Screening for Prediabetes Using Machine Learning Models
Table 1
The weighted characteristics of the data from the Korean National Health and Nutrition Examination Survey (KNHANES) 2010.
Normal ( = 3,681)
Prediabetes ( = 1,004)
Age (years)
41.9 ± 0.5 41.0–42.8
52.5 ± 0.6 51.3–53.7
<0.001
Gender (% men)
46.9 (0.9) 45.2–48.7
58.8 (1.9) 55.0–62.5
<0.001
Family history of diabetes (%)
18.3 (0.9) 16.6–20.2
22.9 (1.7) 19.8–26.4
0.007
Current smoker (%)
27.5 (1.0) 25.5–29.6
26.9 (1.9) 23.4–30.8
0.799
Alcohol intake (drinks/day)
0.8 ± 0.0 0.7–0.9
1.0 ± 0.1 0.9–1.2
<0.001
Physically active (%)
50.6 ± 1.1 48.4–52.9
52.1 ± 2.1 48.0–56.3
0.535
BMI (kg/m2)
23.2 ± 0.1 23.1–23.3
25.1 ± 0.1 24.8–25.3
<0.001
Waist circumference (cm)
79.1 ± 0.2 78.7–79.6
85.8 ± 0.4 85.1–86.6
<0.001
FPG (mg/dL)
89.0 ± 0.1 88.7–89.3
107.4 ± 0.3 106.9–108.0
<0.001
Systolic blood pressure (mmHg)
116.8 ± 0.4 116.0–117.5
127.7 ± 0.7 126.4–129.1
<0.001
Diastolic blood pressure (mmHg)
76.4 ± 0.3 75.8–77.0
81.5 ± 0.5 80.6–82.4
<0.001
Hypertension (%)
16.4 (0.8) 14.9–18.0
41.1 (2.2) 36.8–45.5
<0.001
BMI: body mass index; FPG: fasting plasma glucose. Table values are given as mean ± standard error or % (standard error) [95% confidence interval] unless otherwise indicated. were obtained by -test or chi-square test. Impaired fasting glucose was considered with values ≥ 100 mg/dL and <126 mg/dL.