Fig. 4: Heterogeneity in nonlinear exposure-response relationships between U5M and anomalous temperature by different temperature zones.

a Shows the estimated nonlinear associations between under-5 mortality (U5M) and anomalous temperature by different temperature zones, based on subpopulation analysis using a multicenter design (n = 1,745,132 biologically independent samples). The exposure-response relationships are shown as solid lines with 95%-CI ribbons. b Shows different slices of the two-dimensional exposure-response function estimated in (c) as solid lines with 95%-CI ribbons. c Shows the varied nonlinear associations estimated by a varying-coefficient nonlinear model. The directly estimated effect are presented both as a two-dimensional function of long-term temperature and its anomaly and as a function of temperature and its long-term average. c Illustrates how to infer the effect of temperature by correlating its fluctuations to variation in U5M.