OBJECTIVES: To assess the bias introduced to absolute costs, absolute QALYs and the incremental cost effectiveness ratio (ICER) associated with Markov models, compared with discrete event simulation (DES) models. To investigate how such biases are a function of cycle length and half-cycle correction. METHODS: A hypothetical three health state model was constructed using both Markovian and DES approaches. Costs and utility were assigned to each health state and the ICERs between two treatment strategies were estimated. Six Markov models using different cycle lengths (1 month, 3 month, 1 year), and with and without half cycle correction were constructed. Differences in the absolute costs and QALYs generated between each Markov model were compared with the DES approach and the ICERs generated by each model were compared. RESULTS: Markov model simulation was shown to introduce biases in the absolute costs and QALYs when compared with a DES approach. The bias was related to the duration of the time cycle with the results converging to the DES values as the time cycle was reduced. The initial bias in cost fell from 14% to less than 1%; QALY bias was consistently below 1%. The ICERs  show bias between 2.4% and 9.6% when using a 1 year cycle and between 0.6% - 5.4% when using a 1 month cycle. The half-cycle correction reduced absolute bias between 2% - 10%, the ICERs were not affected. The time cycle duration was the primary parameter in reducing bias. CONCLUSIONS: Markov models introduce bias due to the simplifying assumptions of fixed cycle length and half cycle correction; DES models do not suffer the same biases. It is suggested that when the ICERs produced are close to  the Willingness to Pay threshold, Markov models should be analyzed with shorter cycle length or a DES approach adopted to ensure conclusions are robust.