BACKGROUND: Excess winter deaths, the ratio between average daily deaths in December-March versus other months, is a measure commonly used by public health practitioners and analysts to assess health burdens associated with wintertime weather. We seek to demonstrate that this measure is fundamentally biased and can lead to misleading conclusions about health impacts associated with current and future winter climate. METHODS: Time series regression analysis of 779,372 deaths from natural causes in London over 15 years (1 August 1997-31 July 2012),collapsed by day of death and linked to daily temperature values. The outcome measures were the excess winter deaths index, and daily and annual deaths attributable specifically to cold. RESULTS: Most of the excess winter deaths are driven by cold: The excess winter deaths index decreased from 1.19 to 1.07 after excluding deaths attributable to low temperatures. Over 40% of cold-attributable deaths occurred outside of the December-March period, leading to bias in the excess winter deaths measure. Although there was no relationship between winter severity and annual excess winter deaths, there was a clear correlation with annual cold-attributable deaths. CONCLUSIONS: Excess winter deaths is not an appropriate indicator of cold-related health impacts, and its use should be discontinued. We advocate alternative measures. The findings we present bring into doubt previous claims that cold-related deaths in the UK will not reduce in future as a result of climate change.