Table 2 Predictive performance of the integrated multivariate Cox proportional-hazard regression model with different entry levels for pathway-specific PRSs included for feature selection by LASSO method. The seven clinical risk factors include AFib, BMI, CAD, diabetes, dyslipidemia, hypertension, and smoking.
From: Predicting mortality among ischemic stroke patients using pathways-derived polygenic risk scores
Univariate p value cutoff | Features Input | Features selected (LASSO) | C-Index Mean ± SE [95%CI] | median improvement against base model |
|---|---|---|---|---|
< 0.1 | 31 PRSs + Age | 24 PRSs + Age | 0.705 ± 0.035 [0.637–0.773] | < 0.001 |
< 0.05 | 20 PRSs + Age | 16 PRSs + Age | 0.684 ± 0.033 [0.619–0.749] | < 0.001 |
< 0.025 | 9 PRSs + Age | 8 PRSs + Age | 0.661 ± 0.036 [0.592–0.731] | 0.027 |
< 0.01 | 3 PRSs + Age | 2 PRSs + Age | 0.643 ± 0.034 [0.575–0.710] | 0.233 |
base model | 0 PRSs + Age | 0 PRSs + Age | 0.626 ± 0.024 [0.578–0.674] | NA |
< 0.1 | 31 PRSs + Age + 7 | 16 PRSs + Age + 7 clinical risk factors | 0.754 ± 0.031 [0.693–0.814] | < 0.001 |
< 0.05 | 20 PRSs + Age + 7 | 11 PRSs + Age + 7 clinical risk factors | 0.751 ± 0.030 [0.693–0.809] | 0.02 |
< 0.025 | 9 PRSs + Age + 7 | 6 PRSs + Age + 7 clinical risk factors | 0.740 ± 0.030 [0.680–0.799] | 0.066 |
< 0.01 | 3 PRSs + Age + 7 | 2 PRSs + Age + 7 clinical risk factors | 0.729 ± 0.028 [0.674–0.783] | 0.319 |
base model | 0 PRSs + Age + 7 | 0 PRSs + Age + 7 clinical risk factors | 0.729 ± 0.027 [0.676–0.782] | NA |
