Progression of individuals screened and intervened for diabetes. The hybrid tree combined a decision tree and Markov models. The decision tree (the left side) consisted of five main arms representing five scenarios. The first three scenarios involved screening for undiagnosed diabetes and IGT followed by any of the three active lifestyle interventions (diet, exercise, and duo-intervention), which were applied to the IGT subjects. The fourth scenario involved screening for undiagnosed diabetes and IGT, but without formal interventions, and the fifth scenario involved control group. Nine Markov models represented the nature history of diabetes (the lower right side). Each of them consisted of eight states: IGT, normal glucose tolerance, onset of diabetes, four diabetes complication states and death. The IGT states were tunnel states that included six temporary ones representing 6 years lifestyle interventions. Transition probability, costs, benefits were required for each state. Three separate models were performed for strategies starting at age of 25, 40 and 60 respectively. “2-hour PG” means 2-hour plasma glucose after breakfast. “DM” means diabetes mellitus. “OGTT” means oral glucose tolerance test. “IGT” means impaired glucose tolerance. “NORMAL GT” represented normal glucose tolerance state. “DIABETES COMPLICATIONS” included four different diabetes complications states: cardiovascular disease, retinopathy, nephropathy, and overt neuropathy disease. We numbered the transition paths corresponding to the main transition parameters in Table 1. (#The complement probabilities of one branch. *The life-table information used to model competing causes of death. @The proportion of individuals with normal PG. ¢, £, ¤ Transition parameters which determined whether a subject would receive interventions. §, ¶, ß, ð,&: Transition parameters applied to the Markov models: ð1 to ð4 determined transitions from onset of diabetes state to complications states respectively; &1 and &2 determined transitions from CVD or nephropathy to death state. We did not include the neuropathy-specific and retinopathy-specific mortalities, since these complications are not fatal).
Liu et al. BMC Public Health 2013 13:729 doi:10.1186/1471-2458-13-729