Permanent supportive housing and cardiovascular outcomes among Veterans experiencing homelessness: a difference-in-differences analysis

Abstract

Background:

Veterans experiencing homelessness have a high burden of cardiovascular disease (CVD). Identifying approaches that can improve the health of this population is crucial.

Objective:

The Department of Housing and Urban Development-Veterans Affairs Supportive Housing (HUD-VASH) provides permanent supportive housing (PSH) assistance. We examined whether obtaining PSH through HUD-VASH was associated with changes in outcomes among Veterans with CVD experiencing homelessness.

Design:

Using a difference-in-differences (DID) approach, outcomes among Veterans who received a HUD-VASH voucher and moved into PSH within one month (early group) were compared with those who received a voucher but remained unhoused for at least six months (delayed group).

Participants:

Veterans ≥18 years of age with CVD (coronary artery disease, heart failure, peripheral arterial disease, ischemic stroke/cerebrovascular disease, or atrial fibrillation) who experienced ≥6 months of homelessness and received a HUD-VASH voucher in 2016 to 2019.

Intervention:

Entering PSH after receiving a HUD-VASH voucher.

Main Measures:

Six-month probability of a cardiovascular ED visit or hospitalization.

Key Results:

Among 970 Veterans with CVD experiencing homelessness, 845 were in the early and 125 in the delayed PSH group. There was a significant decrease in the 6-month probability of cardiovascular ED visits or hospitalizations (16.3% [95% CI 13.9% – 18.7%] to 11.9% [95% CI 9.8% – 14.0%]) among Veterans who received a voucher and moved into PSH within one month, but no statistically significant change among Veterans who did not move into PSH for at least 6 months (13.4% [95% CI 8.0% – 18.8%] to 18.1% [95% CI 11.4% – 24.8%]) for a DID estimate of −9.1 (95% CI −17.7 – −0.5) percentage points, p=0.04.

Conclusions:

Obtaining PSH through HUD-VASH was associated with a significant decrease in the probability of CVD related ED visits or hospitalizations among Veterans with chronic CVD experiencing homelessness.

Introduction:

On a single night in 2024, more than 32000 Veterans experienced homelessness.(1) Cardiovascular disease (CVD) is the most common cause of death among Veterans experiencing homelessness, and individuals experiencing homelessness have a high burden of CVD and risk factors.(2–5) Permanent supportive housing (PSH) – rental assistance in conjunction with support services – is associated with improvements in medication adherence and psychosocial stress, factors that may play a role in improving cardiovascular health.(6, 7) However, whether such programs can improve cardiovascular outcomes among individuals experiencing homelessness is not well studied.

The Housing and Urban Development-Veterans Affairs Supportive Housing program (HUD-VASH) is the largest PSH program for Veterans experiencing homelessness. HUD-VASH combines VA case management with Housing Choice Voucher rental assistance. A previous analysis linked HUD-VASH housing to lower inpatient healthcare costs.(8) Other studies of PSH, typically in single cities or states, have noted a reduction in hospitalizations and emergency department (ED) utilization.(9–11) However, it is not known whether PSH is associated with improved cardiovascular health outcomes among Veterans with chronic CVD experiencing homelessness.

To evaluate whether PSH is associated with changes in cardiovascular outcomes, we used national data and a difference-in-differences (DID) design to compare changes in cardiovascular ED visits and hospitalizations among Veterans experiencing homelessness with pre-existing CVD who received PSH through HUD-VASH and those who also received a rental assistance voucher but had a delayed PSH entry.

