We examined the correlation between a sprawl index (based on land-use data from 2000) (^{Ewing et al. 2003a}) and the rate of increase in EHEs over a five-decade period (1956–2005). Each of these variables is defined below, followed by a description of the analytical approach.

An analysis of trends in excessively hot days over the period of 1956–2005 found the frequency of EHEs to be increasing significantly on an annual basis. Extending the trend measured by ^{Gaffen and Ross (1998)} by an additional decade, our analysis found that the mean annual number of EHEs in major U.S. cities increased by 0.20 days/year [95% confidence interval (CI), 0.14–0.26], which is consistent with 10 more events per city, on average, in 2005 than in 1956.

As illustrated in Figures 1 and 2, the rate of increase in annual EHEs over this 50-year period varied significantly by metropolitan form. Although the average annual number of EHEs increased during this period across all cities, the most sprawling cities (top quartile) experienced a rate of increase in EHEs that was more than double that of the most compact cities (bottom quartile). Between 1956 and 2005, the most compact cities experienced an average increase in the number of EHEs of 5.6 days (95% CI, 0.9–10.3), whereas the average annual number of events increased by 14.8 days (95% CI, 7.9–21.7) in the most sprawling cities. Variation in the size or rate of growth in metropolitan populations did not diminish the measured statistical association between land-use patterns and the rate of increase in EHEs in these cities (r = 0.34; p < 0.05). These findings are consistent with the hypothesis that urban sprawl contributes to EHE frequency.

This analysis yields two principal findings. First, the annual occurrence of EHEs continues to increase in large metropolitan regions of the United States. This finding, previously demonstrated by ^{Gaffen and Ross (1998)} for the period 1949–1995, is extended here to 2005 for 53 metropolitan regions for which both apparent temperature and sprawl index values are available. Second, the rate of increase in EHEs is higher in sprawling than in more compact metropolitan regions, an association that is independent of climate zone, metropolitan population size, or the rate of metropolitan population growth.

The mechanisms of EHEs in cities are complex. Cities are typically characterized by lower rates of evapotranspiration and lower albedo than are rural areas, as a result of the reduced vegetative cover and the increased presence of darkly hued bituminous roofing and paving materials. Urban areas are further characterized by higher thermal loads than are rural areas, because of the concentrated presence of generators, air-conditioning units, motor vehicles, and other heat sources. Although our data do not permit an assessment of the relative contribution of each of these factors, the loss of vegetative cover is well established as a principal driver of the urban heat island effect (^{Oke 1982}; ^{Stone and Norman 2006}). The availability of data on rates of deforestation across the continental United States between 1992 and 2001 enables an assessment of the association between changes in regional vegetative cover over time and EHEs during a portion of our study period.

The U.S. National Land Cover Database provides maps of 21 categories of land cover across the continental United States between 1992 and 2001 (^{U.S. Department of the Interior 2009}). For each of the 53 metropolitan regions in our study, we measured the area of forest canopy change over this 10-year period and associated it with the sprawl index and rate of change in EHEs. The results of this analysis indicate that the rate of deforestation in the most sprawling metropolitan regions is more than double the rate in the most compact metropolitan regions. For those regions in the top quartile of the sprawl index, 187 km² (95% CI, 33.4–339.8) of forest were lost during this decade, compared with 72 km² (95% CI, 1.4–143.3) for those regions in the bottom quartile of the sprawl index. This analysis further finds the rate of tree canopy loss to be significantly associated with the rate of increase in EHEs over time, when controlling for metropolitan population size and growth rate (r = 0.30; p < 0.05). Based on this assessment, there is evidence to suggest that sprawling patterns of urban development may be influencing the frequency of EHEs through their effects on regional vegetative land cover.

The mechanisms through which extreme heat translates into human health effects are also complex. The incidence of heat-related illness in the United States has been level or slightly declining despite rising average temperatures since roughly 1980, with significant variability depending on incidence of heat waves [^{Centers for Disease Control and Prevention (CDC) 1995}, ¹⁹⁹⁹, ^2002b]. This relatively stable mortality rate is presumably due to increased prevalence of protective factors such as air conditioning. Differences in the incidence of heat-related morbidity and mortality between sprawling and compact cities have not been examined, and our data do not allow for such a comparison. It is possible that protective factors have thus far outweighed the influence of urban form on the incidence of heat-related illness. Projecting forward, however, the exposure amplification associated with sprawl may be increasingly important as average ambient temperatures continue to climb and eventually outpace physiologic adaptation thresholds in many regions. Indeed, as Shanghai’s urban heat island has grown, heat-related mortality rates have increased. This finding suggests that there is the potential for a similar trend in association with urban sprawl (^{Tan et al. 2010})—a question that deserves further study.

Our findings have clear implications for public health officials and urban planners. Most important, there is a need to incorporate land-use patterns into models that project climate change impacts over time in urban areas. Anticipating the increased exposure to extreme heat in cities, planners can work to control extreme temperatures through such strategies as preservation of regional green space; the installation of street trees, more reflective surfaces on roads and buildings, and green roofs; and replacement of vehicular travel by transit, walking, and bicycling—features all promoted through more compact design. Models suggest that urban albedo and vegetation enhancement strategies have significant potential to reduce heat-related health impacts (^{Silva et al. 2010}). These risk reduction strategies must be complemented by strategies that identify and protect vulnerable populations, standard elements of heat-wave preparedness plans (^{Bernard and McGeehin 2004}).

Such strategies, fortunately, do more than reduce the risk of heat waves. Sprawl is associated with a wide range of adverse exposures, including ozone exceedances (^{Stone 2008}), poor water quality (^{Tu 2007}), and adverse health outcomes from obesity (^{Ewing et al. 2003b}; ^{Lopez 2004}) to decreased physical activity (^{Garden and Jalaludin 2009}; ^{Rashad 2009}) to fatal road traffic injuries (^{Ewing et al. 2003c}; ^{Lucy 2003}; ^{Schlundt et al. 2004}). Interventions that increase density, green space, and public transit offer considerable co-benefits by reducing air pollution levels and the risk of injuries and promoting physical activity (^{Younger et al. 2008}). They also increase urban resiliency to other climate-related risks such as severe precipitation events; trees, for example, play a key role in managing stormwater runoff and flooding. With the increasing frequency and severity of environmental hazards such as heat, urban design strategies will play an important role in reducing vulnerability, promoting health, and building resilience.

This research was supported by the National Center for Environmental Health, U.S. Centers for Disease Control and Prevention (grant 300617201-03).

The authors acknowledge the valuable contributions of J. Vargo in compiling the data used in this study.