Birth-weight is modelled without controlling for gestational age (red lines) and controlling for it (turquoise lines). Shading shows 95% credible intervals. Effects are modelled by splines. Results show a study-region wide 44g drop (CI: −34 to −55) in mean birth-weight from 2006 to 2016, and 55% increase (1.55, from CI: 1.37 to 1.77) in LBW odds, unexplained by other predictors. Controlling for GA reduces these changes to −30 g, (CI: −17 to −44 g), and 1.17 OR (CI 1.02 to 1.33) and model fit becomes poor (evidence of over-fitting of the linear term). PTB odds increased 7.44 (CI: 6.1 to 8.99) during this period. Results indicate increasing PTB risks and no clear evidence of poorer growth. These trends could relate to changes in access to quality healthcare (e.g. leading to fewer miscarriages), methodological changes in GA and birth-weight assessment in provincial hospitals, or broader societal transitions.