Decision models in type 2 diabetes mellitus: A systematic review

Aims To reduce the burden of type 2 diabetes (T2DM), the disease decision model plays a vital role in supporting decision-making. Currently, there is no comprehensive summary and assessment of the existing decision models for T2DM. The objective of this review is to provide an overview of the characteristics and capabilities of published decision models for T2DM. We also discuss which models are suitable for different study demands. Materials and methods Four databases (PubMed, Web of Science, Embase, and the Cochrane Library) were electronically searched for papers published from inception to August 2020. Search terms were: "Diabetes-Mellitus, Type 2", "cost-utility", "quality-of-life", and "decision model". Reference lists of the included studies were manually searched. Two reviewers independently screened the titles and abstracts following the inclusion and exclusion criteria. If there was insufficient information to include or exclude a study, then a full-text version was sought. The extracted information included basic information, study details, population characteristics, basic modeling methodologies, model structure, and data inputs for the included applications, model outcomes, model validation, and uncertainty. Results Fourteen unique decision models for T2DM were identified. Markov chains and risk equations were utilized by four and three models, respectively. Three models utilized both. Except for the Archimedes model, all other models (n = 13) implemented an annual cycle length. The time horizon of most models was flexible. Fourteen models had differences in the division of health states. Ten models emphasized macrovascular and microvascular complications. Six models included adverse events. Majority of the models (n = 11) were patient-level simulation models. Eleven models simulated annual changes in risk factors (body mass index, glycemia, HbA1c, blood pressure (systolic and/or diastolic), and lipids (total cholesterol and/or high-density lipoprotein)). All models reported the main data sources used to develop health states of complications. Most models (n = 11) could deal with the uncertainty of models, which were described in varying levels of detail in the primary studies. Eleven studies reported that one or more validation checks were performed. Conclusions The existing decision models for T2DM are heterogeneous in terms of the level of detail in the classification of health states. Thus, more attention should be focused on balancing the desired level of complexity against the required level of transparency in the development of T2DM decision models.