Abstract

PURPOSE People experiencing homelessness have lower rates of cancer screening than housed people, contributing to later stages at cancer diagnosis and poor outcomes. We examined whether gaining housing increased rates of cancer screening in a cohort of homeless veterans.

METHODS We conducted a retrospective cohort study examining all veterans experiencing homelessness who were eligible for, but not up to date on, colorectal and breast cancer screening from 2011 to 2021. Our exposure was gaining housing in the 24 months after the index clinic visit, conceptualized as a time-varying covariate. Our primary outcome was undergoing breast or colorectal screening in the 24 months after that visit. We performed bivariate analysis and Cox proportional hazards analysis, clustering on the facility level and adjusting for clinical and demographic covariates.

RESULTS Our cohort included 117,619 homeless veterans who were eligible for but not up to date on colorectal cancer screening at their index visit, of whom 57,705 (49.0%) gained housing over 24 months. The cohort included 6,517 homeless veterans who were eligible for but not up to date on breast cancer screening, of whom 3,101 (47.5%) gained housing over 24 months. Compared with peers who remained homeless, veterans who gained housing were more than twice as likely to undergo colorectal cancer screening (adjusted hazard ratio, 2.3; 95% CI, 2.2-2.3; P <.001) and breast cancer screening (adjusted hazard ratio, 2.4; 95% CI, 2.2-2.7; P <.001).

CONCLUSIONS Veterans experiencing homelessness who gain housing have higher rates of cancer screening. This finding supports promotion of housing to improve health outcomes for homeless individuals.

INTRODUCTION

Homelessness interferes with health in myriad ways.1,2 Profound poverty, lack of insurance, competing priorities, discrimination from health care teams, and complex medical and behavioral problems all constitute barriers to navigating complex health systems and interfere with the timely receipt of care.1,3,4

These barriers are evident in cancer screening; as a result, people experiencing homelessness have lower rates of screening when compared with the general population, including in the Veterans Health Administration (VHA).5,6 These lower rates of screening contribute to later stages at cancer diagnosis, higher rates of emergency operations, and poorer survival than seen in the general population.5,7 More than one-half of the population experiencing homelessness is aged 50 years or older,8 which is the age at which many national guidelines recommend initiating cancer screening.9,10 This demographic trend, coupled with the fact that cancer is a leading cause of death among older adults who are homeless,11,12 raises the question of what might improve cancer screening rates in this population. To our knowledge, no studies have examined whether changes in housing status (eg, gaining housing) are associated with changes in cancer screening rates for people experiencing homelessness.

The VHA offers a unique setting in which to study this question as it routinely screens veterans for homelessness and housing stability,13 enabling assessment of changes in veterans’ housing status over time. The VHA has made a sustained commitment to reducing homelessness among veterans since 2009, culminating in a 50% reduction in veteran homelessness.14,15 Our team has previously demonstrated that veterans experiencing homelessness have lower rates of cancer screening than housed veterans in the VHA.6 We sought to build on this analysis and examine whether gaining housing, compared with remaining homeless, is associated with increased rates of screening for cancer among veterans experiencing homelessness who receive care at the VHA.

METHODSStudy Overview

We conducted a retrospective cohort study examining the association between gaining housing and cancer screening among veterans experiencing homelessness who received care at the VHA from 2011 to 2021. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.16 The Stanford Institutional Review Board approved the study with a waiver of informed consent (protocol #66019).17

Analytic Sample

Our analysis included all homeless veterans who were eligible for, but not up to date on, colorectal or breast cancer screening at their first VHA clinic visit of the most recent year in the study period. For each veteran, we selected the most recent calendar year in which they had a visit at the VHA. If a veteran had multiple visits, we selected the earliest visit in that calendar year to ensure adequate follow-up. We termed this the index visit.

