We present our participation in Task~2 of the 2014 ShARe/CLEF eHealth Evaluation Lab.We used a hybrid approach based on a rule-based system and supervised classifiers depending on the properties of the attributes. The rule-based system identified course, severity and body location attributes based on the annotations of the training set and resources obtained from the UMLS.The Heideltime system was used to identify the dates. A MaxEnt model was trained to detect negation and uncertainty based on the disorder and surrounding words. A Decision Tree detected the relation to document time based on the position of the disorder in the document and on the words in the current sentence.Our system obtained a global 5th position out of ten ranked teams (accuracy of 0.804), and ranked 2nd for the detection of the relation to document time (accuracy of 0.322).