The increasing use of social media allows the extraction of valuable information to early prevent some risks. Such is the case of the use of blogs to early detect people with signs of depression. In order to address this problem, we describe k-temporal variation of terms (k-TVT), a method which uses the variation of vocabulary along the different time steps as concept space to represent the documents. An interesting particularity of this approach is the possibility of setting a parameter (the k value) depending on the urgency (earliness) level required to detect the risky (depressed) cases. Results on the early detection of depression data set from eRisk 2017 seem to confirm the robustness of k-TVT for different urgency levels using SVM as classifier. Besides, some recent results on an extension of this collection would confirm the effectiveness of k-TVT as one of the state-of-the-art methods for early depression detection.XVI Workshop Bases de Datos y Minería de Datos.Red de Universidades con Carreras en Informática