Global warming has increased the frequency of meteorological disasters, especially extreme temperature events. Many previous studies have reported that human health risk is very sensitive to temperature and climate change was considered to be the most severe global health threat in the 21st century. Nowadays, the research on the impact of extreme temperature on public health has been a hotspot. Compared to those in developed countries, the related studies have started late in China. Moreover, there are three limitations in these studies. (1) Most of such studies focused only on one city or a few cities and the studies on the whole country are few.(2) The previous studies have not quantitatively identified the influence of temperature on health because the spatial scales were based on administrative regions, not on temperature zones. (3) Comparing with many studies on hot wave, relatively fewer are concerned with the influence of extreme low temperature. To overcome aforementioned problems, we collected the mortality dataset and meteorological variables of 127 communities in China during 2007 to 2012 from China Center for Disease Control and Prevention and pooled the community-specific cold risk in various latitude-effected temperature zones with the meta-analysis method. Then, we utilized the Distributed lag non-linear model (DLNM) at community level to investigate temperature-mortality relationship in different temperature zones and calculated the relative risk (RR) of extreme low temperature on mortality. The results showed that although temperature-mortality curves at the community level appeared huge differences, the pooled curves were generally U-or J-shaped in these five zones. Temperature-mortality curves in three zones (the sub-temperate region, warm temperate region and north subtropical region) were all U-shaped, indicating both low and high temperatures could increase significantly mortality risk. Moreover, the curves appeared J-shaped in other two zones (the middle subtropical region and south subtropical region). The most significant cold effect was observed in middle subtropical, with a RR of 1.93 (95% CI: 1.08-3.60); while the cold effect in north subtropical was not so obvious, with a RR of 1.27 (95% CI: 0.94-1.72). Based on this, an M-shaped curve of the cold risk was found across Chinese mainland. This means the risks of cold-related mortality are high in warm temperature and middle subtropical zone, moderate in sub-temperate and south subtropical zone, and low in north subtropical zone. Low temperature does show significant impact on temperature-mortality risk, but considering the M-shaped risk curve, we believe social-economic factors should also be taken into consideration. To explain this phenomenon, we collected the social-economic data including population and GDP and found that the highest per capita GDP matched with the lowest cold-related risk, while the related lower per capita GDP matched with the highest cold-related risk. Based on these findings, different characteristics of mortality of cold stress highlighted that not only ambient temperature but also social-economic condition can be a main factor controlling health risk. Our findings also suggest that more adaptive and effective measures especially increasing investment on public health are necessary, especially for the middle subtropical zone, to reduce health risks in China.