AI-Based Clinical Handover Assistant: Enhancing Patient Safety through Intelligent Clinical Communication

Shift-to-shift handover is one of the most error-prone moments in hospital care. In most hospitals it is still performed through verbal briefings, handwritten notes or free-text electronic health record (EHR) entries, which is time-consuming, inconsistent and rarely restricted to the nurse who actually receives the patient. This paper presents the AI-Based Clinical Handover Assistant, a secure hybrid web and mobile system that automates nurse-to-nurse handover documentation, with escalation to a doctor when the receiving nurse cannot manage a case. An outgoing nurse records a voice or text handover addressed to specific incoming nurses; the audio is normalised with FFmpeg, transcribed by a speech-to-text service and summarised by the Gemini API into a structured SBAR/ISBAR report from which medications, vital observations and pending tasks are extracted. Every record is encrypted before storage and can be decrypted only by its addressed recipients; on escalation it is re-encrypted for the chosen doctor. A retrieval-augmented generation (RAG) chatbot lets authorised users query their own handovers. The system is built with Django, MySQL and Flutter. We present the literature survey, system architecture and data flow design of the proposed system.