Table 3.
Multivariable lasso logistic regression analysis and scoring system
| Variable | Selection probabilitya | Odds ratio | β regression coefficient (95% CI)b | Pointsc |
|---|---|---|---|---|
| Age (years) | 0.984 | |||
| <75 | Reference | |||
| ≥75 to <85 | 0.77 | −0.26 (−1.44–0.0) | −1 | |
| >85 | 2.08 | 0.73 (0.0–1.49) | 1 | |
| Body mass index < 20 (kg/m2) | 0.864 | 2.39 | 0.87 (0.0–1.75) | 1 |
| Female | 0.918 | 1.70 | 0.53 (0.0–1.40) | 1 |
| Heart failure | 0.765 | 1.52 | 0.42 (0.0–1.29) | 1 |
| Coronary artery disease | 0.809 | 0.77 | −0.27 (−1.66–0.0) | −1 |
| Prior myocardial infarction | 0.731 | 1.65 | 0.50 (0.0–1.60) | 1 |
| Pulmonary hypertension | 0.596 | 1.09 | 0.08 (−0.31–1.19) | 1 |
| COPD | 0.944 | 2.83 | 1.04 (0.0–1.91) | 2 |
| Atrial fibrillation | 0.979 | 0.41 | −0.89 (−1.68 to −0.05) | −1 |
| Prior cardiac surgery | 0.761 | 0.74 | −0.31 (−1.51–0.0) | −1 |
| Dialysis | 0.664 | 1.30 | 0.26 (−0.21–1.79) | 1 |
CI, confidence interval; COPD, chronic obstructive pulmonary disease.
a
Proportion of 1000 bootstrap samples in which the variable was included in the lasso logistic model.
b
Confidence intervals derived from the 2.5th and 97.5th percentile from the distribution of each coefficient across 1000 bootstrap samples. If the variable was not selected in the model for a particular bootstrap sample it was included as having a value of zero.
c
Points were assigned by increasing (or decreasing) β coefficients up or down to the next higher positive or lower negative integer, respectively.