Deaths in supported housing: a cohort study of mortality of previously homeless people moving into supported housing units in Ottawa, Canada Authors: Nicole Wisener, Nicole E Edgar, Nassima Marouf, Brooklyn Ward, Simon Hatcher Published: 2026 ABSTRACT AbstractBackgroundThe transition from chronic homelessness to being housed is a period of increased mortality risk. Examining the post-housing period is important for improving our understanding of this increased risk.ObjectivesTo investigate, in a retrospective cohort study, the mortality rate of people who had been chronically homeless and who moved into supported housing and to identify risk factors associated with mortality.DesignRetrospective cohort study with eligible records reviewed from 1 October 2017 until 14 June 2024.SettingThree supported housing units in Ottawa, Ontario, Canada.ParticipantsParticipants were any individual, alive or deceased, who moved into one of the eligible supported housing units within 6 months of their opening date. 123 participants were included in the study. The mean age was 41.3 (SD 11.9) years, and 76 (62%) were male.OutcomesThe primary outcome was all-cause and cause-specific mortality following transition to supported housing. Differences in demographics and morbidity were also compared between people who were alive as of 14 June 2024 and those who were deceased as of 14 June 2024.ResultsDuring a mean follow-up of 2.6 years, there were 27 deaths recorded (22% of those who moved into supported housing). Most, 11 (40%), died from ‘other causes’, where the cause of death was unclear or unknown and where they were found dead in their room, seven (26%) died of ‘natural causes’ usually in hospital after a medical event, seven (26%) of an opioid overdose and two (8%) where we have suppressed cause of death for privacy. There were no statistically significant differences in psychiatric diagnoses, substance use, medical diagnoses or opioid agonist therapy between those who died and those still alive at the end of the study period. Survival differed significantly across age groups (log-rank χ²=11.05, p=0.011), with residents aged 31–45 years demonstrating better survival than older age groups. No significant differences in survival were observed by sex.ConclusionsThere is a high mortality rate in people who are chronically homeless who move into supported housing. As our findings do not identify any clear baseline risk factor, it supports the idea that the transition into housing must be complemented with comprehensive health and well-being services that can meet this population’s complex needs to reduce the mortality risk. Strengths and limitations of this study This study includes a cohort of all individuals transitioning directly from chronic homelessness into supported housing within the first 6 months of the housing unit opening, reducing selection bias and allowing mortality to be examined specifically from the point of housing entry. This study includes a cohort of all individuals transitioning directly from chronic homelessness into supported housing within the first 6 months of the housing unit opening, reducing selection bias and allowing mortality to be examined specifically from the point of housing entry. Use of real-world clinical and programme data enhances ecological validity and provides policy-relevant insights into causes of death, including a substantial burden of overdose mortality. Use of real-world clinical and programme data enhances ecological validity and provides policy-relevant insights into causes of death, including a substantial burden of overdose mortality. The modest sample size and limited number of deaths reduce statistical power, meaning the absence of statistically significant associations should not be interpreted as evidence that demographic, clinical or substance use factors are unrelated to mortality risk. The modest sample size and limited number of deaths reduce statistical power, meaning the absence of statistically significant associations should not be interpreted as evidence that demographic, clinical or substance use factors are unrelated to mortality risk. Reliance on electronic medical records and incident reports rather than standardised diagnostic instruments or death certificates may have resulted in misclassification of comorbidities, substance use and causes of death. Reliance on electronic medical records and incident reports rather than standardised diagnostic instruments or death certificates may have resulted in misclassification of comorbidities, substance use and causes of death. As a retrospective study conducted in a single city, findings may not be generalisable to other settings, and unmeasured factors—such as trauma severity, substance use patterns over time and engagement with post-housing supports—could not be examined. As a retrospective study conducted in a single city, findings may not be generalisable to other settings, and unmeasured factors—such as trauma severity, substance use patterns over time and engagement with post-housing supports—could not be examined. Introduction People experiencing homelessness have markedly higher mortality than the general population, with