The projected increase in the frequency and intensity of extreme heat events due to climate change means an associated increase in risk of heat-related illnesses and mortality. Public health systems need to be prepared to identify and reduce the susceptibility of vulnerable populations to increased occurrence of heat-related illness and stress. To facilitate this, climate services have begun developing climate change projections for heat-stress indices based on exceedances of thresholds used operationally in meteorological heat warning systems. This task is complicated by the fact that heat-stress indices are generally computed using hourly data whereas climate model outputs are often archived at daily or longer time steps. This study focuses on Humidex, a heat-stress index used in heat alerts issued by the Meteorological Service of Canada. Several potential solutions for computing robust Humidex indices using daily data are examined, including a new approximation method. Indices obtained with the new method are compared with indices obtained using the classic method based on hourly data as well as with other two methods based on average daily values. The new approximation gives good estimations for humidex indices, while the daily-average-value methods present biases with respect to the hourly-value method.