The Charlson Comorbidity Index - Text Mining Techniques for Healthcare Provider Quality Determination - page 194

Figure 24. Lift function for risk adjustment by Charlson index

Table 28. Actual versus predicted mortality values by hospital for patients with COPD using Charlson
index for risk adjustment

Hospital Actual Mortality Predicted Mortality Difference

1 3.94 23.62 19.68

2

7.94 33.33 25.39

3 2.13 44.68 42.55

4 4.78 29.57 24.79

5 4.65 29.07 24.42

6 3.21 27.98 24.77

7 3.85 50.00 46.15

8 No COPD Patients

9 4.11 43.84 39.73

10 0 0 0

have an index higher than 9. It indicates that many of these patients are not identified as having any level
of severity. However, as we review the Charlson Index codes, it is clear that severe heart disease is not
included in the definition of severity, unless it is diagnosed as congestive heart failure.
We look at the ten hospitals performing cardiovascular surgery to examine their Charlson Index levels
(Table 30). Hospital #1, with 50% of procedure number 3615, has almost 45% of its patients in Charlson
category #1; hospital #4 with its divided procedures, has 36% of its patients in Charlson categories #2
and #3. It is hospital #10 that appears to have shifted its patients into higher Charlson catgories, with
1% in category #5. It, too, has its procedures relatively divided between 3611-3614. We use predictive
modeling to investigate the patient outcomes. The results with the Charlson Index should be very similar
to those found in Chapter 4, since the patient conditions that are used to define the Charlson Index were