Several Emergency Medical Service (EMS) agencies across the US have demonstrated that the time spent by EDs on ambulance diversion can be reduced by implementing community-wide policies that restrict the duration and frequency of diversion episodes. However, the mechanisms through which these reductions materialize are not well understood. EDs can respond to such restrictions by improving their patient flow processes to reduce crowding and, thereby reducing the need for frequent and prolonged diversion episodes. Alternatively, they can raise the diversion crowding-threshold, thereby tolerating a higher level of crowding. Paramedics -- who decide whether to comply with an ED's diversion signal by diverting the ambulance or not -- are likely to respond differently to these two strategies. We use the framework of strategic communication between a service provider and customers arriving to a queuing system and obtain differential hypotheses on two outcome variables (diversion probability and ambulance waiting time) depending on which of the above mechanisms are actually operative. We test these hypotheses using evidence from a community-wide intervention to reduce diversion in LA County, California. We estimate a binary choice model for the paramedics' diversion decision as well as a two-part model for ambulance waiting time using data on more than 45000 ambulance transports to a network of seven neighboring EDs for a period of seven years (2003-2009). Our results uncover a multifaceted impact of the policy intervention on relevant operational measures. Specifically, a relatively hands-off intervention that restricts the amount of time spent by EDs on diversion can induce them to improve patient flow processes. However, the intervention is likely to be less effective in reducing the fraction of diverted ambulances and the ambulance waiting time due to the complex network dynamics of ambulance diversion and the strategic signaling between paramedics and EDs.
|Original language||English (US)|
|Publisher||Social Science Research Network (SSRN)|
|Number of pages||28|
|State||Published - Sep 5 2014|