**1. Introduction**

This paper summarizes and discusses key findings on trends over time in the effects of temperature on human mortality. While not intended to be a comprehensive review of the temperature–mortality relationship literature, I start by briefly summarizing main findings from broad literature. I focus on mortality, although there is also some evidence emerging from the morbidity literature on trends in impacts [1].

The effect of temperature on mortality is the most extensively studied topic within the broad domain of climate and health research, with a reference that covers a broad range of methodologies [2]. A limited number of studies have quantified deaths listed as heat-related on death certificates, most often accumulating individual case reports from medical records [3]. However, heat-related deaths identified by medical records tend to capture only cases with clear heat involvement, which represent only a subset of all heat-related deaths [2]. A recent study found that less than 10% of excess heat-related deaths were labeled as such on death certificates from 1997 to 2013 in New York City [4]. Another type of epidemiology study quantifies excess mortality that occurs in a city or region during a clearly identified heat wave event, as compared to during non heat-wave periods in the same locale. This was for example the approach used by researchers to quantify the impacts of the 1995 Chicago heat wave [5] and the 2003 Paris heat wave [6]. However, heat-related deaths also occur during periods when not obvious heat waves happen, such as when occasional temperature spikes that occur in most of summers. To more comprehensively quantify the overall burden of heat on mortality, a third epidemiologic approach uses regression analysis of multi-year daily time series to quantify exposure–response relationships linking temperature and mortality [7,8]. These latter studies usually

include deaths due to all causes, or all causes minus "external" causes such as homicide and suicide, and report substantial impacts in association with high temperatures. Studies that analyze the entire distribution of temperatures also report that cold temperatures are associated with increased mortality risk; however, the extent to which these deaths are caused by low temperature, as opposed to seasonal respiratory infections that co-vary with temperature, has been questioned [9,10]. The specific impact of winter temperature on mortality is a key area of uncertainty in projecting the health effects of climate change. However, the present study only focuses on heat-related deaths.

Multi-city and age-stratified analyses have examined vulnerability factors that can explain differences in heat-mortality effects between cities. Increased heat risk is associated with old or young age, living alone, preexisting chronic diseases, poverty, and low prevalence of air conditioning (A/C) [7,8,11]. Thus, to estimate the quantitative impact of future temperatures on mortality, we need to understand not only how climate may change, but also how these vulnerability factors may evolve in the future.
