Objective: To examine the effect of homelessness on mortality. Methods: This 15-year retrospective longitudinal cohort study compared mortality outcomes of homeless and non-homeless adults attending the emergency department of an innercity public hospital in Melbourne, Victoria between 1 January 2003 and 31 December 2004. Homeless individuals had ≥1 recorded episodes of homelessness within the recruitment period, categorised by type: primary, secondary, tertiary, marginally housed. Non-homeless individuals were stably housed throughout. Results: Over 15 years, homeless individuals had a higher mortality rate (11.89 vs. 8.10 per 1,000 person-years), significantly increased mortality risk (rate ratio 1.47, 95% confidence interval [CI] 1.26-1.71) and younger median age at death (66.60 vs. 78.19 years) compared to non-homeless individuals. Using adjusted Cox proportional hazards models, primary (hazard ratio [HR] 2.05, 95%CI 1.67-2.50), secondary (HR 1.60, 95%CI 1.23-2.10) and tertiary (HR 1.72, 95%CI 1.16-2.56) homelessness were independent risk factors for premature mortality. Conclusion: At least one recorded episode of primary, secondary, or tertiary homelessness was associated with premature mortality and younger age at death over a 15-year period. Implications for public health: Accurately identifying individuals experiencing primary, secondary or tertiary homelessness at the emergency department may enable targeted interventions that could potentially reduce their risk of premature mortality.