Over the past twenty years, the analysis of short-term links between air pollution and health has been the subject of numerous studies. The most commonly used method is time-series analysis. This method aims to quantify the potential links between daily variations in the level of an air pollution exposure indicator and those in the number of occurrences of a health event (deaths, hospitalizations, etc.). To obtain an unbiased estimate of the short-term association between air pollution and health, it is necessary to account in the analysis for all factors that may modify this association (these factors must be related to levels of the air pollution indicator and to the health indicator). The current approach is a Poisson regression based on a generalized additive model (GAM). This model uses smoothing functions that allow the model to be fitted as closely as possible to the shape of the relationships between the health variable and the explanatory variables. Thus, the model incorporates the variable of interest (atmospheric concentrations of a pollutant) as well as long-term trends and seasonality, day of the week, holidays, vacation periods, weather conditions (temperature), flu outbreaks, pollen seasons, etc. Health data often exhibit overdispersion, which is accounted for by assuming a quasi-Poisson distribution for the health variable. The parameters of the smoothing function (penalized spline) used to account for the long-term trend and seasonality are selected to minimize the partial autocorrelation of the residuals. The regression model allows for estimating the coefficient associated with the pollution indicator and calculating a relative risk. Repeating such time-series studies over successive periods is very useful because it enables epidemiological monitoring of short-term links between air pollution and health using routinely recorded data.