Methods:

This study was considered exempt from review by the institutional review board of the Corporal Michael J. Crescenz VA Medical Center. This study follows the reporting requirements of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement.(12)

Data acquisition

Demographic data and information regarding housing status were acquired from the VA Corporate Data Warehouse (CDW) and the VA’s Homeless Operations Management and Evaluation System (HOMES). The HOMES data captured Veterans’ engagement with VA homelessness services, including dates of engagement. Voucher type – tenant-based (voucher can be used to rent any eligible private apartment) vs. project-based (voucher associated with a unit with a long-term contract with state or local housing authority) – was also obtained.(13) Clinical data, including the VA Nosos risk score (a VA risk score used to identify Veterans with high predicted health care costs), were obtained from the CDW.(14) Area-level economic and demographic data were obtained from the US Census American Community Survey for each Veteran based on their ZIP code of residence or location before enrollment into HUD-VASH. Information on non-VA care, that involved VA payment, was obtained from the non-VA care Program Integrity Tools (PIT) and CDW Fee Basis systems.

Study Population

The study population consisted of all Veterans ≥18 years old who experienced homelessness for ≥6 months, had a pre-existing diagnosis of CVD (coronary artery disease (CAD), heart failure (HF), peripheral artery disease (PAD), atrial fibrillation (AF), or stroke/cerebrovascular disease), and received a HUD-VASH voucher in the years 2016 to 2019. Pre-existing CVD was identified if a Veteran had any of the following at least 6 months prior to receipt of a HUD-VASH voucher: 1) ≥2 relevant International Statistical Classification of Diseases diagnostic codes, on separate occasions, for a particular cardiovascular condition, 2) ≥1 relevant Current Procedural Terminology (CPT^®) code, or 3) left ventricular ejection fraction ≤40%. (Supplemental Methods 1) Additionally, Veterans were required to move into PSH within 12 months of voucher receipt, be alive for ≥12 months, and remain in PSH for ≥6 months

Study groups

Veterans who receive a HUD-VASH voucher have 120 days to secure housing but can apply for an extension if unable to find housing.(15) The intervention group consisted of Veterans who received a HUD-VASH voucher and then moved into PSH within 1 month of voucher receipt. The comparison group included Veterans who also received a voucher but moved into PSH between 6 and 12 months after voucher receipt (i.e., after the analysis window). While both groups were engaged with HUD-VASH case management, permanent housing was only achieved after moving into PSH. Veterans in both groups were required to receive PSH for ≥6 months after moving in. The number of Veterans screened and excluded are listed in the Supplemental Figure.

Outcomes

The primary outcomes of interest were the 6-month probability of a cardiovascular ED visit or hospitalization. Secondary outcomes were the probability of all-cause ED visits or hospitalizations, number of all-cause and cardiovascular outpatient visits, and adherence to CVD-related medications as measured by the proportion of days covered (PDC). Encounters were identified using diagnostic and procedure codes (Supplemental Methods 1, Supplemental Tables 1 and 2.). The medication classes used for the adherence outcome are listed in Supplemental Table 3. If a Veteran was identified as having a cardiovascular condition, then PDC was calculated for each of the recommended medications for the diagnosed condition. PDC was calculated for each month in the study period (i.e., the number of days with available medication divided by the total number of days in the month, excluding days in the hospital). For each patient, the mean PDC for each of the recommended medication classes was calculated.

Outcomes for both groups were assessed in the 6 months before and after initial voucher receipt (i.e. the analysis window) such that the intervention group was already housed while the comparison group was not yet housed in the post-voucher period.

Statistical analysis

After identifying Veterans in the intervention and comparison groups, summary measures of baseline variables were compared between the two groups using the Kruskal–Wallis or Fisher’s exact tests. Using a DID design, changes in outcomes were compared from the 6 months before to the 6 months after receipt of a HUD-VASH voucher in the intervention vs. comparison groups. Generalized estimating equation (GEE) models with indicator variables for period (pre vs. post voucher receipt), group (intervention vs. comparison), and the interaction between period and group were fit. The interaction term was the DID estimate of interest and characterized the differential change in an outcome before and after voucher receipt across the two groups. The models also included indicator variables for month and year of voucher issue and primary Veterans Integrated Services Network (VISN) as well as clinical, demographic, and area-level covariates as detailed in Supplemental Methods 2).

