A National Study of Homelessness,Social Determinants of Health,and Treatment Engagement Among Outpatient Medication for Opioid Use Disorder-Seeking Individuals in the United States

Abstract

Background:

Homelessness is an important social determinant of health (SDOH), impacting health outcomes for many medical conditions. Although homelessness is common among people with opioid use disorder (OUD), few studies systematically evaluate homelessness and other SDOH among people enrolled in standard of care treatment for OUD, medication for opioid use disorder (MOUD), or examine whether homelessness affects treatment engagement.

Methods:

Using data from the 2016 to 2018 U.S. Treatment Episode Dataset Discharges (TEDS-D), patient demographic, social, and clinical characteristics were compared between episodes of outpatient MOUD where homelessness was reported at treatment enrollment versus independent housing using pairwise tests adjusted for multiple testing. A logistic regression model examined the relationship between homelessness and treatment length and treatment completion while accounting for covariates.

Results:

There were 188 238 eligible treatment episodes. Homelessness was reported in 17 158 episodes (8.7%). In pairwise analysis, episodes involving homelessness were significantly different from those involving independent living on most demographic, social, and clinical characteristics, with significantly greater social vulnerability in most SDOH variables (P’s < .05). Homelessness was significantly and negatively associated with treatment completion (coefficient = −0.0853, P < 0.001, 95% CI = [−0.114, −0.056], OR = 0.918) and remaining in treatment for greater than 180 days (coefficient = −0.3435, P < 0.001, 95% CI = [−0.371, −0.316], OR = 0.709) after accounting for covariates.

Conclusions:

Patients reporting homelessness at treatment entry in outpatient MOUD in the U.S. represent a clinically distinct and socially vulnerable population from those not reporting homelessness. Homelessness independently predicts poorer engagement in MOUD confirming that homelessness may be an independent predictor for MOUD treatment discontinuation nationally.

Keywords

housing medications for opioid use disorder methadone buprenorphine opioid use disorder homelessness

Introduction

The overdose epidemic is rapidly worsening in the United States. There were more than 105 000 lives lost to overdose in 2021, with a similar number predicted for 2022.¹ Multiple factors contribute to this crisis, including a drug supply that contains increasing quantities of potent synthetic opioids,^2,3 but it is in part driven by untreated opioid use disorder (OUD).⁴ People experiencing homelessness (PEH) are particularly impacted by the overdose crisis. Overdose is the leading cause of death among PEH, comprising up to one fourth of deaths in this group.^5,6 Compared to housed individuals, those experiencing homelessness have a mortality rate 7 to 9 times higher.^6,7 As such, addressing OUD and overdose in this high-risk group should be a priority for treatment organizations, public health entities, and governmental agencies.

The relationship between homelessness and substance use disorders (SUDs) is bidirectional. Homelessness is common among people with SUDs and may exacerbate substance use; likewise, SUDs, including OUD, are common among PEH⁸ and may contribute to housing instability.⁹ Medication for opioid use disorder (MOUD) with methadone, buprenorphine, or naltrexone, is evidence-based treatment for OUD and represents the standard of care.¹⁰ Receipt of MOUD confers many patient benefits, and is the only OUD treatment shown to reduce substance use, infectious disease risk, fatal overdoses, and all-cause mortality.^10
-13 MOUD enrollment is also associated with housing-specific improvements, such as increased retention in supportive housing.¹⁴ While many people with OUD receive non-MOUD treatments,¹⁵ PEH with OUD who enter SUD treatment are less likely than their housed counterparts to receive MOUD.^16
-18 Expanding engagement in evidence-based OUD treatment is even more important in the face of the aforementioned worsening overdose crisis, which disproportionately impacts PEH.

Despite its prevalence among people with OUD, research on homelessness and other social factors that impact health among this population is scant. Housing is a social determinant of health, evidenced by people experiencing homelessness having higher rates of chronic illness, delayed diagnoses of medical conditions, reduced access to treatment and preventive health, and increased mortality compared to those who are housed.^19,20 Social Determinants of Health (SDOH) are individual and systemic factors, such as income, insurance status, or housing, that impact an individual’s physical wellbeing, access to healthcare, and health outcomes.^21,22 Homelessness is an important social determinant that impacts health, but it is intertwined with other factors that may also limit healthcare access, like poverty or insurance status. Investigating SDOH among individuals seeking MOUD may reveal potential areas for intervention to increase engagement in MOUD and improve health outcomes and quality of life among people with OUD. For example, one prior study demonstrated that out-of-pocket costs are associated with treatment discontinuation in buprenorphine.²³ However, literature evaluating patient social characteristics in OUD treatment, particularly MOUD, is generally scarce. To our knowledge, there are no published studied examining SDOH based on housing status in MOUD.

