Table 1 Distributions of non-ATLAS patient demographics and EHR features, stratified by label (N = 97,403)

From: Fair positive unlabeled learning for predicting undiagnosed Alzheimer’s disease in diverse electronic health records

Characteristic

Overall N = 97,403a

Labeled positive N = 4,250a

Unlabeled N = 93,153a

p valueb

Sex (female)

56,013 (58%)

2757 (65%)

53,256 (57%)

<0.001

Race and ethnicity

 NH-white

70,882 (73%)

3056 (72%)

67,826 (73%)

0.2

 NH-AfAm

6003 (6.2%)

350 (8.2%)

5653 (6.1%)

<0.001

 HL

10,782 (11%)

468 (11%)

10,314 (11%)

0.9

 EA

9,736 (10%)

376 (8.8%)

9,360 (10%)

<0.05

Record length

12 (8, 19)

16 (9, 23)

12 (8, 19)

<0.001

Number of diagnoses

51 (31, 80)

77 (52, 110)

50 (30, 78)

<0.001

Number of encounters

28 (16, 47)

39 (25, 61)

28 (16, 46)

<0.001

Record density (per year)

2.06 (1.42, 3.23)

2.42 (1.59, 3.83)

2.04 (1.41, 3.19)

<0.001

Age at last visit

75 (70, 82)

85 (81, 87)

75 (70, 81)

<0.001

  1. ATLAS UCLA ATLAS Community Health Initiative, EA East Asian, HL Hispanic Latino, NH-AfAm non-Hispanic African American, NH-white non-Hispanic white.
  2. an (%); Median (IQR).
  3. bPearson’s Chi-squared test; Wilcoxon rank sum test.
Back to article page