Statistical Methods - Text Mining Techniques for Healthcare Provider Quality Determination - page 121

Table 4. Procedure codes used to predict mortality

pr Procedure Translation Frequency Percent

9904 Transfusion of packed cells 17756 7.05

3893 Venous catheterization, not elsewhere classified 16142 6.41

9671 Continuous mechanical ventilation for less than 96 consecutive hours 10528 4.18

3324 Closed [endoscopic] biopsy of bronchus 8315 3.30

9672 Continuous mechanical ventilation for 96 consecutive hours or more 8243 3.27

3491 Thoracentesis 8118 3.22

3995 Hemodialysis 8083 3.21

9604 Insertion of endotracheal tube 7579 3.01

9921 Injection of antibiotic 6786 2.69

9394 Respiratory medication administered by nebulizer 6309 2.50

9390 Continuous positive airway pressure 7868 1.48

8856 Coronary arteriography using two catheters 7622 1.44

4516 Esophagogastroduodenoscopy [EGD] with closed biopsy 7516 1.42

966 Enteral infusion of concentrated nutritional substances 7203 1.36

3722 Left heart cardiac catheterization 6652 1.25

8853 Angiocardiography of left heart structures 6350 1.20

4513 Other endoscopy of small intestine 6343 1.19

3404 Insertion of intercostal catheter for drainage 5693 1.07

8741 Computerized axial tomography of thorax 5538 1.04

9915 Parenteral infusion of concentrated nutritional substances 5169 0.97

9907 Transfusion of other serum 4962 0.93

9396 Other oxygen enrichment 4937 0.93

4311 Percutaneous [endoscopic] gastrostomy 4831 0.91

3895 Venous catheterization for renal dialysis 4726 0.89

0331 Spinal tap 4362 0.82

3891 Arterial catheterization 3867 0.73

3327 Closed endoscopic biopsy of lung 3776 0.71

9339 Other physical therapy 3492 0.66

311 Temporary tracheostomy 3406 0.64

4523 Colonoscopy 3404 0.64

sification rate. Change to a 5:1 ratio indicates that a false negative is five times as costly compared to
a false positive. A 10:1 ratio makes it ten times as costly. We need to determine if changes to this ratio
result in changes to the optimal model. We will discuss weights in more detail with additional examples
in subsequent sections of this chapter.