The Charlson Comorbidity Index - Text Mining Techniques for Healthcare Provider Quality Determination - page 163

Note that age is the first variable split followed by the Charlson Index, and then age again. We give the
English rules to accompany the decision tree. First, if the patient's age is less than 47.5 and the Charlson
Index is less than 2.5, then the prediction is for non-mortality, which comprises 86% of the prediction.

IF charlson < 2.5
AND Age in years at admission < 47.5
THEN
NODE: 4
N: 36470
1: 14.2%
0: 85.8%

Next, if the Charlson Index is 3 or greater with an age of less than 47.5, the mortality in the subgroup
is almost 70%.

IF 2.5 <= charlson
AND Age in years at admission < 47.5
THEN

Table 7. Odds ratios for logistic regression for mortality outcome

Odds Ratio Estimates
Effect Point Estimate 95% Wald

Confidence Limits

charlson 0.673 0.666 0.679

AGE

0.955 0.953 0.956

FEMALE 0 vs 1 0.862 0.836 0.889

RACE 1 vs 6 1.024 0.923 1.136

RACE 2 vs 6 0.965 0.863 1.079

RACE 3 vs 6 0.960 0.858 1.075

RACE 4 vs 6 0.721 0.626 0.829

RACE 5 vs 6 0.882 0.690 1.127

Income quartile 1 vs 4 1.008 0.964 1.054

Income quartile 2 vs 4 1.017 0.972 1.064

Income quartile 3 vs 4

1.077 1.029 1.128

Table 8. Predicted versus actual outcome

Predicted Value Actual Value Count

0 0 5477484

0 1 15

1 0 117002

1 1 2