Introduction: Hearing impairment is an emerging modifiable risk factor for dementia, but the relative predictive abilities of subjective and objective measures of hearing in dementia risk prediction algorithms are unclear. The objective was to develop and validate prediction models for 3-year incidence of dementia in older Canadians, and to evaluate the independent contribution of self-rated hearing impairment and audiometry-based moderate hearing loss. -- Methods: Baseline (2011 to 2015) and 3-year follow-up data from the Comprehensive cohort of the Canadian Longitudinal Study of Aging were used to build logistic regression models for 3-year incidence of dementia. Individuals who were under 55 years of age, reported physician-diagnosed dementia at baseline, and/or did not have data on dementia status at follow-up were excluded, producing a sample of 19,830 older Canadians. Hearing impairment was defined subjectively as self-reporting fair or poor hearing (versus excellent, very good, or good hearing) and was defined objectively as having a better-ear pure-tone average of the speech-frequencies (500, 1000, 2000, and 4000 Hz) above 40 dB with audiometry. -- Results: Both hearing measures were associated with dementia incidence after adjustment with other risk factors (self-rated fair/poor hearing adjusted odds ratio (aOR) 1.76, CI 0.96-3.23, audiometry-derived hearing loss aOR 2.60, CI 1.38-4.87). Audiometrically-derived hearing loss and self-rated hearing had similar population discrimination (c-statistic of model with self-rated hearing = 0.803, CI 0.752-0.859, c-statistic of model with audiometrically confirmed hearing loss = 0.808, CI 0.762-0.870) and similar calibration. -- Conclusion: Due to the accessibility of the self-reported hearing measure, the use of self-rated hearing in dementia risk prediction tools may have a larger clinical impact than audiometrically-defined hearing ability. Model performance within subgroups (e.g., older age groups, hypertension status, etc.) must be evaluated in future work to assess the magnitude of miscalibration, if any, in the use of self-reported hearing ability compared to audiometry.

La santé est une construction dynamique et multifactorielle qui a une dimension individuelle et une dimension sociale. Cette dernière peut avoir un effet direct ou indirect sur les comportements des individus et sur leurs choix de vie. L’étude de Framingham a révélé que le risque cardiovasculaire était multifactoriel et donc son estimation devait être globale. Mais les méthodes d’estimation du risque cardiovasculaire ne prennent pas en compte l’environnement de vie des individus qui favoriserait le développement des facteurs de risque modifiables. A travers ce travail, nous mettons en lumière les caractéristiques de l'environnement de diffusion des facteurs de risque cardiovasculaires modifiables : métaboliques et comportementales. Depuis des décennies, la région Nord enregistre pour les maladies cardiovasculaires des taux élevés de mortalité globale et prématurée. Nous nous sommes interrogés sur les déterminants de cette situation de la région Nord, en la comparant aux autres régions françaises, et en explorant le lien entre un risque cardiovasculaire élevé et une situation économique défavorable. Dans ce but, nous avons étudié le lien entre la distribution du risque cardiovasculaire et la distribution du revenu, en employant des critères normatifs basés sur la notion de dominance sociale en termes de pauvreté.Ce travail permet d’analyser les connaissances en rapport avec l’environnement de vie des individus et ainsi une meilleure compréhension des mécanismes de diffusion des facteurs de risque modifiables, ce qui s’inscrit dans une double perspective, réduire l’incidence et la prévalence des maladies cardiovasculaires et diminuer les inégalités sociales de santé. / Health is a dynamic and multifactor construction which has both an individual and a social dimension. The latter may have a direct or indirect effect on the behaviour of individuals and their life choices. The Framingham study has revealed that cardiovascular risk is multifactorial and, as such, its estimate should be global. However, the assessment of global cardiovascular risk methods do not take into account the living environment of individuals, which would factor the development of modifiable risk factors. Through this work, we highlight the characteristics of the environment of dissemination of modifiable cardiovascular risk factors: metabolic and behavioural. Since decades, the North region of France has, for cardiovascular diseases, a high level of global and premature mortality. We are asked about the determinants of the situation of this region, by comparing it to other French regions and by exploring the link between a high cardiovascular risk and an unfavourable economic situation. Thus, we are interested in the link between the distribution of cardiovascular risk and the distribution of income, using normative criteria based on the concept of expected social dominance in terms of poverty. Shedding a light on factors favouring the occurrence of cardiovascular problems and analyzing the knowledge about the individual’s life environment allows a better understanding of the mechanisms of diffusion of the modifiable risk factors, with a double objective to lower the incidence and prevalence of cardiovascular diseases and to reduce the social inequalities in health.