Skip to main content

Table 4 Machine learning algorithm identifying the importance predictive performance (%) for death in-hospitalized patients infected with Delta or Omicron*

From: Comparison of the clinical characteristics of SARS-CoV-2 Delta (B.1.617.2) and Omicron (B.1.1.529) infected patients from a single hospitalist service

 

Variable or Dataset

COVID-attributed Deaths

All-Cause Deaths

In-hospital

In-hospital + ≤ 30 days of discharge

In-hospital

In-hospital + ≤ 30 days of discharge

Importance

Initial Rothman Score

16.41

39.66

15.88

36.35

 

News-2 Score

12.56

4.88

13.60

3.90

 

Neutrophils

9.11

9.96

10.86

10.61

 

Level of Care

7.73

3.19

9.47

3.88

 

Days from PCR + to Adm.

6.51

3.41

5.62

3.08

 

O2 Saturation

5.31

7.31

4.19

4.27

 

Creatinine

6.81

3.15

7.97

5.95

 

Age Group

5.63

4.99

9.51

6.41

 

Respiratory Rate

4.44

2.39

4.47

4.84

 

CRP

8.08

6.12

5.08

4.88

 

Lymphocytes

3.44

 

3.11

5.43

 

WBC

4.34

6.08

4.25

4.33

 

Procalcitonin

3.88

4.90

3.49

3.84

 

Initial Severity

2.62

 

2.49

 
 

Days from symptom onset to PCR+

3.11

3.95

  
 

Confusion/altered mental status

   

2.22

AUC

Training

97.1

95.7

98.8

96.8

 

Testing

83.8

81.7

85.5

78.4

  1. *based on the XGBOOST algorithm. Importance scores ≥5 are bolded