Ambient temperature from satellite-hybrid models and preterm birth: A time-stratified case-crossover analysis of 70,000+ preterm births in Central Mexico

Abstract

BACKGROUND AND AIM: Increasing evidence demonstrates that high ambient temperatures (Ta) and fine particulate matter (PM₂.₅) are associated with preterm birth (PTB), but few have studied this relationship in low-and-middle income countries using satellite-based spatiotemporal exposure models. Further, new evidence suggests that studies using the time-stratified case-crossover (TSCCO) may yield biased results if strata are not short enough to account for the changing probability of birth over gestation. METHODS: Preliminary analyses focused on warm season temperatures. We conducted TSCCOs to associate same-day minimum daily temperatures (Ta minimum) with PTB, matched on day of the week, between 2008-2017 in Central Mexico (including Mexico City). Daily exposures were derived from a satellite-hybrid model at a 1-km² resolution, and were assigned using population-weighted centroids of the sub-county locality of residence. Birth data was derived from a national birth registry, restricted to singletons from 20-36 weeks gestation during warm months (May-September). Results compared 14-day and 28-day TSCCOs. Associations were estimated using conditional logistic regressions. RESULTS:Analyses included 70,263 PTBs across 497 sub-counties. The median Ta minimum was 13.1°C (IQR: 12.3-14.2°C, 95th percentile: 15.3°C). Both 14-day and 28-day TSCCOs estimated a positive association between Ta minimum and PTB, with a 1.2% increased odds of PTB for each °C increase (OR: 1.01, 95%CI: 1.00-1.02) with the 14-day TSCCO and a 5.2% increased odds (OR: 1.05, 95%CI: 1.04-1.06) with the 28-day TSCCO. CONCLUSIONS:Our analyses identified an association between the Ta minimum and PTB in Central Mexico, with markedly higher effect estimates for the 28-day TSCCO. This supports our previous simulations that month-long time-strata may yield biased results. Future analyses will assess: 1) interactions with PM₂.₅; 2) full calendar years to assess potential nonlinear relationships; 3) lagged relationships between ambient exposures and PTB; and 4) effect modification by individual-level and community-level factors. KEYWORDS: preterm birth, ambient temperature, particulate matter, short-term exposure, study design and methods