Fig. 2: Country-specific and pooled estimates showing the associations of an increase of temperature the week before birth (lag 0–6 d) from the 75th to the 99th percentile and perinatal mortality of 16 hospitals in Benin, Malawi, Tanzania and Uganda.

This figure depicts the country-specific and pooled estimates presented as ORs and 95% CIs (mean ± 1.96 × s.e.m.) obtained from conditional logistic regression models. The estimates show the association of an increase in temperature (from the 75th to the 99th percentile) the week before birth (lag 0–6 d) on stillbirths (n = 4,886) (a), antepartum stillbirths (n = 2,152) (b), intrapartum stillbirths (n = 2,734) (c) and perinatal deaths (n = 5,880) (d). In the case–crossover design, each case serves as its own control. We defined control days as the same day of the week within the same month as childbirth. The data were obtained from 16 hospitals in four different countries, including Benin (four hospitals), Malawi (four hospitals), Tanzania (four hospitals) and Uganda (four hospitals). The vertical dashed line represents the value of the OR obtained from the meta-analyses, and the size of the square corresponds to the weight contribution to the meta-analyses. Estimates with smaller CIs had larger squares and contributed with more weight to the meta-analyses.