Table 2 Characteristics and results of studies comparing effects of heat-waves on mortality

Kysely et al. 2012 [51]

Czech Republic

1986-2006

≥2 days with temperature >95^th quartile of distribution for given part of the year

All-cause and CVD mortality

Determined whether the deviation of observed deaths significant compared to expected deaths estimated by Monte Carlo method using data drawn from summers between 1986–2006.

Within common definition, length & intensity of HW allowed to vary between years.

Linear test for trend for deviation of mortality for hot spells between 1986 and 2009. Decrease in mortality over time found (significant at p = 0.05 level).

Decline of around 0.4-0.5 % deaths per year.

Hypothesised decreasing mortality due to acclimatisation to heat within a summer season in later years and/or increased adaptive measures such as improved living, health & building standards and increased heat awareness

Kysely et al. 2008 [52]

Czech Republic

2003 HW compared to period 1986-2006

≥3 days with average daily heat index exceeding 95 % quartile of distribution and ≥ 1 day exceeding 98 % quartile

All-cause and CVD mortality

Observed and expected mortality compared. Expected deaths over April-September period computed using smoothed 15 day running means corrected for weekly cycle and annual changes in mortality .

Within common definition, length & intensity of HW allowed to vary between years.

Taken together, the HW effects of 2003 were weaker than HW effects in previous years

Hypothesised that decreased effects of 2003 HW could be due to:

factors unrelated to adaptation – e.g. influenza epidemic affecting European countries in spring 2003 reducing number of susceptible individuals or

improved response to heat

Feuillet et al. 2008 [53]

France (all regions)

2006 compared to previous 29 years

2006 HW defined as period with consecutive days of alert in at least one (of 96) departments of France

All-cause mortality

Observed and expected mortality compared.

Expected mortality derived from baseline deaths predicted by model using data from previous 29 years: model included seasonal control and long-term mortality trend.

Modelled expected deaths from 2006 HW using model & actual deaths from 2006 HW using mortality figures.

4388 fewer deaths than estimated by predictive model for the 2006 HW

Larger decrease in the over 75 years

Hypothesised heat wave plans instigated post 2003 led to a decrease in heat wave related mortality.

Tan et al. 2007 [54]

Shanghai

2003 and 1998

≥3 days where daily maximum temperature exceeds 35 °C

All-cause mortality

Average number of deaths on heat days and non-heat days compared. Linear regressions run for 1998 and 2003 summers including mortality, temperature and air pollution concentrations to assess effect of length of HW, timing in summer and pollution.

Within common definition, length & intensity of HW allowed to vary between years.

Absolute deaths:

1998:

Average number deaths on non-heat days 244, heat days 358

2003

Average number deaths on non-heat days 223, heat days 253

Not adjusted for population size/age

Hypothesised decreased HW effects could be due to:

Urban green area increasing from 19.1 % to 35.2 % over the time period. Increased use of air conditioning and implementation of heat/health watch warning system in 2002

Rey et al. 2007 [55]

France (all regions)

Six Heat Wave periods between 1971 and 2003

≥3 days where max and min temp simultaneously greater than respective 95^th percentile

All-cause and cause-specific mortality

Observed and expected mortality ration (O/E) compared for each HW

Expected mortality calculated from observed mortality in previous 3 years using log-linear Poisson model of mortality rates (by month, year, age, gender, cause of death).

Within common definition, length and intensity of HW allowed to vary between years.

Observed-Expected (O-E) mortality (all cause)

1975 2952

1976 5116

1983 1473

1990 1624

2001 1330

2003 13734

In all six heatwaves, age >75 years were most vulnerable.

Mortality standardised by age and gender

Smoyer et al. 1998 [56]

St Louis, Missouri

1980 and 1995 heat waves

Days with Apparent Temperature > 40.6 °C (cut off for US National Weather service warnings)

All-cause mortality

Mortality –heat relationship modelled using Poisson regression, including terms for HW duration, temperature and interaction between heat wave duration and timing in season. Best models for 1980 and 1995 selected

Only > 65 years studied.

Simulated severe HW using 2 models:

Model 1: deaths estimated using 1980 weather data and 1980 model parameters (adjusted for 1995 population size)

Model 2:deaths estimate using 1980 weather data and 1995 model parameters

For a simulated HW: vulnerability increased using 1995 model parameters

(estimated number of deaths using 1980 parameters 446 (419,465) compared to 1995 model parameters (estimated number of deaths 481 (319,822)

Imprecise estimates make the difference between 1995 and 1980 models difficult to assess

.Between 1980 and 1995 the numbers of persons in the eldest age category and of older persons below the poverty line increased.

Air conditioning prevalence:

1980 64.1 %, 1991 86.7 %