sleep onset (WASO), sleep duration, time in bed, sleep onset latency (SOL), daytime activity, and SRI	1 week	Individuals in the lowest SRI quartile at baseline improved significantly in their SRI through the course of treatment. Diagnosis of mood disorder was associated with lower SRI. There were no significant correlations found between SRI and anxiety disorders, depressive/anxiety symptoms, demographics (race, age, gender, marital status), daytime sleepiness, severity of withdrawal symptoms, pre‐admission drinking amounts and frequency, alcohol craving, self‐reported number of awakenings, and caffeinated drinks. The total duration of daytime naps between week 1 and week 3 was negatively correlated with the change in SRI. SRI improvements were associated with improvements in self reported mental and physical exhaustion.
Alcohol use disorder.	[26]	Actigraphy (wrist worn device) Philips Respironics Actiwatch 2.0	To assess the effect of oxytocin versus placebo on actigraphy‐measured motor activity and sleep during detoxification and compare actigraphy‐recorded motor function between patients and healthy controls.	Randomized Controlled Trial (Setting: Inpatient alcohol dependence treatment program)	74 (40 in RCT (20 in treatment, 20 control), and 34 controls)	None reported	Phosphatidylethanol blood concentration	Oxazepam (benzodiazepine) for alcohol withdrawal symptoms, 24 IU intranasal oxytocin twice daily	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	3 days	No significant differences between the oxytocin group and the placebo group for any of actigraphy variables. No significant correlations between alcohol withdrawal symptoms and motor activity on actigraphy. However, patients had significantly lower motor activity and increased variability than healthy controls in the evening.
At risk for alcohol use disorder.	[35]	Actigraphy (wrist worn device) Actiwatch Spectrum Plus, Philips Respironics	To examine reciprocal, within‐person associations between sleep and alcohol use among young adult drinkers with insomnia.	Observational study (Naturalistic setting)	N = 56 (14 male, 42 female)	Insomnia	Sleep and alcohol diaries	None	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	6 days	Participants did not report significantly different sleep quality or show changes in actigraphy‐measured sleep efficiency on heavier drinking days. Actigraphy‐based sleep efficiency was worse in those who were college students and had more frequent cannabis use.
Alcohol use disorder and post‐traumatic stress disorder.	[25]	Actigraphy (wrist worn device) Respironics Actiwatch 2	To assess sleep in participants who receive treatment for alcohol use and for PTSD.	Randomized Controlled Trial (Naturalistic setting)	N = 119 (90% male, 10% female)	PTSD and Insomnia	Urine samples, Clinician Administered PTSD Scale for DSM‐5 (CAPS‐5); Insomnia Severity Index (ISI)	Participants were randomized to either 1) COPE (integrated treatment using prolonged exposure) 2) Seeking Safety (integrated therapy using cognitive behavioral, interpersonal techniques and case management).	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	seven nights pre treatment, seven nights post treatment	Actigraphy‐derived WASO decreased by 6 min post‐treatment but remaining above the clinical threshold of 30 min. Night‐to‐night variability of sleep duration significantly decreased. Changes in PTSD symptoms significantly predicted improvements in insomnia severity as measured by the ISI, but changes in heavy drinking did not. No significant differences were found between SS and COPE treatments on actigraphy sleep variables. The change in PTSD symptoms and in drinking were able to predict changes in actigraphy. There was no difference between the 2 different treatments in actigraphy sleep variables and self reported sleep.
History of low/moderate alcohol consumption	[36]	Actigraphy (wrist worn device) Ambulatory Monitoring Inc	To investigate the association between elevated cytokines after alcohol consumption and unscheduled napping.	Observational study (Laboratory setting)	N = 10 (four male, six female)	NA	Breathalyzer BACtrack S80 Select, Pittsburg Sleep Quality Index	None	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	1 week (three nights before the study session, three nights after the binge drinking episode in the laboratory)	Baseline sleep duration was 8.0 ± 1.0 h. Sleep in the evening of binge drinking at 8 AM was 8.7 ± 1.8 h in nappers (7 participants) and 8.1 ± 1.8 h in non nappers (3 participants). Naps were 1.3 ± 0.5 h. Both nappers and non nappers in the study had increased sleep duration after binge drinking.
Alcohol use disorder	Lahti et al., 2012	Actiwatch Plus®, Cambridge Neurotechnology Ltd, UK	To assess of the effect of nitrous oxide on sleep in inpatient individuals during alcohol withdrawal.	Randomized Controlled Study (Inpatient setting)	N = 105 voluntary inpatients (80 men and 25 women)	NA	Severity of Alcohol Dependence Data (SADD) questionnaire; Obsessive‐Compulsive Drinking Scale (OCDS); Clinical Institute Withdrawal Assessment of Alcohol Scale (CIWA‐Ar)	(1) nitrous oxide/O2 (O2; starting concentration 30%, maximum 70%, aiming at a 30% EtN2O concentration); (2) normal air/O2 (flow 50%/50%, final O2 concentration 70%); (3) normal air	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	1–3 days (depending on individual inpatient detox stay)	Nitrous oxide did not improve sleep quality in individuals with AWS compared to placebo. The severity of AWS, coffee consumption, alcohol dependence and alcohol cravings were correlated with worse sleep quality in all participants.
One or more substance use disorder(s) including stimulants, opioids, tobacco, crystal meth, alcohol, heroin, crack cocaine, fentanyl, and other opioids.	[37]	Actigraphy (wrist worn device) Fatigue Science	To observe rest and activity patterns and compare these patterns between only stimulant and stimulant and opioid use and psychiatric diagnosis of psychotic and mood disorders.	Observational study (Setting: Inpatient treatment center)	N = 44 (34 male, 10 female)	Schizophrenia or psychotic disorder (79%), bipolar disorder (22%), depressive disorders (26%), anxiety disorders (34%), and ADHD (17%).	Brief Symptom Inventory (BSI)	6‐month inpatient program with pharmacotherapy	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	At least 6 days	All participants had overall long sleep duration, low sleep efficiency, and phase advanced sleep timing. Participants with stimulant and opioid use histories tended to have more sleep disruptions (p = .70) than those with only stimulant use, while patients with psychotic disorders experienced the longer duration of sleep than those with mood disorders. Psychiatric symptom load was associated with high rhythm amplitude but no association with actigraphy assessed circadian phase (midsleep). Sleep duration was not associated with length of abstinence from opioids or stimulants.
One or more substance use disorder(s).	[38]	Actigraphy (wrist worn device) Respironics® Actiwatch Spectrum®	To investigate sleep in the first week of substance use treatment.	Observational study (Naturalistic setting)	N = 23 (61% male, 39% female)	None		The program consists of 3 h of group therapy daily, weekly individual therapy and family participation (as available), regular urine and serum biomarker screening for use of drugs and alcohol, and psychiatric medication management as needed.	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	4 days	More intraindividual variability in total sleep time was associated with higher odds of relapse, but not with treatment completion.
