AI and Peer Coaching for Opioid Use Disorder

Overseen By Babak Toighi, MSc, MD

Age: 18+

Sex: Any

Trial Phase: Academic

Sponsor: Friends Research Institute, Inc.

Must be taking: Buprenorphine

Must not be taking: OUD medications

Disqualifiers: Disability, Medications for OUD, others

Compensation: Varies

Number of Visits: Unknown

292 Participants Needed

Trial Summary

Will I have to stop taking my current medications?

The trial does not specify if you need to stop taking your current medications, but you cannot participate if you've received medications for opioid use disorder in the past 30 days.

What data supports the effectiveness of the AI and Peer Coaching treatment for opioid use disorder?

Research shows that AI can effectively predict opioid use disorder and help manage treatment by analyzing healthcare data, which may improve patient outcomes. Additionally, AI-driven interventions can predict stress and drug cravings, potentially allowing for timely support and reducing relapse risk. ^{1 2 3 4 5}

Is AI and Peer Coaching for Opioid Use Disorder safe for humans?

The research articles reviewed do not provide specific safety data for AI and Peer Coaching interventions for opioid use disorder, but one study evaluated the safety of a digital therapeutic used alongside medication for opioid use disorder, suggesting that digital interventions can be safely integrated into treatment. ^{1 3 4 6 7}

How does the AI and Peer Coaching treatment for opioid use disorder differ from other treatments?

This treatment is unique because it combines artificial intelligence (AI) with peer coaching to help manage opioid use disorder. AI can predict when a person might experience cravings or stress, allowing for timely interventions, while peer coaching provides support from individuals who have experienced similar challenges. ^{1 2 3 5 6}

What is the purpose of this trial?

Black and Latinx people who use opioids are disproportionately impacted by opioid overdose deaths. The proposed study assesses the efficacy of an open source, multimodal artificial intelligence-driven texting tool combined with peer recovery coach-supported text contact that delivers social services, stigma reduction, health habitus, and patient navigation content addressing social determinants of health to enhance receipt of buprenorphine in primary care among emergency department-enrolled Black / Latinx people who use opioids.

Eligibility Criteria

This trial is for Black and Latinx individuals over 18 years old who use opioids, speak English or Spanish, want to start buprenorphine treatment in primary care, have a positive opioid test result, and plan to stay in NYC for at least a year. They must be interested in using their mobile phone with data for the study.

Inclusion Criteria

I haven't used opioids without a prescription in the last 30 days.

I have given my informed consent.

I will be staying in NYC for 12 months or more.

Timeline

Screening

Participants are screened for eligibility to participate in the trial

2-4 weeks

Intervention

Participants receive AI-driven SDH-enhanced text messages, with or without Peer Recovery Coach support, to enhance the receipt of buprenorphine

26 weeks

Ongoing virtual interactions

Follow-up

Participants are monitored for the durability of treatment effect and social services received

26 weeks

Treatment Details

Interventions

• AI and Peer Coaching

Trial Overview The study tests an AI-driven texting tool combined with peer recovery coach support against AI-only texts and usual treatments. It aims to improve access to buprenorphine treatment among emergency department-enrolled Black/Latinx opioid users by addressing social health factors.

Participant Groups

Group I: Intervention Arm-2: AI driven SDH-enhanced text only Experimental Treatment 1 Intervention

Participants in this arm will receive AI-driven SDH-enhanced text messages to enhance the receipt of buprenorphine in Black/ Latin people who use opioids. Unlike Arm-1, this intervention does not include the additional support and coordination provided by Peer Recovery Coaches.

Group II: Intervention Arm-1: PRC supported text+ AI driven SDH-enhanced text Experimental Treatment 1 Intervention

Participants in this arm will receive a combination of Peer Recovery Coaches (PRCs) supported text-based care/services coordination alongside AI-driven SDH-enhanced text messages to enhance the receipt of buprenorphine in Black/ Latin people who use opioids.

Group III: Control Arm 3- Treatment as Usual Placebo Group 1 Intervention

Control Arm-3 will receive treatment as usual (i.e., verbal instructions, NYC Dept of Health pamphlets detailing access to OUD and social services, health system smartphone application EMR patient portal).

Who Is Running the Clinical Trial?

Friends Research Institute, Inc.

Lead Sponsor

Trials

60

Recruited

22,500+

NYU Langone Health

Collaborator

Trials

1,431

Recruited

838,000+

Nathan Kline Institute for Psychiatric Research

Collaborator

Trials

40

Recruited

3,300+

Weill Medical College of Cornell University

Collaborator

Trials

1,103

Recruited

1,157,000+

Findings from Research

A review of gray literature identified 29 unique AI interventions aimed at addressing opioid use disorder (OUD), including smartphone apps and healthcare data-related tools, highlighting the growing interest in using technology to improve OUD care.

