Table 3. Importance of parameters for diabetes prediction from the Heinz Nixdorf Recall cohort.
| Parameter | Mean decrease of Gini impurity* | SD4 |
|---|---|---|
| HbA1c | 189.69 | 5.01 |
| Adiponectin | 67.72 | 3.43 |
| BMI1 | 63.49 | 3.11 |
| GGT2 | 55.47 | 2.70 |
| Vitamin D | 50.98 | 1.63 |
| CRP3 | 46.43 | 1.65 |
| ALT | 41.43 | 1.42 |
| AST | 34.07 | 1.40 |
*The Gini impurity gives an estimate, which parameters are most important for the random forest to predict the condition of interest (in this case: diabetes). A higher decrease of the Gini impurity represents a higher importance for this parameter. 1: Body mass index; 2: gamma-Glutamyltranerase; 3:C-reactive protein; 4: standard deviation.