Skip to main content

Table 2 Validation of administrative data algorithms to identify adults identified with diabetes using clinical data from primary care electronic medical records as a reference standard; all adults, using the most recent year of administrative data (2012–2013)

From: Identifying diabetes cases from administrative data: a population-based validation study

Algorithm

Sensitivity (%) (95% CI)

Specificity (%)

(95% CI)

PPV (%)

(95% CI)

NPV (%)

(95% CI)

Prevalence

Kappa

Probability+

Probability-

Physician claims only

 1 P

76.4(75.8–77.1)

99.2(99.2–99.3)

92.3(91.9–92.7)

97.2(97.1–97.3)

9.01%

0.82

92.1%

3.21%

 2 P in 1 yr

58.5(57.8–59.3)

99.8(99.8–99.8)

97.4(97.1–97.7)

95.2(95.1–95.3)

6.54%

0.71

97.3%

5.48%

 3 P in 1 yr

41.2(40.5–42.0)

99.9(99.9–99.9)

98.3(98.0–98.6)

93.3(93.2–93.4)

4.57%

0.55

98.1%

7.58%

Inclusion of prescription claims

 1 Rx

45.9(45.2–46.7)

100.0(100.0–100.0)

99.3(99.1–99.5)

93.8(93.7–93.9)

5.03%

0.60

7.01%

 1 P or 1 Rx

82.6(82.1–83.2)

99.2(99.1–99.2)

92.6(92.2–93.0)

97.9(97.8–98.0)

9.72%

0.86

92.7%

2.39%

 1 P and 1 Rx

39.7(39.0–40.5)

100.0(100.0–100.0)

99.8(99.7–99.9)

93.1(93.0–93.3)

4.33%

0.54

7.75%

 (2 P in 1 yr) or (1Rx and 1 P)

65.9(65.1–66.6)

99.8(99.8–99.8)

97.6(97.4–97.9)

96.0(95.9–96.1)

7.35%

0.77

97.6%

4.55%

Inclusion of hospital records

 H

9.8(9.4–10.3)

99.9(99.9–100.0)

95.6(94.6–96.6)

90.1(89.9–90.2)

1.12%

0.16

92.3%

11.18%

 H or 1 P

77.9(77.2–78.5)

99.2(99.1–99.2)

92.0(91.6–92.5)

97.3(97.3–97.4)

9.21%

0.83

92.3%

3.01%

 H or 1 Rx

48.6(47.9–49.4)

99.9(99.9–99.9)

98.5(98.2–98.8)

94.1(94.0–94.2)

5.38%

0.62

98.3%

6.69%

 H or (2 P in 1 yr)

61.6(60.9–62.4)

99.8(99.7–99.8)

96.9(96.6–97.3)

95.5(95.4–95.6)

6.92%

0.73

97.4%

5.09%

 H or (3 P in 1 yr)

45.7(44.9–46.4)

99.9(99.8–99.9)

97.6(97.2–97.9)

93.8(93.6–93.9)

5.10%

0.59

98.2%

7.04%

Physician claims, hospital records, prescription claims

 (H or (2 P in 1 yr)) or 1 Rx

73.1(72.5–73.8)

99.7(99.7–99.8)

97.1(96.8–97.4)

96.8(96.7–96.9)

8.21%

0.82

96.8%

3.62%

 (H or 1 P) and (1 Rx or 1 F)

67.3(66.6–68.0)

99.8(99.7–99.8)

97.3(97.0–97.6)

96.1(96.0–96.2)

7.53%

0.78

97.6%

4.37%

 (H or 1 P) and 1 F

56.4(55.7–57.2)

99.8(99.8–99.8)

96.9(96.5–97.2)

94.9(94.8–95.0)

6.34%

0.69

97.2%

5.74%

 (H or 1 P) and 1 Rx

40.6(39.9–41.4)

100.0(100.0–100.0)

99.8(99.7–99.9)

93.2(93.1–93.4)

4.43%

0.55

7.65%

  1. PPV positive predictive value, NPV negative predictive value, H hospital discharge abstracts bearing a diagnosis of diabetes from the Canadian Institute for Health Information Discharge Abstract Database, P physician claims for a diabetes diagnosis (ICD-92 50) from the Ontario Health Insurance Plan Physician Claims Database, Rx prescription for an anti-hyperglycemic medication from the Ontario Drug Benefit Database, F, diabetes specific physician feecode from the Ontario Health Insurance Plan Physician Claims Database, S same day surgery admission bearing a diagnosis of diabetes from the Canadian Institute for Health Information Discharge Abstract Database; yr., year
  2. Reference standard: EMR chart – adult (≥20 years old) with diabetes or one of its synonyms were listed in the cumulative patient profile or they had any of the following: haemoglobin A1c greater than 7%, two abnormal blood glucose tests [fasting blood sugar(s) greater than or equal to 7.0 mmol/L, or a random blood sugar(s) greater than or equal to 11.1 mmol/L], or a prescription for an anti-hyperglycemic medication (insulin or an oral hypoglycemic agent). Patients were excluded if they only had a record of gestational diabetes