MO3 COMPARING METHODS OF DATA SYNTHESIS: UPDATING PARAMETERS OF AN EXISTING PROBABILISTIC COST-EFFECTIVENESS MODEL

Abstract

MO1 ADVANTAGES AND DISADVANTAGES OF MARKOV MODELS VERSUS DISCRETE EVENT MODELS FOR SIMULATING PATIENT FLOWS Al MJ,Welte R Erasmus MC, Rotterdam,The Netherlands, GlaxoSmithKline, Munich, Germany OBJECTIVES: To compare discrete event simulation (DES) models with Markov models for simulating patient flows in an intensive care unit (ICU). METHODS: We developed a Markov model and a DES model to simulate the patient flow in an ICU with a focus on mechanical ventilation (MV) on the basis of the Dutch UltiSAFE trial. In UltiSAFE, different sedation regimens (remifentanil-based sedation vs. conventional sedation) in critically ill patients requiring mechanical ventilation were investigated. We defined 8 states in the models: MV—maintenance, MV—eligible for weaning, MV—actual weaning, MV—eligible for extubation, ICU—extubated, ICU—eligible for discharge, Discharged from ICU, and Dead. Patients move through states 1 to 7 in sequence, unless they die. In the Markov model, a time cycle of 1 hour was applied. We compared the models with respect to results (length of stay in each phase), ease of scenario analysis, stability of results and computation time. RESULTS: The results of both models were very similar. However, differences were found for all times between the eligible and actual event, due to the fact that this time was 0 for around 30% of patients. The Markov model requires a patient to stay in a state for at least the cycle length (1 hour), whereas DES allows instantaneous transitions. The DES model enabled rapid post-hoc subgroup analysis whereas multiple runs of the Markov model were required for this type of analysis. Results were more stable in the Markov model than in the DES model. The computation for both the baseline results and for the probabilistic sensitivity analyses took much less time for the DES model. CONCLUSIONS: DES may be the more (time) efficient model in situations where time to event is essential such as patient flows in an ICU.