Logistic GEE models were used for the occurrence of ED visits and hospitalizations, and negative binomial GEE models for the number of outpatient visits and number of days covered (with necessary medication) as the dependent variable. For the medication adherence models, the number of days in the period was included as an offset term. Robust standard errors were used. Average marginal effects were calculated to estimate the absolute change in each outcome. Pre-intervention trends were evaluated by fitting a model that included a full set of interactions between treatment group and month indicators in the pre-voucher period and using a joint F-test on the interaction terms to determine whether trends differed significantly. The same covariates as in the primary model were also included in these monthly models. We also performed sub-group analyses based on cardiovascular condition and a sensitivity analysis limited to Veterans who received a tenant-based voucher.

To evaluate the change in the primary outcome among the delayed housing (comparison) group after these Veterans also received PSH, we fit an additional model comparing the outcomes in the delayed housing group in the 6 months before and after entering into PSH with the early housing (original intervention) group (the first 6 months after receiving a voucher served as the baseline period and the second 6 months as the follow-up period for the early housing group). Only Veterans from the early housing group who stayed in PSH for 12 consecutive months were included in this analysis. This model was specified the same as the primary DID model.

Statistical tests were 2-tailed and a p-value ≤0.05 was considered statistically significant. Summary measures are presented as means with standard deviations (SD) or 95% confidence intervals (CI) or medians with interquartile range (IQR). Analyses were performed using Stata 18 (StataCorp LP).

Results

A total of 970 Veterans with pre-existing CVD who experienced homelessness for ≥6 months, received a HUD-VASH voucher, and moved into PSH, were identified. Of these, 845 were in the intervention group and moved into PSH within 1 month of voucher receipt and 125 Veterans were in the comparison group and moved into PSH 6 to 12 months after voucher receipt. Veterans in the intervention and comparison groups had similar median age [60 years (IQR 56 – 64) vs. 59 years (IQR 56 – 54), p=0.84] and proportion who were female (3.2% vs. 3.2%, p=1.0) (Table 1). A smaller proportion in the intervention group were non-Hispanic Black (30.9% vs. 44.8%, p<0.001). Veterans in the intervention group had a non-statistically significant higher median VA Nosos score (indicating a higher burden of medical comorbidities) [1.9 (IQR 1.0 – 3.5) vs. 1.5 (IQR 0.9 – 2.9), p=0.053]. A significantly higher proportion of the intervention group received project-based, rather than tenant-based, vouchers (19.8% vs. 4.0%, p<0.001).

Table 1:

