A hidden semi-Markov model is proposed to describe a trend signal for non-invasive mean blood pressure. Based on the model, a Viterbi beam search algorithm working with an adaptive cumulative sum test is proposed to detect change points and recognize change patterns online. Testing results on the simulated signals and clinical signals demonstrate that the algorithm has improved performance over the standard cumulative sum test for change detection, and has the potential to provide a clinically relevant description of trend changes.