SITAR modelling of growth in preterm infants and the effect of antenatal steroids
Neil Lawrence
The effect of antenatal steroids on preterm infant growth
Antenatal steroids are excellent at reducing poor short term outcomes
Neonatal death by 31%
Respiratory distress syndrome by 44%
Intraventricular haemorrhage by 46%
Possible negative impact of ADHD and mental health in later life
Previous guidance for potential preterm deliveries <39 weeks altered to <34 weeks
sitar_model <- sitar(
x = ‘natural_log_age’,
y = ‘height’,
id = ‘id’,
data = ‘data_frame’,
df = 12 )
sitar_model <- sitar(
x = ‘natural_log_age’,
y = ‘height’,
id = ‘id’,
data = ‘data_frame’,
df = 12 )
Log-likelihood: -2941.496
Fixed: s1 + s2 + s3 + s4 + s5 + s6 + s7 + s8 + s9 + s10 + s11 + s12 + a + b + c ~ 1
s1 s2 s3 s4 s5 s6 s7 s8 s9
24.6 35.8 45.1 55.6 61.6 72.8 82.8 94.9 107.9
s10 s11 s12 a b c
128.8 139.9 126.7 33.2 -0.01 0.06
Random effects:
Formula: list(a ~ 1, b ~ 1, c ~ 1)
Level: id
Structure: General positive-definite, Log-Cholesky parametrization
StdDev Corr
a 5.36 a b
b 0.08 0.76
c 0.05 0.74 0.37
Residual 1.43
Population level ‘fixed effects’ – the ‘average’ growth
Distribution of random effects across the population
sitar_model <- sitar(
x = ‘natural_log_age’,
y = ‘height’,
id = ‘id’,
data = ‘data_frame’,
df = 12 )
ranef(sitar_model)
‘a’
Extract model random effects
‘b’
‘c’
SIZE
TIMING
VELOCITY
SIZE
TIMING
VELOCITY
Growth outcomes
Treatment and covariates
Antenatal steroids
Maternal height
Paternal height
Gestational age at birth
Base R!
Girls
Boys
Outcome variable = a : ‘size’
Flextable / officer packages
Problems:
Explainability!
Defining uncertainty around individual patient random effects
Post hoc estimation of individual random effects without the fitting of a new model