While many of these interventions have not been rigorously tested in clinical trials, some have shown potential in identifying patterns of opioid use and detecting overdose risks, suggesting they could play a significant role in managing OUD in the future.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Artificial intelligence interventions focused on opioid use disorders: A review of the gray literature. Beaulieu, T., Knight, R., Nolan, S., et al.[2021]

A study involving 189 outpatients with opioid-use disorder used a randomForest algorithm to predict cravings for heroin and cocaine, achieving high overall accuracy (up to 0.93) in predicting when cravings would not occur, but lower accuracy in predicting actual cravings.

The findings suggest that while passive monitoring through GPS data can provide some predictive insights, accurately predicting subtle events like cravings or stress may require active user input, as high overall accuracy can sometimes hide the prevalence of false alarms.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Prediction of stress and drug craving ninety minutes in the future with passively collected GPS data. Epstein, DH., Tyburski, M., Kowalczyk, WJ., et al.[2023]

A study involving data from 474,208 patients over 10 years demonstrated that an AI-based model can predict opioid use disorder (OUD) more effectively than traditional clinical tools, achieving an area under the curve (AUC) of 0.742.

The AI model outperformed various other predictive models, including logistic regression and random forest, suggesting that integrating AI into clinical practice could enhance risk assessment and management for patients on opioid therapy.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

A Comparative Effectiveness Study on Opioid Use Disorder Prediction Using Artificial Intelligence and Existing Risk Models. Fouladvand, S., Talbert, J., Dwoskin, LP., et al.[2023]

An automated risk-modeling framework was developed using data from 40,005 appointments of 2,742 patients in a medication for opioid use disorder (MOUD) program to predict opioid abstinence and medication adherence at the next appointment.

The XGBoost machine-learning model demonstrated similar accuracy (88%) to traditional logistic regression models but provided better calibration of risk estimates, highlighting its potential as a valuable tool for improving treatment adherence in MOUD patients.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Using Machine Learning to Predict Treatment Adherence in Patients on Medication for Opioid Use Disorder. Burgess-Hull, AJ., Brooks, C., Epstein, DH., et al.[2023]

Computer-assisted learning programs for addiction treatment can be highly adaptive, providing personalized feedback to users, which may enhance their effectiveness in promoting new learning behaviors.

Two specific programs targeting heroin relapse prevention in abstinent prisoners and promoting controlled drinking in alcohol abusers represent innovative approaches to substance abuse treatment, potentially reaching individuals who prefer self-help resources over formal therapy.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Computer-assisted drug prevention. Barber, JG.[2019]

An AI-based digital phenotype using social media language was found to be more effective than traditional psychometric assessments in predicting treatment outcomes for substance use disorder, based on a study of 269 patients.

The deep learning model (BERT) successfully identified high-risk individuals for treatment dropout, achieving a high accuracy (AUC = 0.81), suggesting that social media analysis could be a valuable tool for assessing risk in substance use treatment.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

AI-based analysis of social media language predicts addiction treatment dropout at 90 days. Curtis, B., Giorgi, S., Ungar, L., et al.[2023]

In a study of 170 adults with opioid use disorder, those receiving a digital therapeutic alongside standard treatment had significantly higher rates of opioid abstinence (77.3% vs. 62.1%) during the final weeks of treatment.

The digital therapeutic also improved treatment retention, with participants less likely to leave treatment compared to those receiving standard care alone, while showing no increase in adverse events, indicating it is a safe and effective addition to traditional therapies.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Safety and efficacy of a prescription digital therapeutic as an adjunct to buprenorphine for treatment of opioid use disorder. Maricich, YA., Bickel, WK., Marsch, LA., et al.[2022]

References

1. England pubmed.ncbi.nlm.nih.gov

Artificial intelligence interventions focused on opioid use disorders: A review of the gray literature. [2021]

2. England pubmed.ncbi.nlm.nih.gov

Prediction of stress and drug craving ninety minutes in the future with passively collected GPS data. [2023]

3. United States pubmed.ncbi.nlm.nih.gov

A Comparative Effectiveness Study on Opioid Use Disorder Prediction Using Artificial Intelligence and Existing Risk Models. [2023]

4. Netherlands pubmed.ncbi.nlm.nih.gov

Using Machine Learning to Predict Treatment Adherence in Patients on Medication for Opioid Use Disorder. [2023]

5. United States pubmed.ncbi.nlm.nih.gov

Computer-assisted drug prevention. [2019]

6. England pubmed.ncbi.nlm.nih.gov

AI-based analysis of social media language predicts addiction treatment dropout at 90 days. [2023]

7. England pubmed.ncbi.nlm.nih.gov

Safety and efficacy of a prescription digital therapeutic as an adjunct to buprenorphine for treatment of opioid use disorder. [2022]