Table 3 Association between nutritional patterns and 2 clinical outcomes.

From: Identification of variation in nutritional practice in neonatal units in England and association with clinical outcomes using agnostic machine learning

 

Coefficient

95% Confidence Interval

p value

(a)

Intercept

− 1.68

[− 1.74, − 1.62]

< 0.001

BW z-score

0.72

[0.64, 0.79]

< 0.001

Antenatal Steroids

0.15

[− 0.07, 0.37]

0.185

Apgar 1 min

0.00

[− 0.04, 0.05]

0.861

IMD decile

− 0.01

[− 0.04, 0.02]

0.588

Smoking in Pregnancy

0.05

[− 0.09, 0.19]

0.504

Sex

− 0.06

[− 0.18, 0.06]

0.301

Gestational Age

0.06

[0.01, 0.11]

0.017

Midlands and east

− 0.09

[− 0.27, 0.1]

0.364

North

− 0.09

[− 0.27, 0.09]

0.306

South

0.02

[− 0.16, 0.21]

0.796

Nutritional Treatment

− 0.33

[− 0.46, − 0.2]

< 0.001

Z1

0.03

[− 0.09, 0.15]

0.641

Z2

− 0.04

[− 0.16, 0.07]

0.482

Z3

0.00

[− 0.1, 0.11]

0.940

(b)

Intercept

− 0.15

[− 0.36, 0.05]

0.143

BW z-score

− 0.64

[− 0.92, − 0.37]

< 0.001

Antenatal Steroids

0.27

[− 0.49, 1.03]

0.491

Apgar 1 min

− 0.09

[− 0.24, 0.07]

0.263

IMD decile

0.00

[− 0.15, 0.15]

0.977

Smoking in Pregnancy

0.05

[− 0.48, 0.58]

0.862

Sex

− 0.58

[− 1, − 0.15]

0.008

Gestational Age

− 0.43

[− 0.58, − 0.28]

< 0.001

Midlands and east

0.02

[− 0.62, 0.66]

0.941

North

0.38

[− 0.27, 1.02]

0.254

South

0.42

[− 0.19, 1.04]

0.176

Nutritional Treatment

− 1.16

[− 1.67, − 0.65]

< 0.001

Z1

− 0.26

[− 0.66, 0.15]

0.216

Z2

0.21

[− 0.19, 0.61]

0.309

Z3

− 0.10

[− 0.54, 0.34]

0.648

  1. Regression coefficients, 95% confidence intervals and p values are reported for all tested covariates, and the three latent variables (Z1, Z2 and Z3) of the factor model that facilitate deconfounding. We excluded infants with any missing values for the studied variables and used London as the reference category for geographical location. (a) Linear regression on weight z-score at 36 weeks postmenstrual age (W36 z-score), comparing clusters 13 (n = 371) and cluster 16 (n = 339). (b) Logistic regression on BPD, comparing cluster 17 (n = 376) to 18 (n = 283). Coefficients in (a) measure effect on the outcome by a change of one unit in the corresponding covariate. Coefficients in (b) measure change in log-odds ratio of the outcome by a change of one unit in the corresponding covariate. “Nutritional Treatment” = binary covariate encoding inclusion in one cluster versus the other. Significant (p < 0.05) covariates are highlighted in bold/italic if inducing a negative/positive effect on the outcome variable when the associated covariate is quantitatively positive.
Back to article page