life expectancy reduced by decades in many settings. Deaths among homeless populations occur disproportionately from preventable causes, including drug overdose, untreated or poorly managed chronic disease, violence and injury, reflecting the cumulative effects of social exclusion, trauma exposure, substance use and barriers to healthcare access.1–3 Chronic homelessness is associated with accelerated biological ageing, multimorbidity and prolonged exposure to adverse environments, all of which contribute to sustained excess mortality risk.4 5 A recent population-registry cohort study by Nilsson and colleagues found that among those with at least one homeless shelter contact during the study period, the all-cause and cause-specific mortality risk was higher for both males and females than the general population; this risk increased with higher numbers of shelter contacts, and there was about 15 excess life years lost.6 In response to the growing recognition of homelessness as a public health crisis, supported housing—particularly models informed by the Housing First approach—has become a cornerstone of homelessness policy in Canada and internationally. Supported housing programmes provide permanent accommodation alongside varying degrees of medical, psychiatric and social supports, and there is strong evidence that these models improve housing stability, reduce shelter use and decrease reliance on emergency services.7–9 For many individuals, supported housing represents the first sustained period of safety and stability after years of homelessness. However, improvements in housing outcomes have not consistently translated into improvements in survival. Emerging evidence suggests that mortality remains high among people who have transitioned from homelessness into supported housing. Cohort studies from Canada and the USA have reported crude mortality rates far exceeding those of the general population, even after housing placement, with overdose, cardiovascular disease, liver disease and respiratory illness among the leading causes of death.10–13 Long-term follow-up studies indicate that excess mortality may persist for many years after housing entry, challenging assumptions that housing alone is sufficient to reverse the health consequences of prolonged homelessness.13 Despite these concerning findings, the existing literature has important limitations. Many studies include heterogeneous samples in which participants have been housed for varying—and often lengthy—periods prior to study entry, making it difficult to isolate mortality risk specifically associated with the transition from homelessness into housing. Other studies rely primarily on administrative data, which may lack detailed clinical information and often do not capture the full cohort of individuals entering supported housing. Sample sizes are frequently small, limiting statistical power to identify risk factors for mortality, and cause-of-death information is often incomplete or inconsistently classified.10–13 Additionally, prior research has tended to emphasise housing as the primary intervention, with less attention to the long-term health trajectories of individuals exiting chronic homelessness. Prolonged homelessness is associated with cumulative trauma, severe and persistent substance use and complex medical comorbidity, effects that may not be rapidly reversible with housing alone.4 5 14 Mortality following housing placement may therefore reflect life-course disadvantage and accumulated physiological burden rather than failures of supported housing programmes per se. Understanding mortality in this context requires careful examination of the post-housing period as a distinct and potentially high-risk phase. To address these gaps, we conducted a retrospective cohort study of all individuals who transitioned directly from homelessness into three supported housing programmes in Ottawa, Canada, within 6 months of programme opening. By assembling a cohort of all individuals and following them from the point of housing entry, we aimed to characterise mortality during the post-housing period and to examine whether demographic characteristics, psychiatric diagnoses, substance use or medical comorbidities were associated with risk of death. Unlike studies that include individuals with variable durations of housing exposure, our approach focuses specifically on mortality after the transition from homelessness—a critical and understudied window of vulnerability. By describing both the magnitude and causes of mortality in this cohort and by comparing characteristics of those who died with those who survived, this study seeks to inform the development of supported housing models that not only end homelessness but also better address the persistent health risks faced by people after they are housed. We hypothesised that those with any (i) more chronic histories of homelessness, (ii) more complex physical and/or mental health histories or (iii) increased social isolation may be at higher risk of dying after they move into supported housing. Methods Study design The study design consisted of a retrospective cohort study of people who moved into three different supported