JMIR Publications

ABSTRACT

Background:

Substance use disorders (SUDs) are chronic, relapsing conditions for which effective treatments exist (Volkow & Blanco, 2023). These include medications for opioid use disorder (MOUD), cognitive behavioural therapy (CBT), motivational interviewing and contingency management (CM), in which individuals receive tangible rewards for verified abstinence (Dar et al., 2023; Rosenthal et al., 2022). However, many individuals who could benefit never reach treatment, and many who do begin it leave early. Distance, cost, stigma, limited clinic hours, and workforce shortages restrict access, and the episodic nature of clinic care leaves long gaps between contacts in a condition where relapse can occur at any time (Yazıcı & Bardakçı, 2023). Digital health interventions (DHIs) have been proposed to fill these gaps. The term covers smartphone apps, web programmes, text and chat messaging, telehealth, remote monitoring, digitally delivered CM, clinical decision support for clinicians and, more recently, conversational agents built on large language models (Gentili et al., 2022). DHIs can deliver established treatment components at scale, extend contact between visits and reach people who would not attend a clinic (Businelle et al., 2024; Marsch et al., 2020). Their promise depends on whether they change what matters: substance use, retention in treatment, and access to effective medication. Existing syntheses suggest a qualified benefit. A 2025 systematic review and meta-analysis of 34 randomized trials found that remote or digital interventions added to in-person alcohol and drug treatment appear to reduce relapse and days of use (Kwan et al., 2025). A review of digital interventions for young people found weak evidence overall, with a small, short-lived reduction in alcohol use (O'Logbon et al., 2024). A scoping review of methamphetamine-focused tools described rising attrition in an app study, from 11.5% at 8 weeks to 55.8% at 40 weeks (Fletcher et al., 2025), and a scoping review for opioid use disorder reported reduced opioid use against usual care in several trials but not for the A-CHESS support app (Scialanca et al., 2025). In tobacco control, the evidence is stronger; a vaping-cessation text programme raised 7-month abstinence among young adults from 18.6% to 24.1% (Graham et al., 2021). With increasing technological developments, the field of DHIs has changed. Telehealth prescribing of buprenorphine has been studied at scale in Medicaid, veterans, and commercial populations (Marsch et al., 2020). Cluster-randomized trials have tested monitoring apps and clinical decision support across dozens of clinics (Shahmoradi et al., 2021). Tailored and adaptive messaging has been trialled for alcohol and cannabis, smartphone CM has moved into feasibility testing, and AI chatbots have entered pilot evaluation. However, significant uncertainties remain. Previous syntheses are often limited to one substance or delivery mode, and therefore recent primary studies on DHIs have not been appraised together across substance classes. It is therefore unclear which delivery models have credible effects, for which substances, and where evidence is still insufficient to guide practice. This review addresses that gap by systematically identifying and appraising primary studies that examined DHIs. The review addresses the following research question: In people with substance use disorder or problematic substance use, do digital health interventions improve substance use outcomes, treatment retention, or treatment receipt compared with usual care, active or minimal controls, or no intervention?

Objective:

The review addresses the following research question: In people with substance use disorder or problematic substance use, do digital health interventions improve substance use outcomes, treatment retention, or treatment receipt compared with usual care, active or minimal controls, or no intervention?

