You are viewing the site in preview mode

Skip to main content

Table 3 Validation results for the BMI categorization algorithm in the 6-month preoperative (Cohort 2) and 1-year postoperative periods (Cohort 3)

From: Suitability of administrative claims databases for bariatric surgery research – is the glass half-full or half-empty?

BMI categories, kg/m2 ICD-9-CM or ICD-10-CM Diagnoses Last preoperative BMIa Last postoperative BMIb
Sensitivity Specificity PPV NPV Kappa (95% CI) Sensitivity Specificity PPV NPV Kappa (95% CI)
≤19.9 V85.0, Z68.1 100 100 80 100 0.78
(0.76, 0.79)
100 100 100 100 0.84 (0.80,0.87)
20.0–24.9 V85.1, Z68.20-Z68.24 60 100 100 100 75 100 100 100
25.0–29.9 V85.21-V85.25, Z68.25-Z68.29 70 100 100 100 > 82c > 98c > 82c 98
30.0–34.9 V85.30-V85.34, Z68.30-Z68.34 30 100 > 62c 99 84 > 97c > 93c 93
35.0–39.9 V85.35-V85.39, Z68.35-Z68.39 64 97 81 93 84 94 81 95
40.0–44.9 V85.41, Z68.41 76 87 71 90 76 93 68 95
45.0–49.9 V85.42, Z68.42 71 90 65 92 72 97 73 97
50.0–59.9 V85.43, Z68.43 79 93 76 94 > 65c 97 61 > 98c
60.0–69.9 V85.44, Z68.44 70 98 67 98 100 99 17 100
≥70.0 V85.45, Z68.45 53 100 63 99 --d 100 0 100
≤29.9 V85.0, V85.1, V85.21-V85.25, Z68.1, Z68.20-Z68.29 87 100 100 100 0.77
(0.76, 0.79)
> 84c > 98c > 84c 98 0.82
(0.78, 0.87)
30.0–39.9 V85.30-V85.39, Z68.30-Z68.39 67 98 90 93 89 93 94 87
40.0–49.9 V85.41, V85.42, Z68.41, Z68.42 89 81 82 88 86 93 80 95
50.0–59.9 V85.43, Z68.43 79 93 76 94 > 65c 97 61 > 98c
≥60.0 V85.44, V85.45, Z68.44, Z68.45 75 98 74 98 100 99 14 100
≤19.9 V85.0, Z68.1 100 100 80 100 0.94
(0.89, 0.99)
100 100 100 100 0.90
(0.84, 0.96)
20.0–24.9 V85.1, Z68.20-Z68.24 60 100 100 100 75 100 100 100
25.0–29.9 V85.21-V85.25, Z68.25-Z68.29 70 100 100 100 > 82c > 98c > 82c 98
≥30.0 V85.30-V85.39, V85.41-V85.45, Z68.30-Z68.39, Z68.41-Z68.45 100 87 100 100 > 98c > 81c 98 > 81c
  1. Abbreviations: BMI body mass index, ICD international classification of diseases, PPV positive predictive value, NPV negative predictive value, CI confidence interval
  2. a The last available weight-related diagnosis code in claims in the 6-month preoperative period was compared with the last available BMI measurement in the electronic health records (EHR) during the same period in Cohort 2 patients
  3. b The last available weight-related diagnosis code in claims in the 1-year postoperative period was compared with the last available BMI measurement in the EHR during the same period in Cohort 3 patients
  4. c Cells with 10 or fewer patients have been suppressed to maintain the de-identification nature of the database
  5. d Sensitivity was not calculated because no patients had relevant BMI measurement at this level in the EHR