DefinitionsEligibility for Cancer Screening

We classified veterans as eligible for colorectal cancer screening if they were aged 50 to 75 years and did not have any prior cancer diagnosis or inflammatory bowel disease, and had not undergone colectomy. Veterans were eligible for breast cancer screening if they were women aged 50 to 75 years and did not have any prior cancer diagnosis and had not undergone lumpectomy or mastectomy. We selected these eligibility criteria to be consistent with other studies evaluating screening in the VHA.18 We assumed that all individuals in our cohort remained eligible for cancer screening during the 24 months following their index clinic date.

Up-to-Date Cancer Screening

We defined being up to date on cancer screening as a binary variable consistent with VHA quality measures and US Preventive Services Task Force Recommendations.10,19 For colorectal cancer, veterans were considered up to date if they had undergone colonoscopy in the past 10 years, a flexible sigmoidoscopy in the past 5 years, computed tomography colonography or barium enema in the past 5 years, or a stool-based study in the past year. For breast cancer, veterans were considered up to date if they had had a mammogram in the past 24 months. We captured screening examinations that were provided or paid for by the VHA using Current Procedural Terminology codes, laboratory test results, and radiology results (Supplemental Table 1). As in other studies using VHA administrative data, we could not access screening procedures completed elsewhere that were not paid for by the VHA.

Housing Status at the Index Clinic Visit

We identified veterans experiencing homelessness using a previously published operational definition of homelessness20 that leverages multiple data sources within the VHA, including International Classification of Diseases codes, clinic reminders, outpatient visit screens, and the Veterans Affairs’ Homeless Operations, Management, and Evaluation System (HOMES) (Supplemental Table 2). Use of multiple sources identifies a greater proportion of homeless veterans than does use of a single indicator.20 We classified veterans as homeless if they had any indicator of homelessness at the index visit or in the preceding 12 months with no subsequent indicator of stable housing.

Exposure

Our exposure of interest was gaining housing in the 24 months after the index clinic visit. We classified veterans as having gained housing from 3 measures: (1) response to the Homeless Screening Clinical Reminder, (2) response to assessments by the HOMES program, or (3) presence of a move-in date with the Department of Housing and Urban Development–VA Supportive Housing (HUD-VASH) program.

For the Homeless Screening Clinical Reminder, we classified veterans as having gained housing if they answered yes to the question “In the past 2 months, have you been living in stable housing that you own, rent, or stay in as part of a household?” This reminder is administered to veterans in both inpatient and outpatient settings every 6 to 12 months and allows self-reporting of stable housing.

For the HOMES assessment, we classified veterans as having gained housing if they had an answer of anything other than “literally homeless” to the question “How would you [evaluator] describe the veteran’s housing status?” at any time in the 24 months after the index clinic visit.

Finally, we assumed that veterans who had a move-in date with the HUD-VASH program had gained housing on that date.

In this analysis, we assumed that veterans gained housing in the month indicated via these mechanisms. We attributed cancer screenings during and after that month to the group who gained housing, conceptualizing housing status as a time-varying covariate.

Outcome

Our primary outcome was undergoing colorectal or breast cancer screening in (or paid for by) the VHA in the 24 months following the index clinic visit. For colorectal cancer screening, screening examinations included colonoscopy, flexible sigmoidoscopy, computed tomography colonography, barium enema, or a stool-based study. For breast cancer screening, screening examinations were mammograms.

Covariates

We obtained covariates including sex, age, marital status, race, ethnicity, smoking status, VHA clinic region, VHA clinic urban-rural location, mental health diagnoses, and the Charlson comorbidity index from the VHA’s Corporate Data Warehouse. We determined whether veterans had any primary care visits in the past 2 years to measure engagement with the VHA system, conceptualizing this measure as a binary variable. We used the HOMES system to obtain information on whether a veteran was chronically homeless as defined by HUD (“homeless and [residing] in a place not meant for human habitation, a safe haven, or in an emergency shelter, and has been homeless and residing in such a place for at least 1 year or on at least four separate occasions in the last 3 years”21), also classified as a binary variable.