Methods:

Methods The review followed the guidelines of the PRISMA 2020 statement (Page et al., 2021). The protocol was not prospectively registered; the eligibility criteria, search approach, appraisal tools, and synthesis plan below were developed before screening began. 2.1 Eligibility Criteria Criteria are summarized in Table 1. Table 1. Inclusion and exclusion criteria Domain Inclusion criteria Exclusion criteria Population People of any age with SUD or problematic or hazardous use of alcohol, opioids, stimulants (including methamphetamine and cocaine), cannabis, or tobacco and nicotine (including vaping). Prevention studies in people who do not use substances; behavioral addictions (gaming, gambling, internet use); prescribed stimulant use for ADHD without SUD outcomes. Intervention Digital health interventions: smartphone apps, web programs, text or chat messaging, telehealth or telemedicine, remote or digital contingency management, remote monitoring or wearables, clinical decision support, conversational agents or large language models. Non-digital interventions (for example brain stimulation, medications alone, in-person care navigation); digital tools used only for research data collection. Comparator Usual care, in-person care, waitlist, minimal, educational or active control, or no intervention. Uncontrolled designs were accepted and appraised separately. None. Comparator type was not used to exclude studies. Outcomes Substance use or abstinence, heavy drinking, relapse, treatment retention, linkage to or receipt of treatment, health service use, overdose or mortality, feasibility and engagement. Development-only reports with no participant outcome data. Study design Primary studies: randomized trials (individual, cluster, stepped-wedge, micro-randomized), non-randomized studies, cohort studies, single-arm pilots and case series. Systematic reviews, meta-analyses, scoping, narrative and umbrella reviews, health technology assessments, commentaries, protocols and registry entries, qualitative-only studies, surveys, preprints, secondary and mediation analyses. Date and language Published 1 January 2023 to 29 September 2026; English. Published before 2023; languages other than English. 2.2 Information Sources Searches were run on 25 September 2026 using a web search tool that retrieves journal articles from publisher websites including Addiction, JAMA Network, JMIR, Lancet, Elsevier, Wiley and Springer, PubMed and PubMed Central records, ClinicalTrials.gov entries and open repositories. Licensed bibliographic databases, MEDLINE, Embase, PsycINFO, CENTRAL, and Web of Science were not searched directly. Reference lists and linked records of retrieved articles guided follow-up searches for named trials. 2.3 Search Strategy Searches combined three concept blocks: a substance or SUD term, a digital intervention term, and a study design or outcome term, with substance- and modality-specific variants. Twenty-three strings were applied (Table 2). Year terms (2023 to 2026) were added to many strings, and date eligibility was confirmed during screening. Later strings were targeted at named trials or topics identified in earlier results to retrieve abstracts and outcome data for those studies. Table 2. Full search strings applied (web search tool) Search string Records retrieved 1 systematic review meta-analysis digital health interventions substance use disorder 2025 9 2 randomized controlled trial smartphone app opioid use disorder buprenorphine retention 2025 10 3 digital contingency management stimulant methamphetamine randomized trial results smartphone 2024 2025 10 4 randomized clinical trial app-based intervention alcohol use disorder heavy drinking results 2024 2025 JAMA OR Addiction OR JMIR 10 5 cannabis use digital intervention randomized controlled trial 2025 app OR web-based OR text messaging outcomes 10 6 telehealth opioid use disorder treatment retention overdose cohort study 2024 2025 buprenorphine telemedicine 10 7 chatbot OR large language model OR conversational agent substance use disorder randomized trial or pilot 2025 9 8 Efficacy and comparative effectiveness of telephone and smartphone remote continuing care interventions alcohol use disorder randomized controlled trial Addiction 2024 10 9 opioid use disorder mobile app randomized trial results opioid-negative urine or retention 2024 digital