Optimizing Emergency Supply for Mass Events

Abstract

In this work we provide an approach for simulating and optimizing the emergency supply in a run-up to a mass event.For a given set of hospitals, transport vehicles and injured suffering from common injuries, our algorithm simulates the workload of provided transport and medication capacities, e.g., doctors. In addition to standard methods, our algorithm considers how a patient's individual waiting time until medication impacts the corresponding course of disease. We use Simulated Annealing with transition probabilities favoring a balanced workload of vehicles and doctors as optimization strategy. We show that using this strategy speeds up convergence and leads to better results compared to Greedy and standard Simulated Annealing with an underlying equal transition probability. Finally, we briefly discuss how different initialization strategies affect the performance of the provided algorithm.

Online strategies for optimizing medical supply in disaster scenarios

“…The subsequent explanations follow our past work [7]. The main difference between the data setup used there and the present work is that we now assume casualties to arrive in small groups at different points in time and not as a whole at one specific point in time.…”

“…In this context the goal is to allocate a given number of ambulances among a set of bases to minimize the fraction of calls that are not reached within a time standard. An adjusted version of Simulated Annealing to solve the same assignment problem, but with the assumption that the total set of casualties is known beforehand has been applied in [7]. We will use the adjusted Simulated Annealing approach in this work to test whether a significant result improvement can be achieved when applying it on our online assignment problem.…”

“…We will use the adjusted Simulated Annealing approach in this work to test whether a significant result improvement can be achieved when applying it on our online assignment problem. The crucial difference between [7] and this work is that we want to provide a tool that can be employed during the chaos phase instantly after a major incident has occurred. In this setting the total number of casualties is unknown and the decision about assignment of a casualty has to be taken instantaneously with incomplete information.…”

“…In this setting the total number of casualties is unknown and the decision about assignment of a casualty has to be taken instantaneously with incomplete information. However, in [7] the introduced techniques are used as a tool for planning security measures in advance to a mass event and thus before a (possible) incident and with complete information. We are not aware of approaches similar to our that improve care of the injured in disasters through a mapping to online Job Shop Scheduling problems.…”

“…For modeling the health situation of a casualty we adopt the concept of penalty functions from [7], since ordinary static triage concepts like in [4] which are mainly used in practice do not take into account that casualty states of health can change over time (usually gets worse) when the affected person is not medically treated immediately. As in [7] we define 23 different penalty functions which differ both in their initial priority (t = 0) and in the positive functional course.…”