Cannabis use disorder and insomnia	[39]	Actigraphy (wrist worn device) Actiwatch Spectrum, Philips, OR	To investigate the effect of cognitive behavioral therapy on inflammatory markers and insomnia in people who have cannabis use disorder.	Observational study (Naturalistic setting)	N = 19 (four lost to follow up; 15 participants after follow up)	No other SUDs or SCID diagnoses	Insomnia severity index (ISI), blood sample, daily sleep diaries	cognitive behavioral therapy for insomnia (CBTi).	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	1 week before treatment, 1 week after treatment	Sleep onset latency significantly decreased from 28.6 min before CBTi to 22 min after. Sleep efficiency increased from 82.2% to 83.9%. At 3 months, 80% of participants reported reduced cannabis use, with 53.3% maintaining this reduction and 26.7% relapsing by 6 months. Additionally, 66% of participants reported reduced cannabis use at 6 months, with 46.6% attributing their improved sleep to helping reduce cannabis use.
Cannabis use disorder.	[30]	Actigraphy (wrist worn device) Respironics Actiwatch	To assess the effect of bupropion on cannabis withdrawal symptoms in individuals with cannabis use disorder.	Randomized double blind controlled trial (Naturalist setting with daily lab visits)	N = 22 who were randomized; nine completers (five men; four women)	No other SUDs or SCID diagnoses	Marijuana Withdrawal Checklist; Beck Depression Inventory; Beck Anxiety Inventory; Urine test for THC	2 weeks of placebo or bupropion (days 1–3:150 mg/day; days 4–14: 300 mg/day)	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	14 days	There were no significant differences in sleep between the bupropion group and the placebo group's sleep before or after participants stopped using cannabis.
Cannabis use disorder.	[31]	Actigraphy (wrist worn device) Actiwatch Activity Monitoring System	To assess the effectiveness of quetiapine for reducing cannabis withdrawal symptoms in individuals with cannabis use disorder.	Within subject, double blind, placebo controlled randomized trial (Part naturalist setting part inpatient laboratory setting)	N = 20 enrolled, 14 completers (12 male, two female)	No other SUDs or SCID diagnoses	7 Item sleep quality visual analog scale	200 mg quetiapine per day	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	30 days	Subjective sleep quality was worse in both the treatment and the placebo groups during the cannabis withdrawal phase of the study than during the cannabis self‐administration phase. Quetiapine increased participants subjective sleep quality but there were no objective actigraphy measured improvements in sleep quality and other sleep measures.
Cannabis use disorder.	[32]	Actigraphy (wrist worn device) Actiwatch Activity Monitoring System	To assess the effectivness of lithium carbonate for reducing cannabis withdrawal symptoms in individuals with cannabis use disorder.	Double blind placebo controlled randomized trial (Inpatient Setting)	N = 38 (Placebo = 20; Lithium = 13)	No other SCID diagnoses	Cannabis withdrawal scale; Athens Insomnia Scale	Lithium carbonate (500 mg twice daily) versus placebo	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	8 days	Individuals in the lithium group experienced significantly less fragmented sleep than the placebo group. No other sleep measures differed.
Prescription opioid use disorder.	[21]	Actigraphy (wrist worn device) Philips Respironics®	To examine sleep disturbances among prescription opioid (PO) dependent individuals.	Observational study (Inpatient Hospital Setting)	N = 68 non‐treatment seeking individuals (33 prescription opioid dependent, 35 healthy controls).	Pain condition	Breathalyzer, and urine drug screen, cigarette smokers were provided with a nicotine patch upon admission.	None	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	One night	Actigraphy data showed prescription opioid participants had significantly lower total sleep time and efficiency, and greater sleep latency and wake time, but no significant correlations were found between pain or subjective sleep scores and actigraphy data.
Opioid use disorder.	[22]	Actigraphy (wrist worn device) Philips Respironics	To measure sleep duration and timing in opioid dependent participants treated with methadone or buprenorphine.	Observational study (Setting: Outpatient treatment/research clinic)	N = 48 enrolled; 37 completed (28 men, eight women)		Urinalysis, EMA self report drug use, daily sleep diaries, Pittsburgh Sleep Quality Index (PSQI)	Methadone (100 mg/day) or Buprenorphine (16–24 gm/day)	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	6 sets of 3 nights	Sleep duration was lower when using opioids or both opioids and cocaine. Sleep duration and timing were different for different drugs and for different clinic hours. When using, patients slept less, went to bed later, and woke later when using opioids or opioids and cocaine.
Prescription opioid use disorder.	[40]	Actigraphy (wrist worn device) GT3x+ Actigraph, Pensacola, Florida	To investigate clinical symptoms associated with allostatic dysregulation in patients dependent on prescription opioids at 2 time points after withdrawal, compared with healthy control participants.	Observational study (Setting: inpatient treatment center)	(1) Recently withdrawn prescription opioid dependent patients (RW): N = 7 (2) Patients in extended care for prescription opioid dependence (EC): N = 7 (3) Healthy controls: N = 7	3/7 RW and 2/7 EC patients had a history of intravenous heroin use. Some patients had past or current mood, anxiety, and or alcohol disorders.	None	None	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	8 days	Actigraphy showed less total time asleep in recently withdrawn patients compared to extended care and healthy control patients.
Opioid use disorder.	[24]	Actigraphy (wrist worn device) Micro Motionlogger Sleep Watch	To assess the changes in sleep in methadone maintained individuals with opioid use disorder who are treated with mirtazapine, zolpidem, a combination of mirtazapine and zolpidem, and placebo.	Randomized, double blind crossover with repeated latin square design (Outpatient setting)	N = 10 (four male, six female)	No other SCID diagnoses	Pittsburg Sleep Quality Index; daily sleep diaries	3 active medications (zolpidem CR 12.5 mg, mirtazapine 30 mg, and zolpidem 10 mg plus mirtazapine 30 mg) and placebo	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	8 weeks	Mirtazapine was associated with the largest increase in actigraphy measured total sleep and sleep efficiency, and decrease in sleep latency compared to all other treatments including placebo.
Cocaine use disorder.	[28]	Actigraphy (wrist worn device) Phillips Respironics	To assess the effect of 8 weeks of citicoline treatment on sleep in individuals with cocaine use disorder.	Randomized double blind controlled trial (Naturalist setting)	N = 12 (7 male; five female)	Adults with cocaine use disorder, some with cannabis use disorder, alcohol use disorder or both.	Trail Making B Test; Block Design Task; Digit Symbol Substitution Test; Wisconsin Card Sorting Task; Daily Sleep Diaries	Citicoline (1000 mg/day)	Movement in x, y, z directions (minute by minute)	Total Sleep Time (TST), Sleep Efficiency (SE), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL)	Worn for 10 weeks (except when showering or charging): 2 weeks baseline, 8 weeks treatment	Citicoline treatment did not affect sleep onset latency, sleep efficiency, waking after sleep onset, sleep quantity as measured by the number of sleep episodes, the time asleep per episode, and the amount of time spent immobile compared to placebo. Daily activity and sleep quality and quantity were also not affected by Citicoline.
Cocaine use disorder.	[29]	Actigraphy (wrist worn device) Misfit Flash, Fossil Group	To assess the changes in sleep, stress, and cue‐reactivity in individuals treated with suvorexant, an orexin receptor antagonist, in individuals with cocaine use disorder.	Placebo controlled randomized trial (Naturalist setting)	N = 23 (placebo = 10; suvorexant = 10. Three dropped within first 3 days)	No other SCID diagnoses	Pittsburg Sleep Quality Index; daily sleep diaries	Suvorexant (10 mg/day during week 1, 20 mg/day during week 2) versus placebo	Movement in x, y, z directions (minute by minute)	Sleep duration, awake bands, restful sleep, and sleep/wake switching	2 weeks	The actigraphy watch was utilized to differentiate sleep measures between the treatment and placebo group.

