Temporal data mining is becoming an important tool for health care providers and decision makers. The capability of handling and analyzing complex multivariate data may allow to extract useful information coming from the day-by-day activity of health care organizations as well as from patients monitoring. In this paper we review the main approaches presented in the literature to mine biomedical time sequences and we present a novel approach able to deal with "point-like" and "interval-like" events. The methods is described and the results obtained on two clinical data sets are shown.