| Falls in the elderly represent a major problem. It is therefore not surprising that the identification of factors that may increase the risk of falls has received much attention. Frailer seniors who need support to live in the community remained nonetheless poorly documented, although more recently, the Quebec authorities have given high priority to interventions that target this population.;In this thesis, we review the existing methods and propose a Cox hazards extension. We apply it in the study of potential fall-risk factors associated with 959 community-dwelling seniors using home-care services. Finally, we compare the results of the proposed Wei, Lin, & Weissfeld (WLW) method with those of several other techniques, notably the conventional logistic regression, the pooled logistic regression, the negative binomial regression and the Andersen & Gill regression. At baseline and every six months thereafter, participants were visited at home in order to ascertain information about potential risk factors. Falls were monitored by use of a calendar and monthly phone calls. Baseline exposure variables and updated time-varying exposures include socio-demographic characteristics, BMI, nutritional risk, gait and balance, alcohol consumption, home hazards, and medications.;Almost all (99.6%) of participants showed at least one high risk factor. Exposure to multiple risks was frequent, with an average of 2.7 different high-risk factors per participant. The risk factors significantly associated to the risk of falling include male sex, age, history of falling, Berg balance score, BMI, use of benzodiazepines, number of home hazards and residential facility for seniors. Results demonstrate that the usual methods of analyzing risk factors for falling (any sort of fall as well as those leading to medical consultations) are inappropriate, as they produce considerable biases relative to the WLW model using time-dependent covariates. Bias for the considered effect measures comes from the manner in which the observed data (both measured exposures and health outcomes) was measured and defined as well as the way in which the statistical analysis took into account this information.;An additional part of the thesis was undertaken to identify risk profiles of subjects regarding the recurrence of falling, defined as participants who reported at least two falls within six months of initial assessment at entry in the study. A classification and regression tree analysis classified the population into five groups differing in risk of recurrent falling, based on history of falls in the three months prior to the initial interview, Berg balance score, type of housing, and usual alcohol consumption in the six months preceding study entry. The relative risks varied from 0.7 to 5.1. A subsequent survival analysis showed that the length of time before becoming a recurrent faller varies among risk profiles.;Risk factors for falls are usually identified by observational prospective studies. Their optimal identification is however complicated by the fact that exposure may vary during the follow-up, and that an individual may experience more than one event. Twenty years ago, some researchers attempted to sensitize their peers in this respect, but their efforts were vain. Researchers continue today to neglect these considerations and to use improper statistical techniques, focusing on the proportion of fallers or the time to first fall. In doing so, we discard a significant amount of relevant information.;This thesis discusses highly topical subjects about a target population and a fall-risk screening activity which are priorities in the public health sector in Quebec. We encourage researchers interested in the identification of risk of falls among the elderly to use the statistical method of Wei, Lin and Weissfeld because it takes into account updated time-varying exposures and multiple events. More research will be necessary to determine the best screening test for a given risk-factor in this setting and population.;Keywords : Accidental falls; Cox model; elderly; environmental hazards; home care services; multiple classification analysis, gait and balance; public health intervention; risk factors, survival analysis. |