Table 3.

Devices used to measure clinical symptoms of substance use, stress, and craving.

Population	Reference	Device	Research question	Study design and setting where the device is used	Sample size	Comorbid psychiatric and medical diagnoses	Other measures of substance use and clinical symptoms	Therapeutic intervention	What the device measures	What the device aims to measure	Duration of wear for intervention	Clinical outcomes
Tobacco use disorder.	[41]	Personal Automatic Cigarette Tracker [PACT] worn under or over regular clothing	To investigate the feasibility of a prototype sensory system to detect smoking behavior such as breathing patterns and smoke exposure.	Observational study (Research Center Setting)	N = 20 (10 male, 10 female)	None reported	None	Not Applicable	Respiratory inductance plethysmography, proximity sensor	Measures breathing during smoking and hand to mouth gestures to detect smoking and breathing patterns during smoking.	1 day	2.5% of the total breaths measured in the study were cigarette smoke inhalations. The proximity sensor was able to identify most of the hand movements associated with smoking.
Individuals with tobacco use disorder	[42]	The device is worn around the chest and wrist. Borgwaldt Körber Solutions	To investigate whether the CReSS (Clinical Research Support System for Laboratories) pocket altered smoking behavior in individuals also wearing the Personal Automatic Cigarette Tracker v2 (PACT).	Observational study (Laboratory and naturalistic setting)	N = 45 (sex breakdown not provided)	None Reported	None	None	The tracker consists of a lighter that timestamps each lighter button press and release, a chest module to measure breathing, and an ECG sensor, accelerometer, and RF receiver in the device worn around the wrist.	Alterations in smoke exposure based on use of the mouthpiece based device.	When smoking a cigarette	Significant differences with and without CReSS were observed in puff duration, inhale‐exhale duration and volume, volume over time, smoke hold duration, and inter‐puff interval smoke exposure metrics. There were no statistically significant differences in number of puffs per cigarettes and smoking duration.
Alcohol use disorder	[43]	Wrist worn biosensor Empatica	To investigate the relationship between craving, drinking relapses, and physiological activity in individuals recovering from alcohol use disorder.	Observational study (Outpatient Setting)	N = 10 (six male, four female)	None reported	Electronic momentary assessments every 3 h.	Not described.	EDA, heart rate, movement	Physiologic measures of stress and craving that may predict alcohol use.	Worn for 100 days, charged at night.	The association between drinking lapses and craving varied between the participants. The association between self‐report craving and heart rate was negligible. The association between craving and EDA was also negligible.
Individuals with cocaine use disorder.	[27]	Wristwatch biosensor Q sensor, Affectiva	To identify biophysiometric patterns associated with cocaine use in natural environments.	Observational study (Naturalistic setting)	N = 15 (53% male, 47% female)	None Reported	Urinalysis, time line follow back	None	EDA, skin temperature, and movement in the x, y, z directions (minute by minute)	Physiologic measures of the sympathetic nervouse system associated with substance use.	daytime for 30 days, removed at night	Distinct episodes with biometric parameters consistent with cocaine use were identified on biosensor data. Eleven potential cocaine use episodes were identified by biosensors that were missed by both self‐report and drug screening. Episodes of cocaine use were detected by analysis of biosensor data in instances when self‐report or drug screening or both failed.
Individuals with cocaine use disorder	[44]	Zephyr BioHarness™ 3 chest band	To determine if electrocardiogram sensors and features could effectively differentiate cocaine use from other sympathomimetic substances.	Observational study (Inpatient research center)	N = 11 (nine male, two female)	None reported	None	None	The RR interval measuring heart rate, ECG interval proxies estimating PR, QRS, QT and QTc intervals, and full ECG waveforms involving complete RR data traces.	Cardiac measurements that may be associated with drug use	Participants wore the band for a few minutes to 6 h, depending on the condition (methylphenidate intake, exercise, tobacco use, and routine daily activities).	Cocaine use was discriminated from other conditions (methylphenidate intake, exercise, tobacco use, and routine daily activities) using ECG waves.
Individual with cocaine use disorder	[45]	Wristwatch biosensor Q sensor, Affectiva	To identify biophysiometric patterns associated with human behaviors in natural environments.	Case study (Naturalistic setting)	N = 1 male	None Reported	None	None	EDA, skin temperature, and 3D acceleration	Physiologic measures of the sympathetic nervous system associated with substance use.		The patient injected cocaine 40 min after device placement. Around the time of injection, there was a 10 μS increase in elecrodermal activity, an increase in acceleration, and skin temperature decrease of 1–2 degrees.
