This paper presents an adaptive system for post-discharge physiological monitoring that addresses the challenges of heterogeneous and imperfect data collected in home environments. The proposed framework integrates sensing layer adaptation with a case-based reasoning algorithm recommender (CBR-AR) to enable context-aware selection of data analysis strategies. Instead of relying on a fixed processing pipeline, the system characterizes incoming physiological data and dynamically selects appropriate preprocessing and detection methods based on prior cases stored in a meta-dataset. Rough set theory is incorporated to determine attribute importance and improve similarity-based case retrieval under uncertainty. A system architecture is developed to connect wearable sensing, mobile data management, and cloud-based analytical modules. The sensing layer includes adaptive mechanisms for signal prioritization and sampling rate control, allowing the system to balance data quality, energy consumption, and bandwidth constraints. The analysis layer operationalizes adaptive strategy selection through the CBR-AR framework and executes the selected methods for abnormality detection. A case study using ECG data demonstrates the workflow of the system, where an LSTM-based model is selected and applied to detect arrhythmia-related patterns. The study serves as a functional illustration of how the system adapts to specific data conditions rather. The proposed approach provides a foundation for future integration with higher-level reasoning to assist clinical decision-making in post-discharge care.