Stephen W Hwang1. 1. Inner City Health Research Unit, St. Michael's Hospital, University of Toronto, Toronto, ON. hwangs@smh.toronto.on.ca
Abstract
BACKGROUND: Homeless people suffer from high levels of morbidity and mortality, but there is surprisingly little empiric evidence that homelessness has a direct adverse effect on health. METHODS: This study examined the relationship between shelter use and risk of death using longitudinal data in a cohort of 8,769 homeless men in Toronto, Ontario. Shelter use was modelled as a time-dependent covariate in a Cox regression analysis. RESULTS: In a model adjusted for age and previous pattern of homelessness, the risk of death during months in which homeless shelters were used was significantly increased (hazard ratio, 1.84; 95% confidence interval, 1.27-2.67). CONCLUSIONS: Among men, periods of homeless shelter use are associated with higher mortality. There are three reasons why this finding does not necessarily mean that homelessness itself increases the risk of death. First, the hazard of death associated with shelter use compared to non-shelter use may be significantly different from that associated with homelessness compared to non-homelessness. Second, the association between shelter use and risk of death may be confounded by other variables such as alcohol and drug use. Finally, because the mechanism and time-course of the putative effect of homelessness on health is uncertain, appropriate modelling of the time-dependent covariate is difficult to ensure. Further research into the possible adverse effects of homelessness on health is needed and would have important implications for public policy.
BACKGROUND: Homeless people suffer from high levels of morbidity and mortality, but there is surprisingly little empiric evidence that homelessness has a direct adverse effect on health. METHODS: This study examined the relationship between shelter use and risk of death using longitudinal data in a cohort of 8,769 homeless men in Toronto, Ontario. Shelter use was modelled as a time-dependent covariate in a Cox regression analysis. RESULTS: In a model adjusted for age and previous pattern of homelessness, the risk of death during months in which homeless shelters were used was significantly increased (hazard ratio, 1.84; 95% confidence interval, 1.27-2.67). CONCLUSIONS: Among men, periods of homeless shelter use are associated with higher mortality. There are three reasons why this finding does not necessarily mean that homelessness itself increases the risk of death. First, the hazard of death associated with shelter use compared to non-shelter use may be significantly different from that associated with homelessness compared to non-homelessness. Second, the association between shelter use and risk of death may be confounded by other variables such as alcohol and drug use. Finally, because the mechanism and time-course of the putative effect of homelessness on health is uncertain, appropriate modelling of the time-dependent covariate is difficult to ensure. Further research into the possible adverse effects of homelessness on health is needed and would have important implications for public policy.
Authors: W R Breakey; P J Fischer; M Kramer; G Nestadt; A J Romanoski; A Ross; R M Royall; O C Stine Journal: JAMA Date: 1989-09-08 Impact factor: 56.272
Authors: Sandra K Schwarcz; Ling C Hsu; Eric Vittinghoff; Annie Vu; Joshua D Bamberger; Mitchell H Katz Journal: BMC Public Health Date: 2009-07-07 Impact factor: 3.295