Opioid use disorder, receiving treatment.	[5]	Wrist worn biosensor Garmin vívosmart 4	To investigate the combined use of wearable technologies and EMA as a potential mechanism for identifying physiological and behavioral biomarkers of drug use recurrence.	Observational study (Outpatient Setting)	N = 77 (31 male, 46 female)	None reported	Urine toxicology, electronic momentary assessments about stress, fatigue, mood, pain, substance craving.	A structured group‐based, program in the Comprehensive Opioid Addiction Treatment Program (COAT)	Monitors HR and HRV via photoplethysmography, blood oxygen saturation, sleep, activity, and the Garmin Stress score (the root mean square of successive HRV)	Physiologic measures of substance use	Daily wear for at least 14 days	Participants who had drug use recurrence were more likely to have difficulty concentrating, encounter substance related triggers, and were less likely to go outside than participants who stayed abstinent. Drug use recurrence occurred among 44.2% of the sample. In the week before and in the week during the drug recurrence, participants had higher Garmin stress scores compared to abstinent participants.
Individuals with opioid use disorder	[23]	Spire® Health Tag, attached to underwear waistband or wing of bra.	To investigate respiratory data related to overdose events.	Observational study (Naturalistic setting)	N = 15	Average drug injection history 19 years (SD = 7.5). The majority (87%) had an opioid‐related overdose and 27% believed they were likely to overdose again within the next year.	Not Applicable	None	Tag collects respiratory rate (RR), physical movement, and heart rate.	Collect physiologic measures to indicate overdose.	5‐day period, removing for bathing.	No cases of acute respiratory depression were detected. There were 385 drug administrations, with 91% involving fentanyl. Of these, 6.8% were classified as “too much,” 85% of which were fentanyl‐related. On average, participants had 4.7 fentanyl injections daily (SD = 2.8). Seven participants (46%) only used fentanyl, while the rest (54%) reported polydrug use. No opioid overdoses were reported.
Individuals with substance use disorder (primarily opioid use disorder) who are treatment seeking.	[6]	AutoSense chestband with two‐lead ECG	To demonstrate the feasibility and effectiveness of physiological monitoring to inform the consequences of drug use and to explore potential for a sensor‐triggered mobile intervention that detects behaviorally risky situations.	Observational study (Naturalistic setting)	N = 40 (30 men, 10 female)	Cocaine dependence (50%), opioid dependence (92.5%), alcohol dependence (17.5%), cannabis dependence (2.5%), obsessive compulsive (5%), antisocial personality (50%)	Urinanlysis, electronic momentary assessment of craving and drug use	Methadone (oral) or buprenorphine (sublingual) maintenance from enrollment to 36 weeks, individual counseling available 1/week	Heart rate, heart‐rate variability, respiration, skin conductance, temperature, and movement	Cardiac measurements that may be associated with drug use	Four 1‐week periods separated by a week each, up to 7 weeks	HR was higher than baseline around the time of cocaine use (mean: 85.8 bpm) and lower around the time of heroin use (mean: 77.5 bpm), with these differences evident 2 h after drug use. HR was significantly higher with cocaine compared to heroin use and increased with higher doses of cocaine. HR was significantly increased 30 min before and after EMA reports of craving heroin or cocaine, and varied with different activities and emotions, being lower when relaxed or happy, and higher when stressed or walking.
Heavy (70%; chronic daily use/opioid maintenance therapy/diagnosis of opioid use disorder) and non‐heavy opioid users (30%).	[46]	Affectiva Wrist worn biosensor	To investigate the utility of using a wearable biosensor to detect physiological data after opioid use.	Observational study (Setting: Emergency department)	N = 30 (47% male, 53% female)	None Reported	None	Opioid	EDA, skin temperature, movement	physiological changes associated with opioid use	5 min prior to opioid administration and 30 min post administration	In both groups of people who use opioids heavily and non‐heavily there was an increase in skin temperature and a decrease in movement post opioid administration but there was no significant change in EDA compared to pre opioid administration. There was no difference in skin temperature change or EDA between the groups. There was a greater decrease in movement in the heavy user group.
Individuals with one or more SUDs	[7]	E4 wrist worn biosensor	To determine the accuracy of a wearable sensor to detect and differentiate episodes of self‐reported craving and stress in individuals with substance use disorders, and to assess the feasibility of sensor based monitoring in this population.	Observational study (Outpatient setting)	N = 30 (57% male, 43% female)	Half of the participants identified multiple primary substances of use. History of alcohol use was reported by 73% of individuals with a single substance use disorder, with heroin use disorder making up the remaining 27%. A history of a psychiatric disorder was reported in 70%.	Self‐reported stress and craving in their wearable device.	Medication assisted treatment (33% of sample).	EDA, skin temperature, three‐dimensional acceleration, and heart rate.	Physiologic measures of sympathetic nervous system arousal associated with stress and craving.	Daily wear for 4 days, charged during the night.	The three machine learning models generated were able to differentiate between: (1) stress versus non‐stress (accuracy 74.5%; AUC 0.82), (2) craving versus no‐craving (accuracy 75.7%; AUC 0.82) and (3) craving versus stress (accuracy 76.8%: AUC 0.8) primarily based on accelerometry data.