Baseline characteristics by treatment group

	Intervention or early HUD-VASH housinga (n=845)	Comparison or delayed HUD-VASH housinga (n=125)	p-valueb
	Median (IQR)
Age (years)	60 (56 – 64)	59 (56 – 64)	0.84
Income ($)	8992 (0 – 12864)	7700 (0 – 12456)	0.13
VA Nosos risk scorec	1.86 (1.05 – 3.5)	1.49 (0.91 – 2.88)	0.05
ZIP code level CDC Social Vulnerability Index d
Overall Index	80.63 (60.69 – 92.51)	84.39 (66.13 – 93.66)	0.13
Socioeconomic Status Sub-Index	74.38 (51.63 – 89.17)	79.14 (56.98 – 91.49)	0.04
Housing Type and Transportation Sub-Index	81.88 (64.54 – 93.13)	84.74 (67.32 – 93.88)	0.27
Household Composition and Disability Sub-Index	54.97 (28.46 – 76.09)	47.29 (26.24 – 74.98)	0.32
Minority Status and Language Sub-Index	77.9 (61.46 – 90.75)	83.73 (69.32 – 94.92)	0.002
	Number (Proportion, %)
Female	27 (3.2)	4 (3.2)	1.00
Duration of homelessness prior to HUD-VASH enrollment			0.65
6–12 months	210 (24.85)	36 (28.8)
1–2 years	261 (30.89)	37 (29.6)
≥2 years	374 (44.26)	52 (41.6)
VA Service Connection			0.75
<50%	204 (24.14)	28 (22.4)
50% - 100%	205 (24.26)	30 (24)
Non-service connected	312 (36.92)	52 (41.6)
Other	124 (14.67)	15 (12)
Non-VA health insurance coverage			0.60
Medicare	205 (24.26)	23 (18.4)
Medicaid	43 (5.09)	10 (8)
Commercial	25 (2.96)	3 (2.4)
No Insurance	562 (66.51)	88 (70.4)
Other	10 (1.18)	1 (0.8)
ZIP code level Rural-Urban commuting area e			0.009
Metropolitan area core: primary flow within an urbanized area	682 (80.71)	116 (92.8)
Metropolitan area high commuting: primary flow 30% or more to an urbanized area	48 (5.68)	2 (1.6)
Metropolitan area low commuting: primary flow 10% to 30% to an urbanized area	3 (0.36)	0 (0)
Non-metropolitan area	112 (13.25)	7 (5.6)
Race/ethnicity			<0.001
Hispanic (any race)	34 (4.02)	5 (4)
Non-Hispanic Black	261 (30.89)	56 (44.8)
Non-Hispanic other race	31 (3.67)	11 (8.8)
Non-Hispanic white	519 (61.42)	53 (42.4)
Cardiovascular diseasesg
Coronary artery disease	526 (62.25)	88 (70.4)	0.09
Heart Failure	334 (39.53)	58 (46.4)	0.15
Ischemic Stroke/Cerebrovascular disease	254 (30.06)	40 (32)	0.68
Peripheral arterial disease	121 (14.32)	18 (14.4)	1.00
Atrial Fibrillation	201 (23.79)	21 (16.8)	0.09
Married	60 (7.1)	10 (8)	0.71
Is primary caregiver of children	30 (3.55)	3 (2.4)	0.79
Condition present or treatment needed at time of HUD-VASH enrollment h
Has alcohol use disorder	93 (11.01)	9 (7.2)	0.22
Has substance abuse disorder	232 (27.46)	36 (28.8)	0.75
Has mental health disorder	568 (67.22)	83 (66.4)	0.84
Is a danger to themselves	11 (1.3)	1 (0.8)	1.00
Is in danger from others	11 (1.3)	0 (0)	0.38
Needs mental health disorder treatment	460 (54.44)	64 (51.2)	0.50
Needs substance use disorder treatment	309 (36.57)	49 (39.2)	0.62
Needs medical treatment	643 (76.09)	91 (72.8)	0.44
Is receiving Supplemental Security Income	331 (39.17)	43 (34.4)	0.33
Is receiving Supplemental Nutrition Assistance Program benefits	230 (27.22)	33 (26.4)	0.91
Location prior to HUD-VASH enrollmenth
Emergency shelter	35 (4.14)	5 (4)	1.00
Outdoors	54 (6.39)	9 (7.2)	0.70
Transitional housing	47 (5.56)	1 (0.8)	0.02
Hospital	21 (2.49)	0 (0)	0.10
Project-based voucher	167 (19.76)	5 (4.03)	<0.001
Received HUD-VASH voucher previously	249 (29.47)	38 (30.4)	0.83
Had eviction from HUD-VASH housing previously	38 (4.5)	7 (5.6)	0.65

^a.Intervention group consisted of Veterans with pre-existing cardiovascular disease, who had experienced homelessness for at least 6 months, and who moved into permanent supportive housing (PSH) within 1 month of receiving a HUD-VASH voucher. Comparison group consisted of Veterans with the same characteristics, who moved into PSH 6–12 months after receiving a HUD-VASH voucher.

^b.Based on Kruskal–Wallis test for continuous variables and Fisher’s exact test for categorical variables

^c.Higher score indicates higher level of medical comorbidities.

^d.Higher score indicates greater degree of social vulnerability. Calculated using ZIP code level data from the 2015–2019 5-year American Community Survey (Supplemental Methods 2).

