Clinic Queue

Local healthcare clinics face significant operational friction due to information asymmetry in waiting rooms. Paper token systems and verbal calling lead to ambient noise, missing patients, and high anxiety regarding unknown wait times. Traditional WebSocket or heavy cloud-database solutions introduce unwanted cost and setup complexity for local small-scale medical centers. The objective was to build a zero-hardware, instant-sync digital queue matrix that eliminates verbal shouting and delivers transparent, data-driven wait-time forecasting. Process I architected a dual-screen web app using Python and Streamlit, backing state via an atomic JSON storage engine. Initially, I calculated wait times using a simple positional index formula, but discovered a critical edge case: the next patient in line read an inaccurate 0-minute wait time despite the active doctor consultation block. I updated the model to shift calculations dynamically to include ongoing visits. To solve the morning data cold-start problem, I built a hybrid lifecycle pipeline. The application boots up utilizing a receptionist-adjusted manual baseline (Tokens 0-3) and transitions entirely to an autonomous closed-loop state machine powered by a 3-point rolling moving average telemetry engine from Token 4 onward. Results The dual-screen queue prototype completely eliminates information asymmetry by displaying real-time wait telemetry on a 1-second auto-polling loop. Operational overhead for receptionists dropped to a single click, allowing patients to be checked in or called immediately. Most importantly, the hybrid time-series algorithm achieves fully autonomous wait-time prediction within 4 patient cycles, rendering completely data-driven clinic lines. Reflection While file-based state storage using an atomic JSON system works excellently for an isolated single-doctor prototype, it presents a horizontal scaling ceiling. If expanding to a multi-doctor hospital ecosystem with high concurrent usage, write-collisions or file-locking latency could occur. Next time, I would decouple the state layer entirely by moving the queue pipeline onto an in-memory database like Redis or utilizing row-locked SQLite tables to handle high concurrent traffic seamlessly.