Statistical Approach

For each cancer type, we conducted bivariate analyses to examine differences between the housing status groups (gained housing and remained homeless). We then constructed Cox proportional hazards models, clustering on the facility level and adjusting for year of visit, age, sex, marital status, race, ethnicity, smoking status, Charlson comorbidity index, mental health diagnoses, VHA clinic region and urban-rural location, evidence of chronic homelessness in the past year, and primary care use in the past 2 years. We included race as a covariate because of well-described racial inequities in cancer screening.22 We confirmed the Cox proportional hazards model assumption of proportionality by visual inspection of Kaplan-Meier curves.

We analyzed data from August 2023 to September 2024 in R version 4.1.2 (R Foundation for Statistical Computing). Statistical tests were 2-sided with a significance level of .05.

RESULTSCohort Characteristics

A total of 117,619 homeless veterans were eligible for but not up to date on colorectal cancer screening at their index visit, of whom 57,705 (49.0%) gained housing in the subsequent 24 months. A total of 6,517 homeless veterans were eligible for but not up to date on breast cancer screening, of whom 3,101 (47.5%) gained housing in the subsequent 24 months.

We identified 93% of the cohort who gained housing by using the Homeless Screening Clinical Reminder, 23% by using the HUD-VASH move-in date, and 14% by using responses on HOMES assessments (Supplemental Table 3). We identified some of this group through multiple mechanisms.

Table 1 shows characteristics of the veterans eligible for, but not up to date on, each type of cancer screening, stratified by their housing status. In both cohorts, compared with peers who remained homeless, veterans who gained housing were older and had a higher level of comorbidity, and they were more likely to be married, to be former smokers, to have had a primary care visit in the past 2 years, and to have a mental health diagnosis. Those gaining housing were less likely to have evidence of chronic homelessness.

Table 1.

Characteristics of Homeless Veterans Eligible for, but Not Up to Date on, Cancer Screening, Stratified by Cancer Type and Housing Status

Housing Status and Up-to-Date Cancer Screening

Figure 1 demonstrates results of the Kaplan-Meier analysis, which assesses time to cancer screening over the 24 months of observation with stratification by housing status. During this period, 36.1% and 47.9% of the veterans who gained housing underwent colorectal cancer screening and breast cancer screening, respectively, compared with 18.8% and 23.7% of the veterans who remained homeless.

Figure 1.

Time to Cancer Screening, Stratified by Veteran Housing Status

Note: Homeless veterans who were eligible for, but not up to date on, cancer screening were enrolled in the cohort at their index clinic visit. Housing status and cancer screenings were assessed over the subsequent 24 months.

The median time (interquartile range) between the index visit and cancer screening was 8 months (4-15 months) for colorectal cancer screening and 8 months (3-14 months) for breast cancer screening. For patients who gained housing, the median time from gaining housing to screening was 4 months (1-9 months) for colorectal cancer screening and 3 months (1-8 months) for breast cancer screening.

In the Cox proportional hazards models, compared with eligible but not up-to-date veterans who remained homeless, veterans who gained housing had 2.3 times the adjusted hazard (95% CI, 2.2-2.3; P <.001) of undergoing colorectal cancer screening and 2.4 times the adjusted hazard (95% CI, 2.2-2.7; P <.001) of undergoing breast cancer screening in the 24 months after their index clinic visit (Table 2).

Table 2.

Association of Housing Status With Likelihood of Undergoing Cancer Screening

DISCUSSIONKey Findings

Among veterans experiencing homelessness who were eligible for but not up to date on colorectal and breast cancer screening, gaining housing was associated with more than twice the hazard of undergoing screening when compared with remaining homeless. As timely cancer screening may reduce rates of later-stage cancer diagnoses, need for more intensive therapy, emergency operations, and death, providing housing may improve a variety of health outcomes.

People experiencing homelessness have lower rates of cancer screening than the general population.5 These inequities persist in the VHA system, although they are attenuated when compared with other settings.6 By examining a narrow cohort of veterans experiencing homelessness who were eligible for, but not up to date on, cancer screening, we demonstrated that gaining housing was associated with veterans effectively catching up with overdue screening.

