Cold snaps raise stroke deaths days after they hit

Extreme temperature is one of the few environmental exposures that arrives with a forecast, which makes its timing genuinely useful to understand. Stroke is a plausible target: circulatory strain, blood pressure and fluid balance all shift with the weather. But the average daily temperature, a sustained cold spell and a heatwave are three different exposures, and they may not act on the same timescale. Knowing whether risk spikes immediately or builds over days determines when the burden on people and health services actually lands.

Researchers assembled daily stroke mortality and meteorological data from 12 counties across Hubei province, China, for 1 January 2009 to 31 December 2012. A distributed lag nonlinear model was fitted for each county to estimate the association between daily mean temperature, cold spells, heatwaves and stroke deaths. Multivariate meta-analysis then pooled the county-specific exposure-response relationships and the lag-response patterns. The study population totalled 6.7 million people, with an average annual mean temperature of 16.6 °C and a mean of 2.7 stroke deaths per county per day.

Temperature and stroke mortality followed an inverse J-shaped curve at provincial level, meaning risk rose at both ends but more steeply in the cold. The pooled cold-spell mortality risk at lags of 3-14 days was 1.180, with a 95% confidence interval of 1.043 to 1.336, and county estimates spanning 0.968 to 1.523. Heatwaves produced a pooled risk of 1.114 at lags of 0-2 days, confidence interval 1.012 to 1.227, with county estimates from 0.675 to 2.066. The timing differed clearly: cold effects were delayed by 2-3 days and persisted for around 10 days, while heat effects were acute and attenuated quickly.

The inverse J-shape and the delayed cold response are consistent with what temperature-mortality research has reported elsewhere, and this study adds stroke-specific estimates from a subtropical Chinese province. The wide spread of county-level results, with some estimates below 1, is a reminder that pooled figures conceal considerable local variation. As an ecological time-series analysis it cannot identify who is vulnerable or why, and it covers a single province over four years. Whether the same lag structure applies in different climates or to non-fatal stroke is untested here.

This is a reason to think of cold weather risk as something that unfolds over the following week rather than on the coldest day itself, and heat risk as more immediate. The estimates are population-level associations from death records, so they describe how many deaths occur in a region, not any individual's odds. For people with existing cardiovascular risk, the timing pattern is the interesting part: the days after a cold snap begins were when pooled mortality was elevated. Nothing here identifies a protective action or threshold.