Analysis

We calculated the absolute risks of neonatal death, infant death and preterm birth, and the mean birth weights across characteristics and overall. Crude and adjusted risk ratios for neonatal death, infant death and preterm birth were estimated using binomial regression, and the mean differences in gestational age-adjusted birth weight were estimated using linear regression, with UK-born groups as the reference category. We adjusted for maternal age (including a quadratic term for age to account for departures from linearity), birth registration type, year of birth, sex of infant, IMD decile and for Black African infants only, fragile state status of mother’s country of birth. Gestational age was fitted as a continuous variable in models that adjusted for it. The robust estimator of variance was used to allow for non-independence of birth outcomes among infants born to the same mother. All regression models were fitted separately for each ethnic group although pooled estimates and tests for effect modification by ethnic group are also reported. The analysis was conducted using Stata V.15.