An Early Warning System For Risk Management
isk management in healthcare has solved a wide range of healthcare-related issues in Saudi Arabia. However, the limitation of risk management teams working under special conditions (needing to solve critical health-related issues) has highlighted the urgent need for an early risk warning system (ERWS) in healthcare. The influences of changing weather conditions demand that diabetic patients and doctors in Saudi Arabia have a continuous check on health conditions. The number of diabetic patients is increasing rapidly in Saudi Arabia. Hence, risk management teams in healthcare must be supported with a system that alerts to changes before the changes become a significant risk/problem. Our proposed approach does the following: 1) predicts changes in BP and blood sugar level within hospital environment at runtime. 2) Continually checks patient health status with respect to health condition at runtime. 3) Alerts to the changes as detected (e.g. risk or unknown parameter), and also provides feedback for patient and doctor. We present a computational model that defines the interaction and communication of the system components and describes the prediction and checking process in our proposed approach. We designed the architecture for our proposed approach with respect to the computational model. The thesis proposes an early risk warning system approach, which predicts and checks patient health conditions with respect to the ideal conditions according to medical standards. The health status of a patient will be communicated to doctors and patients on an emergency note if the predicted values are outside normal conditions. In this way, the risk can be mitigated before the occurrence of damage to patient health at runtime. To implement the proposed approach, neural networks is used for developing the prediction component using Java programming. The results of this research successfully predicted the health condition of a patient by checking outputs against medical standards. The risks defined in this research include hyperglycaemia, hypoglycaemia, hypertension and hypotension. Appropriate results were obtained for almost every patient when checked with four input parameters for 200 patients. Consistent results were produced by the risk prediction component and the alerts were generated after every five (5) seconds to communicate to the patients and doctors at runtime. Health status of all 200 patients can also be seen to check the changes in health conditions in the hospital environment. Finally, a case study with different scenarios based on changes in patient health status with respect to ideal conditions revealed evaluated the approach.
- PhD