Information Technology Reference
In-Depth Information
Table 10. Hospital by resource demand level: disease staging
Table of DSHOSPID by DS_RD_Level
Hospital
DS_RD_Level(Disease Staging: Resource Demand Level)
Total
Frequency
Row Pct
Col Pct
2
3
4
5
1
0
0.00
0.00
78
61.42
11.56
42
33.07
10.82
7
5.51
29.17
127
2
3
4.76
4.11
35
55.56
5.19
23
36.51
5.93
2
3.17
8.33
63
3
0
0.00
0.00
30
63.83
4.44
17
36.17
4.38
0
0.00
0.00
47
4
14
6.09
19.18
130
56.52
19.26
83
36.09
21.39
3
1.30
12.50
230
5
9
10.47
12.33
66
76.74
9.78
11
12.79
2.84
0
0.00
0.00
86
6
4
1.83
5.48
77
35.32
11.41
135
61.93
34.79
2
0.92
8.33
218
7
0
0.00
0.00
46
88.46
6.81
6
11.54
1.55
0
0.00
0.00
52
9
0
0.00
0.00
143
65.30
21.19
68
31.05
17.53
8
3.65
33.33
219
10
43
36.44
58.90
70
59.32
10.37
3
2.54
0.77
2
1.69
8.33
118
Total
73
675
388
24
1160
For patients with COPD, the levels of 1 and 2 are not assigned. Hospital #6 again has the highest
proportion of patients in category 4; very few of the patients are in category #5. Hospital #10, again with
no mortality has 94% of its patients in level 3; hospital #5 has 80% of its patients in that level.
Figure 31 shows the results of the predictive model to predict mortality rates. It shows that the memory
based reasoning node is the most accurate with 28% misclassification. Figure 32 gives the decision tree
for the results.
Note that the tree uses the patient's age only and none of the disease staging levels. This tree indicates
that the results are highly questionable since the severity level is not used to predict the patient outcome
of mortality. The ability of the model to predict mortality is extremely poor (Table 11).
Even though the predictive ability is extremely poor, we can still define a ranking, regardless of the
accuracy of that ranking. In this case, because the demand resource levels skew toward the higher levels
in comparison to the APRDRG levels, the threshold will have a high proportion of predicted mortality.
Search WWH ::
Custom Search