Most of the literature linking the living conditions of the elderly with their health outcomes refers to the population living in private dwellings, while studies dealing with the topic from a broader perspective, including the population living in institutions, are sparse. This can be ascribed to the fact that nation-wide surveys on health do not generally include the population living in institutions, and to a strong selection of the institutionalised population, which calls for specific techniques to deal with the induced bias in estimators.

Our study aims at highlighting the role of family support on elderly survival looking at private vs collective households.

We examine the relationship between several aspects of living conditions and mortality in a cohort of 16,263 individuals aged 55 and over, living in community and in institutions in France. To disentangle the role of the family on survival while dealing with possible selection bias due to the peculiar characteristics of individuals that enter institutions, a two-stage weighted proportional hazards model is used. At the first stage, weights are estimated for institutionalised individuals through a logistic regression model fitted to the whole sample; at the second stage, a weighted Cox model is fitted to the institutionalized population.