In clinical treatment, a variety of data is collected to assess a patient's condition and determine necessary interventions. The common practice is to periodically observe a selection of features and compare these measurements to defined thresholds. The question arose, if there is a benefit in considering not only the newest updates of features, but taking prior observations into account as well. To examine possible advantages, a dataset at the Lorenz Böhler intensive care unit was collected, where the progression over time of biomarker levels of trauma patients were documented. Motivated by a patient monitoring, this work deals with the problem of classification for this longitudinal data. The dataset demands a classification based on knowledge of a positive class only, hence it was approached by a one class classification. Further, the classifier has to deal with both, the unbalanced data as well as its updating nature. Based on a linear mixed effects regression model, characteristics of the class of survival patients are estimated. Deviations of observation from this estimations are penalized and evoke a negative classification. The mixed model approach allows not only for estimating class characteristics, but further features, e.g. it can expand the classification by a visual support. The gold standard method of brain injury assessment for trauma patients is used to benchmark the proposed longitudinal classifiers. Different views on evaluation show slight improvements to the benchmark, which however are in conflict with additional effort.