Background: Mortality during sustained periods of hot weather is generally regarded as being in excess of what would be predicted from smooth temperature-mortality gradients estimated using standard time-series regression models. However, the evidence for an effect of continuous days of exceptional heat ("heat wave effect") is indirect. In addition, because some interventions may be triggered only during forecasted heat waves, it would be helpful to know what fraction of all heat-related deaths falls during these specific periods and what fraction occurs throughout the remainder of the summer.
Methods: Extended time-series data sets of daily mortality counts in 3 major European cities (London, 28 years of data; Budapest, 31 years; Milan, 18 years) were examined in relation to hot weather using a generalized estimating equations approach. We modeled temperature and specific heat wave terms using a variety of specifications.
Results: With a linear effect of same-day temperature above an identified threshold, an additional "heat wave" effect of 5.5% was observed in London (95% confidence interval = 2.2 to 8.9), 9.3% in Budapest (5.8 to 13.0), and 15.2% in Milan (5.7 to 22.5). Heat wave effects were reduced slightly when we relaxed the linear assumption and these effects were reduced substantially when temperature was modeled as an average value of lags 0 to 2 days. In London, fewer than half of all heat-related deaths could be attributed to identified heat wave periods. In Milan and Budapest, the fraction was less than one fifth.
Conclusions: Heat wave effects were apparent in simple time-series models but were reduced in multilag nonlinear models and small when compared with the overall summertime mortality burden of heat. Reduction of the overall heat burden requires preventive measures in addition to those that target warnings and responses uniquely to heat waves.