“…Older HE models estimated their own prediction models (e.g., Eastman, Archimedes, UKPDS and EAGLE), and newer HE models used or re-estimated existing prediction models, with a few exceptions (e.g., BRAVO and JJCEM). The UKPDS risk engine (n = 20, 59%) is the most frequently used set of prediction models (Table 2), followed by the Eastman [33,87], Caro [76,77], IQVIA-CORE [93], Michigan model [97], Tilden [104], ODEM [98], Sheffield [101], Syreon [103], PROSIT [99], and ECHO [88], and PREDICT [ GDM [16], UKPDS-OM1 [25], EAGLE [86], Cardiff [75], JADE [5], DMM [85], UKPDS-OM2 [11], TTM [105], SPHR [102], BRAVO [74], COMT [82], RAMP-DM [100], Cornerstone [83], CHIME [81], and PRIME [53] Individual trajectories were simulated as a sequence of events by repeatedly applying prediction models, using a discrete time cycle, usually a fixed annual cycle Risks for events were based on existing or self-developed prediction models, reflecting the risk to develop a certain complication ADVANCE and QRisk2) were only applicable for macrovascular disease. In many models, microvascular disease was estimated using diabetes duration stratified constant hazard ratios, assuming implicit exponential survival models.…”