Fig 4. Correlation between radiographic findings and all-cause mortality.

(A-B): Forest plot for the Cox model with all 14 variables and selected 9 variables. Numbers represent hazard ratios. The range of lines indicates 95% confidence intervals. Color blue implies significant variables with p<0.05. (C) Variable importance plot of the Random Survival Forest model. Large positive values indicate the dependency of the outcome to get high predictive power. Values closer to zero represent a lower contribution to improved predictive accuracy. Negative numbers indicate the predictive accuracy would improve when the variables were unspecified.