Statistical analysis
We summarized descriptive characteristics of the study cohort using
means with standard deviations and frequencies with percentages. We
calculated descriptive characteristics for all births in the province
>28 weeks to help describe the source population from which
our tertiary care cohort was drawn.
We estimated the relationship between fetal size percentiles and EDI
score using generalized additive modelling, which allowed us to describe
associations using smooth, non-linear patterns in risk. All models
included age at testing (range 4 to 7 years, mean = 5.6 years) as a
covariate. We then categorized estimated fetal weight into five
categories (<3rd, 3-9th,
10-19th, 20-30th, 31st-90th), and
compared EDI scores in each category with the reference category of
31-90th percentile using absolute mean differences
with 95% confidence intervals. We repeated these calculations for the
two EDI subdomains, and for our two binary outcomes (using generalized
additive modelling and predicting risks and risk ratios for these). For
comparative purposes, we calculated EDI scores for all other children in
the province born at or beyond 28 weeks, overall and in gestation age
categories of 28-33, 34-36, and 37-42 weeks.
To evaluate the charts’ ability to predict developmental vulnerability
at the individual level (i.e., discriminatory ability), we calculated
the area under the receiver operating characteristic curve (AUC), as
well as sensitivity, specificity, positive predictive value, negative
predictive value, and likelihood ratios with 95% confidence intervals
at select percentile cut-points of the chart (3rd,
10th, 20th percentile). Because the
10th percentile of both charts classified considerably
less than 10 percent of our population as small-for-gestational age, we
additionally identified the percentile on each chart that corresponded
to the 10th percentile weight in our cohort, and
evaluated predictive ability at this cut-off. Optimal thresholds for low
and high risk were established using Youden’s Index.