Published models for schizophrenia in literature are generally Markov models, which entail many methodological disadvantages. OBJECTIVE: To build a disease progression model for schizophrenia that circumvents the problems of Markov modelling. The model is designed to simulate individual histories of schizophrenic patients from the age of onset until death, but it can easily be adapted to shorter time horizons. A patient enters the model when he visits the psychiatrist because he suffers from a relapse. At each visit the psychiatrist will re-evaluate the visiting scheme, the patient's treatment and his location. Furthermore, the psychiatrist will estimate the patient's severity of disease (PANSS, QALY, Danger). At any time the patient can decide to be non-compliant. METHODS: The model is characterised by four steps: 1) definition and description of patient characteristics, 2) calculation of the progression of schizophrenia for that individual patient in each of the compared treatment strategies, 3) calculation of the costs and effects generated in each of the strategies; and 4) when previous steps are performed for a specific patient population, calculation of the average medical and economical outcomes per strategy. RESULTS: The model is able to simulate individual patient histories with patient-specific probabilities. Unlike Markov models, we do not assume a direct link between a health state and costs. Moreover, we individualise the relations between what happens to a patient and the expected costs and effects. Also, we take account of unobserved heterogeneity. CONCLUSIONS: The outlined model is complex and needs much input. Moreover, many estimates are surrounded by uncertainties. However, it needs to be stressed that simpler models implicitly need the same input, but they hide assumptions that are explicit in this model. Therefore, the presented model is subtler, more transparent and closer to clinical practice.