OBJECTIVES: To assess the published literature on the clinical drivers and design of health economic models in schizophrenia, as well as utility mapping algorithms, in order to support the construction of health economic models.

METHODS: As part of a systematic literature review, studies describing health economic models in schizophrenia were searched in MEDLINE. A targeted literature review was also conducted regarding utility mapping algorithms in schizophrenia. Both literature searches were conducted on December 1, 2015.

RESULTS: Eighty-nine studies, from 483 screened abstracts, were included in the assessment of models. Models found were published from the early 1990s. Markov models were the most prevalent modelling method, followed by decision tree models, then simulation models, though models of the last category were more frequent in more recent studies. Of the assessed models, 37.8% used quality-adjusted life years (QALYs), and an additional 8.1% used disability-adjusted life years (DALYs) as benefit measures. The rest of the models used various clinical endpoints as value drivers, for example relapses, annual number of stable days, or percentage of patients in remission. The targeted literature review identified four main approaches for utility mapping algorithms, mainly based on Positive and Negative Syndrome Scale (PANSS) scores: the Mohr-Lenert, Chouinard, Siani, and Briggs approaches and their sub-types. These methods showed various levels of detail, for example Mohr and Lenert categorizing categorized patients into three or eight different health states, while the Chouinard approach only used three health states.

CONCLUSIONS: Health economic models developed for schizophrenia showed great variability, with simulation models becoming more frequently used in the last decade. While nearly half of models used QALYs or DALYs for value drivers, the others used a great variety of clinical outcomes. Several methods to convert PANSS scores to utility values were also identified and assessed.