Statistical Methods - Text Mining Techniques for Healthcare Provider Quality Determination - page 128

IF pneumonia < 0.5
AND immune < 0.5
AND Died during hospitalization < 0.5
AND septicemia < 0.5
THEN
NODE: 10
N: 25361
AVE: 4.00122
SD: 5.98604
IF 0.5 <= pneumonia
AND immune < 0.5
AND Died during hospitalization < 0.5
AND septicemia < 0.5
THEN
NODE: 11
N: 1075
AVE: 5.70233
SD: 5.00537
IF pneumonia < 0.5
AND 0.5 <= immune
AND Died during hospitalization < 0.5
AND septicemia < 0.5
THEN
NODE: 12
N: 3878
AVE: 5.91594
SD: 7.19594
IF 0.5 <= pneumonia
AND 0.5 <= immune
AND Died during hospitalization < 0.5
AND septicemia < 0.5
THEN
NODE: 13
N: 435
AVE: 7.07586
SD: 6.33627
The final split occurs for pneumonia, given different values of immune disorder, septicemia, and
mortality. If the patient has pneumonia and immune disorder, but not septicemia, then the average stay
is 7 days. If the patient has immune disorder and septicemia but not pneumonia, then the stay is slightly
longer at 7.2 days. Figure 20 gives the regression results.
All of the variables are statistically significant, but the r 2 value is less than 5%, indicating that most
of the variability in length of stay is not accounted for by these variables. If we cut the size of the sample
in half, the results are virtually the same as in the larger sample (Figure 21).