The aim of this work was to determine the criteria for heat loads associated with an increase in mortality in different climatic regions of Croatia. The relationship between heat stress and mortality was analysed for the period 1983-2008. The input series is excess mortality defined as the deviations of mortality from expected values determined by means of a Gaussian filter of 183 days. The assessment of the thermal environment was performed by means of physiologically equivalent temperature (PET). The curve depicting the relationship between mortality and temperature has a U shape, with increased mortality in both the cold and warm parts of the scale but more pronounced in the warm part. The threshold temperature for increased mortality was determined using a scatter plot and fitting data by means of moving average of mortality; the latter is defined as the temperature at which excess mortality becomes significant. The values are higher in the continental part of Croatia than at the coast due to the refreshing influence of the sea during the day. The same analysis on a monthly basis shows that at the beginning of the warm season increased mortality occurs at a lower temperature compared with later on in the summer, and the difference is up to 15 °C between August and April. The increase in mortality is highest during the first 3-5 days and after that it decreases and falls below the expected value. Long-lasting heat waves present an increased risk, but in very long heat waves the increase in mortality is reduced due to mortality displacement.