The All Patient Refined Diagnosis Related Group - Text Mining Techniques for Healthcare Provider Quality Determination - page 223

Figure 22. Partial diagram starting at branch 3
PredIctIon of aPrdrg usIng PatIent dIagnoses
We want to consider the relationship between primary and secondary diagnoses, and levels of severity.
We consider the condition with the primary diagnosis of congestive heart failure. We next consider the
co-morbidities that patients with the disease will also have. Then we take the most frequent co-morbidities
to see if the severity and mortality levels can be predicted based upon these co-morbidities. In order to
do this, we also look to the primary procedure to divide it into medical and surgical categories.
To investigate this issue, we filter down to all patients with a primary diagnosis of congestive heart
failure, code 428. Then we find the most frequently occurring secondary codes using the following SAS
code:
data nis.chf;
set nis.addseverity;
where rxmatch('428',dx1)> 0 ;
run;
data nis.dx2;
set nis.chf;