Skip to main content

Table 2 Baseline characteristics and performance of candidate algorithms among people in the derivation cohort, stratified by diabetes type. Candidate algorithms developed using encounter codes (“code algorithms”) or prescriptions (“prescription algorithms”) are also shown. For each algorithm, values in the Type 1 and 2 columns indicate the number and percentage of individuals satisfying the algorithm (sensitivity). Positive predictive values for classifying type 1 diabetes are shown in the right column. The best 4 algorithms are indicated by the letters in parentheses (A–D; see text for selection criteria)

From: Development and validation of algorithms to classify type 1 and 2 diabetes according to age at diagnosis using electronic health records

 

Type 1 (n = 60)

Type 2 (n = 10,136)

Positive Predictive Value (%)

Demographic Characteristics

 Age at diagnosis (years; mean, standard deviation)

22.7 (12.6)

55.9 (11.4)

 

   < 18

16 (26.7)

20 (0.2)

  18–39

33 (55.0)

768 (7.6)

   ≥ 40 years

11 (18.3)

9348 (92.3)

 Female

32 (53.3)

4456 (43.9)

Coding Characteristics*

 Number of type 1 codes (median, interquartile range)

3.0 (4.0)

0.0 (0.0)

 

  Total number of type 1principal codes (median, interquartile range)

2.0 (2.5)

0.0 (0.0)

  Total number of type 1 mixed codes (median, interquartile range)

1.0 (2.0)

0.0 (0.0)

 Number of type 2 codes (median, interquartile range)

0.0 (1.0)

1.0 (3.0)

  Total number of type 2 principal codes (median, interquartile range)

0.0 (0.0)

1.0 (2.0)

  Total number of type 2 mixed codes (median, interquartile range)

0.0 (0.0)

1.0 (2.0)

Candidate Code Algorithms*

 At least 1 type 1 code

50 (83.3)

142 (1.4)

26.0

  At least 1 type 1 principal code

42 (70.0)

65 (0.6)

39.2

  At least 1 type 1 mixed code

38 (63.3)

108 (1.1)

26.0

 At least 1 type 2 code

43 (71.7)

92 (0.9)

31.8

  At least 1 type 2 principal code

29 (48.3)

25 (0.2)

53.7

  At least 1 type 2 mixed code

21 (35.0)

66 (0.6)

24.1

 Ratio of type 1 to type 2 codes ≥ 0.5 (A)

50 (83.3)

97 (1.0)

34.0

 Ratio of type 1 to type 2 codes ≥ 0.75

49 (81.7)

81 (0.8)

37.7

 Ratio of type 1 to type 2 codes ≥ 1

49 (81.7)

78 (0.8)

38.6

 Ratio of type 1 to type 2 codes ≥ 2

47 (78.3)

47 (0.5)

50.0

 Ratio of type 1 to type 2 codes ≥ 3

46 (76.7)

38 (0.4)

54.8

 Ratio of type 1 to type 2 codes ≥ 4 (B)

43 (71.7)

32 (0.3)

57.3

Candidate Prescription Algorithms

 At least 1 insulin prescription

59 (98.3)

3408 (33.6)

1.7

  within 90 days (C)

58 (96.7)

615 (6.1)

8.6

  within 180 days

58 (96.7)

715 (7.1)

7.5

  within 365 days

58 (96.7)

844 (8.3)

6.4

 At least 1 insulin prescription with no other glucose-lowering medication prescription

36 (60.0)

80 (0.8)

31.0

  within 90 days

43 (71.7)

362 (3.6)

10.6

  within 180 days

44 (73.3)

483 (4.8)

8.4

  within 365 days

45 (75.0)

653 (6.4)

6.4

 At least 1 insulin prescription with no other glucose-lowering medication prescription except metformin

50 (83.3)

275 (2.7)

15.4

  within 90 days

53 (88.3)

451 (4.4)

10.5

  within 180 days

54 (90.0)

566 (5.6)

8.7

  within 365 days

54 (90.0)

727 (7.2)

6.9

 Multiple daily injections

47 (78.3)

273 (2.7)

14.7

  within 90 days

7 (11.7)

5 (0.0)

58.3

  within 180 days

8 (13.3)

9 (0.1)

47.1

  within 365 days

12 (20.0)

13 (0.1)

48.0

 Multiple daily injections with no other glucose-lowering medication prescription (D)

32 (53.3)

9 (0.1)

78.0

  within 90 days

7 (11.7)

2 (0.0)

77.8

  within 180 days

8 (13.3)

6 (0.1)

571

  within 365 days

12 (20.0)

11 (0.1)

52.2

 Multiple daily injections with no other glucose-lowering medication prescription except metformin

42 (70.0)

26 (0.3)

61.8

  within 90 days

7 (11.7)

2 (0.0)

77.8

  within 180 days

8 (13.3)

6 (0.1)

57.1

  within 365 days

12 (20.0)

11 (0.1)

52.2

 At least 1 metformin prescription

22 (36.7)

8979 (88.6)

0.2

 Other glucose-lowering medication prescription excluding insulin and metformin

10 (16.7)

7673 (75.7)

0.2

  1. *In our dataset, encounter codes were classified as “principal” (principal diagnoses from hospital discharge abstracts) or “mixed” (including secondary diagnoses from hospital discharge abstracts and encounter codes from hospital outpatient clinics). Type 1 codes are defined as International Classifications of Diseases Ninth Revision (ICD-9) codes 250.x1 or 250.x3; type 2 codes are defined as ICD-9 codes 250.x0 or 250.x2
  2. We only included long-term insulin prescriptions (duration ≥ 28 days). See Appendix Table 4 for prescriptions algorithms using renal function criteria. All indicated durations are counted from the diagnosis date. If no duration is indicated, all data were utilized (up to 2016)
  3. Multiple daily injections: defined as prescriptions for long-acting and short-acting insulin (both initiated within the same time frame as specified)
  4. See Supplementary Table 4 (Additional File 1) for algorithms using renal function criteria