housing units within 6 months of the housing opening. Residents who died during follow-up were compared with those who survived to the end of follow-up. Reporting of this study follows the guidance from Reporting of studies Conducted using Observational Routinely-collected Data (RECORD), an extension of Strengthening the Reporting of Observational Studies in Epidemiology (STROBE).15 The RECORD checklist is available in online supplemental file 1. Setting Three supported housing units in Ottawa, Ontario, were selected. These units were selected as these locations received physical and mental healthcare from one organisation, Ottawa Inner City Health (OICH). As such, the study team had access to the comprehensive medical record information needed to complete the study. Further, given the number of units in each supported housing site, we anticipated this would provide an adequate number of participants and deaths so that results would not require suppression for privacy protection. The three supported housing units included in this study are Richcraft Hope Residence (42 units, opened in April 2021), 1095 Merivale Road (57 units, opened April 2023) both operated by the Shepherds of Good Hope and Carruthers Residence (36 units, opened April 2018) operated by the John Howard Society. These housing programmes serve adults who have experienced chronic homelessness, defined as experiencing homelessness for a period of 6 months or longer. The supported housing units offer similar supports including both physical and mental healthcare services, substance use supports, recreation activities, life skills support, opportunities for social connection and case management. At Richcraft Hope Residence, this also includes managed alcohol services which provide a dose of wine at regular intervals to stabilise alcohol use disorder as well as access to opioid agonist therapy (OAT). Participants A roster of all residents who moved into the three participating residences during the eligible enrolment windows was obtained from OICH records. Vital status and clinical information were then determined through the Open Source Clinical Applications and Resources (OSCAR) electronic medical record system and OICH death incident reports. Eligible participants were all adults (aged 18 or over) who moved into one of the three participating residences within 6 months of the residence’s opening. Follow-up began on each participant’s individual move-in date and ended on the date of death or 14 June 2024, whichever occurred first. All individuals who moved into an eligible residence during the enrolment window were included, regardless of whether they subsequently remained in or left the residence. Age was determined as age at moving into housing. Clinical records from the 6 months prior to each participant’s move-in date were accessed to provide baseline information (earliest date was 1 October 2017). Person-years of observation were calculated by summing the duration of observation from moving in to death or 14 June 2024, whichever came first. Data source For each person in the cohort, data were abstracted on age, sex, cause of death (deceased participants only), substance use, psychiatric and physical diagnoses and opioid agonist prescription. All data were abstracted from the participant’s OICH medical record, including any incident reports. Death information was obtained from the OICH Death Incident Reporting. As death data were extracted from incident reports and not an official coroner ruling, we used the information provided there and used Ontario Coroner terminology to categorise the deaths.16 Deaths were classified in four categories: (1) natural (defined as a death due to disease or where injury did not significantly contribute to the death), (2) overdose (any suspected overdose based on information available during incident reporting), (3) suicide (where suicide is suspected based on available information at time of incident reporting) and (4) other (where cause of death was unknown or unclear at time of incident reporting). Following appropriate privacy considerations, data fields with a cell count of less than five where data could potentially be re-identified, including cause of death, were suppressed or re-coded to something that would not be readily identifiable. Data fields where multiple options could be selected (ie, proportions could be more than 100%) cell counts were not suppressed. Relevant data were systematically extracted using a Research Electronic Data Capture (REDCap) database. Data entry was routinely reviewed by a staff member not directly involved in data extraction for accuracy and completeness. Review of missing data and logic checks were completed as data cleaning prior to analysis. No data linkage was completed for this study. Sample size A formal sample size calculation was not undertaken as all eligible participants were included in the study. Statistical analysis Descriptive statistics were used to summarise demographic characteristics, mortality outcomes and causes of death. Continuous variables are presented as means and standard deviation (SD), while categorical variables are presented as counts and percentages. Residents