Objective: To describe the patterns of patients' attendance over the first three quarters of a year under methadone maintenance treatment (MMT) clinics in Yunnan Province, China. Methods: Data were obtained from drug abuse treatment databases from five MMT clinics in Yunnan Province. Patients registered between April 2007 and December 2007 were included in the analysis. The study period was divided into three phases consisting of 3 months each. Logistic regression was used to predict factors determining whether the patients attended the clinic on each day with repetition of the patients taken into account. Results: The median number of days attending the clinics was 61 in the total treatment period, and the likelihood of a patient attending the clinic in the second and third phases was significantly less [odds ratio (OR) = .44 and .30, 95% confidence interval (CI) = .40-.49 and .26-.34] compared with the first phase. The predictors for attendance at clinic were being unmarried (OR = .63, 95% CI = .49-.82), self-employed (OR = .18, 95% CI = .11-.28), having a history of syringe sharing (OR = .67, 95% CI = .48-.92), and having been in a detoxification program (OR = 1.59, 95% CI = 1.09-2.33). Conclusion: Heroin users attended MMT programs irregularly and their periods of attendance significantly declined by duration of the time they stayed in the treatment program. Additional interventions should be employed to help patients adhere to this long-term treatment program.