Table 1.
Results of the different models across the 3 databases
| Database | Optum EHR | Optum DOD | IBM CCAE | |||
|---|---|---|---|---|---|---|
| No. of patients | 37,011,188 | 5,280,836 | 6,332,087 | |||
| No. index BMI readings | 343,711,980 | 16,316,746 | 15,147,663 | |||
| No. rows/columns in training and testing datasets | Model 1 | Model 2 | Model 1 | Model 2 | Model 1 | Model 2 |
| 6,800,000/123 | 6,800,000/111 | 3,300,000/123 | 3,300,000/111 | 3,400,000/123 | 3,400,000/111 | |
| Considered patient cases out of the patient cohort | 2% | 20% | 22% | |||
| Oversampling ratio in training data | 50/50, 60/40, 70/30 | |||||
| Age group | ||||||
| < 21 years | 19% | 7% | 27% | |||
| 21–30 years | 13% | 4% | 11% | |||
| 31–45 years | 20% | 12% | 22% | |||
| 46–60 years | 23% | 22% | 30% | |||
| > 60 years | 25% | 55% | 10% | |||
| BMI classification | ||||||
| ≥ 30 kg/m2 | 51% | 40% | 45% | |||
| ≥ 35 kg/m2 | 29% | 20% | 27% | |||
| ≥ 40 kg/m2 | 16% | 10% | 16% | |||
| US region | ||||||
| South | 24% | 50% | 51% | |||
| Midwest | 50% | 21% | 22% | |||
| West | 9% | 20% | 10% | |||
| Northeast | 13% | 9% | 16% | |||
| Others | 4% | 0% | 1% | |||