Information Technology Reference
In-Depth Information
Table 2. Frequency for class variable
class
Frequency
Percent
Cumulative
Frequency
Cumulative
Percent
1
930
0.01
930
0.01
2
649478
9.32
650408
9.33
3
950622
13.64
1601030
22.97
4
867899
12.45
2468929
35.43
5
801427
11.50
3270356
46.92
6
744191
10.68
4014547
57.60
7
651332
9.35
4665879
66.95
8
578229
8.30
5244108
75.25
9
539992
7.75
5784100
82.99
10
1185273
17.01
6969373
100.00
We start with the information in the National Inpatient Sample, and divide the patient population
by the number of ICD9 diagnosis codes for each patient. Then we compare the patient severity indices
that are included with the data to see if there is a difference based just upon the number of codes. We
will then examine the least significant codes and compare their severity to the most significant codes
(recall that the codes are to be listed in their order of importance). We will truncate the number of codes
to compare the severity of the other Indices based upon the number of codes used.
First, we define a class based just upon the number of codes listed in the National Inpatient Sample.
We use the following code:
data nis.numberofcodes;
set nis.addseverity;
if (dx10 eq ' ') then class= 10 ;
if (dx9 eq ' ') then class= 9 ;
if (dx8 eq ' ') then class= 8 ;
if (dx7 eq ' ') then class= 7 ;
if (dx6 eq ' ') then class= 6 ;
if (dx5 eq ' ') then class= 5 ;
if (dx4 eq ' ') then class= 4 ;
if (dx3 eq ' ') then class= 3 ;
if (dx2 eq ' ') then class= 2 ;
if (dx1 eq ' ') then class= 1 ;
run;
The results are given in Table 2. Next, we perform a table analysis of the number of codes without
any reference to the severity and compare it to the severity indices as discussed in Chapters 5-7.
The distribution of these class values is relatively uniform. Note, however, that the largest class con-
tains ten diagnosis codes. We restricted our attention to just the first ten codes. Tables 3 and 4 compare
the classes defined above in Table 2 with the APRDRG mortality and severity indices. We want to see
 
Search WWH ::




Custom Search