Journal of clinical epidemiology

A comprehensive model of treatment participation in chronic disease allowed prediction of opioid substitution treatment participation in Zurich, 1992-2012.

PMID 26073899


Chronic diseases are often associated with cycling in and out of treatment. We used data of a large opioid substitution treatment case register to (1) identify associated factors and (2) integrate retention and readmission into a model of overall participation over subsequent treatment episodes of various groups. Data of all 9,407 patients undergoing 26,545 methadone or buprenorphine substitution treatment episodes between 1992 and 2012 in the canton of Zurich, Switzerland, were analyzed. We used extended survival analysis to estimate the duration of, and time between, treatment episodes, with the number of episodes, gender, nationality, administration route, age at onset of first regular heroin use, and provider type as independent variables. A similar analysis was applied to estimate overall participation (the probability of being in treatment at a given day after first entry independent of current number of treatment episode) and to test for group differences. The time between treatment episodes shortened with the increasing number of episodes. Retention slightly increased after the first episode and then shortened for later treatment episodes. Effect sizes were generally rather weak (odds ratio ≤ 1.47). Effects were usually equal for all episodes, and if changing, weakened for later episodes. The complex process of leaving and entering treatment as well as the daily probability of being in treatment independent of treatment episode can be predicted by comprehensible statistical models applied to patient-period data sets.