Clinical and health care processes are often summarised through sequences of ordinal data describing patient’s state over time. Identifying patterns in these sequences can provide valuable insights into patient progression trajectories for the purposes of clinical monitoring and quality assurance. However, both the variation in the length of each sequence and the ordinal nature of observable states present challenges to pattern identification. In this paper, we address these challenges by presenting a novel measure of dissimilarity for comparing two or more variable-length ordinal sequences that can be used in conjunction with conventional clustering methods to identify patterns in patient progression trajectories. We provide practical guidance on how this can be achieved, and demonstrate it in the context of identifying patterns in post-stoke recovery trajectories.