This report responds to a request from the Directorate General for Health. It presents the current state of the literature on the relationship between cold weather and health (physiological mechanisms, risk factors, diseases), contextual information on past cold waves, the InVS’s activities in this area, and measures taken in other countries. An assessment of measures to protect the population against cold weather was conducted, distinguishing between routine and emergency actions. An indicator and predictive thresholds for health impacts were identified. The study, conducted in the inner suburbs of Paris and Marseille (1984–2003), is based on time-series analysis to define cold-related excess mortality. A sensitivity/specificity analysis revealed that a dual indicator combining minimum and maximum temperatures over three days, associated with a cumulative excess mortality of 15% from Day 4 to Day 6, was the most relevant, with predictive thresholds for excess mortality of -9/-2 °C (Tmin/Tmax) in Paris, and -3°/+4 °C in Marseille. If these thresholds were used to trigger alerts, this would result in an average of 2.2 alerts per year in Paris and 1.4 in Marseille, with 1 false alert for every 3 true alerts in both cities, thereby avoiding an excessive number of false alerts that could undermine the system’s credibility. However, implementing an alert system is not appropriate because the actions needed to reduce the population’s vulnerability are primarily long-term measures. Short-term measures, which require alerting the public and local authorities, can be carried out using Météo-France’s alert map according to procedures to be defined in consultation with them. (R.A.)