Skip to main content

Table 3 Description of contributing variables and results of variable reduction post-processing, by data type

From: A machine learning model to predict the risk of 30-day readmissions in patients with heart failure: a retrospective analysis of electronic medical records data

Data Type

Major Variables

Number of feature categories

Variable reduction from âž” to

Demographics

Marital status, education, gender, language

2

39 âž” 15

Admissions

Total cost of index admission, age at admission, cumulative number of 30-day readmissions, length of stay

2

217 âž” 53

Diagnoses

ICD-9 codes, WKF

4

8101 âž” 1297

Labs

WKF at admission, WKF at discharge

2

94 âž” 58

Medications

RXCUI, Medication name, WKF

7

16,779 âž” 1107

Procedures

ICD-9 codes

1

1833 âž” 95

Notes

Words from: social history, hospital course, hospital reason, allergies

7

7558 âž” 887

Total

 

25

34,621 âž” 3512