Improving treatment engagement for substance use disorders by leveraging digital interventions post-emergency care

Background
Substance use disorders (SUD) are common, preventable disorders characterized by persistent substance use despite adverse consequences. Although few affected individuals engage in treatment, many present for acute and hospital care for issues related directly or indirectly to their substance use. As such, acute and hospital care constitutes a critical healthcare access point, which ideally could link individuals with sustained treatment. Nonetheless, the treatment gap persists, with few individuals pursuing post-acute SUD treatment, resulting in missed opportunities to prevent disease progression, increase remission and, by extension reduce mortality. In the current thesis, the overall aim was to contribute actionable insights on how to increase post-acute treatment engagement for SUD in routine healthcare, with a special emphasis on digital interventions as a potential bridge to treatment.

Methods
To address our aim, we conducted four studies within a multi-method framework, all in preparation for developing and evaluating a novel digital aftercare guide titled the BRIDGE. The four studies utilized cohort (Study I), qualitative (Study II), secondary analyses, and psychometric (Study IV) study designs to address knowledge gaps in prior literature. Our practical objectives were to inform intervention development (Study II and III) and outcome selection (Study I and IV) for the intervention study. For studies I (n = 9,771) and II (n = 23), as well as the subsequent intervention study, participants were sampled from an emergency department (ED) specifically dedicated to the care of SUD, the Stockholm SUD-ED. Study III and IV included participants seeking a digital intervention for their alcohol use (n = 1,169) and study III additionally included online help-seekers for regular cannabis use (n = 303). In study IV, data was collected from two further sources: k = 15 brief intervention trials and k = 20,000 simulated samples based on general-population statistics. To examine post- acute treatment engagement and subgroup heterogeneity, we conducted systematic care-flow mapping and multinomial logistic regression modelling in Study I. Study Il analyzed patients' perceived treatment barriers and views on digital aftercare within a reflexive thematic analysis framework. Study III used arithmetic means and parametric confidence intervals to analyze between- group (cross-substance) differences in attitudes, and k-means clustering to explore within-group heterogeneity. In Study IV, we applied several psychometric techniques (including Pearson correlation and non-parametric item response theory) to examine the internal structure of AUDIT-C in clinical trials and simulated samples. We additionally utilized quantile regression to examine right censoring at baseline and simple slopes analysis to investigate responsiveness to change.

Results
Findings from Study I and II suggested that a purpose-built ED for SUD (the Stockholm SUD-ED) can achieve relatively high rates of post-acute treatment (50.1%) compared to prior literature, yet several preventable barriers persist even within this model. As shown in Study I, SUD-ED patients arriving by law- enforcement, and with probable milder levels of SUD, are subgroups with lower odds of engaging in post-acute treatment, signaling the need for both targeted efforts and broadening existing aftercare approaches.

Importantly, patient interviews in Study II revealed overall sentiments of having benefited from digital aftercare guides, should such options have been available for them alongside regular aftercare. Particularly, the opportunity for self- directed support was highlighted as a desirable feature of digital interventions. Study III revealed key attitudinal differences toward digital interventions, between and within groups of participants with alcohol and cannabis use. Together, these findings signal the need for flexible digital interventions.

Study IV revealed that most brief intervention trials sampled (68%) exhibited non-positive associations between AUDIT-C's frequency and quantity items, a sign shift in the frequency-quantity association that simulations indicated were due to selection on problematic alcohol use. Non-parametric item response theory analyses further demonstrated that these non-positive correlations disrupt ordinal measurement. Furthermore, responsiveness analyses demonstrated that a one-unit AUDIT-Ct2-ti change represented greater average SUt2-ti at higher versus lower baseline-consumption levels. Based on these findings, we recommend caution in repurposing the AUDIT-C screening measure for outcome monitoring in populations with problematic alcohol use. Within the context of this specific thesis, this led us to exclude AUDIT-C from the list of potential outcome measures for the main intervention study.

Building on these studies, we developed a three-module digital aftercare guide, which was provided to SUD-ED patients during approximately fourteen months, ending patient recruitment on January 31, 2025. By that date, n = 195 users had created accounts for the guide, and the landing page of the aftercare guide had received 2,989 page views. A retrospective research evaluation is scheduled to commence in autumn 2025.

Conclusions
Several barriers to post-acute treatment persist, even within a fully integrated SUD-ED model. Patients with indicators of lower SUD severity had lower odds of engaging in post-acute treatment. Aftercare options may need to be broadened to better engage this group and improve outcomes. Digital interventions may serve this purpose. Preliminary numbers from the clinic suggest that a proportion of patients are interested in using the guide. A natural next step is to evaluate the effectiveness of digital aftercare guides in increasing engagement in post-acute treatment. To improve the quality of evidence, both these intervention studies and the broader SUD field should exclude AUDIT-C from the set of outcome measures.

List of scientific papers

I. Romero, D., Kåberg, M., Carlbring, P., Berman, A., Franck, J., & Lindner, P. Care pathways and predictors of post-acute outpatient treatment engagement among individuals visiting an emergency department dedicated to substance use disorders: A cohort study. [Manuscript]

II. Romero, D., Rozental, A., Carlbring, P., Johansson, M., Franck, J., Berman, A. H., & Lindner, P. (2024). From alcohol detoxification to treatment: A qualitative interview study on perceived barriers and assessed potential of mHealth among individuals postdetoxification. Psychology of Addictive Behaviors, 38(8), 879- 890. https://doi.org/10.1037/adb0001008

III. Romero, D., Johansson, M., Hermansson, U., & Lindner, P. (2021). Impact of users' attitudes toward anonymous internet interventions for cannabis vs. alcohol use: A secondary analysis of data from two clinical trials. Frontiers in Psychiatry, 12, 730153- 730153. https://doi.org/10.3389/fpsyt.2021.730153

IV. Romero, D., Johansson, M., Berman, A. H., & Lindner, P. (2025). Questionable generalizability of AUDIT-C scoring warrants caution when used for outcome monitoring: Evidence from simulated and real-world trial data. Addiction. Advance online publication. https://doi.org/10.1111/add.70074