Changes in population susceptibility to heat and cold over time: assessing adaptation to climate change

Background In the context of a warming climate and increasing urbanisation (with the associated urban heat island effect), interest in understanding temperature related health effects is growing. Previous reviews have examined how the temperature-mortality relationship varies by geographical location. There have been no reviews examining the empirical evidence for changes in population susceptibility to the effects of heat and/or cold over time. The objective of this paper is to review studies which have specifically examined variations in temperature related mortality risks over the 20th and 21st centuries and determine whether population adaptation to heat and/or cold has occurred. Methods We searched five electronic databases combining search terms for three main concepts: temperature, health outcomes and changes in vulnerability or adaptation. Studies included were those which quantified the risk of heat related mortality with changing ambient temperature in a specific location over time, or those which compared mortality outcomes between two different extreme temperature events (heatwaves) in one location. Results The electronic searches returned 9183 titles and abstracts, of which eleven studies examining the effects of ambient temperature over time were included and six studies comparing the effect of different heatwaves at discrete time points were included. Of the eleven papers that quantified the risk of, or absolute heat related mortality over time, ten found a decrease in susceptibility over time of which five found the decrease to be significant. The magnitude of the decrease varied by location. Only two studies attempted to quantitatively attribute changes in susceptibility to specific adaptive measures and found no significant association between the risk of heat related mortality and air conditioning prevalence within or between cities over time. Four of the six papers examining effects of heatwaves found a decrease in expected mortality in later years. Five studies examined the risk of cold. In contrast to the changes in heat related mortality observed, only one found a significant decrease in cold related mortality in later time periods. Conclusions There is evidence that across a number of different settings, population susceptibility to heat and heatwaves has been decreasing. These changes in heat related susceptibility have important implications for health impact assessments of future heat related risk. A similar decrease in cold related mortality was not shown. Adaptation to heat has implications for future planning, particularly in urban areas, with anticipated increases in temperature due to climate change. Electronic supplementary material The online version of this article (doi:10.1186/s12940-016-0102-7) contains supplementary material, which is available to authorized users.

1. I think adding explanations of RR at all figures(2-4) would be better. For example, RR associated with 1 degree Celsius. Self-explanatory figures would be good.
2. Adding effects of influenza would be good at page 11 just after the statements of ambient air pollution, first paragraph. I think influenza would be a strong confounder for estimating effects of cold temperature on mortality.
3. Two statements on page 4, "In a random-effects meta regression of studies, the relative risk of heat related mortality was found to increase in countries closer to the equator (with higher summertime mean temperatures) and …." and page 11, "A review of these studies [23]used meta-regression to establish city-level characteristics associated with the heat-mortality relationship, demonstrating thresholds were generally higher in communities living closer to the equator." seem to be contradictory. On pare 4, hightemperature area has higher RR but higher threshold for high-temperature area. Higher threshold generally means smaller vunerability because people suffer for smaller window of temperature range. 4. I think risk of high temperature has been decreasing over years. But the rate of change has also been decreasing. It will eventually saturated to a constant because of the physiological as well as infrastructural limits. Please comment on this idea.

Reviewer 1
EnvironmentalHealth1476-069X-14-S1-S7 manuscript.pdf review Comments to the editor I think the authors conveyed important messages out of the reviewed papers. Especially it is very important that no study demonstrated that winter excess mortality will increase along with the global warming.
For this reason, I think this paper should be published.
That said, I was a bit frustrated to read this review paper, because they just list up differences among the studies; though I understand that the part of the problem would be diverse methods, observation periods and areas. This may be just my complaint, but if the editor thinks the same way and has ideas to improve this situation, letting them know the ideas would be helpful for the readers. 2) Because this is a review paper, external consistency is more important than statistical significance. In this regard, I think the authors put too much weight on statistical significance. For example, in p.11, the Reviewer 2 last sentence of the first paragraph, they mentioned only statistical significance. The effect size and its alteration after adding air pollution term would be more of interest for the readers. I would suggest to revise the whole manuscript along this line. 4) Regarding the two types of the studies, i.e., moving minimum mortality temperature (MMT) and fixed MMT, the authors just described the difference and abandoned to think further. However, given that the assumption of fixed MMT would make the cold risk appear increasing and the heat effect decreasing, there would be room for discussion in comparing these two types of studies.
Let me explain using the hypothetical example. The blue 3 V-shapes are the moving MMT type results.
If the fixed MMT is assumed, the middle MMT can be used for all the 3 periods. So, the relation of the earliest period would be like the solid red V and that of the latest period would be like the dotted red V. In this example, the moving MMT study find no risk alteration but the fixed MMT study can.