Surveys of hard to reach populations (rare, no known sampling frames) have been, for some years, the object of methodological reflection. Various methods aiming at the production of an ‘extrapolable’ sample of these populations have been proposed: time-space sampling (TSS) or time-location sampling (TLS), respondent driven sampling (RDS), or the ‘capture - recapture’ method. After defining what a hard-to reach-population is, this article provides an outline of these various approaches before going on to briefly consider the papers contained in this special issue.
Keywords: Hard-to-reach populations; surveys; time-location sampling (TLS); time-space sampling (TSS); respondent driven sampling (RDS); capture-recapture

Les enquêtes auprès des populations difficiles à joindre (rares, dépourvues de base de sondage) font, depuis quelques années, l‟objet de réflexions. Différentes méthodes d‟investigations permettant d‟obtenir un échantillon extrapolable de ces populations ont été proposes : l‟échantillonnage de lieux-moments (TSS ou TLS en anglais), l‟échantillonnage dirigé par les répondants (RDS en anglais), ou la méthode de « capture-recapture ». Après avoir défini ce qu‟est une population difficile à joindre, cette introduction dresse un panorama de ces méthodes d'enquêtes, puis présente les différentes contributions réunies dans ce numéro spécial.
Mots-clés : population difficile à joindre ; enquêtes ; échantillonnage lieux-moments (TSS) ; échantillonnage dirigé par les répondants (RDS) ; capture-recapture

Since the mid-1990s, Time-Location Sampling methods have been widely used in France in quantitative research about the homeless (Marpsat and Firdion, 2000; Brousse, 2002) and drug users (Emmanuelli et al. 2003; Jauffret-Roustide et al. 2006). Adapted from methods of indirect sampling initially developed in the United States (Burt and Cohen, 1989; Dennis and Iachan, 1993), they enable the design of representative samples of hard-to-reach populations for whom there are no sample frames, by basing it on support facilities for the populations of interest: day centres, soup kitchens, accommodation services, health care, mobile teams etc. However, in transition from the theoretical protocol to its implementation, various difficulties emerge. In particular, the necessary involvement of these support services; the constraints associated with the administering of questionnaires in sometimes very particular conditions (at night, in public places or while following mobile teams) and the singular organisation of each of the services that has to be involved means that challenges occur as regards coverage of the field being investigated and the representativeness of the sample. These challenges mean that the method has to be appropriately and pragmatically adapted to the various configurations that are met with. In this perspective, we will show how the integration of qualitative methods, such as the observation of the services and the realisation of in-depth interviews with the persons in charge of them, have contributed, along the various data collections and with the accumulation of experiences, to improve the methodology of and the scope covered by quantitative surveys.
Keywords: Time-location sampling, qualitative methods, fieldwork, homeless, drug users

Depuis le milieu des années 1990, les méthodes d’échantillonnage espace-temps (ou Time-Location Sampling) ont été largement utilisées en France dans les recherches quantitatives auprès des populations sans domicile (Marpsat et Firdion, 2000 ; Brousse, 2002) et d’usagers de drogues (Emmanuelli et al. 2003 ; Jauffret-Roustide M. et al. 20061). Adaptées de méthodes d’échantillonnage indirect initialement développées aux Etats-Unis (Burt et Cohen, 1989 ; Dennis et Iachan, 1993), elles permettent de construire des échantillons représentatifs de populations difficiles à joindre, pour lesquelles il n’existe pas de base de sondage, en s’appuyant sur les dispositifs d’aide aux populations concernées : services d’accueil, de restauration, d’hébergement, de soins, équipes itinérantes… Toutefois, du protocole théorique à sa mise en oeuvre pratique, diverses difficultés émergent. En particulier, la nécessaire implication de ces services d’aides, les contraintes associées à la passation de questionnaires dans des conditions parfois très particulières (de nuit, dans des espaces publics ou au sein d’équipes mobiles…), l’organisation singulière de chacun des services dans lesquels on est amené à enquêter, invitent à réfléchir aux enjeux de couverture du champ investigué mais aussi de représentativité de l’échantillon, et à adapter la méthode aux différentes configurations rencontrées. Dans cette perspective, nous montrerons comment l’intégration de méthodes d’enquêtes qualitatives, telles que l’observation des services et la réalisation d’entretiens auprès des responsables, a contribué, au fil des collectes et des expériences capitalisées, à améliorer la méthodologie et le champ couvert par les enquêtes quantitatives.
Mots-clés : Echantillonnage espace-temps, méthodes qualitatives, travail de terrain, sans domicile, usagers de drogue.

The French ANRS-Coquelicot survey was conducted between 2004 and 2007 to follow the dynamics of the epidemic of hepatitis C among drug users (DU) in France. Having taken place in 5 cities (Lille, Strasbourg, Paris, Bordeaux, Marseilles), the survey was constituted of two parts, one epidemiological and the other socio-anthropological. The target population of this survey was both difficult to define and hard to reach, because of diversified social profiles, relations to drug use, and of the stigmatization of this practice. The studied DU population is the one that frequents specialised services, represents a sub-population having a ‘problematic relationship to drugs’ and is rather socially precarious. The methodology used in the epidemiological part of the study implemented a strategy of random sampling via a sampling scheme integrating the Generalized Weight Share Method (GWSM) and relies on the collection of biological data to estimate the prevalence of hepatitis C. The use of innovative methodological tools used for the first time in France (sampling plan with GWSM, biological tests) in the ANRS-Coquelicot survey has enabled to produce valid estimates of the prevalence of the HCV in the DU population, a particularly high prevalence which raises to 60%. The methodology implemented in this survey enabled a generalisation of its results to the studied target population. Using a mixed methods approach, the epidemiological data and qualitative data (in-depth interviews and ethnographical observations) were combined to enable a better understanding of the social context of risk and the meaning that the DUs gave it. The use of both quantitative and qualitative methodologies thus allowed a more sophisticated analysis of the risk factors involved in the exposure to hepatitis C among drug users in France.
Keywords: Methodology, drug use, infectious risk, epidemiology, socio-anthropology