MOUD retention rates range widely²⁴ with most individuals staying in treatment fewer than 6 months.^25,26 Few studies examine MOUD retention among PEH. Studies investigating predictors of retention in methadone,^27
-29 buprenorphine,³⁰ and MOUD generally,²⁵ demonstrate associations between housing and treatment engagement, but few prior studies directly compare MOUD outcomes based on housing status.^18,31
-35 A recent study found that homelessness is a significant independent predictor of reduced 12-month retention in low-barrier outpatient methadone treatment, even after adjusting for sociodemographic and clinical characteristics.³¹ Studies examining potential associations between housing and engagement in buprenorphine have mixed results.^30,33
-35 One prior study utilized the 2018 Treatment Episode Dataset Discharges (TEDS-D), a national database of SUD treatment episodes across the United States in a given year, to examine factors associated with MOUD retention length greater than 6 months and found homelessness was one of several associated factors.²⁵ A recent study utilizing the 2015 to 2017 TEDS-D database found that, among all outpatient OUD treatment episodes, those experiencing homelessness were less likely to enter MOUD and had reduced treatment completion and 90-day retention in all treatment types (ie, non-pharmacologic and MOUD).¹⁸ However, no published studies have examined whether there are differences in longer-term retention once patients experiencing homelessness enter MOUD.

Prior work suggests that demographic characteristics (eg, age), clinical characteristics (eg, substance use), and treatment-related characteristics (eg, methadone dose) affect retention in MOUD.^23,24,36,37 However, there are surprisingly few studies examining whether such characteristics differ in OUD treatment between PEH and their housed counterparts. Two prior studies compared patient demographic and clinical characteristics based on housing status using the TEDS database^17,18; Han and colleagues studied all OUD treatment modalities whereas Friesen and Young examined episodes from all outpatient OUD treatment types. They both found significant differences in demographic (eg, sex) and clinical variables (eg, co-occurring psychiatric and substance use disorders) based on housing status, but whether these differences persist for individuals receiving standard of care OUD treatment (MOUD) is unknown.

We used a national database to investigate patient characteristics, social determinants of health, and retention in outpatient MOUD treatment based on housing status. Using data on outpatient MOUD treatment episodes from the 2016 to 2018 Treatment Episode Dataset Discharges (TEDS-D), we compared patient demographics, SDOH, and clinical characteristics based on housing status and examined 2 measures of engagement: 180-day treatment retention of time in treatment and treatment completion. We hypothesized that housing status would be significantly and negatively associated with treatment engagement and that this association would remain after accounting for other covariates.

Materials and Methods

Data Source

Data were retrieved from the 2016 to 2018 Substance Abuse and Mental Health Services Administration (SAMHSA) Treatment Episode Data Set—Discharges (TEDS-D), a national, publicly available database containing information about all SUD treatment episodes that were either publicly funded or conducted at facilities receiving public funding in the U.S. SAMHSA requires all states to report this information for any treatment episodes with providers receiving public funding. The TEDS dataset represents the majority of all SUD treatment episodes conducted in the U.S. in a given year and the TEDS-Discharges (TEDS-D) contains information on both admissions and discharges for these treatment episodes. These data reflect treatment episodes from all states with the exception of Georgia (2016-2018), Oregon (2016-2018), Washington (2018 only), and West Virginia (2016-2018), which were excluded due to a lack of sufficient data. The data reflect treatment episodes rather than individual patients; thus, an individual patient may be included more than once in the dataset if they had multiple treatment episodes. Study investigators contacted the Human Investigations Committee at the Yale School of Medicine to determine whether submission was appropriate for this study, who determined a study utilizing this publicly available dataset did not require review.

Population

We limited our study population to treatment episodes for outpatient MOUD treatment. We restricted the analysis to treatment episodes coded as ambulatory, non-intensive outpatient, then further limited to those where patients reported homelessness or independent living for their living arrangements at admission. We excluded those with dependent living arrangements due to the potential heterogeneity of these settings. We then further restricted the analysis to episodes whose completion reason was listed as: “completed,” “dropped out,” or “terminated by facility.” We excluded episodes whose reason for treatment discontinuation were coded as “transferred to another treatment program or facility,” “incarcerated,” “death,” or “other.” Patient transfers were excluded as the definition or requirements for a transfer may differ across states and individual treatment facilities.