^e.Based on the US Department of Agriculture ZIP code level 2010 Rural-Urban Commuting Area Codes.

^f.Including Asian, Native American/Alaska Native, Hawaiian/Pacific Islander, more than one race

^g.Based on diagnostic codes listed in Supplemental Methods 1.

^h.Based on intake form completed on enrollment into VA homelessness services.

Primary outcomes

After adjusting for baseline covariates, among Veterans in the intervention group, 16.3% (95% CI 13.9% – 18.7%) had a cardiovascular ED visit or hospitalization in the 6 months prior to voucher receipt and 11.9% (95% CI 9.8% – 14.0%) in the 6 months after voucher receipt (Table 2, Figure 1). In the comparison group, the proportion with a cardiovascular ED visit or hospitalization was 13.4% (95% CI 8.0% – 18.8%) in the pre-voucher period and 18.1% (95% CI 11.4% – 24.8%) in the post-voucher period. Moving into PSH after receiving a voucher was associated with a −9.1 (95% CI −17.7 – −0.5, p=0.04) percentage point (pp) differential change in the 6-month probability of having a cardiovascular ED visit or hospitalization. Pre-voucher trends in the monthly probability of ED visits or hospitalization did not differ significantly between the two groups (Supplemental Table 4).

Table 2.

Difference-in-differences analysis^a

	Intervention or early HUD-VASH housingb Estimate (95% CI)		Comparison or delayed HUD-VASH housingb Estimate (95% CI)		Difference-in-differences estimate Estimate (95% CI), p-value
	Pre-voucher period	Post-voucher period	Pre-voucher period	Post-voucher period
Proportion of Veterans with any encounter during period	Percentage points
Cardiovascular ED visit or hospitalizationc	16.3 (13.9 – 18.7)	11.9 (9.8 – 14.0)	13.4 (8.0 – 18.8)	18.1 (11.4 – 24.8)	−9.1 (−17.7 – −0.5), p=0.04
All-cause ED visit or hospitalization	57.6 (54.6 – 60.6)	45.2 (42.0 – 48.4)	51.6 (43.8 – 59.5)	54.1 45.9 – 62.3)	−14.8 (−25.7 – −4.0), p=0.01
Number of outpatient visits	Number
All visits	4.6 (4.3 – 4.9)	4.2 (3.9 – 4.4)	4.3 (3.5 – 5.0)	4.5 (3.9 – 5.2)	−0.7 (−1.5 – 0.2), p=0.12
Cardiovascular disease related visitsd	0.7 (0.7 – 0.8)	0.7 (0.6 – 0.8)	0.8 (0.6 – 1.0)	0.9 (0.7 – 1.2)	−0.2 (−0.4 – 0.1), p=0.22
Mean Proportion of Days Covered with recommended cardiovascular medicationse	0.30 (0.28 – 0.33)	0.36 (0.33 – 0.39)	0.32 (0.25 – 0.39)	0.35 (0.27 – 0.42)	0.03 (−0.03 – 0.08), p=0.35

^a.Estimates based on generalized estimating equation models with indicator variables for period, group, and the interaction (DID term) between period and group. Models also adjusted for month and year of voucher receipt, primary Veterans Integrated Services Network (VISN), baseline clinical and demographic variables, income, duration of homelessness, use of other social services, type of habitation prior to enrollment, prior HUD-VASH enrollment, prior eviction, and area-level measures of social vulnerability, and rural-urban status. A binomial distribution with a logit link was used for models with the probability of emergency department visits and hospitalizations as the dependent variable and a negative binomial distribution with a log link was used for models with the number of outpatient visits and number of days covered (with necessary medication) as the dependent variable. Average marginal effects were calculated to estimate the absolute change in the predicted probability or number of each outcome.

^b.Intervention group consisted of Veterans with pre-existing cardiovascular disease, who had experienced homelessness for at least 6 months, and who moved into permanent supportive housing (PSH) within 1 month of receiving a HUD-VASH voucher. Comparison group consisted of Veterans with the same characteristics, who moved into PSH 6–12 months after receiving a HUD-VASH voucher.