therapeutic OR recovery support 10 10 smoking cessation smartphone app OR text messaging randomized clinical trial 2025 abstinence results 10 11 automated tailored adaptive mobile messaging reduce alcohol consumption help-seeking adults randomized controlled trial Addiction 2024 9 12 DynamiCare OR "digital contingency management" randomized trial results 2025 abstinence stimulant OR opioid OR alcohol 9 13 randomized clinical trial telehealth OR app stimulant use disorder cocaine methamphetamine abstinence outcomes published 2024 2025 10 14 vaping cessation text message OR app randomized clinical trial young adults 2024 2025 results 10 15 recovery support app OR text messaging opioid use disorder randomized trial 2025 outcomes emergency department OR jail OR pregnancy 9 16 Gustafson A-CHESS smartphone app opioid use disorder buprenorphine methadone naltrexone randomized clinical trial opioid abstinence results Am J Psychiatry 10 17 Clinical Decision Support System for Primary Care of Opioid Use Disorder randomized clinical trial JAMA Netw Open results buprenorphine prescribing 10 18 Smoke Free app pragmatic randomized trial ASCET results quit rates offer of app 6 months primary outcome JMIR 2024 10 19 Östh breathalyser smartphone application alcohol dependent Sweden randomised controlled trial heavy drinking days 26 weeks IRR results abstract 9 20 Digital cognitive behavioral therapy program adults alcohol use disorder randomized clinical trial JAMA Network Open 2024 e2435205 percent days abstinent 99 participants 9 21 Can Internet-based intervention reduce cannabis use randomized controlled trial results Dutch frequent cannabis users Addiction OR "Journal of Medical Internet Research" 2025 9 22 Kawasaki Lehman Kong ReSET-O digital therapeutic mobile app adjunct medication opioid use disorder randomized clinical trial opioid treatment program Journal of Addiction Medicine retention 9 23 Effects of a text messaging smoking cessation intervention amongst online help-seekers and primary health care visitors randomised controlled trial abstinence odds ratio results Sweden 10   2.4 Selection and Data Collection Process Two reviewers screened all records in two stages against the criteria in Table 1: title and abstract screening, followed by detailed assessment of the retrieved record. The reviewers extracted data into a structured form: study identifier, design and sample size, substance, digital component, comparator, follow-up, primary outcome, effect estimates as reported, and engagement or attrition where available. Where several records described one study, values were cross-checked across them. 2.5 Risk of Bias Assessment Randomized studies were appraised with the Cochrane Risk of Bias 2 tool (RoB 2), judged Low, Some concerns, or High. Non-randomized studies were appraised with ROBINS-I, judged Moderate, Serious, or Critical. The tool was adapted for the micro-randomized pilot, and single-arm studies and case series were rated Critical for lack of a comparison group. Domains considered were randomization and allocation, deviations from intended interventions, missing outcome data, outcome measurement (self-report versus objective or biochemical verification, assessor blinding), and selective reporting. When the retrieved text did not report a domain, the rating did not assume the best case: self-reported outcomes in unblinded designs were rated Some concerns at minimum. 2.6 Synthesis Methods Studies differed in populations, comparators, outcome metrics, and follow-up, so meta-analysis was not appropriate. Synthesis followed the Synthesis Without Meta-analysis (SWiM) guidance (Campbell et al., 2020). Studies were grouped by substance and delivery model, effect estimates were reported as published, and the direction of effect on each study's primary outcome was tallied.

Results:

Results 3.1 Study Selection The 23 searches retrieved 221 records. After removing 61 duplicates, 160 unique records were screened, of which 117 were excluded: 44 were protocols, registry entries, or trial descriptions; 25 were published before 2023; 21 were secondary studies, including systematic and scoping reviews, health technology assessments, commentaries, and secondary or mediation analyses; 18 did not concern a digital SUD intervention; and 9 were non-scholarly pages. The remaining 43 records were assessed in detail. Thirteen were excluded because they were qualitative studies (n = 6), a survey (n = 1), a development-only report (n = 1), preprints (n = 2), or opinion papers and editorials (n = 3). A total of 30 studies were therefore included in the review (Figure 1). Figure 1. PRISMA 2020 flow diagram of study identification, screening, eligibility and inclusion 3.2 Study Characteristics The 30 included studies comprised 22 randomized studies and 8 non-randomized studies. By substance, 7 concerned alcohol, 10 opioids, 2 stimulants, 5 cannabis, and 6 tobacco or nicotine. Sample sizes ranged from fewer than 10 in a pilot phase to more than 91,000 in a claims-based cohort, and the total exceeded 137,000 participants where sample size was retrieved. Most studies were conducted in the United States and Europe; one trial was from Japan. Interventions included monitoring and feedback apps, digital CBT, CM apps, text and chat messaging, telehealth delivery of buprenorphine, clinical decision support, and a conversational agent. Comparators were usual care, active or educational controls, waitlist, or external benchmarks. Primary outcomes were self-reported or biochemically verified abstinence, heavy drinking days, treatment retention, treatment receipt, and feasibility. Follow-up ranged from 30 days to 24 months (Table 3). Table 3: Summary of Included Studies Study Design and sample Digital intervention Comparator Main result Alcohol 1 Muench et al.(2024) RCT, single-blind; N = 723 help-seeking adults Daily tailored static or tailored adaptive text messages for 6 months Weekly drink-tracking texts Tailored messaging reduced weekly drinking more than drink tracking; adaptive messaging was slightly better early on. 2 Östh et al.(2025) RCT, 3-arm; N = 110 alcohol-dependent outpatients Breathalyser-coupled app; drink-counting app Treatment as usual (TAU) Heavy drinking days at 26 weeks lower with breathalyser (IRR 0.67, 95% CI 0.49 to 0.91); no effect of drink-counting app; no effects at 12 weeks 3 Kiluk et al.(2024) RCT, 3-arm; N = 99 (180 planned) CBT4CBT digital CBT with brief weekly clinical monitoring TAU; clinician-delivered CBT Percent days abstinent rose in all arms (digital 47.6 to 75.1%; TAU 49.3 to 69.3%; CBT 53.7 to 68.1%) with no clear between-arm difference. 4 Cunningham et al.(2024) RCT; N = 761; 81% follow-up Full-feature smartphone app Educational-only app Weekly consumption lower with full app (IRR 0.89, 0.80 to 0.98); no secondary outcome differences 5 So et al.(2025) RCT, multicentre, open-label Therapeutic app alongside clinic care Control group (details not retrieved) Heavy drinking days at week 12 favoured the app; HDD response 39.8% at week 12 in the app arm. 6 Gustafson, et al.(2024) Pilot RCT; 33 couples Couple-focused app for 6 months A-CHESS Fewer heavy drinking days not significant (beta -17.4; 95% CI -36.1 to 1.4; P = .07); app use declined over time 7 Suffoletto et al.(2025) Single-arm pilot; N = 8, then 37 (ages 18 to 25) GPT-4 motivational interviewing chatbot None Acceptable with high MI fidelity; efficacy not tested Opioids 8 Dhillon et al.(2026) RCT, non-inferiority Telemedicine follow-up after Day-1 in-person buprenorphine induction In-person induction over 3 days One-week retention 82.35% vs 75.71% (difference 6.64%; 95% CI -6.88 to 20.16); non-inferior on a 15% margin 9 Palacios et al.(2025) Stepped-wedge cluster RCT; 9 clinics, 8 clusters; 1,524 eligible patients Remote monitoring app in blended MOUD care Usual care before implementation 30-day retention 75.5% vs 63.6% (P < .001); model-based 74.8% vs 69.5% 10 Kawasaki et al.(2026) RCT; N = 100 in an opioid treatment program CBT plus CM mobile app TAU 24-week retention 70.8% vs 51.9% (OR 2.28; 0.99 to 5.24; P = .053); no significant difference in secondary outcomes 11 Gustafson et al.(2024) RCT, unblinded; N = 414 A-CHESS smartphone support plus MOUD MOUD alone No difference in illicit-opioid abstinence; greater counselling attendance and less emergency or urgent care use 12 Rossom et al.( 2025) Cluster RCT; 92 primary care clinics; N = 10,891 EHR clinical decision support Usual care More naloxone orders and MOUD prescription or referral (14.0% vs 9.4%; OR 1.48, 1.05 to 2.08); no effect on OUD diagnosis, MOUD days covered, overdose, or death 13 Marino et al.