who died during follow-up were compared with those who survived to the end of follow-up. Differences in proportions between groups were calculated with corresponding 95% confidence intervals (CI). χ2 tests or Fisher’s exact test was used to compare proportions between deceased and surviving residents for selected demographic characteristics, including age-group distributions. Overall survival following entry into supported housing was examined using the Kaplan–Meier survival analysis. Survival curves were compared using the log-rank test. Survival analyses were conducted for sex (male vs female), age (≤40 years vs >40 years) and four age categories (18–30, 31–45, 46–60 and ≥61 years). Age was operationalised to be aligned approximately by different stages of adulthood for age groupings including youth (18-30), young adulthood (31-45), middle adulthood (46-60) and older adults (61+). The binary age groupings for the Kaplan–Meier curves reflect the approximate median age of the cohort. All statistical tests were two-sided, and statistical significance was defined as p<0.05. Analyses were conducted using Microsoft Excel and Statistical Package for the Social Sciences (SPSS). Patient and public involvement statement Patients and public were not involved in the conception or operations of this study. Results 123 participants moved into the three residences within 6 months of their opening. Of these, 27 died (22%) over the observation period of the study. The mean length of follow-up was 2.6 (SD 2.0) years. Table 1 shows the characteristics of the individuals in the cohort divided into those who died and those still alive at the end of the study. The mean age of participants was 41.3 years (SD 11.9), and 62% were men. Table 1Comparison of characteristics with differences and 95% CIVariableDeceased (n=27)Alive (n=96)Difference and 95% CIDemographics Female10 (37.0%)37 (38.5%)−1.5 (−2.0 to 2.0) Male17 (62.9%)59 (61.4%)+1.5 (2.0 to −2.0) 18–306 (22.2%)21 (21.9%)+0.3 (−17.4 to 18.1) 31–456 (22.2%)45 (46.9%)−24.7 (−43.2 to −6.1) 46–6011 (40.7%)25 (26.0%)+14.7 (−5.8 to 35.2) 61+4 (14.8%)5 (5.2%)+9.6 (−4.5 to 23.7)Psychiatric diagnoses Depression9 (34.6%)26 (26.8%)+7.8 (−12.5 to 28.1) Bipolar disorder4 (15.4%)20 (20.6%)−5.2 (−21.3 to 10.8) PTSD7 (26.9%)14 (14.4%)+12.5 (−5.9 to 30.9) Psychotic disorder9 (34.6%)31 (31.9%)+2.7 (−17.9 to 23.2) Anxiety/panic9 (34.6%)30 (30.9%)+3.7 (−16.8 to 24.2) Personality disorder4 (15.4%)11 (11.3%)+4.0 (−11.2 to 19.3) Alcohol abuse10 (38.5%)22 (22.7%)+15.8 (−4.7 to 36.3) Drug abuse18 (69.2%)64 (66.0%)+3.2 (−16.8 to 23.3) Eating disorder1 (3.9%)1 (1.0%)+2.8 (−4.8 to 10.5) OCD1 (3.9%)5 (5.2%)−1.3 (−9.9 to 7.3) Other disorder6 (23.1%)21 (21.6%)+1.3 (−16.7 to 19.6) No mental disorder0 (0%)1 (1.0%)−1.0 (−3.0 to 1.0)Substance use Alcohol use8 (30.8%)17 (17.5%)+13.2 (−6.0 to 32.5) Cannabis7 (26.9%)16 (16.5%)+10.4 (−8.2 to 29.0) Opiates12 (46.1%)44 (45.4%)+0.70 (−20.8 to 22.4) Benzodiazepines2 (7.7%)12 (12.4%)−4.70 (−16.8 to 7.5) Stimulants12 (46.1%)64 (66.0%)−19.8 (−41.2 to 1.5) Hallucinogens0 (0%)1 (1.0%)−1.0 (−3.0 to 1.0) Other drugs1 (3.9%)5 (5.2%)−1.3 (−9.9 to 7.3)Physical diagnoses Chronic pain4 (15.4%)11 (11.3%)+4.0 (−11.2 to 19.3) Cardiovascular8 (30.8%)30 (30.9%)−0.1 (−20.1 to 19.8) Diabetes3 (11.5%)10 (10.3%)+1.2 (−12.5 to 14.9) Active cancer2 (7.7%)0 (0%)+7.7 (−2.6 to 17.9) HIV/AIDS4 (15.4%)10 (10.3%)+5.1 (−10.1 to 20.2) Hepatitis12 (46.1%)27 (27.8%)+18.3 (−2.8 to 39.5) Musculoskeletal4 (15.4%)29 (29.9%)−14.5 (−31.1 to 2.1) Neurological5 (19.2%)14 (14.4%)+4.8 (−11.9 to 21.5) Other GI/liver9 (34.6%)24 (24.7%)+9.9 (−10.3 to 30.1) Kidney/renal1 (3.9%)11 (11.3%)−7.4 (−17.2 to 2.2) Other infection3 (11.5%)6 (6.2%)+5.3 (−7.8 to 18.5) Respiratory10 (38.5%)31 (32.0%)+6.5 (−14.4 to 27.4) Thyroid/endocrine1 (3.9%)2 (2.1%)+1.8 (−6.1 to 9.7) Ophthalmologic3 (11.5%)6 (6.2%)+5.3 (−7.8 to 18.5) Stroke history0 (0%)1 (1.0%)−1.0 (−3.0 to 1.0) Hearing impairment1 (3.9%)1 (1.0%)+2.9 (−4.8 to 10.5) Other medical12 (46.1%)40 (41.2%)+4.9 (−16.6 to 26.4) No medical1 (3.7%)6 (6.2%)−2.5 (−11.1 to 6.5)GI, gastrointestinal; OCD, obsessive-compulsive disorder; PTSD, post-traumatic stress disorder. Comparison of characteristics with differences and 95% CI GI, gastrointestinal; OCD, obsessive-compulsive disorder; PTSD, post-traumatic stress disorder. Most participants had more than one psychiatric diagnosis, and the most prevalent diagnosis was drug abuse or dependence (66%). The most common physical medical conditions reported by participants were respiratory disorders (33%), cardiovascular-related issues (31%) and hepatitis (31%). Musculoskeletal issues, chronic pain, diabetes and neurological conditions were also prevalent. Most participants had co-occurring medical, psychiatric and substance use disorders; however, a small proportion (5.7%) reported no medical conditions. The Kaplan–Meier curves for sex and age are shown in figure 1A,B. There was a gradual decline in survival over 6 years with most deaths occurring between years 2 and 6. Of the 27 participants who died after moving into supported housing, most, 11 (40%), died from ‘other causes’ where they were found dead in their room, seven (26%) died of ‘natural