^c.Cardiovascular visits or hospitalizations identified as those with primary diagnosis corresponding to cardiovascular condition listed in Supplemental Methods 1

^d.Cardiovascular disease related outpatient visit identified if any cardiovascular disease diagnosis code was listed for the encounter.

^e.Medication adherence measured by proportion of days covered (PDC) which is the ratio of the number of days in a period with medication available divided by the total number of days in the period. For each pre-existing cardiovascular condition, PDC was calculated for each guideline recommended medication classes as listed in Supplemental Table 3. The mean PDC was then calculated for each medication class patient was recommended to be taking.

Figure 1: Adjusted six-month probability of cardiovascular emergency department visits or hospitalizations in the pre and post-HUD VASH voucher period by intervention group^a,b,c.

a. Estimates based on generalized estimating equation models after adjusting for month and year of voucher receipt, primary Veterans Integrated Services Network (VISN), baseline clinical and demographic variables, income, duration of homelessness, use of other social services, type of habitation prior to enrollment, prior HUD-VASH enrollment, prior eviction, and area-level measures of social vulnerability, and rural-urban status.

b. Error bars indicate 95% confidence intervals of the estimates.

c. Intervention group consisted of Veterans with pre-existing cardiovascular disease, who had experienced homelessness for at least 6 months, and who moved into permanent supportive housing (PSH) within 1 month of receiving a HUD-VASH voucher. Comparison group consisted of Veterans with the same characteristics, who moved into PSH 6–12 months after receiving a HUD-VASH voucher.

d. Cardiovascular ED visits or hospitalizations identified as those with primary diagnosis corresponding to cardiovascular condition listed in Supplemental Methods 1

Secondary outcomes

Moving into PSH was also associated with a −14.8 (95% CI −25.7 – −4.0) pp differential change in the 6-month probability of all-cause ED visits or hospitalizations (Table 2, Figure 2). The DID estimates for the number of outpatient visits for any cause or for CVD related visits and the mean PDC for CVD related medications was not statistically significant [−0.7 (95% CI −1.5 – 0.2) pp, −0.2 (95% CI −0.4 – 0.1) pp, and 0.03 (95% CI −0.03 – 0.08) pp respectively].

Figure 2: Adjusted six-month probability of all-cause emergency department visits or hospitalizations in the pre and post-HUD VASH voucher period by intervention group^a,b,c.

a. Estimates based on generalized estimating equation models after adjusting for month and year of voucher receipt, primary Veterans Integrated Services Network (VISN), baseline clinical and demographic variables, income, duration of homelessness, use of other social services, type of habitation prior to enrollment, prior HUD-VASH enrollment, prior eviction, and area-level measures of social vulnerability, and rural-urban status.

b. Error bars indicate 95% confidence intervals of the estimates.

c. Intervention group consisted of Veterans with pre-existing cardiovascular disease, who had experienced homelessness for at least 6 months, and who moved into permanent supportive housing (PSH) within 1 month of receiving a HUD-VASH voucher. Comparison group consisted of Veterans with the same characteristics, who moved into PSH 6–12 months after receiving a HUD-VASH voucher.

Sub-group and sensitivity analysis

In sub-group analyses based on cardiovascular condition, the DID estimate for the primary outcome among Veterans with CAD, HF, and stroke was −11.5 (95% CI −21.9 – −1.2) pp, −10.2 (95% CI −23.8 – 3.3) pp, −19.4 (95% CI −35.8 – −3.0) pp, respectively (Supplemental Table 5). Due to small sample size, models for Veterans with PAD and AF did not converge. The DID estimate for Veterans with a tenant-based voucher was −9.8 (95% CI −20.2 – 0.6) pp.

