Skip to main content

Table 2 Changes in feature distribution and Kolmogorov-Smirnov-Test results of the most important cohort-describing pre-pandemic features

From: Susceptibility of AutoML mortality prediction algorithms to model drift caused by the COVID pandemic

 

pre pandemic validation

n = 102,666

pre pandemic test

n = 4,802

KS-test§

p-value

in pandemic

n = 2,251

KS-test#

p-value

Deceased

834

(0.8%)

49

(1.0%)

1

22

(1.0%)

0.458

Agea

56

(38, 70)

56

(38, 70)

0.292

58

(40, 72)

< 0.001

Sex

    

1

  

0.720

    Male

55,640

(54.2%)

2,546

(53.0%)

 

1,186

(52.7%)

 

    Female

47,026

(45.8%)

2,256

(47.0%)

 

1,065

(47.3%)

 

BMI missing

33,445

(32.6%)

831

(17.3%)

0.711

484

(21.5%)

< 0.001

BMI

25.2

(22.5, 28.6)

25.2

(22.5, 28.7)

0.914

24.9

(22.2, 28.4)

0.034

Lab request RPC’s ordered

27,727

(27.0%)

1,485

(30.9%)

1

849

(37.7%)

< 0.001

No of ordered RPC’sa

4.00

(2.00, 4.00)

4.00

(2.00, 4.00)

1

4.00

(2.00, 4.00)

< 0.001

Laboratory CRP missing

47,588

(46.4%)

1,878

(39.1%)

1

879

(39.0%)

< 0.001

Laboratory CRPa

0.30

(0.10, 1.00)

0.30

(0.10, 1.10)

0.558

0.30

(0.10, 1.60)

< 0.001

Number of consults before surgerya

2.00

(1.00, 3.00)

2.00

(1.00, 4.00)

0.211

3.00

(2.00, 4.00)

< 0.001

ASA missing

30,624

(29.8%)

783

(16.3%)

0.999

440

(19.5%)

< 0.001

ASAa

    

1.000

  

< 0.001

    I

20,202

(28.0%)

1,002

(24.9%)

 

415

(22.9%)

 

    II

37,049

(51.4%)

1,950

(48.5%)

 

889

(49.1%)

 

    III

14,102

(19.6%)

1,011

(25.2%)

 

474

(26.2%)

 

    IV

642

(0.9%)

53

(1.3%)

 

33

(1.8%)

 

    V

47

(0.1%)

3

(0.1%)

 

0

(0.0%)

 

Reason for admissiona

    

0.513

  

< 0.001

    Obstetric

20,369

(19.8%)

911

(19.0%)

 

553

(24.6%)

 

    Normal case

56,961

(55.5%)

2,935

(61.1%)

 

1,233

(54.8%)

 

    Organ donor

21,239

(20.7%)

753

(15.7%)

 

359

(15.9%)

 

    Accident

4,001

(3.9%)

197

(4.1%)

 

106

(4.7%)

 

    Full inpatient with pre-inpatient treatment

96

(0.1%)

6

(0.1%)

 

0

(0.0%)

 

Treatment days pre-inpatienta

1.00

(0.00, 1.00)

1.00

(0.00, 2.00)

0.204

1.00

(0.00, 1.00)

< 0.001

Laboratory INR missing

24,796

(24.2%)

647

(13.5%)

1

283

(12.6%)

< 0.001

Laboratory INRa

1.00

(0.90, 1.00)

0.90

(0.90, 1.00)

0.902

0.90

(0.90, 1.00)

< 0.001

  1. The Kolmogorov-Smirnov-Test was performed between the pre-pandemic validation and the pre-pandemic test set(§), and the pre-pandemic validation and the in-pandemic test set(#), a p-value < 0.05 indicates that the distributions of the respective parameter differ significantly from each other
  2. The table shows the common parameters of the best pre-pandemic model of each family which are marked with a. Sex and body mass index (BMI) were included in the table to better describe the patient collective. The number and percentage of missing values are referred to as "missing" and are also listed in the table. The absolute number and the percentage share are shown for discrete variables and the median and interquartile range for continuous variables in brackets
  3. BMI Body mass index, RPC’s Red packed cells, CRP c-reactive protein, ASA American Society of Anesthesiologists physical score [16], INR International normalized ratio
Back to article page

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us