Fig. 1

Models to undertake hypothesis testing. A Model A1 and Model A2 assume additive and independent effects from component causes to low Apgar at 5-min (Ap5) and stillbirth (STB). B An unbiased causal path between exposure, low birth weight at delivery (LBW) or maternal anaemia (ANM) in the perinatal period, and each respective outcome is assumed in Models B1 and B2. The unbiased causal path can be measured after controlling effect from the exposure on the outcome for perinatal malaria (MalPeri), malaria during pregnancy (MalPreg), P. falciparum during pregnancy (MalFalc), living in rural area (Rural), and perinatal fever (Fever). Mediation is assumed in Models B1 and B2 when ancestors of the exposure modulate the effect on the outcome. The minimal and valid adjustment set for each exposure–outcome relationship is composed by {MalPeri, ANM} in Model B1 and {Rural, Fever} in Model B2. Unobserved variables (UnOb VAR) are factors (e.g., nutritional status, alcoholism, or smoking) that could modulate anaemia, but were not assessed