Brief Addiction Monitor-Taiwan version as a simple tool for the measurement-based care of people who use drugs in the era of harm reduction: predictive validity in prospective follow-ups of patients nationwide

The Brief Addiction Monitor (BAM) is a simple tool for the periodic assessment of substance use disorders, but whether it can be applied in non-US settings is unknown. This study aimed to evaluate the reliability of a translated version of the BAM, named the BAM-Taiwan (BAM-T), and examine whether the baseline BAM-T score was associated with being lost to follow-up or with changes in the score at 3 months and 6 months, respectively. A total of 2336 participants recruited from people who use drugs enrolled in treatment programs nationwide were followed up at 3 months and 6 months. Within this cohort, intraclass correlation reliability (ICCR) of interview responses was used to assess interrater reliability in 127 participants and test–retest reliability in 102 participants. To evaluate predictive validity, we examined associations between baseline BAM-T total score and follow-up outcomes, including loss to follow-up and subsequent BAM-T severity, using multivariable regression models. To account for informative loss to follow-up, stabilized inverse probability–weighted (IPW) linear regression models were applied. The BAM-T total score demonstrated excellent interrater (ICCR = 0.855) and 1-week test–retest reliability (ICCR = 0.789). Higher baseline BAM-T total score was associated with increased odds of loss to follow-up and with higher BAM-T total scores at both 3-month and 6-month follow-up. These associations remained robust in IPW-weighted analyses accounting for informative loss to follow-up. Each 1-SD increase in baseline BAM-T total score (7.9 points) was associated with a 5.21-point (robust SE = 0.18) higher BAM-T total score at 3 months (95% CI, 4.85–5.57; p < 0.0001) and a 4.95-point (robust SE = 0.21) higher score at 6 months (95% CI, 4.54–5.36; p < 0.0001) after adjustment for key sociodemographic and clinical covariates. Our results provide empirical support for the utility of BAM-T as a simple tool for the measurement-based care of patients who use various types of drugs. Not applicable.