Recently, the health care industry has observed a growing demand of continuous monitoring of patients to timely diagnose the diseases. This is attributed to a remarkable growth in uncertain deaths caused by diseases like heart attack and malignant neoplasm. While the existing sensor network technology has been envisaged as a solution to the real time patient monitoring, the focus of such efforts has been to deploy a sensing network and passively deliver the physiological readings from patients to the hospital infrastructures. In this paper, we introduce a framework that addresses a drawback in existing experimental test beds: the lack of feedback from the diagnosis infrastructure to the sensor network so that the network can reconfigure itself to monitor the patient in a more useful manner, while concomitantly maximizing the network reliability, throughput and life.