CGRAs with high parallelization/customization capabilities are viable platforms for ever-evolving health-monitoring applications. We propose BioCare, an area/energy-efficient CGRA for health-monitoring edge devices, which exploits approximations across HW/SW stack while still provides acceptable final QoR. BioCare offers different energy-accuracy trade-off levels through the plasticity of its PEs, each can support functional-versatility and precision-adaptability with a SIMD manner. BioCare surpasses SoAs, by achieving up to 32%/67% area/energy savings and 3.6x higher throughput, when analysed on multiple widely-used ECG/EEG kernels/application. Our implementations will be open-sourced to springboard future research for bio-signal processing/reconfigurable/approximate computing communities.