causes’ usually in hospital after a medical event, seven (26%) of an opioid overdose and two (8%) remaining deaths with cause suppressed. Overdose and unexplained deaths together accounted for two-thirds of all mortality events. The crude mortality rate in the cohort was 8.6 deaths per 100 person years of observation. Mean age at death was 46.0 years (SD 14.6). (A) The Kaplan–Meier survival curve of supported housing residents by sex (M=male, F=female) (log-rank χ²=0.09, p=0.77). (B) The Kaplan–Meier survival curve of supported housing residents by age (log-rank χ²=0.71, p=0.40). Of the 27 deceased residents, 13 (48.1%) were being prescribed OAT compared with 42 of the 96 non-deceased residents (43.8%) (χ²=0.03, p=0.85). These findings do not indicate a meaningful difference in OAT receipt between groups. Among those who died, 17 (63.0%) were male and 10 (37.0%) were female. The proportion of males did not differ significantly between deceased and surviving residents (63.0% vs 61.5%; χ²=0.00, p=1.00; difference 1.5%, 95% CI −19.0% to 22.0%). The Kaplan–Meier survival analyses demonstrated no significant difference in survival between male and female residents (log-rank χ²=0.09, p=0.77) (figure 1A). Survival did not differ significantly between residents aged 40 years or younger and those aged over 40 years (log-rank χ²=0.71, p=0.40) (figure 1B). However, when age was examined using four categories (18–30, 31–45, 46–60 and ≥61 years), survival differed significantly across groups (log-rank χ²=11.05, df=3, p=0.011). Residents aged 31–45 years demonstrated the highest probability of survival over follow-up, whereas residents aged 46–60 years and those aged 61 years and older experienced poorer survival. Discussion In this retrospective cohort of adults transitioning from chronic homelessness into supported housing in Ottawa, over one in five residents died during a mean follow-up of 2.6 years. Mortality remained substantially higher than that expected in the general population despite access to supported housing. Importantly, we found little evidence that routinely collected demographic, psychiatric or medical characteristics distinguished those who died from those who survived, suggesting that mortality risk may reflect complex cumulative vulnerabilities rather than easily identifiable baseline risk factors. The crude mortality rate of 8.6 deaths per 100 person-years is substantially higher than that observed in the general population and comparable to rates reported in other cohorts of formerly homeless individuals living in supported housing.10 12 Notably, at least one quarter of deaths were attributable to opioid overdose, underscoring the continued lethality of substance-related risk after housing placement. We did not identify statistically significant differences in mortality by sex, psychiatric diagnosis, substance use, medical comorbidity or OAT treatment. Survival did not differ significantly by sex or when age was dichotomised as ≤40 versus >40 years. In an exploratory analysis using four age categories, however, survival differed across groups, with the highest observed survival among residents aged 31–45 years. This finding should be interpreted cautiously given the small subgroup sizes, limited number of deaths and absence of adjusted pairwise comparisons. Given the modest sample size and limited number of deaths, the study was underpowered to detect anything other than large effects, and the absence of statistically significant associations should not be interpreted as evidence that these factors are unimportant determinants of mortality. Rather, our findings highlight the difficulty of identifying individual-level predictors of death using routine clinical data in small cohorts, and they raise the possibility that mortality risk in this population is shaped by complex, interacting factors that are not well captured by binary diagnostic categories. The causes of death observed in this cohort are informative. While some deaths occurred following medical events in hospital, a substantial proportion involved individuals found dead in their rooms, and over a quarter were attributed to opioid overdose. These findings are consistent with emerging evidence that housing, although essential for stability and dignity, does not in itself mitigate the cumulative physiological burden of long-term homelessness, trauma, substance use and chronic illness.17 The persistence of overdose deaths after housing placement also reflects the ongoing risks posed by a toxic and unpredictable drug supply and18 19 the fact that individuals could now use alone in their rooms,18 19 as well as limitations in existing harm-reduction and clinical supports within housing environments.18 A notable proportion of deaths could not be confidently classified. This highlights the limitations of relying on incident reports rather than coroner records and raises the possibility that overdose mortality may have been underestimated. Our results align with prior studies demonstrating elevated mortality among people who are formerly homeless but now housed. For example, Ramsay and colleagues reported a crude mortality rate of 5.1 deaths per 100 person-years