Patient Transfer Under Ambulance Offload Delays: An Approximate Dynamic Programming Approach

Ambulance offload delay (AOD) occurs when emergency medical services (EMS) transferring a patient to a busy hospital emergency department (ED). It is a crucial bottleneck for patient cares on both patient and ambulance sides that delays the transfer of a patient from an ambulance to the emergency department, which will negatively influence the care quality and efficiency. This paper formulates the AOD problem as a Markov decision process (MDP) and develops an approximate dynamic programming (ADP) approach to overcome the curse of dimensionality. We formulate the problem based on the post-decision states and derive two approximate dynamic programming solution algorithms using linear regression and neural network frameworks. The numerical result shows that the transfer policy obtained from our solution approach outperforms the myopic policy, which always transfers patients to the closest hospital. Our findings suggest that our approach can effectively alleviate the AOD problem, improve health care quality, and improve the overall system efficiency.