Variables

Treatment Completion

The TEDS database provides data on episode reason for discharge. We considered a patient to have completed treatment if their reason for treatment completion was coded as “Completed.” Participants were considered to have been prematurely terminated from treatment for those whose reasons for discharge were “dropped out,” or “terminated by facility,” representing patient-initiated and program-initiated discharges, respectively.

Treatment Length

Treatment episode length was reported as an interval censored length variable. In the dataset, treatment episodes lasting up to 30 days have the actual number of days in treatment coded, whereas longer lengths of stay are coded as 31 to 45, 46 to 60, 61 to 90, 91 to 120, 121 to 180, 181 to 365, and greater than 365 days. We dichotomized this variable such that episodes were categorized as lasting longer than 180 days (a composite variable composed of the 181 to 365 and greater than 365 days groups) or 180 days or less (a composite variable composed of the remaining 5 groups). This time point was selected to align with other studies of MOUD treatment retention that utilize between 180 and 365 days as a marker of longer-term retention^24,25 including one prior study which illustrated significantly decreased retention at 12-months between individuals enrolled in methadone treatment who reported experiencing homelessness at treatment entry compared to those who were living independently.³¹

Demographic, Social, and Clinical Characteristics

A full codebook for the TEDS-D dataset is available at https://www.datafiles.samhsa.gov/dataset/teds-d-2018-ds0001-teds-d-2018-ds0001. Race was compressed to American Indian or Alaska Native/Asian, Pacific Islander, or Native Hawaiian/Black/other/2 or more races/white. Ethnicity was compressed to Hispanic/Not Hispanic. Marital status was compressed to Now married/Never married/Previously married. Education was compressed to high school or greater/less than high school.

Statistical Analysis

Housing Status Versus Covariates Pairwise Analysis

To understand associations between housing status and the other covariates, difference in distribution tests were run for each covariate comparing the housed to the unhoused group. Categorical variables were evaluated with a chi-squared test. The ordinal “age” variable was evaluated with a Cochran-Armitage test.^38,39 After computing raw P-values for each test, we controlled the False Discovery Rate with the Benjamini-Hochberg (BH) procedure to obtain adjusted P-values.⁴⁰

Housing Status Versus Dropout Reason

For the subset of 156 840 treatment episodes where the non-completion reason was listed as “Dropped out” or “Terminated by facility” we tested for an association between non-completion reason and housing status using a chi-squared test.

Retention Regression, Confirmatory Analysis: Housing Status Association With Retention

We tested the hypothesis that the association between housing status and retention would remain after accounting for other covariates by fitting a logistic regression model regressing the retention outcome on 61 covariates. For this analysis we report the p-value for housing status. The logistic regression model was fit using the statsmodels in python.⁴¹ We completed this regression twice to examine 2 outcomes for retention: length of time and reason for treatment termination. For length of time, we compared whether an episode of treatment lasted at greater than 180 days or not. For reason for treatment termination, we compared completed treatment to premature treatment discontinuation, a combination variable containing episodes coded as “dropped out” or “terminated by facility.”

Software and Reproducibility

The statistical analysis was performed using Python and R. A software repository reproducing the entire analysis can be found at https://github.com/idc9/repro_teds_d_retention. In addition to the packages mentioned above, the analysis made use of matplotlib,⁴² numpy,⁴³ pandas,⁴⁴ and sklearn.⁴⁵

Results

The overall sample contained 188 238 treatment episodes, within which the patients reported independent living in 171 115 episodes and experiencing homelessness in 17 158. Results of the pairwise analysis comparing characteristics of patients for each treatment episode based on housing status are shown in the tables: Table 1 shows demographic characteristics; Table 2 shows social characteristics; and Table 3 shows clinical and treatment-related characteristics.

Table 1.

Demographic Characteristics Based on Housing Status Among 188 238 Treatment Episodes for Outpatient Medication for Opioid Use Disorder Treatment in the United States From 2016 to 2018.