( 2024) Cohort; N = 600 (300 vs 300) App-based CM plus MOUD MOUD alone Retention 61.7% vs 39.3% (P < .001); fewer opioid-use days at end of treatment (8.4 vs 12.0); patients chose their arm 14 Vakkalanka et al.(2025) Retrospective cohort; US Veterans Health Administration Telehealth during buprenorphine care In-person care only Fewer ED visits (aHR 0.81, 0.77 to 0.85) and admissions (aHR 0.71, 0.67 to 0.76); lower mortality 15 Monico et al.(2026) Retrospective cohort; N = 22,064 Telehealth-only OUD program External national benchmarks 6-month retention 71.0% vs 25.9% (Medicaid) and 29.9% (commercial) benchmarks 16 Hammerslag et al.(2023) Cohort; N = 91,914 Telemedicine initiation of buprenorphine Non-telemedicine initiation Higher odds of retention; no change in odds of nonfatal overdose 17 Saad et al.(2025) Real-world comparison Telehealth buprenorphine visits In-person visits Retention comparable; lower transport and indirect costs with telehealth Stimulants 18 Hallgren et al.(2023) Single-arm pilot; N = 28 consented Smartphone CM with prompted substance tests (methamphetamine) None 15 completed the welcome phase; abstinent tests averaged 12% of prompted and 31% of completed tests 19 Forster et al.(2024) Case series Video sessions with self-administered oral fluid tests None Most participants joined sessions, tested and shared results; interpretation of near-cutoff results was unreliable. Cannabis 20 Mason et al.(2025) RCT; N = 1,078 young adults Text-based peer network counselling Waitlist No direct effect on cannabis use; indirect effects via readiness to change and protective strategies 21 Teeters et al.(2026) Phase II RCT, 3-arm; emerging adults Personalized feedback with or without interactive MI text messaging Control condition Driving after cannabis use reduced (feedback plus texts: IRR 0.34 at 3 months, 0.40 at 6 months; feedback alone: IRR 0.25 at 6 months); no effect on combined alcohol and cannabis driving. 22 Olthof et al.(2023) RCT, single-blind; N = 744 recruited Guided interactive digital CBT and MI intervention Non-interactive educational modules Both groups reduced use; no advantage for ICan (d = 0.15 for grams at 3 months); no change in treatment-seeking attitudes. 23 Coronado et al.(2025) Pilot RCT; N = 101; first-episode psychosis Cannabis harm-reduction app Control Retention and acceptability supported feasibility of a full trial 24 Coughlin et al.(2024) Micro-randomized pilot; 30 days; emerging adults Adaptive (reinforcement-learning) messaging Within-person no-message randomization Feasible and acceptable; messages increased proximal engagement Tobacco and nicotine 25 Luk et al.(2026) RCT; N = 590 recent quitters Mobile chat messaging for relapse prevention Text messaging 6-month biochemically validated abstinence 45.9% vs 35.5% 26 Cheung et al.(2025) RCT; N = 459 daily smokers unwilling to use aids Smartphone EMA profiling guiding telephone counselling and instant messaging EMA only Higher biochemically validated abstinence at 3 months 27 Jackson et al.(2024) Pragmatic RCT; N = 3,143; 41.9% effective follow-up Offer of the Smoke Free app Follow-up only No detectable benefit of the offer on 6-month continuous abstinence; among those who downloaded the app, 12.7% vs 7.0% (OR 1.80; 1.30 to 2.45) 28 Blomqvist et al.(2023) RCT; N = 1,012 12-week text messaging TAU (national quitline) 8-week prolonged abstinence at 3 months 32.9% vs 18.9% (OR 2.15; 1.51 to 3.06) 29 Graham et al.(2024) RCT; N = 1,503; 70.8% retained Automated text message program Assessment only 7-month abstinence 37.8% vs 28.0% (missing coded as vaping) 30 Heffner et al.(2026) Pilot RCT App plus text messaging Incentivized text control Self-reported cessation 12.5% vs 0% (preliminary) RCT, randomized controlled trial; TAU, treatment as usual; MOUD, medication for opioid use disorder; CM, contingency management; IRR, incidence rate ratio; OR, odds ratio; aHR, adjusted hazard ratio; HDD, heavy drinking days; EHR, electronic health record; ED, emergency department. 3.3 Risk of Bias Of the 22 randomized studies, 3 were judged at low risk of bias, 19 had some concerns, and none were at high risk. The main reasons for concern were reliance on self-reported outcomes in designs that cannot blind participants, open-label delivery, substantial loss to follow-up, and small or pilot samples. Studies with objective retention outcomes or biochemically validated abstinence fared better. None of the 8 non-randomized studies was at moderate risk: 4 were at serious risk, mainly from confounding in telehealth cohorts and patient-chosen arms, and 4 were at critical risk because they had no comparison group (Table 4). Table 4. Risk-of-bias judgments and principal reasons Study Tool Overall judgment Principal reasons Muench et al.