Death rates become progressively higher when outdoor air temperature rises above or falls below 20-25 degrees C. This study addresses the question of whether this relation is largely attributable to the direct effects of exposure to heat and cold on the human body in general, and on the circulatory system in particular. The association between daily mortality and daily temperatures in the Netherlands in the period 1979-1987 was examined by controlling for influenza incidence, air pollution, and "season"; distinguishing lag periods; examining effect modification by wind speed and relative humidity; and distinguishing causes of death. Important direct effects of exposure to cold and heat on mortality were suggested by the following findings: 1) control for influenza incidence reduced cold-related mortality by only 34% and reduced heat-related mortality by 23% (the role of air pollution and "season" was negligible); 2) 62% of the "unexplained" cold-related mortality, and all heat-related mortality, occurred within 1 week; and 3) effect modification by wind speed was in the expected direction. The finding that 57% of "unexplained" cold-related mortality and 26% of the "unexplained" heat-related mortality was attributable to cardiovascular diseases suggests that direct effects are only in part the result of increased stress on the circulatory system. For heat-related mortality, direct effects on the respiratory system are probably more important. For cold-related mortality, the analysis yielded evidence of an important indirect effect involving increased incidence of influenza and other respiratory infections.

Health expectancy is a widely used measure for monitoring trends in the health of a population and assessing differences in health among population groups. However, no decomposition method is available to examine the contribution made by causes of death and disability to differences in health expectancy among population groups or periods. We present a method for decomposing differences in health expectancy, based on the Sullivan method. This method is an extension of the decomposition method for life expectancy developed by Arriaga. We illustrate the method and its added value by decomposing male-female differences in health expectancy for the Netherlands.

The aim of this study was to investigate the importance of 'cultural/behavioural' and 'materialist/structuralist' explanations for socio-economic inequalities in health, and to examine the interrelationship between them. We used data from a survey among a sample of the population in the southeastem part of the Netherlands. When analysed separately, both behavioural and structural factors contributed substantially to observed inequalities in health. In a simultaneous analysis, both groups of factors had a substantial part of their contribution in common. We defined the overlap as an indirect contribution of structural conditions, through behaviour. If that overlap is ignored, this could lead to an overestimation of the behavioural explanation. In our analysis, the total (direct plus indirect) contribution of structural factors is larger than that of behavioural factors. However, because of, in particular, the cross-sectional character of the data, these analyses must not be considered a. final answer as to the question of the relative contribution of behavioural and stmctural factors. Instead, they are an illustration of the way their importance could be assessed, taking the effect of structural conditions on lifestyle into consideration.

Study objective-The aim was to describe the pattern of seasonal variation in all cause mortality in The Netherlands, and to analyse the contribution of specific causes of death to the winter excess of all cause mortality.Design-Daily numbers of deaths in The Netherlands, by cause, were obtained for the period [1979][1980][1981][1982][1983][1984][1985][1986][1987]. Patterns ofvariation were analysed using Poisson regression. The model related the observed number of deaths to (1) the number expected for that day on the basis of person-days at risk by age and sex, (2) secular trend, and (3) first and higher order cosine terms where appropriate.Main results-All cause mortality has a bimodal peak in the first months of the year. After that it declines to reach a plateau in late spring. Mortality is lowest at the end of August, after which it rises steeply again. The winter excess of all cause mortality is primarily due to cardiovascular diseases (66%) and respiratory conditions (13%). Cardiovascular mortality peaks before respiratory mortality, suggesting different lag times in the effects of winter. There was an episode of exceptionally high mortality (above the normal winter excess) in early 1986, which was primarily due to cardiovascular diseases (39%) and respiratory conditions (25%). This episode was probably caused by a severe influenza epidemic, and was not followed by a compensatory lowering of mortality.Conclusions-The pattern of variation of mortality within the year suggests that it is not based on a simple relationship with climatological circumstances, because the latter fluctuate according to a less complex pattern. Cause specific data suggest an instantaneous effect of "winter" on the cardiovascular system, and a delayed effect mediated by respiratory infections.