Abstract:
HealthSync was built with Next.js 15, React 19, PostgreSQL and Prisma ORM. The system has been designed in such a way that they have four distinct user roles such as Super Administrator, Clinic Administrator, Healthcare Provider, and Patient. The roles offer a tailored dashboard display and distinct workflows that are user specific. Moreover, the system also enforces strict control over access and authorization using Role-Based Access Control (RBAC) and JWT Authentication. To achieve these two, HealthSync will be in a position to integrate directly with the Fitbits of patients through OAuth 2.0 to get the real-time and continuous vital signs information. It is through this connection that we are able to get minute-by-minute heart rate, total daily steps taken, sleep patterns, peripheral oxygen saturation (SpO2), Heart Rate Variability (HRV), and respiratory rate. After collection of the data, it is processed by a three-phase AI Engine. Phase 1 uses rule-based constraints depending on the age and medical history of the patient and the medication he/she has been prescribed now. Phase 2 performs a statistical analysis to determine the trend of the resting heart rate identification, determines abnormal data changes by calculating Z-score, and traces the data by relying on the calculation of a linear regression. Phase 3 runs the Cerebras Llama 3.3-70B Large Language Model (LLM) to generate clear and concise medical summaries, assess the risk to health of the identified problem(s), and provide actionable next-steps. To get access to continuous and real-time vital sign readings of patients, HealthSync can directly interoperate with the Fitbits of patients via OAuth 2.0. It is through this relationship that we are able to gather minute by minute heart rate, total daily steps taken, sleep patterns. After collecting the data, it is processed by a three-phase AI Engine. Phase 3 applies the Cerebras Llama 3.3-70B Large Language Model (LLM) to generate clear and concise medical summaries, assess the risk to health presented by the detected problem(s), and suggests actionable next steps.