Additional 6-month follow-up

Among the delayed housing group, the adjusted 6-month probability of having a cardiovascular ED visit or hospitalization, declined significantly from 15.2% (95% CI 9.3% – 21.1%) in the 6 months prior to moving into PSH to 6.3% (95% CI 2.3% – 10.4%) in the 6 months after moving in (Figure 3). For the early housing group, among the 672 Veterans who remained in PSH for at least 12 months, the probability of a cardiovascular ED visit or hospitalization was 11.6% (95%CI 9.2% – 13.9%) in the first 6 months after receiving a voucher and 12.0% (95% CI 9.7% – 14.4%) in the subsequent 6 months. During this additional follow-up period, the delayed housing group had a −9.3 (95% CI −16.1 - −2.5) pp change in the probability of cardiovascular ED visits or hospitalizations compared to the early housing group.

Figure 3: Adjusted six-month probability of cardiovascular emergency department visits or hospitalizations for an additional 6 months by intervention group^a,b,c,d,e.

a. Intervention group consisted of Veterans with pre-existing cardiovascular disease, who had experienced homelessness for at least 6 months, and who moved into permanent supportive housing (PSH) within 1 month of receiving a HUD-VASH voucher. Comparison group consisted of Veterans with the same characteristics, who moved into PSH 6–12 months after receiving a HUD-VASH voucher. Only intervention group Veterans who were in PSH for ≥12 months were included.

b. For comparison group Veterans, initial 6 months refers to 6 months prior to moving into PSH and follow-up 6 months refers to 6 months after moving into PSH. For Intervention group Veterans, initial 6 months refers to first 6 months after HUD-VASH voucher receipt and follow-up 6 months refers to the 6 to 12 month period after voucher receipt.

c. Estimates based on generalized estimating equation models after adjusting for month and year of voucher receipt, primary Veterans Integrated Services Network (VISN), baseline clinical and demographic variables, income, duration of homelessness, use of other social services, type of habitation prior to enrollment, prior HUD-VASH enrollment, prior eviction, and area-level measures of social vulnerability, and rural-urban status.

d. Error bars indicate 95% confidence intervals of the estimates.

e. Cardiovascular ED visits or hospitalizations identified as those with primary diagnosis corresponding to cardiovascular condition listed in Supplemental Methods 1

Discussion

In this DID study of Veterans with CVD experiencing homelessness, among Veterans who obtained a HUD-VASH voucher, those who moved into PSH within 1 month of voucher receipt experienced a greater reduction in the 6-month probability of ED visits or hospitalizations, compared to the 6-month change among Veterans who had not yet moved into PSH after receiving a voucher. Therefore, moving into PSH may be associated with a meaningful short-term reduction in the probability of cardiovascular ED visits or hospitalizations.

Individuals experiencing homelessness are disproportionately impacted by CVD.(3, 17) However, the evidence that PSH improves health, particularly related to CVD, is limited.(18,19) As the largest PSH program for Veterans experiencing homelessness, understanding whether HUD-VASH can improve the cardiovascular health of this population is important. A 2010 analysis found that HUD-VASH participation was associated with lower healthcare costs.(8) However, since this study, HUD-VASH has expanded significantly, with a greater focus on Veterans experiencing chronic homelessness. To our knowledge, this is the first study to examine the relationship between HUD-VASH and changes in CVD outcomes using contemporary, national data and contributes to the evidence that PSH may play a role in improving health outcomes.

HUD-VASH participants may differ in meaningful ways from non-participants, and Veterans enrolled in HUD-VASH have previously been noted to have higher care utilization than Veterans experiencing homelessness not in the program.(20) To account for these differences, this study examined Veterans who received a HUD-VASH voucher and compared those who moved into PSH earlier vs. later after voucher receipt. Although there were some differences with respect to race and ethnicity between the intervention and comparison groups, pre-intervention trends did not differ significantly between the two groups, suggesting that the DID design helped isolate changes in outcomes linked to the immediate receipt of PSH. It is possible that there may have been unmeasured differences in psychosocial factors, not fully captured in our model, that could explain some of the differences noted, such as the stress of being unable to find housing after receiving a voucher leading to an increase in adverse health events. However, in the delayed housing group, a significant decrease in the primary outcome was also noted after moving into PSH.

