On some approaches to calculation of health risks caused by temperature waves

The paper dwells on techniques applied for assessing impacts exerted by environmental factors on population health which have become conventional all over the world over recent years. The greatest attention is paid to up-to-date approaches to calculating risks of additional mortality which occurs in big population groups during cold and hot temperature waves. The authors consider basic stages in direct epidemiologic research: temperature waves definition; statistics hypotheses formulation; models specification; statistical criteria sensitivity, and statistical validity of the obtained results. As per long-term research performed by us in various Russian cities, we constructed logistic curves which show probability of obtaining significant risk assessment results for small samplings. We recommend to apply percentiles of long-term average daily temperature distributions as temperature thresholds when identifying temperature waves; in our opinion, such thresholds correspond to perceptions of extreme (for this or that region) temperatures and provide comparable results in terms of expected waves quantity in different climatic zones. Poisson's generalized linear model for daily mortality is shown to be the most widely spread technique for calculating risks caused by hazardous environmental factors. It is advisable to allow for an apparent correlation between mortality and time and air contamination in any regression model. We can allow for meteorological conditions which influence heat balance (air humidity and wind speed) either via including them apparently into a model or via bioclimatic indexes application; research in this sphere is going on. When calculating risks, it is advisable to allow for time lags between extreme temperatures waves and changes in mortality. We revealed that minimal population of a typical city for which it is possible to obtain statistically significant assessment of risks caused by heat waves ensembles is about 200 000 people.