Hospital Length-of-Stay Prediction

Hospitals run on predictions: how long will each patient stay, and how many will arrive tomorrow? At HealthcareLogic I designed and developed a machine learning model for hospital length-of-stay (LoS) prediction.

• LoS model - outperformed the existing model by at least 47% in a baseline scenario, enabling more precise allocation of beds and clinical resources.
• Arrival-rate model - a patient arrival rate prediction model that improved operational efficiency and resource management.
• Rigorous evaluation - experimented, analysed and evaluated several machine learning models for regression to select the most reliable predictor.

Proprietary system - architecture and results shared at a high level; details available on request.