Abstract
ObjectiveTo describe the results of the new organization of influenzasurveillance in France, based on a regional approach.IntroductionIn France, until winter 2014-2015, management and preventiveactions for the control of the flu epidemic were implemented whenthe national incidence of influenza-like illness (ILI) consultationsin general practice was over an epidemic threshold. The 2014-2015influenza epidemic had a major public health impact, particularly inthe elderly, and caused a severe overloading of the health care system,in particular emergency departments (ED) [1]. The epidemic alertemitted by the French National Public Health Agency at the nationallevel was too late for the hospitals to prepare themselves in manyregions.After a national feedback organized in April 2015 with allpartners involved in influenza surveillance and management, it wasrecommended to improve influenza surveillance in France following3 axes: 1) regionalize surveillance so that healthcare structures canadapt to the particular situation of their region; 2) use a pre-epidemicalert level for better anticipating the outbreak; 3) use multiple datasources and multiple outbreak detection methods to strengthen thedetermination of influenza alert level.MethodsA user-friendly web application was developed to provide commondata visualizations and statistical results of outbreak detectionmethods to all the epidemiologists involved in influenza surveillanceat the national level or in the 15 regional units of our agency [2].It relies on 3 data sources, aggregated on a weekly time step: 1) theproportion of ILI among all coded attendances in the ED participatingto the OSCOUR Network [3] ; 2) the proportion of ILI among allcoded visits made by emergency general practitioners (GPs) workingin the SOS Médecins associations [3]; 3) the incidence rate of ILIestimated from a sample of sentinel GPs [4].For each region each week, 3 statistical outbreak detection methodswere applied to the 3 data sources, generating 9 results that werecombined to obtain a weekly regional influenza alarm level. Basedon this alarm level and on other information (e.g.virological data),the epidemiologists then determined the epidemiological status ofeach region as either 1) epidemic-free, 2) in pre/post epidemic or 3)epidemic.The R software was used for programming algorithms and buildingthe web interface (package shiny).ResultsThe epidemiological status of influenza at the regional level wascommunicated through maps published in the weekly influenzareports of the Agency throughout the surveillance season [5].In week 2016-W03, Brittany was the first French region to declarethe influenza epidemic, with nine other regions in pre-epidemic alert.The epidemic then spread over the whole mainland territory. The peakof the epidemic was declared in week 11, the end in week 16.ConclusionsThis regional multi-source approach has been made possible bythe sharing of data visualizations and statistical results through a webapplication. This application helped detecting early the epidemicstart and allowed a reactive communication with the regionalhealth authorities in charge of the organization of health care, themanagement and the setting up of the appropriate preventivemeasures.