Visit-level clinical correlates of next-visit heroin-positive urine in methadone maintenance treatment: A repeated-events analysis from Taiwan

INTRODUCTION: Clinicians providing methadone maintenance treatment (MMT) care must repeatedly decide, at each visit, whether a given patient faces heightened short-term risk of imminent heroin use. However, research has incompletely characterized visit-level correlates of near-term heroin-positive urine, particularly when the analysis explicitly considers the variable interval between visits. The present study characterized within-patient, visit-level clinical correlates of next-visit heroin-positive urine in MMT care. METHODS: A retrospective longitudinal design drew on 1585 visit-level events from the MMT program at a tertiary referral hospital in Taiwan (January 2023-December 2024). The primary analysis used conditional logistic regression with subject-level stratification and focused on three clinically accessible index-visit variables-past-month self-reported heroin use, methadone dose (per 10-mg increment), and depressed mood severity (BSRS-5 q4). A sensitivity analysis used subject-stratified Cox proportional-hazards models, with inter-visit interval as the time scale. Supplementary Material presents descriptive clinical profiles. RESULTS: Among 1585 visit-level events, 195 (12.3%) preceded a heroin-positive urine result at the next visit. In the adjusted conditional logistic regression, each 10 mg increase in methadone dose was associated with lower odds of heroin-positive urine (aOR = 0.72; 95% CI, 0.57-0.91; p < 0.001), and past-month self-reported heroin use was associated with higher odds (aOR = 2.19; 95% CI, 1.35-3.53; p = 0.001). Depressed mood showed a graded risk-increasing pattern, which adjustment attenuated in the primary model. In the subject-stratified Cox sensitivity analysis, the depressed mood association became clearer: mild, aHR = 1.95 (95% CI, 1.05-3.62; p = 0.035); moderate-or-above, aHR = 4.10 (95% CI, 1.49-11.28; p = 0.006). CONCLUSIONS: In this MMT cohort, recent heroin use, lower methadone dose, and depressed mood were complementary short-term correlates of next-visit heroin-positive urine, with the mood association becoming more evident once the model explicitly incorporated inter-visit interval. These three clinically accessible variables may support short-term risk recognition during MMT encounters. These findings remain exploratory and require external validation before any formal clinical prediction use.