Using a modified NIH quality assessment tool (8‐point scale), all 33 studies were rated as good quality. ¹⁷ All studies clearly reported their research objectives and population characteristics, but fewer included sample size justifications. SUD assessment tools varied across studies: some used validated tools, some relied on prior clinical diagnosis without clearly reporting independent diagnostic confirmation, and others enrolled participants based on substance‐specific risk criteria such as at‐risk drinking.

Feasibility and acceptability

Feasibility and acceptability of remote opioid overdose detection

This subsection included three opioid overdose‐detection acceptability studies, including two hypothetical devices and one wrist‐watch device that measured heart rate and oxygen saturation. ¹⁸ , ¹⁹ , ²⁰ We found three studies specifically investigating the perceived acceptability of hypothetical wearable devices to detect opioid overdose and alert others via semi‐structured interviews. ¹⁸ , ¹⁹ , ²⁰ These studies found that interest in hypothetical overdose‐detection devices was moderate to high, and willingness was associated with past overdose experiences. ¹⁸ , ¹⁹ , ²⁰ Evidence from studies using actual wearable devices was limited but generally suggested positive user experiences alongside notable adherence challenges, including reduced compliance during periods of active substance use and substantial attrition in some settings. ⁵ , ⁷ , ²¹ Concerns were raised about privacy and location tracking, which could put the wearer at risk of harm or trigger police involvement. ¹⁸ Another consideration that arose was the requirement of manual activation of the device, which burdens the wearer and may therefore result in missed overdose events. ¹⁸ , ²⁰ Participants were excited about the potential to reverse opioid overdose through administration of a reversal agent and found it would be particularly applicable for unhoused individuals and for individuals with low tolerance, such as those who have reduced their use due to incarceration or residential treatment, who may be more likely to overdose. ²⁰ However, one study found that unhoused individuals, compared to those in stable housing, were less likely to be interested in wearing a device that detects an opioid overdose and alerts others for help. ¹⁹ There were also concerns related to the risk of a false positive overdose detection, which could lead to unnecessary administration of a reversal agent or the risk of detecting an overdose caused by a different substance. ²⁰ Overall, studies of hypothetical opioid overdose detection wearables found that individuals with opioid use disorder (OUD) were generally interested in discreet, comfortable wrist‐worn devices that could monitor vital signs and alert others during an overdose. These studies assessed hypothetical overdose‐detection devices, therefore their findings were interpreted as evidence of acceptability, perceived feasibility, and design preferences. Acceptability was shaped by concerns about privacy, location tracking, manual activation, false detections, and potential risks for unhoused individuals, highlighting the need for careful device design and implementation.

Acceptability of a wearable device to track stress in patients with various substance use disorders

This subsection included two physiological monitoring devices: one wrist‐worn biosensor and one chest band. ⁶ , ⁷ Wearable devices used to track physiological markers of stress and craving were generally acceptable among individuals with SUDs, with participants reporting ease of use, comfort, and willingness to wear the devices beyond the study period. ⁶ , ⁷ Majority of participants (93%) who wore a wrist worn device that measured skin conductance, heart rate, and movement for 4 days were willing to wear the device for longer than the study period and enjoyed the device's user interface. ⁷ Moreover, more than 66% of participants in that study reported being more mindful of their stress and cravings while wearing the sensor. ⁷ A chest band measuring heart rate and other physiological proxies of stress in 40 polydrug users was reported to be comfortable and easy to use. ⁶ It should be noted that 28% dropped out of the trial, and none did so for reasons related to device acceptability. ⁶ These findings suggest that stress‐monitoring wearables may be feasible in outpatient and polydrug‐use populations, although evidence remains limited to a small number of short‐term studies.

Adherence and compliance

This subsection summarized adherence across ten wrist‐worn actigraphy studies, two wrist‐worn physiological biosensor studies, and one clothing‐attached respiratory/heart‐rate monitoring tag. ⁵ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ⁴⁷ Overall, there were few wearable device‐related dropouts reported, particularly in RCTs using actigraphy to assess sleep‐related outcomes. ²¹ , ²² , ²⁴ , ²⁵ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ⁴⁷ However, adherence varied by population and context, with lower data contribution among individuals with longer opioid use, polysubstance use, recurrence of drug use, or brief or unclear monitoring protocols, highlighting the importance of wear duration, participant burden, and substance‐use patterns when interpreting wearable data. ⁵ , ²³ In a study investigating sleep using a wrist‐worn actigraphy device in inpatient individuals with alcohol use disorder (AUD) and post‐traumatic stress disorder (PTSD), 49% wore the device throughout the study period which was 1 week before and 1 week after 12 sessions of behavioral therapy. ²⁵ In another study looking at sleep disturbances in individuals with prescription opioid dependence in one overnight hospital stay, ~50% of the eligible participants withdrew from the study. ²¹ In another actigraphy sleep study of 48 participants, only two individuals with OUD withdrew from the study because of the device‐wearing requirement. ²² Adherence to both EMA and wrist worn biosensor monitoring in individuals with OUD declined around drug use recurrence, and some participants were excluded due to insufficient wearable or recurrence‐status data, highlighting that compliance may be most challenging during clinically important periods. ⁵ Wrist‐worn biosensors were found to have high protocol compliance among individuals with cocaine use disorder, and approximately half of participants reported increased mindfulness and reduced likelihood of drug use while wearing the device. ²⁷ Seven RCTs investigating pharmacotherapy treatments for sleep regulation in various substance use disorders did not report any device related study drop outs or device non‐adherence, suggesting that wrist‐worn actigraphy is a feasible device to study sleep. ²⁴ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ⁴⁷

Clinical outcomes

Measuring sleep

All sleep studies used wrist worn actigraphy to measure sleep. The included studies reported sleep‐related outcomes rather than daytime physical activity; therefore, conclusions about waking activity were not reported.

Sleep and alcohol use

Six studies investigating sleep in individuals with alcohol use disorder or at risk of developing alcohol use disorder were found (Table 2). ²⁵ , ²⁶ , ³⁴ , ³⁶ , ⁴⁷ These studies suggest that alcohol use and treatment are associated with measurable changes in sleep duration, regularity, activity, and variability, although findings differed by population, drinking context, and treatment setting. In a study looking at young adults at risk for AUD, there were no significant changes in sleep efficiency on heavy drinking days. ³⁵ In another study measuring sleep after a morning binge drinking episode among low to moderate drinking individuals with no history of AUD, actigraphy revealed that participants slept longer compared to baseline and most of the sample napped as their blood alcohol content was decreasing. ³⁶ A study looked at sleep regularity index (SRI) scores, which are a measure of sleep consistency, among inpatients in an AUD treatment, using actigraphy. The average SRI improved over the course of treatment, suggesting improved sleep. ³⁴ AUD inpatients monitored for acute alcohol withdrawal in a treatment center found that these patients had lower motor activity than healthy controls, who were not reported to sleep in the inpatient setting, in the evening. ²⁶ In a study looking at veterans with comorbid AUD and PTSD, actigraphy was used to monitor sleep and found that there was some improvement in sleep quality, including less variability in total sleep time after cognitive behavioral therapy and exposure therapy. ²⁵ Another study used actigraphy to compare sleep outcomes across treatment arms for alcohol withdrawal syndrome and found no significant differences, indicating that actigraphy can be incorporated into SUD treatment research to objectively monitor sleep parameters, although this study did not demonstrate treatment‐related differences in actigraphy‐derived sleep outcomes. ⁴⁷

Sleep and cannabis

Four studies investigating sleep in individuals with cannabis use disorder were found. ³⁰ , ³¹ , ³² , ³⁹ In a study observing individuals with cannabis use disorder and insomnia, actigraphy revealed a significant decrease in sleep onset latency (SOL) following cognitive behavioral therapy for insomnia, providing objective evidence of improved sleep patterns (Table 2). ³⁹ Three RCTs used actigraphy to evaluate sleep‐related outcomes during cannabis withdrawal and found mixed evidence for pharmacotherapy effects, with some improvements in objective sleep measures but no consistent benefit across sleep outcomes or subjective sleep quality. ³⁰ , ³¹ , ³²

Sleep and opioids

Four studies investigating sleep in individuals with OUD were identified (Table 2). ²¹ , ²² , ²⁴ , ⁴⁰ Overall, actigraphy studies in opioid‐use populations suggest disrupted sleep timing, sleep duration, and objective sleep quality, although findings varied by treatment status, recent withdrawal, polysubstance use, and comparison group. In a study comparing healthy controls to individuals with prescription OUD, actigraphy revealed significant impairments in objective sleep quality among individuals with opioid dependence, showing lower sleep efficiency, more time awake, and increased movement during the night compared to controls. ²¹ In another study, observing sleep in individuals with OUD, sleep onset and waking times was significantly influenced by opioid and cocaine use, with later bedtimes and wake‐up times associated with active drug use. ²² Another study found less sleep among recently withdrawn individuals with OUD, compared to individuals in extended care for opioid use and healthy controls. ⁴⁰ One RCT used actigraphy to evaluate sleep‐related outcomes during treatment with mirtazapine, zolpidem, a combination of mirtazapine and zolpidem, and placebo. ²⁴

Sleep and other SUDs

Table 2 outlines the studies that used wearable devices to study sleep in other SUDs. In a study observing patients with complex substance use histories, a wrist accelerometer revealed that participants with severe concurrent disorders exhibited long sleep durations, frequent nighttime awakenings, and phase‐advanced rest‐activity rhythms, with significant disruptions in sleep quality, particularly in those with a history of opioid use and psychotic disorders. ³⁷ Night‐to‐night variability in sleep duration, as measured by actigraphy, predicted relapse more strongly than average sleep patterns in individuals undergoing treatment for SUDs. ³⁸ In a study comparing sleep in young adult smokers and non‐smokers, actigraphy revealed that nicotine dependence in young adult smokers was linked to poorer sleep continuity, even though smokers may not report subjectively worse sleep quality. ³³ Two RCTs used actigraphy to evaluate sleep‐related outcomes in individuals with cocaine use disorder receiving treatments to improve sleep and found mixed results, with citicoline showing no disruption or improvement in sleep/wake cycles, while suvorexant showed preliminary evidence of improved actigraphy‐based sleep outcomes. ²⁸ , ²⁹

