Background: Differences in baseline mortality, age structure, and influenza epidemics confound comparisons of cold-related increases in mortality between regions with different climates. The Eurowinter study aimed to assess whether increases in mortality per 1 degree C fall in temperature differ in various European regions and to relate any differences to usual winter climate and measures to protect against cold.

Methods: Percentage increases in deaths per day per 1 degree C fall in temperature below 18 degrees C (indices of cold-related mortality) were estimated by generalised linear modelling. We assessed protective factors by surveys and adjusted by regression to 7 degrees C outdoor temperature. Cause-specific data gathered from 1988 to 1992 were analysed by multiple regression for men and women aged 50-59 and 65-74 in north Finland, south Finland, Baden-Württemburg, the Netherlands, London, and north Italy (24 groups). We used a similar method to analyse 1992 data in Athens and Palermo.

Findings: The percentage increases in all-cause mortality per 1 degree C fall in temperature below 18 degrees C were greater in warmer regions than in colder regions (eg, Athens 2.15% [95% CI 1.20-3.10] vs south Finland 0.27% [0.15-0.40]). At an outdoor temperature of 7 degrees C, the mean living-room temperature was 19.2 degrees C in Athens and 21.7 degrees C in south Finland; 13% and 72% of people in these regions, respectively, wore hats when outdoors at 7 degrees C. Multiple regression analyses (with allowance for sex and age, in the six regions with full data) showed that high indices of cold-related mortality were associated with high mean winter temperatures, low living-room temperatures, limited bedroom heating, low proportions of people wearing hats, gloves, and anoraks, and inactivity and shivering when outdoors at 7 degrees C (p < 0.01 for all-cause mortality and respiratory mortality; p > 0.05 for mortality from ischaemic heart disease and cerebrovascular disease).

Interpretation: Mortality increased to a greater extent with given fall of temperature in regions with warm winters, in populations with cooler homes, and among people who wore fewer clothes and were less active outdoors.