Establishing the therapy of the juvenile diabetic patient at onset is a challenge. We propose a new way for building medical guideline to support physicians. The course of the treatment is mainly described by the sequence of insulin dosage. However the daily insulin dose is prescribed based on the glycemia levels from the previous day/days and on the other hand is verified by the glycemia levels in the following day. To generate medical guidelines we discover sequential patterns from the treatment sequences and supplement them with patterns of medical examinations and interventions. Before mining sequential patterns we group the sequences with respect to patient's medical data influencing the course of the disease by applying k-means clustering.