Text Mining and Patient Severity Clusters - Text Mining Techniques for Healthcare Provider Quality Determination - page 292

Table 3. Summary statistics by cluster

Cluster
Number

N Obs Variable Mean Std Dev Minimum Maximum

1 23213 LOS

TOTCHG

4.9532589
31049.85

6.2661966
41923.65

0
75.0000000

160.0000000
878999.00

2 27095 LOS

TOTCHG

2.7629083
6709.54

4.3595463
17419.82

0
29.0000000

142.0000000
690811.00

3 17657 LOS

TOTCHG

5.6405392
31784.15

8.8703964
60821.73

0
35.0000000

219.0000000
993107.00

4 5665 LOS

TOTCHG

5.1419241
21622.08

12.6491378
53211.10

0
35.0000000

265.0000000
838707.00

5 27663 LOS

TOTCHG

5.3739426
27290.85

6.3360912
40271.16

0
35.0000000

154.0000000
966378.00

6 14615 LOS

TOTCHG

4.0107431
26859.63

4.3202349
31417.12

0
35.0000000

191.0000000
919723.00

7 12956 LOS

TOTCHG

4.5065998
17698.05

6.2346689
22699.33

0
35.0000000

149.0000000
755327.00

8 7850 LOS

TOTCHG

6.5688623
23245.37

8.2627672
32148.28

0
263.0000000

315.0000000
797968.00

9 14412 LOS

TOTCHG

5.2954892
22256.87

7.0770023
34559.13

0
35.0000000

233.0000000
705743.00

10 8775 LOS

TOTCHG

2.8694017
9835.06

3.3940094
12656.74

0
494.0000000

122.0000000
740929.00

Note that the orders are slightly different for the two outcomes. As means are susceptible to outliers, we
can avoid this susceptibility by looking at the kernel density probabilities as well (Figures 1 and 2).
Cluster 2, for treating a newborn, has both the lowest costs and total charges, followed by cluster
10 for labor and delivery. Cluster 7, for patients with severe mental disorders, very likely has low costs

Figure 1. Probability density for length of stay