EAGER: Advanced Capacity Allocation Methodology: Time-sensitive Appointments in Congested Service Systems

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

Healthcare and other appointment-based service industries tend to struggle with long waits for service visits. It is increasingly important to model pathways, or itineraries, of patient visits over time from various service providers. Current practice emphasizes the basic first-come-first-served rule when setting appointment dates. This fails to enable appointment setting that gives relatively lesser waits to relatively more urgent patients. This EArly-concept Grant for Exploratory Research (EAGER) project addresses this gap by creating methods that allow organizations to service multiple classes of patient types while offering each the expectation that their waiting time will be close to some target level (with high probability) that is appropriate to the patient's urgency. In settings with service resources shared in common, the methods created will optimize trade-offs involving the utilization of key resource, staff overtime, volumes of services fulfilled, and waiting times for appointments. The potential impact of these new methods includes improved cost control through efficiency, better ability to coordinate a patient's care across providers over time, and improved health outcomes as a result of more timely visits. The work will help broaden participation of underrepresented groups in research and improve the content of engineering courses. The research will address the above challenges surrounding appointment scheduling with emphasis on optimization based planning models for resource allocation. The methods will optimize complex admission control policies for multi-class stochastic queueing network models. Itineraries of care involving multiple visits from multiple service types over time will be treated to increase the value and relevance of the models. The research advances approaches that include mixed integer programming optimization methods that exploit effective linear approximations to model controlled multi-class queueing networks. The approaches will estimate performance metrics such as means, variances, and delay constraint violation probabilities. The methods will optimize the admission control plans over a finite or an infinite time horizon. The approach addresses the above challenges with new methods which can treat realistic problem features that include, but are not limited to (1) the incorporation of historical system data, (2) itinerary flowtime metric models allowing for service bundles and stochastic visit itinerary processes over time, and (3) linear approximations of key performance metrics such as the means and standard deviations of waiting times and delay constraint violation probabilities.