Design of a Model to Predict Surge Capacity Bottlenecks for Burn Mass Casualties at a Large Academic Medical Center
Published in: Prehospital and Disaster Medicine, v. 28, no. 1, Feb. 2013, p. 23-32
Posted on RAND.org on February 01, 2013
OBJECTIVES: To design and test a model to predict surge capacity bottlenecks at a large academic medical center in response to a mass-casualty incident (MCI) involving multiple burn victims. METHODS: Using the simulation software ProModel, a model of patient flow and anticipated resource use, according to principles of disaster management, was developed based upon historical data from the University Hospital of the University of Michigan Health System. Model inputs included: (a) age and weight distribution for casualties, and distribution of size and depth of burns; (b) rate of arrival of casualties to the hospital, and triage to ward or critical care settings; (c) eligibility for early discharge of non-MCI inpatients at time of MCI; (d) baseline occupancy of intensive care unit (ICU), surgical step-down, and ward; (e) staff availability—number of physicians, nurses, and respiratory therapists, and the expected ratio of each group to patients; (f) floor and operating room resources—anticipating the need for mechanical ventilators, burn care and surgical resources, blood products, and intravenous fluids; (g) average hospital length of stay and mortality rate for patients with inhalation injury and different size burns; and (h) average number of times that different size burns undergo surgery. Key model outputs include time to bottleneck for each limiting resource and average waiting time to hospital bed availability. RESULTS: Given base-case model assumptions (including 100 mass casualties with an inter-arrival rate to the hospital of one patient every three minutes), hospital utilization is constrained within the first 120 minutes to 21 casualties, due to the limited number of beds. The first bottleneck is attributable to exhausting critical care beds, followed by floor beds. Given this limitation in number of patients, the temporal order of the ensuing bottlenecks is as follows: Lactated Ringer's solution (4 h), silver sulfadiazine/Silvadene (6 h), albumin (48 h), thrombin topical (72 h), type AB packed red blood cells (76 h), silver dressing/Acticoat (100 h), bismuth tribromophenate/Xeroform (102 h), and gauze bandage rolls/Kerlix (168 h). The following items do not precipitate a bottleneck: ventilators, topical epinephrine, staplers, foams, antimicrobial non-adherent dressing/Telfa types A, B, or O blood. Nurse, respiratory therapist, and physician staffing does not induce bottlenecks. CONCLUSIONS: This model, and similar models for non-burn-related MCIs, can serve as a real-time estimation and management tool for hospital capacity in the setting of MCIs, and can inform supply decision support for disaster management.