Gaining housing may facilitate cancer screening through multiple mechanisms. Housing may lower barriers to health care by promoting consistent access to communication, reducing competing priorities, and decreasing vulnerability to external forces such as encampment clearing or theft.3,23 In the VHA system, housing may be accompanied by other supportive services that facilitate access to care. In non-VHA settings, clinicians and patients face constraints in making decisions around screening for homeless patients.24 Colorectal cancer screening may be logistically challenging for patients experiencing homelessness who may not have access to private bathrooms for colonoscopy preparation or stool-based studies.25 Gaining housing may make this type of screening more feasible. Further research, however, including qualitative analysis, should be done to clarify the mechanisms by which gaining housing may improve rates of cancer screening. This research could inform efforts to mitigate the impact of homelessness on screening rates.

The VHA provides universal coverage to its beneficiaries, which eliminates financial concerns as a potential barrier to cancer screening. Additionally, there are many supportive programs within the VHA for veterans experiencing homelessness or unstable housing.15 Even in this setting, we observed a large association between gaining housing and cancer screening. Our findings may underestimate or overestimate the association between gaining housing and receipt of cancer screening in other settings that do not have these supports for people experiencing homelessness.

Implications

These findings support promoting housing to improve health for homeless individuals. Although the VHA setting is unique in its robust programmatic and financial commitment to ending veteran homelessness, other health systems and insurers have made investments in housing directly or housing supports to address upstream social determinants of health.26 Clinicians, professional organizations, and health systems can advocate for policies that increase the supply of affordable housing, reduce poverty, and end homelessness to advance equity for patients. In tandem with these efforts, clinicians can explore other mechanisms to support patients experiencing homelessness in obtaining cancer screening. For example, medical respite, which provides a place for patients experiencing homelessness to recuperate, or other noncongregate shelters may assist patients in completing colon preparation for colonoscopy.

Limitations

This study has limitations. First, there is likely residual unmeasured confounding between housing status groups in this observational study. Veterans experiencing homelessness who can navigate the system to gain housing may also be more likely to pursue preventive care such as cancer screening. At the index visit date, however, both groups were not up to date with cancer screening, which may minimize this source of unmeasured confounding.

We used multiple sources to ascertain homelessness as recommended by the National Center on Homelessness Among Veterans.20 This is a strength of this study given that housing status is not well documented in other health systems.27 We may be undercounting or overcounting veterans as homeless, however. We also used multiple sources to evaluate whether veterans gained housing. The Homeless Screening Clinical Reminder, the mechanism by which we identified most of our cohort who gained housing, was not designed to evaluate changes in housing status and may not be administered to veterans already identified as homeless. Other studies, however, have shown that this screener can be used to examine transitions into and out of homelessness, and it has been used in other studies despite its limitations.28,29

In our cohort, both housing status groups had high rates of screening via the Homeless Screening Clinical Reminder after the index clinic visit date; thus, there may be misclassification of gaining housing in our study that would bias our results toward the null. We assume that veterans gained housing the month that they reported stable housing; however, those who do gain housing likely gain it before reporting it to the VHA. We believe this classification is the most conservative approach because in the group gaining housing, we considered only cancer screenings that were completed after the date they reported stable housing.

Consistent with other studies completed in the VHA system,30 we can measure only screening examinations performed or paid for by the VHA. Veterans with Veteran Affairs’ health care and private insurance who underwent screening examinations outside of the VHA paid for by private insurance were not captured in our sample. We anticipate, however, that the proportion of homeless veterans having private insurance is likely low. Finally, our population included only homeless veterans who had contact with the VHA system. Screening outcomes among those without contact are unknown.

Conclusions

Gaining housing is associated with higher rates of cancer screening among veterans experiencing homelessness who are eligible for screening but not up to date. This finding supports promotion of housing to improve health outcomes for homeless individuals.