The objective of this study was to propose a statistical method for detecting local outbreaks of seasonal influenza based on data from SOS Médecins. Introduction: At the regional level, the Regional Health Surveillance Units (Cire) verify the presence of influenza-like illness outbreaks by cross-referencing multiple sources, primarily diagnoses made by general practitioners (the unified GROG-Sentinelles-InVS network, SOS Médecins). Methods: The inclusion criterion for SOS Médecins associations in the study was a diagnosis coding rate of at least 75% during the three epidemics from 2010 to 2013. A weekly alert threshold was constructed using a periodic regression model (Serfling) to identify seasonal clinical influenza epidemics as observed by SOS Médecins. The percentage of clinical influenza cases among coded SOS diagnoses was selected as the surveillance indicator. The consistency of this indicator with local indicators and with the spread of the epidemic was verified over the three winters of 2010–2013. Results: Twenty-one associations were included out of 59. Good agreement was observed between the influenza epidemic periods defined by consensus within the CIREs based on several indicators and the epidemic periods defined using the SOS Médecins indicator with the weekly threshold (94% agreement for detection and 78% for the end of epidemics). The spread across the 12 regions lasted 3 to 4 weeks, and the first regions affected were also those reported by the unified Sentinelles-GROG-InVS network. Conclusion: The satisfactory agreement with other surveillance systems allowed the percentage of clinical influenza cases among SOS Médecins diagnoses to be selected as an indicator for detecting the presence of seasonal influenza epidemics at the regional level. (R.A.)