in permanent supportive housing, though many participants had already been housed for several years at study entry.12 Similarly, long-term follow-up data from Boston indicate that mortality remains high even after prolonged periods of housing stability.13 Together with our findings, this literature suggests that supported housing does not eliminate excess mortality risk,20 as may be expected by increased access to health services and decreased exposure to violence and extreme weather conditions, and that the period following housing placement may represent a critical window of vulnerability. There are several methodological limitations that affect the interpretation of our findings. Diagnoses of psychiatric illness, substance use and medical conditions were derived from electronic medical records rather than standardised assessments, and substance use information is likely incomplete in the context of widespread polysubstance exposure and drug contamination. Causes of death were based on incident reports and clinical records rather than coroner or vital statistics data, which may have resulted in misclassification, particularly among individuals found deceased in their rooms. Additionally, our analyses relied on relatively crude categorizations of exposures and did not account for severity, chronicity or changes over time, all of which may be more relevant to mortality risk. Lastly, given our modest sample size, we were underpowered to detect anything other than large effects, and so findings should be interpreted cautiously. Despite these limitations, this study has important implications. The high mortality observed after housing placement challenges assumptions that housing alone is sufficient to substantially improve survival among people exiting chronic homelessness. While supported housing remains a critical and evidence-based intervention for improving housing stability and quality of life, our findings add to growing evidence that additional, sustained health-focused interventions are necessary to address premature mortality in this population. Supported housing has largely expanded in recent years because of the impact of the ‘Housing First’ policy where there is robust evidence that providing housing regardless of substance use or psychiatric disorder improves housing outcomes.17 However, the expansion of supported housing as currently practised is not the same as delivered in various Housing First studies which placed people in individual dwellings located in diverse neighbourhoods. Supported housing as currently delivered risks creating mini ‘shelters’. However, there is equally good evidence that Housing First, even when it adheres to the original model of housing, does not improve mental health, substance use or mortality outcomes.17 20 The results of this study highlight the need to move beyond simple comparisons of housed versus unhoused status and toward a more nuanced understanding of post-housing trajectories. The transition into housing may expose individuals to new risks, including social isolation, changes in substance use patterns and supply and reduced informal supervision from their street community, while persistent trauma and multimorbidity continue to exert their effects. Rather than replacing Housing First, we suggest an evolution toward a ‘Housing First 2.0’ model that combines permanent housing with sustained trauma-informed medical care, overdose prevention, chronic disease management, social connection and recovery-oriented rehabilitation. Possible interventions include a need to focus on the long-term effects of trauma on health; support for basic living skills and addressing meaning in life once day-to-day survival is not such a pressing issue. However, these hypotheses cannot be tested within the present study and should be viewed as directions for future research rather than conclusions drawn from our data. Future studies should prioritise larger, multisite cohorts to improve statistical power, linkage with coroner and vital statistics databases to enhance accuracy of mortality classification and longitudinal measurement of substance use patterns, trauma exposure, engagement with care and functional recovery following housing placement. Evaluating targeted interventions—such as enhanced overdose prevention, integrated medical and psychiatric care and trauma-informed, recovery-oriented supports—will be essential to determining how supported housing can better reduce mortality, not merely improve housing outcomes. In conclusion, this study demonstrates that mortality among people transitioning from chronic homelessness into supported housing remains high, with overdose accounting for a substantial proportion of deaths. While we were unable to identify clear individual-level predictors of mortality, our findings reinforce the urgent need to complement housing with comprehensive, long-term health and harm-reduction strategies, for example, as a ‘Housing First 2.0’ approach. Addressing premature mortality among formerly homeless individuals will require not only housing but sustained, integrated approaches that recognise the enduring health consequences of homelessness. Supplementary material