OBJECTIVES: Health care claims data offer the possibility to track patients over extended time periods. However, most widely used data mining techniques cannot incorporate the temporal component of this data. To analyse health care patterns over time, the need for sequence analysis arises. It is especially important to (a) detect and analyze event sequences like therapy lines to (b) identify common patterns and (c) visualize patient or treatment behavior. METHODS: We present an innovative approach to infer therapy lines from prescription data. Using a sliding window approach we extract disease and therapy states per patient. The resulting event sequences are further analyzed with an approach originally developed in the field of computational biology. We adjusted the Smith-Waterman local alignment algorithm to meet the needs of claims data. Using the resulting pairwise distance measure, we performed clustering to group similar treatment sequences and find and visualize the most common sequence per cluster. RESULTS: The resulting sequence analysis method is exemplarily applied to multiple sclerosis (MS) therapy lines. In the first stage of clustering, we obtained 7 different characteristic therapy lines for MS patients. For 78% of sequences we could identify a clean cluster mapping, with the highest fraction in mono-therapies (34% IFN-β-1a, 22% Glatiramer acetate and 17% IFN-β-1b) and a switch from IFN-β-1a to Dimethyl fumarate in 5% of cases. Additionally, we could identify different health conditions and costs associated with each therapy cluster. CONCLUSIONS: This work outlines a clustering approach for the analysis of sequential patterns in claims data. The sequence analysis method allows for new insights into the behavior of a heterogenic population over time. Finally, the visualization of treatment patterns enables a better understanding of treatment reality.