Table 1 Results of the regression analysis studies and RCT

Rogot et al. 1992 [65]

Comparing mortality during heat in people with air conditioned homes to those with no air conditioning

Central air condition compared to no air condition: OR below 1 for all groups, significant (p = 0.03 Mantel-Haenszel). Room air condition compared to no air condition: OR 0.96 for total group, p = 0.71). RR for central air condition vs. no air condition 0.58 for total group, RR for room air condition to no air condition 0.41 for total group

Smoyer 1998 [67]

Comparing mortality rates of 1980 and 1995

The average elderly mortality rate on heat wave days went down from 2.36 (SD 1.20) to 1.65 (SD 0.52), the average elderly mortality rate on non- heat days went down from 1.56 (SD 0.45) to 1.46 (SD 0.55)

Palecki et al. 2001 [64]

Comparing excess deaths in 1995 and 1999

Mortality rates in Chicago and St Louis both 1.4 per 100.000 in 1999, if not using core cities but counties. In 1995, 700 died in Chicago and 27 in St Louis

Weisskopf et al. 2002 [68]

Changes in population vulnerability

Model 1: predicted mortality rate of 1.80 per heat-index degree above 80 °F. 42.3 expected deaths, actual deaths in 1999 were 10. Model 2: RR for heat-related death in 1999: 0.17-0.24, RR for emergency medical services in 1999 0.32-0.46

Davis et al. 2003 [60]

Comparing temperature mortality relationship from 1964 to 1998

The threshold for 1960s-1970s is no longer connected to an increased mortality in the 1980s in Northeastern cities, and in the 1990s 10 show no elevated mortality above threshold and of the remaining 18 cities 12 show a decline in mortality rate. Six cities remain with an increased mortality rate above the threshold: Atlanta, Buffalo, Dallas, Denver, Seattle, San Francisco

Delaroziere and Sanmarco 2004 [52]

Comparing mortality before and after implementation of warning system

Mean index of daily excess mortality has dropped from 3.27 in the years 1986 to 1982, down to 1.32 in the years 1984–1997, p = 0.008)

Marinacci et al. 2009 [57]

Comparing no. of hospitalizations and deaths in summer 2004, RCT

Males: in intervention group Odds to be emergency hospitalized: OR 0.33, 95% CI: 0.11; 0.96. Females: in intervention group odds to be hospitalized overall: OR 0.96, 95% CI: 0.93; 0.98

Tan et al. 2007 [70]

Comparing daily excess mortality in 1998 and 2003.

Correlation coefficient between daily deaths and weather and air pollution parameters: death and time of heat wave: 0.34 in 1998 and 0.41 in 2003, Tmax in 1998 0.51 to 0.62 in 2003. Heat related deaths in 1998: 358 (absolutes), 253 in 2003 (absolutes)

De’Donato et al. 2008 [51]

Daily excess mortality before (reference period) and after implementation of heat warning system

J-shape temperature-mortality curve in all cities. In Milan and Rome in 2007 there was a weaker association between high temps and mortality. In Bari and Catania there was a greater impact of high temp on mortality in 2007 (all compared to 2003). In 2007 excess mortality occurred during three heat waves, with impacts on mortality of +10-41% in the center and 11-56% in the South

Fouillet et al. 2008 [53]

Comparing excess daily mortality in 2003 to 2006

During summers 2004 and 2005, observed no. of deaths was 2-8% lower than predicted no. of deaths. In 2006 2065 excess deaths occurred, predicted for that temperature were 6452 excess deaths, 4400 fewer deaths than predicted

Kysely and Kriz 2008 [55]

Comparing excess mortality in the 1990s and 2003

Excess daily mortality in 1990s: 98 deaths in 1992, 113 deaths in 1994; 50 deaths in 2003. Aggregated: 1992 718 excess deaths, in 1994 919 excess deaths, in 2003 236 excess deaths

Bargagli et al. 2009 [49]

Mortality rate among patients with active surveillance and those without = comparison of mortality rate with and without intervention

Excess mortality on heat days vs. non-heat days in controls: RR 1.20, 95% CI: 1.14-1.27; excess mortality on heat days vs. non-heat days in intervention patients: RR 0.95, 95% CI: 0.65-1.34

Chau et al. 2009 [69]

Comparing associations between hot weather warning and mortality rates from ischemic heart disease and stroke from 1997 to 2005.

Absence of warning system was associated with an increase of 1.23 deaths from IHD (95% CI 0.32; 2.14), an increase of 0.97 deaths from stroke (95% CI: 0.02; 1.92) per day

Ostro et al. 2010 [63]

Comparing hospitalization among those with air conditioning to those without

Reduction in excess risk of hospitalization with 10% increase in A/C ownership: respiratory disease: relative reduction 19.9% (95% CI 0.7;39.), CVD relative reduction: 49.1% (95% CI 19.9;78.3), heat stroke relative reduction 4.0% (95% CI 1.9;6.0)

Kysely and Plavcova 2012 [78]

Comparing temperature mortality relationship from 1986 to 2009

Significant trends in deviation of mortality on lag days from 1986 to 2009: all ages D + 1 -0.61, D + 2 -0.55; 70- years: D + 1 -0.66; 70+ years: D + 2 -0.66. Relative deviations of mortality declined by 0.4% to 0.5% in all age groups until 2009. Overall decline of mortality by 10% for all groups

Morabito et al. 2012 [58]

Comparing mortality before and after implementation of warning system

Odds Ratios for mortality by age group pre- and post-2003: only significant in 75 years+, OR for average apparent temperature before 2003 1.18 (CI 1.10-1.26), 2004 to 2005: 1.24 (CI 1.14-1.35), 2006–2007: 1.20 (CI 1.09-1.31). Also significant for maximum temperature

Schifano et al. 2012 [59]

Comparing daily mortality in 1998–2002 (before) and from 2006 to 2010 (after) implementation of prevention program

Weaker relationships between heat and mortality in all 16 cities post-intervention. Percentage change in mortality per 3°C increase in max apparent temperature MAT (pooled results): for 0 to 3% increase of 3°C increase: 1998–2002: 5.65%, for 2006 to 2010: 5.65%; 3 to 6% MAT increase: in 1998–2002 6.72% change, in 2006 to 2010: 7.79% change. Largest results: 12 to 15% MAT increase, 41.76% change from 1998–2002; 5.65% change from 2006 to 2010