The prevalence of schizophrenia varies between 0.2-1% and the cost to treat patients is substantial. Modeling a schizophrenic population is demanding since schizophrenia is a life long disease during which patients go through many states and the transition from one state to another is dependent on the history of patients. A previously built 1st order Monte Carlo DES model was adapted to enable 2nd order Monte Carlo simulation. OBJECTIVE: This abstract describes why and how the model was upgraded from a 1st order Monte Carlo (MC) simulation to a 2nd order MC model. The abstract will describe the choices made in the design of the model, the internal validity and evaluation of its strengths and weaknesses. METHODS: Internal validity of the model has been explored using data from several patient databases, as well as literature on a list of variables, including PANSS, the proportion of patients institutionalized and costs. Pert, beta, lognormal and uniform distributions have been used to describe 2nd order uncertainty of relevant variables, such as PANSS, QALY, risk and costs. RESULTS: The model was programmed to reflect the PANSS at 0 year, 1 year, and 5 years for patients from the considered databases. The modeled annual cost per patient and the location distribution were similar to published data. Outcomes were expressed in terms of direct medical costs, number and duration of episodes, PANSS, QALY, GAF, CGI, SF36 and the SF6 mental component. The uncertainty surrounding the outcomes of costs and effect measures were assessed with acceptability curves and ellipses. DISCUSSION: The original DES model was vastly improved with the use of additional database analyses, additional correlation analyses and 2nd order Monte Carlo simulations. This has resulted in less emphasis on expert opinion, yielding a partially validated probabilistic model which can be adapted for numerous health care settings.