Ability to detect predictors of substance use

Table 3 outlines the studies that observed craving and stress in substance users. Five studies examined wearable‐derived predictors of craving, stress, or substance‐use recurrence using four wrist‐worn biosensors and one chest‐worn band. ⁵ , ⁶ , ⁷ , ⁴³ , ⁴⁵ Across studies, wearables measured physiological and behavioral signals including heart rate, heart rate variability, electrodermal activity, skin temperature, movement, blood oxygenation, and sleep‐related measures to examine whether these signals corresponded with stress, craving, substance use, or recurrence risk. Overall, findings suggest that wearable‐derived data may capture clinically relevant changes, particularly for stress and substance‐use‐related states, although results were not consistent across studies. Machine learning models showed promise in distinguishing stress from no stress, craving from no craving, and craving from stress, while other studies found associations between heart rate changes, stress scores, drug use, craving, or recurrence risk. ⁵ , ⁶ , ⁷ A case study also demonstrated acute physiological changes following cocaine injection, including changes in skin temperature, electrodermal activity, and increased movement as detected by an accelerometer. ⁴⁵ However, one longitudinal study in individuals recovering from AUD found no significant association between self‐reported craving and relapse or wearable‐derived cardiovascular and electrodermal activity, highlighting the need for further validation against subjective reports and clinical outcomes. ⁴³ Another study found decreased movement after an opioid administration, which was greater among individuals with heavier opioid use, detected by a wrist‐worn biosensor with an accelerometer, suggesting the utility of this method as a way to detect opioid intoxication or overdose. ⁴⁶ Together, these studies suggest that wearables may support real‐time identification of stress, craving, or substance‐use risk states, but the evidence remains preliminary, heterogeneous, and dependent on the specific substance, wearable signals, and analytic methods used.

Ability to detect substance use

Six studies examined whether wearable devices could detect substance use or overdose‐related events using physiological or behavioral signals using two wrist‐worn biosensors, one chest‐worn ECG band, two smoking‐detection systems incorporating respiratory or gesture sensors, and one clothing‐attached respiratory/heart‐rate monitoring tag (Table 3). ²³ , ²⁷ , ⁴¹ , ⁴² , ⁴⁴ , ⁴⁶ One study used a wrist worn accelerometer to detect smoking‐related hand‐to‐mouth movements and inhalation patterns of cigarettes. ⁴¹ Overall, these studies suggest that wearable devices may be able to detect substance‐related events through changes in respiratory rate, heart rate, electrodermal activity, skin temperature, movement, and substance‐specific gestures. Evidence was strongest for detecting discrete use‐related behaviors, such as cocaine use or smoking episodes, where wearable signals distinguished cocaine use from other activities or substances and identified smoking‐related exhalations and hand‐to‐mouth movements. ²⁷ , ⁴¹ , ⁴² , ⁴⁴ Wearable detection of opioid‐related events was more limited: one study found increased skin temperature and decreased movement after opioid administration, while another overdose‐detection study reported no overdose events or respiratory depression during monitoring, which limits any conclusions about real‐world overdose detection using this biosensor. ²³ , ⁴⁶ Together, these findings suggest that wearables may support detection of substance‐use‐related physiological and behavioral patterns, but further validation is needed in real‐world settings and during clinically significant events such as overdose.

DISCUSSION

Our systematic review synthesized the emerging literature on the feasibility, acceptability, and potential clinical applications of wearable devices among individuals with substance use disorders (SUDs) or at‐risk substance use. Overall, wearable devices appear feasible and generally acceptable in these populations, but evidence for their clinical utility remains preliminary. ⁴ , ¹² The strongest evidence was for actigraphy‐based sleep monitoring, while findings related to stress, craving, substance use, and overdose detection were promising but limited by small samples, heterogeneous devices and populations, short monitoring periods, and limited longitudinal outcome data. There were only three studies that reported daytime movement data. Most of the studies identified in our search did not include daytime physical activity, and step‐count outcomes were not represented among the eligible included studies; future reviews may need to specifically examine wearable physical‐activity monitoring in SUD populations as this literature expands.

These findings suggest that wearables may be useful for capturing real‐world physiological and behavioral changes, but additional work is needed before these measures can be used to guide clinical decision‐making. ⁴⁸

Device user perspectives highlight that individuals value wearable devices that are comfortable, discreet, durable, privacy‐preserving, and suitable for sustained use in real‐world substance‐use contexts. ⁷ , ¹⁸ , ¹⁹ , ²⁰ The large representation of OUD studies likely reflects the urgent need for overdose prevention and the suitability of wearable devices for capturing opioid‐relevant physiological changes, such as respiratory depression, oxygen desaturation, heart‐rate changes, and immobility. ⁷ , ²³ Although overdose events may be infrequent during study monitoring periods, overdose remains an important target for wearable research because timely detection and response could be life‐saving. ²⁰ , ²³ Future studies should evaluate overdose‐detection wearables in populations and settings where overdose risk is elevated and where clear safety protocols, consent procedures, and response pathways are in place. Studies should also assess device durability, false alarms, privacy risks, and unintended harms, particularly for people who use opioids alone or are unstably housed.

Smaller and less visible devices were generally preferred, although some individuals may also benefit from optional access to their own wearable data through smartphones or other platforms. ¹⁸ , ¹⁹ , ²⁰ However, access to data should be paired with education and clinical support so that patients and providers can interpret wearable‐derived information appropriately. ¹⁰

The prominence of sleep‐focused studies in this review likely reflects that sleep disturbance is common during substance use, withdrawal, and recovery, and that actigraphy provides an established and feasible method for objectively measuring sleep and rest‐activity patterns. ⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ However, this emphasis also highlights a gap in the literature, as wearable research in SUD has focused more on measurable sleep outcomes than on longer‐term clinical outcomes such as relapse, treatment response, overdose risk, or recovery trajectories. Across studies, individuals with SUDs often showed poorer or more variable sleep, including lower sleep efficiency, greater wake after sleep onset, longer sleep onset latency, and increased night‐to‐night variability. ²¹ , ³³ , ³⁷ Several studies also showed a mismatch between subjective sleep complaints and actigraphy‐measured sleep, indicating that wearable‐derived sleep outcomes should be interpreted alongside self‐report, sleep diaries, psychiatric symptoms, substance type, and duration of abstinence. ²⁶ , ³⁵ This is particularly important because actigraphy estimates sleep from movement and may not fully capture perceived sleep quality or insomnia symptoms. ⁸ , ²⁶ , ³⁵

