Table 2.
Variables associated with risk of cognitive impairment.
| RFSA: Irel | Cox PH | |||||
|---|---|---|---|---|---|---|
| Rank | Variable | Mean | HR | 95% CI | p-value | |
| 1 | African American | 1.00 | 2.52 | [2.18, 2.91] | 0.000 | |
| 2 | Wealth | 0.59 | 0.98 | [0.91, 1.06] | 0.634 | |
| 3 | Education | 0.57 | 0.80 | [0.76, 0.84] | 0.000 | |
| 4 | BMI Slope | 0.33 | 1.16 | [1.09, 1.22] | 0.000 | |
| 5 | Subjective Health | 0.27 | 0.87 | [0.83, 0.92] | 0.000 | |
| 6 | Emotional Distress | 0.27 | 1.38 | [1.24, 1.53] | 0.000 | |
| 7 | Ethnicity (Hispanic) | 0.22 | 1.65 | [1.35, 2.01] | 0.000 | |
Note. Random forest survival analysis (RFSA, n = 4,990) and Cox proportional hazard analysis (Cox PH, n = 4,989) were conducted in different subset of participants. The data were randomly divided into two subsamples so that each of the survival analyses could be conducted independently. The simple splitting was based on the outcome using the function “createDataPartition” in R. Relative importance (Irel) refers to the relative importance in predicting risk of cognitive impairment. The relative importance of the strongest predictor in RFSA is expected to be equal 1.00. HR = Hazard ratio; 95% CI = 95% confidence intervals.