Decision-analytic models have an important role in healthcare funding decisions in the UK and internationally. However, errors have been reported in published models, which may indicate poor modelling practices, potentially leading to sub-optimal recommendations on cost-effectiveness. Little in-depth research has been undertaken to investigate the processes used by modellers in model development. The objective of this research was to explore the modelling methods used by modellers, with particular focus on problems encountered. This work involved two qualitative phases of research. In the first phase, twenty-four in-depth interviews with modellers were undertaken. Constant comparative analysis was used to compare informant practices, and identify common issues in model development. The second phase involved two separate model-building case studies with teams of modellers and clinicians. Methods of non-participant observation, qualitative interviews, and think-aloud were used to investigate model development. The findings of the case studies were compared using framework analysis. Important themes emerging from both phases of the research concerned the diversity of practices in structural development, problems with clinician involvement in modelling, and a lack of time and resources to carry out good practice methods. This work offers important recommendations for modelling practice, and suggestions for future research to improve modelling methods.