Decision Analysis for Transporting Critically Ill Patients with Cardiovascular Diseases
Abstract
Transporting critically ill patients with cardiovascular disease is often difficult due to patient’s health condition and the remoteness of the cardiac centers. In countries with large territories, cardiovascular patients often require transportation to specialized clinics located in large cities. However, transportation should satisfy specific medical conditions, and the decision on the transportation mode may not be made quickly. We collected statistics from the National Coordination Center for Emergency Medicine (NCCEM) of Kazakhstan and experts in the field on five alternative modes of transport: airplane, helicopter, ambulance, train, and private clinical cars. Our model is based on a multiattribute utility (MAU) theory and considers the latter two alternatives in addition to those used in NCCEM. The novelty of our study is that it employs a MAU function U(X1, X2, X3) that captures the decision maker’s preferences and uses three main attributes: transportation cost saving (X1), transportation time saving (X2), and the health effect of transportation (X3). For the latter, we recommend using an internationally recognized scoring system (APACHE II) to assess patients' health status rather than a triage system (red/yellow/green) currently used in the country. APACHE II has a larger range (0-71) and gives more flexibility, but is more complex in assessment. An anesthesiologist with many years of experience provided an assessment for our model. The assessment showed utility interdependence among the attributes. The model ranked the alternatives in the following order: (1) airplane, (2) helicopter, (3) ambulance, (4) clinical cars, and (5) train. In practice, ambulances and clinical cars can be used over distances of up to 200 km (~124 miles). Finally, we compared the results of our model with the model based on the assumption of utility independence among the attributes (i.e., multilinear form) as well as with the ranking of the alternatives based on only one attribute. To illustrate the use of our model, we presented two cases. The model introduced in this paper can be adapted for use in other large countries.