Table 1 Impact on predictive performance of exams on Model 3
From: Towards stroke prediction using electronic health records
Model Inputs | AUC | 95% CI |
|---|---|---|
with everything | 0.623 | (0.614, 0.632) |
w/o Cholesterol Exam (HDL-C, LDL-C, triglyceride) | 0.618 | (0.610, 0.627) |
w/o Diabetes Exam (HbA1c JDS & NGSP, blood glucose) | 0.608 | (0.600, 0.617) |
w/o Kidney Exam (creatinine) | 0.624 | (0.615, 0.632) |
w/o Liver Exam (GGT, GOT, GPT) | 0.615 | (0.606, 0.623) |
w/o Platelet count | 0.607 | (0.599, 0.616) |
w/o ALP | 0.616 | (0.607, 0.625) |
w/o Blood count (Hb, hematocrit, MCH, MCHC, MCV, RBC) | 0.600 | (0.591, 0.609) |
w/o Uric acid | 0.627 | (0.618, 0.635) |
w/o WBC | 0.608 | (0.600, 0.617) |