Online System for Primary Care to Prevent and Address Teen Substance Use

The primary goals of this project are to develop and test an innovative teen substance use (SU) module in CHADIS integrating the cSBI questionnaire and decision support into CHADIS along with teen strength and goal data to be called CHADIS-cSBI. The CHADIS-cSBI will facilitate evidence-based SU screening, severity assessment and initiation of primary care management and care coordination for any referrals to reduce morbidity, mortality and costs of SU. Inclusion of strength and goal data is anticipated to foster the doctor-teen relationship and make SU advice more effective.

Primary Research Questions

1. Of those who reported any substance use at their annual well-visit, is there a lower rate of reported use at 3- and 12- months follow-up among those exposed to CHADIS-cSBI as compared to those who received treatment as usual?
2. Of those who reported no substance use at their annual well-visit, is there a lower rate of reported use at 3- and 12- months follow-up among those exposed to CHADIS-cSBI as compared to those who received treatment as usual? Secondary Research Question

1. Among all patients, is there are lower rate of reported driving after using or riding with a driver who has been using substances (risky driving/riding) at 3- and 12-months follow-up among those exposed to CHADIS-cSBI as compared to those who received treatment as usual? We will employ a cluster randomized two-group pre-test-post-test design, clustered at the provider level. Given the nature of the enhanced care and the measurement it is not possible to randomize at the patient level, and it is not necessary to randomize at the clinic level. The use of a cluster-randomized design adds complexity, and generally requires larger samples to achieve sufficient statistical power as compared to simple randomized control trials. Sample size estimates consider the design effect resulting from a cluster randomized design as well as other assumptions.

A total of 40 PCPs reporting at least 2 teen well visits per week, either in person or by telemedicine visit, will be recruited and randomly assigned in equal numbers to the enhanced care and standard care arms of the study. We will randomize clinicians stratified by whether they have completed Adolescent Medicine fellowships, as those with Adolescent Medicine training have been shown to have higher rates of addressing substance use in their practice. The authors' prior experience showed that, in a trial where the patients were randomized, individual PCPs being trained to deliver the counseling did not give similar levels of counseling to intervention and control patients. Their studies also show wide variation in patient-reported receipt of counseling across participating providers; e.g., within the intervention group, there was a range from 65%-93% across providers within one site. Such heterogeneity within a site reduces concern about contamination of practice across providers. All 12-18 year old patients in the practice will complete pre-visit screens dependent on the randomization of their own attributed PCP. While there are, in fact, three levels (i.e., patient -> practitioner -> practice), we confine the fundamental design herein to two levels (i.e., patient -> practitioner). Intra-class correlations will be used to adjust sample size estimates during power analyses, and also used to account for the covariance that is presumed to exist wherein patients from the same cluster (i.e., PCP) are likely to be more similar to each other than to patients from other clusters. The study design is depicted by the following research design notation figure: