Algorithm | PPV training seta | PPV validation setb | Sensitivity training set | Sensitivity validation set | F-Scorec training set | F-Score validation set |
---|---|---|---|---|---|---|
Any Clinician Coded | ||||||
ICD-9 code only | ||||||
≥1 count of the ICD-9 code | 57% | 66% | 100% | 88% | 73% | 75% |
≥4 counts | 77% | 80% | 87% | 69% | 82% | 74% |
ICD-10 codes only | ||||||
≥1 count of the ICD-10 code | 77% | 67% | 95% | 98% | 85% | 80% |
≥4 counts | 94% | 73% | 71% | 83% | 81% | 78% |
ICD-9 or ICD-10 code counts | ||||||
≥1 ICD-9 or ICD-10 code | 56% | 65% | 100% | 93% | 72% | 77% |
≥4 ICD-9 or ICD-10 codes | 81% | 76% | 95% | 75% | 87% | 75% |
ICD-9 or ICD-10 code counts AND antimalarials ever used | ||||||
≥1 ICD-9 or ICD-10 code | 68% | 72% | 85% | 41% | 76% | 52% |
≥4 ICD-9 or ICD-10 codes | 83% | 77% | 83% | 38% | 83% | 51% |
ICD-9 or ICD-10 code AND labs ever checked | ||||||
≥1 count | 62% | 72% | 95% | 86% | 75% | 78% |
≥4 counts | 84% | 85% | 93% | 54% | 88% | 66% |
ICD-9 or ICD-10 code AND ANA+c | ||||||
≥1 count | 64% | 70% | 76% | 63% | 69% | 66% |
≥4 counts | 79% | 77% | 76% | 52% | 77% | 62% |
Rheumatology Coded | ||||||
ICD-9 code only | ||||||
≥1 ICD-9 code | 78% | 86% | 86% | 73% | 82% | 79% |
≥4 ICD-9 codes | 88% | 94% | 70% | 48% | 78% | 64% |
ICD-10 codes only | ||||||
≥1 ICD-10 code | 70% | 82% | 70% | 83% | 70% | 82% |
≥4 ICD-10 codes | 88% | 92% | 45% | 71% | 60% | 80% |
ICD-9 or ICD-10 codes | ||||||
≥1 ICD-9 or 10 code | 78% | 85% | 76% | 80% | 77% | 82% |
≥4 ICD-9 or 10 codes | 85% | 93% | 71% | 55% | 77% | 69% |
ICD-9 code: 710.0. ICD-10 codes: M32.1,* M32.8, M32.9 |
aThe training set consisted of 100 subjects from the Vanderbilt EHR.
bThe validation set consisted of 545 subjects from the Duke Autoimmunity in Pregnancy Registry and Duke EHR
cF-score: measure of performance of an algorithm, harmonic mean of PPV and sensitivity calculated as 2 x [(PPV x Sensitivity)/(PPV +Sensitivity)]
dANA positive ≥1:160.