Optimizing hospital patient flow and resource allocation
To improve quality of care and relieve clinicians and staff, hospital operations need to be coordinated and optimized across all services in real-time. In this project, we first develop and implement predictive machine learning models to predict patient flows at a major hospital (Bertsimas, Pauphilet, Stevens and Tandon). Then, we integrate these predictions into an overall bed recommendation engine, that unifies the bed assignment process across the entire hospital, and accounts for request and availability of beds in all units, currently and through the rest of the day. On simulations, we demonstrate that our optimization model can substantially reduce off-service placements and delays (Bertsimas and Pauphilet). We now integrate this patient flow optimization model into a broader resource allocation model to account for appointment and surgery scheduling and doctors and nurses staffing as well (Bertsimas, Na, and Pauphilet). Altogether, our frameworks offer a unique opportunity to revitalize healthcare delivery with optimization and data at the core.
This project connects to SDG 3, Good Health and Well-being.
Researchers
- Dimitris Bertsimas (MIT, US)
- Liangyuan Na (MIT, US)
- Jean Pauphilet (London Business School, UK)