The French Sentinelles network in primary care
A collaboration between general practionners, pediatricians and researchers
The French National Institute of Health and Medical Research (Inserm) and a university in Paris (Sorbonne Université) have developed, beginning in 1984, an information system based on a computer network of physicians in France, which is called the French Sentinelles network (Réseau Sentinelles in french). It allows compilation of large databases on communicable diseases, with anonymous individual description of cases for surveillance and research. This network is currently constituted of approximately 1300 volunteers general practitioners (GPs) and 100 volunteers community-based pediatricians, working throughout the metropolitan regions of France (2% of the total GPs and 4% of the total community-based pediatricians in these regions). This network is regulated by the Pierre Louis Institute of Epidemiology and Public Health (IPLESP), in collaboration with the French Public health agency (Santé publique France). It has obtained a research autorisation from the french independent administrative authority protecting privacy and personal data (CNIL), n°471 393.
1) Continuous surveillance of health indicators
This national system of clinical surveillance collects real-time epidemiological data, originating from the participating general practitioners and community-based pediatricians, to be used in analysis, forecasting and redistribution.
In 2019, the French Sentinelles surveillance program focuses on 10 health indicators (nine infectious diseases and one non-infectious indicator). Asthma attacks, hospitalizations, measles, HIV test and hepatitis are no longer included in Sentinelles because a more efficient surveillance system, mandatory notification, covers those conditions. The electronic system has been developed to be flexible and is able to abandon or add a new condition easily according to the evolution of the situation in France.
Each week, Sentinelles physicians transmit data from their patient consultations on the selected indicators via secure Internet connections. This site follows the current requirements in France on confidentiality in information systems. From these data, it is possible to estimate the weekly incidence rate for each indicator and trace the evolution in time and geography. To estimate the weekly or yearly national incidence rate, the mean number of cases per Sentinelles GP (standardized according to their participation and their geographical distribution) is multiplied by the total number of GPs in France and the result is divided by the French population from that year (more details in french).
Epidemiological data in real time
In the French Sentinelles system, information feedback is organized as a platform allowing for data query. The interface is user-friendly in the sense that any end-user may explore the database without any computer language or without any SQL programming. It is possible to download maps, time-series or tables on any of the health indicators monitored by the Sentinelles physicians since 1984. The database is updated weekly and it contains the whole series without any interruption in time or space location. Maps are built "on-the-fly" therefore, nothing is stored into the MySQL database except the raw data. The site receives more than 200 000 hits per month. French media, newspapers and TV channels, download maps during epidemic periods for influenza, gastroenteritis, and varicella. In addition, any Internet user has free access without any restriction.
To help the user interpret the data, the site provides a weekly epidemiological bulletin in French and English, called Sentiweb-Hebdo, published on the Sentinelles website and emailed to all registrants. Detection of national and regional influenza or acute diarrhea outbreaks are published weekly, as are forecasts of epidemic trends for the three weeks ahead (see below sections "Detection of epidemics of influenza and gastroenteritis" and "Epidemic forecasting").
An annual report is published in French on the website with an analysis of the health indicators under surveillance. These reports are available for online viewing.
2) A platform for mathematical research
Sentinelles physicians have contributed epidemiological data on selected communicable disease in France since 1984. These data have enabled national and international collaborations on epidemiologic evaluations and research; below are some examples:
Modeling infectious diseases to predict the effect of proposed interventions and to assist in decision making: By reproducing a theoretical distribution of infection and its consequences, it is possible to estimate the basic parameters of the transmission, to assess a priori the medical, economic and social impact of control strategies and interventions, and to identify information needed to be collected.
Detection of epidemics of influenza and gastroenteritis. The detection of outbreaks of influenza-like illness and acute diarrhea is made by a periodic regression model applied to the historical incidence rates series. It achieves an epidemic threshold for the current week of calculation. By comparing the incidence rate observed and the epidemic threshold, we can determine if an outbreak has begun.
- Epidemic forecasting. The French Sentinelles network uses two epidemic prediction models:
Model based on historical incidence rate: This method, derived from a model used in meteorology, has been adapted by Viboud et al to forecast epidemic trends. This method uses vectors selected from historical ILI time series that match current activity. Looking at the incidence rate that followed selected historical sequences, we can predict the incidence rates that will follow in the next three weeks.
Model based on drug sales: In addition to physician data, the Sentinelles network receives weekly data on sales of drugs provided by IMS Health, France. Research by Vergu et al. showed a high correlation between sales of certain classes of drugs and the ILI incidence, with a time lag: the increase in drug sales occurs before the increase of ILI incidence rate. A Poisson regression model predicts the ILI incidence rate at week t +1, t +2 and t +3 using sales of drugs at week t and ILI incidence rate at week t.
Assessment of the field effectiveness of influenza vaccine. Since 1995, the field vaccine effectiveness (FVE) of influenza vaccines is monitored inside this surveillance system using the screening method. The screening method compares the vaccine coverage of ILI cases with the coverage in the population from where the cases are drawn. The vaccine coverage of ILI cases is usually estimated based on ILI cases reported by Sentinelles GPs ?vaccination status is reported with each ILI case. Data on vaccine coverage in the population are obtained, at no cost to the program, from an external source (observational study). Then they can easily be integrated to the surveillance system and used for quick estimates of FVE. Thus, the process can be reproduced each year, allowing for year-to-year comparison.
Website for the detection and measurement of epidemics. A freely accessible analysis tool has been provided online to facilitate the processing of epidemiological time series data (Periodic). The tool can detect outbreaks in a series of observations, measure the magnitude, or establish an epidemic threshold for prospective surveillance. It is the result of a synthesis of periodic regression models. A generic method to construct periodic regression models has been developed. A user-friendly Web-tool is provided to facilitate analysis by non-statisticians.
3) A platform for epidemiological studies
The French Sentinelles network is regularly involved in epidemiological studies. Study results are posted as final reports on the website of the Sentinelles network, presented in international conferences, and published (specific section in french).