Overview and Purpose
Illicit drug use remains a global health challenge, with millions seeking effective treatments. Digital psychosocial interventions have emerged as scalable options to reach diverse populations, offering flexible access via apps, websites, and messaging. This study conducts a systematic review and meta-analysis to quantify dropout rates in digital interventions for adults with illicit drug use and to identify factors that predict attrition. Understanding these patterns is essential to improve engagement, adherence, and ultimately treatment outcomes.
What Was Studied and Why It Matters
Dropout undermines the evidentiary value of trials and safety of patients, especially in digital formats where engagement can waver. This review differentiates three concepts: engagement (behavior during use), adherence (alignment with intervention expectations), and dropout (leaving or being lost to follow-up before outcome assessment). By focusing on dropout in randomized controlled trials, the analysis aims to guide design choices that minimize attrition and tailor interventions to at-risk groups.
Methods at a Glance
Following Cochrane and PRISMA guidelines, the team registered the protocol in PROSPERO (CRD42024534389). A broad search across five databases (Web of Science, PubMed, PsycINFO, Embase, Cochrane Central) identified randomized trials of digital psychosocial interventions for adults (18+) with illicit drug use. Eligible studies reported dropout counts and total participants, enabling calculation of dropout proportions for intervention and control groups at posttreatment and longest follow-up. Data extraction and bias assessment were performed by independent reviewers, with ROB 2.0 used to gauge study quality.
Key Findings: Dropout Rates and Comparisons
The meta-analysis included 41 studies with 9,693 participants, spanning 82 intervention groups and 48 dropout data points. At posttreatment, the pooled dropout rate in the digital intervention arm was 22%, compared with 26% in control conditions. At the longest follow-up, the intervention group showed a dropout rate of 28.2% versus 27.8% in controls. Heterogeneity was substantial in both timeframes (I² well above 90%), underscoring variability across trials, populations, and intervention designs.
Patterns by Time Point
Posttreatment dropout was influenced by a mix of demographic, clinical, and intervention-related factors. Higher employment (a proxy for life demands) was surprisingly associated with greater attrition. More frequent intervention contact (higher frequency) tended to reduce dropout, suggesting that ongoing engagement supports retention. Clinically, baseline diagnoses and cocaine use at baseline predicted higher dropout risk, highlighting substance-specific adherence challenges.
During the longest follow-up, being single emerged as a protective factor in some analyses, while higher baseline drug use frequency increased dropout risk. Recruitment source mattered as well: web-based recruitment linked to higher dropout, whereas campus-based recruitment showed lower dropout. The degree of digitalization (fully digital vs. partially digital) yielded inconsistent results, potentially affected by reporting gaps.
Clinical and Design Implications
Several actionable insights emerge for researchers and clinicians designing digital psychosocial interventions:
– Increase contact frequency judiciously to strengthen the therapeutic alliance and reduce dropout.
– Tailor interventions for high-risk subgroups, such as individuals with baseline cocaine use or co-occurring mental health conditions.
– Consider recruitment strategies that combine online reach with stable, structured settings (e.g., campuses) to improve retention.
– Improve reporting of intervention characteristics (digitalization level, human support) to enable clearer guidance on what keeps participants engaged.
Limitations and Future Directions
Despite rigorous methods, the analysis faced notable limitations: high heterogeneity, incomplete reporting on digital features and dropout reasons, and potential publication bias. Future work should prioritize standardized reporting, preregistration, and sharing of participant-level data. Incorporating machine learning could help predict dropout risk, while design innovations—such as gamification and adaptive feedback—may boost engagement and adherence over time.
Conclusion
Digital psychosocial interventions hold promise for adults with illicit drug use, with a modestly lower short-term dropout rate than traditional modalities. However, attrition remains a major challenge, especially over longer follow-ups and among certain subpopulations. By integrating multidimensional predictors into trial design and prioritizing consistent reporting, researchers can optimize digital tools to sustain engagement, improve adherence, and enhance real-world outcomes in substance use treatment.
