Skip to main content

Table 3 Predictive models for composite adverse outcome developed using data from ANZNN, CNN and SNQ 2008–2011

From: Comparing very low birth weight versus very low gestation cohort methods for outcome analysis of high risk preterm infants

 

(a) Model based on BW cohort

(b) Model based on GA cohort

Covariates

Estimate

SE

P-value

Estimate

SE

P-value

Intercept

18.66

0.363

<0.0001

17.8

0.285

<0.0001

GA (weeks)

−0.682

0.013

<0.0001

−0.649

0.01

<0.0001

BW z score

1.386

0.272

<0.0001

−0.15

0.022

<0.0001

GA x BW z score

−0.057

0.01

<0.0001

#

  

Country (sweden vs CA)

−0.427

0.07

<0.0001

−0.326

0.065

<0.0001

Country (Australia vs CA)

−0.416

0.045

<0.0001

−0.371

0.042

<0.0001

Gender (male)

0.192

0.042

<0.0001

0.257

0.039

<0.0001

Antenatal steroid use

−0.259

0.08

0.0012

−0.237

0.075

0.002

Singleton

#

  

#

  

Cesarean

0.3

0.044

<0.0001

0.241

0.04

<0.0001

Presentation (Vertex)

#

  

#

  

Area under ROC curve

0.837

0.0033

 

0.835

0.003

 

Hosmer-Lemeshow test

  

0.19

  

0.373

  1. Multiple logistic regression models were applied to obtain the final predictive models using stepwise variable selection procedure with inclusion and exclusion criterion of 0.05; # = excluded by the variable selection procedure
  2. Notes: BW z score Birth weight z score, GA x BW z score interaction between GA and Birth weight z score, Estimate estimated coefficient of the covariate, SE standard error