L'étude de l'insomnie en population générale requiert une méthodologie d'enquête rigoureuse afin de s'assurer de la fiabilité des données recueillies. Ce travail présente la méthodologie utilisée pour effectuer une étude épidémiologique sur l'insomnie en France : 5 622 répondants représentatifs de l'ensemble de la population française ont participé à cette étude. Le rationnel motivant les choix méthodologiques est passé en revue : le tirage de l'échantillon et la sélection des sujets dans chaque foyer. Le système expert, Sleep-EVAL, a été utilisé pour réaliser les entretiens téléphoniques. Son fonctionnement, ses capacités et les types de données recueillies sont décrits.

The study of sleep disorders in the general population involves several methodological issues that need to be defined prior to proceeding to the epidemiological study. The rigor of the methodology is an important issue since it will determine the reliability of the data gathered. This paper describes the methodology used in an epidemiological study performed in the French general population using telephone interviews with the help of Sleep-EVAL, an expert system designed for this purpose. The study aimed to investigate the prevalence of insomnia disorders according to the DSM IV classification and the use of psychotropic medications in the general population. The methodological choices for this study were based on several considerations. First, the sample had to be representative of the French population. Second, the study had to be conducted in the shortest period of time. Third, the interviews had to be conducted with respect to a strict standardization and fourth, the realization costs had to be minimal for a maximum of data collected. The telephone interview procedure was chosen over postal and face-to-face interviews because it offered the possibility of conducting all the interviews from the same site. Supervision was easier. It also offered an absolute control in the application of the selection procedure. To draw the sample, a two-stage procedure was adopted. At the first stage, we pulled a random series of telephone numbers in each Nielsen region with respect to the size of the settlement. At the second stage, during the initial telephone contact, a household member was chosen using the Kish selection procedure. This method is based on the utilization of eight tables of selection that allows for the choice of the person to interview in a given household and keeps the representativeness of the sample. This technique is little used in telephone surveys because of its burden and its intrusive nature : the interviewer must collect the age and gender of all eligible subjects, to classify men from the oldest to youngest and then to classify women. However, it is the most rigorous selection method for epidemiological surveys. To reduce the refusal rates and to alleviate the work of interviewers, the Kish method was implanted in the computer software used for this study. For this study, the exclusion criteria were minimal. Only individuals younger than 15 years of age, those with a speech or hearing impairment and those who were too ill to perform the interview were not included. Subjects who refused to participate, those who hung up without speaking to the interviewer and those who hung up before completing at least half of the interview were tabulated as refusal. The participation rate was calculated by dividing the number of completed interviews by the number of eligible participants (completed interviews, refusals and telephone numbers where the interviewer was unable to determine if the individuals met an exclusion criterion). In this study, the participation rate was 80.8 % (5 622 completed interviews/6 966 eligible households). The diagnostic tool used for this study was the Sleep-EVAL system, an expert system designed to conduct epidemiological studies in the general population. It is a level 2, non-monotonic system endowed with a causal reasoning able to provide sleep and mental disorders diagnoses according to the DSM IV classification for this study. Subsequent versions of Sleep-EVAL also included the International Classification of Sleep Disorders. System symbolic representation of the classifications was put in a compiled knowledge base. This knowledge base was read and interpreted by the inference engine at the beginning of the interview. During the interview, this interpretation changed as a function of the answers provided by the interviewee and by deductions made from the analysis of information the system already knew. All interviews began with a standard questionnaire about sociodemographic information and sleep habits. From these first answers, the Sleep-EVAL system emitted a series of diagnostic hypotheses that were confirmed or rejected with supplementary questions. The interview ended once all diagnostic possibilities were exhausted. The validity of the Sleep-EVAL system was demonstrated in different studies performed in sleep disorders clinics. There were several advantages in using such a tool to conduct epidemiological surveys. No special skills from the interviewers nor specific knowledge of sleep and mental disorders were required. All the questions were chosen and formulated by the Sleep-EVAL system. The interviewer had simply to read the questions as they appeared on the monitor screen and enter the interviewee's responses by clicking the appropriate answer or typing it on the keyboard. Missing answers were non-existent because there was no possibility of skipping a question or entering inconsistent answers. It also ensured the uniformity of the interviews. Furthermore, it allowed the exploration of infrequent diagnoses. In summary, the methodology used for this study allowed for the investigation of the sleep pathology of the French population in a short period of time : only three months were necessary to complete the 5 622 interviews. The use of a computerized tool greatly facilitated the training of the interviewers and also their work. Furthermore, it ensured a standardized administration of the interviews and the exploration of a broad range of disorders that could hardly be realized with traditional paper-pencil questionnaires.