Screening for Prediabetes Using Machine Learning Models

Table 3

Performance of the screening score model (Lee et al. [8]) in predicting prediabetes and undiagnosed diabetes using the data from the Korean National Health and Nutrition Examination Survey (KNHANES) 2010 and 2011.

		AUC	Accuracy (%)	Sensitivity (%)	Specificity (%)
Prediabetes	KNHANES 2010*(internal validation)	0.734	63.4	76.1	60.0
	KNHANES 2011*(external validation)	0.712	59.9	74.3	56.4
Undiagnosed diabetes	KNHANES 2010^†(internal validation)	0.772	66.6	76.5	66.4
	KNHANES 2011^†(external validation)	0.751	64.6	74.4	64.3

AUC: area under the curve; KNHANES: Korean National Health and Nutrition Examination Survey.
Prediabetes was defined as fasting plasma glucose, with values ≥100 mg/dL and <126 mg/dL. *Internal and external validation sets to evaluate the screening score for prediabetes (𝑛 = 1,551 for KNHANES 2010 and 𝑛 = 4,566 for KNHANES 2011). ^†Internal and external validation sets to evaluate the screening score for undiagnosed diabetes (𝑛 = 1,585 for KNHANES 2010 and 𝑛 = 4,683 for KNHANES 2011).