OBJECTIVES To describe the patterns of medication discontinuation in a retiree population. Premature discontinuation of medications adversely affects patients' outcomes and may require use of additional healthcare resources. METHODS The study included pharmacy claims from a retirement system for the period January 2000-September 2005. The unit of analysis was the course of drug therapy (CDT), representing a unique combination of a patient and a drug product (i.e. generic name, formulation and strength). CDTs initiated between August 1, 2000 and July 31, 2001 and discontinued before March 1, 2005 were included in the analysis. Days in therapy for each CDT were calculated as the difference between the date of the first and last prescription of the CDT. RESULTS The study included 1.1 million CDTs representing 5.9 million claims. 37.0% of CDTs were discontinued with less than a month in therapy, 50.0% with less than 6 months, and 76.0% within one year. Maintenance therapy comprised 740,788 CDTs (70.30%) of which 27.0% had a single claim. Maintenance therapies had the following cumulative utilization patterns: 36.7% CDTs discontinued in less than 3 months of therapy, 45.0% in less than 6 months, 74.7% within one year, and 87.2% within two years. 21.6% of non-maintenance CDTs were continued for more than a year and 11.8% for more than 2 years. CONCLUSIONS Premature discontinuation of therapy intended for long-term use is highly prevalent with more than one-fourth of all maintenance therapies discontinued at the first prescription, and nearly three-fourths discontinued within the fist year of therapy. In the other hand, over one-fifth of non-maintenance therapies were used for over a year. The assessment of compliance using claims data should account for discontinuation of therapy prior to the potential manifestation of positive patient outcomes and for the short-term usage of maintenance therapies and prolonged use of non-maintenance therapies.