(2024) RoB 2 Some concerns Self-reported drinking; participants cannot be blinded Östh et al.(2025) RoB 2 Some concerns Self-reported primary outcome by interview; low attrition; missing-data sensitivity analysis done Kiluk et al.(2024) RoB 2 Some concerns Underpowered (99 of 180); differing treatment completion; weekly testing supports verification Cunningham et al.(2024) RoB 2 Some concerns Self-reported consumption; 19% lost to follow-up So et al.(2025) RoB 2 Some concerns Open-label; subjective heavy drinking outcome Gustafson, et al.(2024) RoB 2 Some concerns Pilot with feasibility purpose; declining use Dhillon et al.(2026) RoB 2 Some concerns Open-label; objective retention outcome; wide non-inferiority interval Palacios et al.(2025) RoB 2 Some concerns Unequal exposure (1,205 vs 319); linkage analyses confounded; objective retention outcome Kawasaki et al.(2026) RoB 2 Some concerns Unblinded; objective retention; small sample Gustafson et al.(2024) RoB 2 Some concerns Unblinded; abstinence from urine screens plus verbal report Rossom et al.( 2025) RoB 2 Low Cluster randomization; objective EHR outcomes at scale Mason et al.(2025) RoB 2 Some concerns Waitlist control; self-reported use Teeters et al.(2026) RoB 2 Some concerns Self-reported behaviour; blinding not possible Olthof et al.(2023) RoB 2 Some concerns Single-blind; self-reported use; online recruitment Coronado et al.(2025) RoB 2 Some concerns Pilot with feasibility endpoints Coughlin et al.(2024) RoB 2 (adapted) Some concerns Micro-randomized pilot; short follow-up Tobacco and nicotine RoB 2 Low Biochemically validated primary outcome Luk et al.(2026) RoB 2 Low Biochemically validated primary outcome Cheung et al.(2025) RoB 2 Some concerns 41.9% effective follow-up; self-report; sensitivity analyses done Jackson et al.(2024) RoB 2 Some concerns Self-reported abstinence; attrition Blomqvist et al.(2023) RoB 2 Some concerns Self-report; 29% lost to follow-up; missing coded as vaping Graham et al.(2024) RoB 2 Some concerns Pilot with small sample Marino et al.( 2024) ROBINS-I Serious Patient-chosen arms; MOUD type differed between groups Vakkalanka et al.(2025) ROBINS-I Serious Retrospective; confounding by indication Hammerslag et al.(2023) ROBINS-I Serious Claims-based cohort; residual confounding Saad et al.(2025) ROBINS-I Serious Real-world allocation Monico et al.(2026) ROBINS-I Critical No concurrent comparison group Suffoletto et al.(2025) ROBINS-I Critical Single-arm pilot Hallgren et al.(2023) ROBINS-I Critical Single-arm, small sample Forster et al.(2024) ROBINS-I Critical Case series 3.4 Synthesis of Results The Findings from the included studies were synthesized by substance below Alcohol Of seven randomized alcohol studies, four reported positive and favoured the digital arm (Cunningham et al., 2024; Muench et al., 2024; Östh et al., 2025; So et al., 2025), and three were inconclusive (Gustafson, 2024; Gustafson et al., 2024; Kiluk et al., 2024). Benefits centred on monitoring with feedback and tailored, adaptive messaging. A breathalyser-coupled app lowered 26-week heavy drinking days in alcohol-dependent outpatients (IRR 0.67), while a drink-counting app did not, and no effects appeared at 12 weeks (Östh et al., 2025). A full-feature app modestly reduced weekly consumption compared with an educational app (IRR 0.89) (Cunningham et al., 2024), and tailored adaptive text messaging outperformed drink tracking in 723 help-seeking adults (Muench et al., 2024). A multicentre open-label trial from Japan supported a digital therapeutic added to clinic care (So et al., 2025). In treatment-seeking adults, digital CBT with brief monitoring improved abstinence over time, but all arms improved, and the trial was underpowered (Kiluk et al., 2024). A couple-focused app showed a non-significant trend with declining use (Gustafson et al., 2024). A GPT-4 counselling chatbot showed acceptability and motivational interviewing fidelity only (Suffoletto et al., 2025). Opioids Telehealth delivery of buprenorphine showed retention at least equal to in-person care. A