The mechanisms underlying the association between PSH and cardiovascular outcomes are not completely elucidated in this study. No significant differences in outpatient care and medication adherence were noted. Other mechanisms that might explain the association between PSH and lower ED visits and hospitalizations may be related to the alleviation of psychosocial stress. Individuals experiencing psychosocial stress, including housing instability, have elevated inflammatory markers and increased activation of the sympathetic nervous system and hypothalamic-pituitary-adrenal axis.(21–24) These pathways in turn have been associated with the development of CVD and acute events.(25–27) Furthermore, one previous study noted that PSH was associated with a reduction in perceived stress.(7) With the association of housing insecurity with poor sleep and dietary quality, both of which are linked with adverse cardiovascular health, improvements in these factors may also a play a role in the association of PSH with improved cardiovascular outcomes.(28–30) These and other mechanisms, including the coordination and intensity of case management and supportive services accompanying PSH, warrant investigation. Within the VA, engagement with supportive services and primary care have been associated with improved housing stability among HUD-VASH participants.(31) Such factors, for HUD-VASH specifically, may potentially improve the efficacy of PSH in improving health outcomes.

The specific reasons why each Veteran who received a HUD-VASH voucher had a delay in moving into housing is unclear. However, one difference between the two groups was a higher proportion of people who received project-based, rather than tenant-based, vouchers in the intervention group. As project-based vouchers do not require the recipient to actively seek out eligible housing, this can shorten the time to moving into PSH. In a sensitivity analysis limited to Veterans with tenant-based vouchers, the DID estimate was qualitatively similar to the primary model, suggesting that differences in voucher type is likely not the primary reason for the noted association. Other possible reasons for differences in time to housing include discrimination faced by individuals based on race and ethnicity when seeking housing that have been described previously.(32,33)

Limitations

This study has some limitations. First, although a quasi-experimental design was used for this analysis, unmeasured differences between the intervention and comparison groups (who moved into PSH earlier vs. later) introduce bias if these differences are correlated with outcome trends. However, pre-intervention parallel trends suggest that unmeasured confounding, conditional on the included covariates, was minimized. Second, we were unable to identify specific reasons why each Veteran in the comparison group had a delayed entry into PSH after receiving a HUD-VASH voucher. However, in a secondary analysis examining outcomes among the delayed entry group, the association of housing with the primary outcomes was similar to that in the early entry group. Third, findings may not generalize to non-Veterans or other PSH programs. However, Veterans compose a significant proportion of the overall population experiencing homelessness and these findings provide some insight into the role PSH may play in improving health outcomes among individuals experiencing homelessness. Fourth, as being in the comparison group required a Veteran to survive for 6 months before entering PSH, Veterans who died during the study period were not included. This analysis therefore cannot account for potential differences in out-of-hospital mortality. However, the outcomes evaluated – cardiovascular hospitalizations and ED visits – are associated with mortality.(34) Additionally, as Veterans in the delayed entry group were required to survive for a longer period of time while still unhoused, this may select against less healthy Veterans in this group. However, this is likely to bias the results towards the null and underestimate the true effect of PSH. Fifth, although our analysis included outcomes that occurred in the VA health system and those that were paid for by the VA, it did not include non-VA healthcare utilization which was not paid for by VA. However, prior studies suggest that engagement with VA homelessness services, as all of the Veterans in the study population had, is associated with a shifting of care utilization to VA sources.(35) Finally, potential miscoding of CVD related visits is possible. However, our analysis found similar trends with all-cause outcomes.

Conclusions

In this nationwide analysis of HUD-VASH for Veterans with pre-existing CVD, entry into PSH was associated with fewer ED visits and hospitalizations. Further study of policies that expand access to PSH is needed to determine if such policies may play a role in improving the long-term cardiovascular health of individuals experiencing housing instability.