Information Technology Reference
In-Depth Information
Table 15. APRDRG severity by hospital
Table of DSHOSPID by APRDRG_Severity
HOSPID
APRDRG_Severity(All Patient Refined DRG: Severity of Illness Subclass)
Total
Frequency
Row Pct
Col Pct
1
2
3
4
1
41
12.62
32.54
166
51.08
24.74
91
28.00
17.70
27
8.31
17.42
325
2
2
18.18
1.59
6
54.55
0.89
3
27.27
0.58
0
0.00
0.00
11
3
9
8.04
7.14
44
39.29
6.56
42
37.50
8.17
17
15.18
10.97
112
4
23
9.09
18.25
136
53.75
20.27
78
30.83
15.18
16
6.32
10.32
253
5
3
3.41
2.38
30
34.09
4.47
44
50.00
8.56
11
12.50
7.10
88
6
9
5.14
7.14
65
37.14
9.69
83
47.43
16.15
18
10.29
11.61
175
7
14
4.95
11.11
116
40.99
17.29
109
38.52
21.21
44
15.55
28.39
283
8
2
4.44
1.59
24
53.33
3.58
15
33.33
2.92
4
8.89
2.58
45
9
11
14.10
8.73
40
51.28
5.96
18
23.08
3.50
9
11.54
5.81
78
10
12
12.50
9.52
44
45.83
6.56
31
32.29
6.03
9
9.38
5.81
96
Total
126
671
514
155
1466
The APRDRG severity index was not used at all in the decision tree prediction. In this model, we did
not use the Charlson Index in combination with the APRDRG indices, although we did use the specific
diagnosis codes related to the Charlson Index. We will also want to determine whether the addition of
the Charlson Index will improve the model.
Table 17 shows the actual and predicted values by hospital. Note that this model (in contrast to the
Charlson Index in the previous section) has predictive models that are relatively reasonable compared to
the actual mortality levels, which is surprising given the proportion of patients in the highest mortality
and severity levels. Clearly, the first split based upon the condition of congestive heart failure modifies
the threshold value with respect to the APRDRG index. Hospital #3 with the highest actual mortality
has the highest rank defining quality. Hospital #2 with zero actual mortality ranks the lowest. We then
wanted to see how the results are modified if we include the Charlson Index with the APRDRG indices.
Search WWH ::
Custom Search