Apparatus and associated methods relate to invoking an alert based upon a behavior of a patient as determined by a machine-learning model operating on a video stream of the patient. Video data, audio data, and semantic text data are extracted from a video stream of the patient. The video data are analyzed to identify first, second, and third features sets of video, audio, and semantic text features, respectively, which have been identified by a computer-implemented machine-learning engine as being indicative of at least one of a set of alerting behaviors corresponding to a patient classification of the patient. Using a computer-implemented machine-learning model, a patient behavior of the patient is determined based on the first, second, and/or third features sets. The patient's behavior is compared with the set of alerting behaviors, and, when the patient's behavior is determined to be included therein, the alert is automatically invoked.