Predicting diabetes clinical outcomes using longitudinal risk factor trajectories

Table 4 Estimating diabetes risk in subpopulations that developed pre-diabetes at different time points. Some groups returned to normoglycemia thereafter. The table shows the number of patients (N), mean predicted diabetes risk as the expected number of incidents in 10 years (Pred) and the estimation error (Error) by the Baseline, Cumulative Exposure, and the Extreme Value models

Subpopulation	N	Baseline		Cumulative		Extreme
Subpopulation	N	Pred	Error	Pred	Error	Pred	Error
PreDM in 2000–2001, normal from 2003 onwards	2181	.135	.006	.145	.001	.181	−.018
PreDM in 2002–2003, normal before and after	1193	.157	.026	.175	.019	.262	−.034
Normal before 2004, PreDM in 2004	1065	.332	−.057	.235	.001	.282	−.234
Normal before 2002–2003, PreDM since then	889	.411	−.029	.384	−.008	.402	−.019
Normal throughout 2000–2004	14,387	.092	−.016	.080	−.007	.069	−.001

ISSN: 1472-6947