An important new study in lower-level health facilities in low- and middle-income countries found an increased risk of neonatal deaths with corticosteroid use in pregnant women with imminent preterm birth, in contrast with the positive results previously found in high-income countries. The surprising finding demonstrates that context matters. The increase appears largely due to steroids administered in cases that were not actually preterm, probably due to inaccurate pregnancy dating and challenges with diagnostic capacity. Promoting public health often requires decisions based on less-than-perfect evidence, but we must be vigilant about gathering and assessing new evidence and ready to change strategies.