The sleep findings suggest that sleep instability may be particularly relevant to SUD recovery. Improvements in sleep regularity during alcohol treatment were associated with reduced mental and physical exhaustion, while greater variability in total sleep time early in treatment was associated with higher odds of relapse. ³⁴ , ³⁸ However, sleep disruption can persist for weeks or months into abstinence, making it difficult to determine whether sleep changes reflect acute substance use, withdrawal, psychiatric comorbidity, medication effects, or broader recovery‐related dysregulation. ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² Future studies should establish substance‐specific sleep trajectories and examine whether wearable‐informed sleep assessment or intervention can improve treatment outcomes. Given the importance of sleep disturbance in recovery, future studies should test whether wearable‐informed sleep assessment can improve treatment engagement, relapse prevention, or recovery outcomes before these tools are integrated into routine SUD treatment models.

Beyond sleep, wearable devices were used to detect physiological and behavioral signals associated with stress, craving, substance use, and overdose risk. Some findings suggest that wearable signals may differ by substance class, such as increased heart rate, movement, and electrodermal activity after stimulant use, compared with decreased movement and altered temperature or cardiorespiratory patterns after opioid use. ⁶ , ²⁷ , ⁴⁴ Sleep and circadian disruption may be especially relevant for alcohol, cannabis, and stimulant use, whereas respiratory and heart‐rate changes may be more clinically relevant for opioid overdose detection. ³⁸ , ⁴⁹ , ⁵¹ , ⁵² However, these signals are not substance‐specific on their own and may overlap with stress, emotion, physical activity, medication effects, psychiatric symptoms, or environmental context. ⁶ , ⁴⁸ Future studies should therefore examine substance‐specific wearable profiles rather than assuming a single physiological pattern of substance use or relapse risk.

Building on existing mobile health just‐in‐time adaptive interventions in addiction science, wearable devices may extend smartphone‐based approaches by adding passive physiological or movement‐based signals to help identify clinically relevant states such as craving, stress, sleep disruption, substance use, or overdose risk and trigger tailored feedback or support. ⁵³ , ⁵⁴ For example, a wearable device could eventually alert a patient to rising stress or sleep disruption, prompt a coping strategy, or notify a support person during a potential overdose event. However, this will require validation of substance‐specific algorithms that combine passive physiological data with brief self‐report, EMA, or sleep diaries to improve real‐time clinical interpretation. ³ , ⁴⁸

Studies varied widely in device type, monitoring duration, clinical population, and outcome definition, making cross‐study comparison difficult. Many studies also relied on short‐term feasibility data and did not consistently report adherence, acceptability, or reasons for dropout. Brief wear periods, including single‐night actigraphy, may provide useful preliminary cross‐sectional data when comparing clinical and control groups, but should be interpreted cautiously. Stage of recovery is also important, as signals collected during active use, withdrawal, early abstinence, or maintenance recovery may reflect different clinical states. ⁵⁵ Future studies should define the clinical purpose of monitoring in advance and justify the duration and frequency of recording based on that purpose.

Control group selection and comorbidity also affect interpretation. Some studies compared individuals with SUDs to healthy controls, while others compared treatment groups within the same clinical population. Although excluding psychiatric comorbidity may reduce confounding, it limits generalizability because mental health comorbidities are common in SUD populations and may independently affect wearable data. ⁵⁶ Future studies should consider whether wearable‐derived signals are specific to substance use or reflect broader clinical or physiological processes. This distinction is important if wearable data are intended to inform treatment decisions. From a policy perspective, wearable use in SUD care will require guidance on privacy, data ownership, clinical responsibility, device validation, and clear response pathways, particularly for overdose‐alert systems. ⁴⁸

This review has several limitations. The available literature was small and heterogeneous, and most studies did not evaluate whether wearable monitoring improved substance use outcomes over time. Few studies addressed equity or contextual factors influencing adoption, and many focused on individuals with OUD, limiting generalizability across other substances. Hypothetical or conceptual wearables were included only when they assessed feasibility, acceptability, or design preferences; however, because these studies may be inconsistently indexed, some relevant studies may not have been captured. Although direct biochemical drug‐detection biosensors were outside the scope of this review, emerging devices such as wrist‐worn transdermal alcohol sensors may be promising for future alcohol‐use research, particularly for detecting drinking patterns or problematic alcohol‐use behaviors in real‐world settings. ¹² , ¹⁴ , ¹⁵ , ¹⁶ Wearable technologies are an emerging field and are described using many partially overlapping terms with different scopes and meanings, which may have limited the sensitivity of a systematic search focused specifically on wearable devices.

Overall, wearable devices hold promise for monitoring physiological and behavioral indicators relevant to SUDs, particularly sleep disruption, stress, craving, substance use, and overdose risk. ⁴ , ¹² At present, the strongest evidence supports the feasibility of passive monitoring of sleep and rest‐activity patterns, while their use for real‐time detection of craving, relapse, or overdose remains preliminary. ⁸ , ²¹ , ²² , ²⁵ , ³⁴ , ³⁸ Future research should move beyond feasibility by testing whether wearable‐informed monitoring can improve clinical decision‐making, patient self‐management, treatment engagement, and long‐term recovery outcomes. ⁴⁸ To support clinical translation, studies should prioritize user‐centered design, transparent reporting of adherence and acceptability, substance‐specific validation, integration with patient‐reported outcomes, and careful consideration of the clinical context in which wearable data would be used. ⁴⁸ , ⁵⁴

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supplementary Table 1. Risk Of Bias Tool. Supplementary Table 2. Risk of Bias Assessment for Each Study.

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.