On Dedicated versus Pooled Service in the Presence of Triage Errors

Many service systems, such as emergency departments (EDs), differentiate among customer's types to meet specific service level (SL) targets. To meet these targets, some EDs serve customers using dedicated resources (e.g., in different zones) while others use pooled ones. We study this choice using stylized queueing models with two customer's types where the decision maker minimizes capacity while meeting SL targets. We compare the performances of two dedicated single-server systems with these of a pooled server that prioritizing the high priority customers or serving all customers in a first- come-first-serve (FCFS) fashion. We consider the servers', customers', and system's perspectives. We explain how the optimal system design depends on the system's parameters, including SL targets and workloads. Moreover, when a service system cannot observe customer's types, it triages customers to service classes. However, in many settings, this triage process is prone to errors. We highlight that triage errors change the pattern of workload arriving to servers and the pattern of service to customers. We show that such errors significantly impact the optimal dedicated systems, slightly impact the optimal priority system, and have no impact on the FCFS system. Finally, while, as expected, triage errors typically harm performance, we demonstrate that errors may improve performance (in comparison to the no error case) from all three perspectives simultaneously: servers may face a lower utilization (even though one capacity increases), customers are served faster (one type is strictly faster), and costs are lower.