Background: In 2004, the Health Resources and Services Administration (HRSA) required U.S. hospitals to build surge capacity systems that permit “the triage, treatment and initial stabilization of 500 adult and pediatric patients per 1,000,000 awardee jurisdiction,” a mandate that is interpreted to mean 500 additional beds per million population. We modeled the relationship between surge capacity beds and surge treatment capacity in order to clarify this preparedness goal.

Methods: We developed a Markovian simulation model of emergency hospital activity based on a modified Open Jackson Network framework. All patients are admitted through an emergency department (ED), from which they can be routed to an observation area, hospital floor, intensive care unit (ICU), or comfort care (CC) area based on triage category. Acceptable patient movements include ICU->Floor->home or ->CC and Floor->home, ->ICU, or ->CC. The ED station operates as a standard M/M/s queuing process, but all transfers to subsequent stations are limited by an imposed restriction on queue length of ~1, meaning that patients are not transferred unless a bed is available. Inputs include length of stay (LOS) at each station, routing probabilities, and available resources. The model first iteratively derives the optimal distribution of resources among stations given these constraints and then simulates patient flow through the whole hospital for a specified response period. The model does consider bed cleaning or any other impediment to rapid bed turnover and was created in Microsoft Excel with a Visual Basic macro running the simulation.

Results: We ran 100 iterations of the model for two LOS scenarios (ED=6h/3h, Floor=168h/84h, ICU=48h/24h, and CC=12h/6h) and two triage scenarios (“mild”=10% to ICU immediately and “severe” with 30% to ICU immediately) for a 500-bed hospital. Overall bed utilization ranged from 92.0% for fast/”mild” to 76.6% for slow/”severe.”

Conclusion: Number of surge capacity beds does not equal number of patients served, even in an ideal model that permits immediate filling of free beds. Mismatch between the stochastic processes of station-specific LOS, influenced by both length of stay and type of patient routing through the hospital, among other factors, accounts for at least part of this underutilization. These results have important implications for regional hospital preparedness planning, suggesting that substantially more than 500 beds are needed to provide emergency care for 500 individuals.