Meta-actions as a Tool for Action Rules Evaluation - Feature Selection for Data and Pattern Recognition - page 187

Table 9.5 Mapping between CCS single level diagnosis codes and their description [ 3 ]
Diagnosis code

Description

101 Coronary atherosclerosis and other heart disease

55 Fluid and electrolyte disorders

62 Coagulation and hemorrhagic disorders

106 Cardiac dysrhythmias

99 Hypertension with complications and secondary hypertension

158 Chronic kidney disease

114 Peripheral and visceral atherosclerosis

108 Congestive heart failure; nonhypertensive

59 Deficiency and other anemia

58 Other nutritional; endocrine; and metabolic disorders

117 Other circulatory disease

105 Conduction disorders

155 Other gastrointestinal disorders

663 Screening and history of mental health and substance abuse codes

257 Other aftercare

259 Residual codes; unclassified

96 Heart valve disorders

253 Allergic reactions

211 Other connective tissue disease

As mentioned in the dataset description, the flexible attributes represented by
diagnosis are not ordered by attribute column (there are no fixed attribute columns).
For this reason, we represented each visit with a set of diagnoses instead of a set
of fixed attributes. However, as you can note in Table 9.6 , to simplify the reporting,
we assume the domain of each diagnosis is in , where 1 means that the patient

is diagnosed with that particular diagnosis at the current visit, and 0 means that the
patient is not diagnosed with that specific diagnosis at the current visit.
We first grouped our dataset by procedures that are represented here by the first
primary procedure attribute PRCCS1 . We also grouped the visits by patient ID Vis-
itLink and ordered each patient's visits by the DaysToEvent attribute. We then built
pairs of visits for the meta-action applied, and extracted the action terms from the
pairs. Finally, we computed the likelihood and term confidence for each action term.
As you can see from Table 9.6 , the likelihood and term confidence of the action
terms extracted are very high. The higher the likelihood of the action term, the more
important the diagnoses and the more likely the diagnoses involved are the main
reason of the procedure. In addition, the higher the term confidence the more stable
the meta-action and procedure result.

{ 0 , 1 }