Information Technology Reference
In-Depth Information
Table 13. COPD text cluster to mortality
Table of _CLUSTER_ by DIED
CLUSTER
DIED
Total
Frequency
Row Pct
Col Pct
0
1
1
231
96.65
20.70
8
3.35
18.18
239
2
280
93.65
25.09
19
6.35
43.18
299
3
59
100.00
5.29
0
0.00
0.00
59
4
55
98.21
4.93
1
1.79
2.27
56
5
491
96.84
44.00
16
3.16
36.36
507
Total
1116
44
1160
Figure 21. Kernel density of length of stay by cluster
However, this model still misclassified almost 30% of the time, suggesting that there is still consider-
able random chance in the mortality of patients unaccounted for by the model.
We next use the example of cardiovascular surgery using procedure 36.1 to compare the results using
Text Miner to those results from other measures. We use the diagnosis codes to define a severity index,
again using just 5 classes. Once we define the classes, we need to determine how to rank the classes.
 
Search WWH ::




Custom Search