The emergence of the Internet of Things (IoT) and the increasing number of cheap medical devices enable geographically distributed healthcare ecosystems of various stakeholders. Such ecosystems contain different application scenarios, e.g., (mobile) patient monitoring using various vital parameters such as heart rate signals. The increasing number of data producers and the transfer of data between medical stakeholders introduce several challenges to the data processing environment, e.g., heterogeneity and distribution of computing and data, lowlatency processing, as well as data security and privacy. Current approaches propose cloud-based solutions introducing latency bottlenecks and high risks for companies dealing with sensitive patient data. In this paper, we address the challenges of medical IoT applications by proposing an end-to-end patient monitoring application that includes NebulaStream as the data processing system, an easy-to-use UI that provides ad-hoc views on the available vital parameters, and the integration of ML models to enable predictions on the patients' health state. Using our end-to-end solution, we implement a real-world patient monitoring scenario for hemodynamic and pulmonary decompensations, which are dynamic and life-threatening deteriorations of lung and cardiovascular functions. Our application provides ad-hoc views of the vital parameters and derived decompensation severity scores with continuous updates on the latest data readings to support timely decision-making by physicians. Furthermore, we envision the infrastructure of an IoT ecosystem for a multi-hospital scenario that enables geo-distributed medical participants to contribute data to the application in a secure, private, and timely manner.