Medication change is very important to know the medical history of a patient. Most of the clinical notes are in unstructured format and in addition to that due to its narrative nature expert human annotators are needed to interpret the events, which is quite expensive. In this work, we present an end-to-end model for the task of automatic extracting and classifying the medication change events from a clinical note. We propose a joint learning model trained with adaptive sample weighting loss which incorporates the use of clinical contextual embedding and static embeddings. Our proposed system obtained competitive performance on CMED dataset (n2c2 challenge 2022) for contextual medical event detection and classification.