non-inferiority RCT found telemedicine-assisted induction non-inferior for one-week retention (82.35% versus 75.71%) (Dhillon et al., 2026). Cohorts in Medicaid (Hammerslag et al., 2023), the Veterans Health Administration (Vakkalanka et al., 2025), a telehealth-only provider (Monico et al., 2026), and a real-world comparison (Saad et al., 2025) reported similar findings, with fewer emergency visits and admissions and lower mortality in the veterans cohort. Results for adjunct apps diverged. Recovery Connect improved 30-day retention in a stepped-wedge trial (Forster et al., 2024). reSET-O showed higher 24-week retention (70.8% versus 51.9%) that was not statistically significant (P = .053) (Kawasaki et al., 2026). A-CHESS did not improve illicit-opioid abstinence but increased counselling attendance and reduced emergency use (Gustafson, 2024). An app-based CM cohort showed better retention and fewer opioid-use days but with patient-chosen arms (Marino et al., 2024). A decision-support tool raised naloxone orders and MOUD prescribing or referral in 92 clinics without improving days covered, diagnoses, overdose, or death (Rossom et al., 2025). Stimulants A small single-arm pilot of smartphone CM for methamphetamine use found modest engagement, with 15 of 28 completing the welcome phase and abstinent tests averaging 12% of those prompted and 31% of those who completed (Hallgren et al., 2023). A telehealth CM case series showed that participants could join video sessions, self-administer oral fluid tests, and share results, but interpretation of results near assay cutoffs was unreliable (Forster et al., 2024). Cannabis Narrowly targeted interventions fared better than broad reduction programs. Personalized feedback with interactive text messaging reduced driving after cannabis use (IRR 0.34 at 3 months and 0.40 at 6 months), with no effect on combined alcohol and cannabis driving (Teeters et al., 2026). A large text-based treatment for young adults had no direct effect on use, only indirect effects through readiness to change and protective strategies (Mason et al., 2025), and the ICan intervention did not outperform education because both groups reduced use (Olthof et al., 2023). Pilots supported the feasibility of a harm-reduction app in first-episode psychosis (Coronado et al., 2025) and of adaptive reinforcement-learning messaging (Coughlin et al., 2024). Tobacco and nicotine This group showed the most consistent randomized benefit. Chat messaging raised 6-month biochemically validated abstinence in recent quitters (45.9% versus 35.5%) (Luk et al., 2026). Smartphone profiling to guide telephone counselling and messaging raised validated abstinence at 3 months among people unwilling to use cessation aids (Cheung et al., 2025). Twelve weeks of text messaging raised prolonged abstinence compared with quitline care (32.9% versus 18.9%) (Blomqvist et al., 2023), and an automated text program improved 7-month vaping abstinence in adolescents (37.8% versus 28.0%) (Graham et al., 2021). Offering the Smoke Free app showed no detectable overall benefit, with higher quit rates only among people who downloaded it (12.7% versus 7.0%), a comparison open to selection bias (Jackson et al., 2024). The ACT on Vaping pilot was preliminary (Heffner et al., 2026).

Conclusions:

Recent evidence indicates that digital health interventions are most useful when they deliver, extend, or strengthen established care rather than replace it. In answer to the review question, digital interventions improved substance use outcomes, treatment retention, or treatment receipt in some circumstances, but effects were inconsistent across intervention types and populations. In practice, digital health should therefore be integrated into established treatment pathways rather than viewed as a universal substitute for clinical care. Services can reasonably consider telehealth for buprenorphine initiation and follow-up, monitoring with clinician feedback as an adjunct to alcohol treatment, and interactive messaging for people seeking to stop smoking or vaping. However, the availability of an app or digital platform should not be interpreted as evidence of added clinical benefit over effective usual care. Treatment retention is an important outcome, but it should not be treated as a proxy for reduced substance use, sustained abstinence, or overdose prevention.