OBJECTIVES: Modelling-based economic evaluations have been widely used to inform healthcare decision making. There are diverse modelling approaches each with merits and challenges. This study compared three modelling approaches, cohort Markov model (CM), Microsimulation (MS) and Discrete Event Simulation (DES) in assessing treatments for chronic hepatitis C (HCV) infection.   METHODS: Using our published CM model of long-term HCV treatment as the base-case model, we developed MS and DES models. Model structures were framed as similar to each other as possible in terms of time horizon, discounting and clinical pathway, while providing details to demonstrate the distinctive features of each model. We measured the impact of cohort heterogeneity and intermediate events relevant to model conceptualization and validity. We did not consider resource constraints, dynamic population or interactions within the system. The models were developed and analysed with TreeAge Pro 2015. RESULTS: In terms of model construction, CM required relatively more simplifying assumptions which might be potential sources of bias and threats to validity. CM also needed more health states which had to be organized sequentially giving rise to possible deviation from clinical reality. On the other hand, MS and DES were statistically intensive which would increase the possibilities of technical errors.  Absolute values of cost and quality adjusted life years were similar among the three models. However, relative rankings based on incremental cost effectiveness ratios of alternative treatments varied. There appeared to be a positive relationship between number of evaluated interventions and difference in results. Increasing the complexity of clinical pathway increased the likelihood of generating different outputs by the three models.   CONCLUSIONS: Our study showed that the choice of modelling method would affect model outputs leading to different conclusions and recommendations. Limitations of various modeling methods and contextual issues need to be considered to improve validity and usefulness of model results.