What question did this study set out to answer?

The aim is to develop an AI platform that improves patient flow and reduces wait times in government healthcare systems.

May 19, 2026Open Access

QueueIQ: Real-Time Hospital Queue Intelligence Using LSTM Prediction, Multilingual NLP, and RAG-Powered Health Education

Key Points

The aim is to develop an AI platform that improves patient flow and reduces wait times in government healthcare systems.
Developed an ensemble model using LSTM and LightGBM for real-time queue predictions with 94% accuracy.
Utilized multilingual voice-based intake for symptoms in 12 Indian languages with an 8.9% word error rate.
Implemented emergency triage using gradient-boosted decision trees achieving 93% physician concordance.
Real-time queue prediction achieved 94% accuracy within a ±3 patient window.
Health education module based on WHO and ICMR knowledge attained 96% factual accuracy.
System maintained sub-2-second latency under 1,000 concurrent users, supporting efficient operations.

Abstract

QueueIQ is an AI-powered hospital queue intelligence platform designed to address critical patient flow challenges in Indian government healthcare systems, where wait times of 2–4 hours are common despite appointment systems. The platform integrates four core AI modules: (1) real-time queue prediction using an LSTM–LightGBM ensemble achieving 94% accuracy within a ±3 patient window; (2) multilingual voice-based symptom intake supporting 12 Indian languages via Whisper ASR and IndicTrans2, with 8.9% average Word Error Rate; (3) a RAG-powered health education module grounded in WHO and ICMR knowledge bases using Gemini 2.0, achieving 96% factual accuracy with full source attribution; and (4) emergency triage classification using gradient-boosted decision trees with 93% physician concordance. The system is built on a FastAPI + React stack with PostgreSQL, ChromaDB, Firebase, and Kubernetes, achieving sub-2-second P99 latency under 1,000 concurrent users. It is fully compliant with India's DPDP Act 2023 and ABDM standards, making it suitable for production deployment in government hospital infrastructure. Training data includes 4,320 hourly hospital arrival records from a Karnataka district hospital, a 4,400-sample multilingual clinical speech corpus across 12 Indian languages, and 4,200 curated medical knowledge chunks from 47 WHO guidelines and 18 ICMR protocols. Future work includes federated learning across hospital networks, EHR integration via ABDM FHIR APIs, wearable sensor fusion for real-time triage, and pilot deployment in 2–3 government hospitals.

QueueIQ: Real-Time Hospital Queue Intelligence Using LSTM Prediction, Multilingual NLP, and RAG-Powered Health Education

Key Points

Abstract

Cite This Study