	Overall	Housed	Experiencing homelessness	Test statistic	Adjusted P value
	Overall	F	T	Test statistic	Adjusted P value
N	188 238	171 115	17 158
Biological sex				183.96	<0.01
Female, n (%)	78 361 (39.9)	69 099 (40.4)	6015 (35.1)
Male, n (%)	1 17 970 (60.1)	101 987 (59.6)	11 137 (64.9)
Ethnicity				346.82	<0.01
Non-Hispanic, n (%)	1 55 872 (81.5)	138 060 (81.8)	12 874 (76.0)
Hispanic, n (%)	35 271 (18.5)	30 669 (18.2)	4072 (24.0)
Age at admission, mean (SD)	38.1 (11.8)	38.0 (11.9)	39.3 (11.4)	1 09 689.0	<0.01
Race				693.82	<0.01
American Indian or Alaska native, n (%)	2508 (1.3)	2042 (1.2)	311 (1.8)
Asian, Pacific Islander, or native Hawaiian, n (%)	1566 (0.8)	1400 (0.8)	133 (0.8)
Black, n (%)	24 594 (12.8)	21 298 (12.6)	2641 (15.6)
Other, n (%)	23 993 (12.5)	20751 (12.2)	2916 (17.2)
2 or more, n (%)	3498 (1.8)	3045 (1.8)	434 (2.6)
White, n (%)	135 849 (70.8)	120 913 (71.4)	10 515 (62.0)
Region				1560.45	<0.01
Midwest, n (%)	23 455 (11.9)	20 953 (12.2)	1851 (10.8)
Northeast, n (%)	106 866 (54.4)	96 894 (56.6)	8756 (51.0)
South, n (%)	21 756 (11.1)	18 280 (10.7)	992 (5.8)
US territories, n (%)	1022 (0.5)	654 (0.4)	168 (1.0)
West, n (%)	43 267 (22.0)	34 334 (20.1)	5391 (31.4)

Table 2.

Social Determinants of Health Based on Housing Status Among 188 238 Treatment Episodes for Outpatient Medication for Opioid Use Disorder Treatment in the United States From 2016 to 2018.

	Overall	Housed	Experiencing homelessness	Test statistic	Adjusted P value
	Overall	F	T	Test statistic	Adjusted P value
N	188 238	171 115	17 158
Relationship status				323.74	<0.01
Currently married, n (%)	24 980 (16.4)	23 251 (17.0)	1335 (10.8)
Never married, n (%)	98 673 (64.7)	87 621 (64.1)	8343 (67.7)
Previously married, n (%)	28 839 (18.9)	25 782 (18.9)	2643 (21.5)
Education experience				475.42	<0.01
No high school diploma, n (%)	51 192 (27.2)	44 625 (26.4)	5818 (34.2)
High school diploma, n (%)	137 210 (72.8)	124 361 (73.6)	11 191 (65.8)
Employment history				3429.49	<0.01
Any employment, n (%)	50 113 (26.6)	48 043 (28.6)	1302 (7.7)
Not in labor force, n (%)	72 554 (38.5)	62 515 (37.2)	8254 (48.9)
Unemployed, n (%)	65 890 (34.9)	57 330 (34.1)	7328 (43.4)
Veteran experience				5.62	0.0191
Not veteran, n (%)	178 410 (97.4)	159 912 (97.4)	15 744 (97.7)
Veteran, n (%)	4678 (2.6)	4247 (2.6)	367 (2.3)
Primary source of income				3984.86	<0.01
Wages/salary, n (%)	33 230 (28.3)	32 268 (30.5)	699 (7.2)
Public assistance, n (%)	17 595 (15.0)	14 034 (13.3)	2703 (27.7)
Retirement/pension, disability, n (%)	9213 (7.8)	8370 (7.9)	585 (6.0)
Other, n (%)	31 200 (26.6)	28 591 (27.1)	2234 (22.9)
None, n (%)	26 129 (22.3)	22 397 (21.2)	3530 (36.2)
Criminal-legal history				204.12	<0.01
No arrests in the 30 days prior to admission, n (%)	180 500 (95.1)	160 498 (95.4)	15 776 (93.0)
Arrests in the 30 days prior to admission, n (%)	9256 (4.9)	7704 (4.6)	1195 (7.0)
Health insurance				238.33	<0.01
Private insurance, Blue Cross/Blue Shield, HMO, n (%)	3513 (6.2)	3233 (6.8)	112 (2.7)
Medicaid, n (%)	37 648 (66.0)	30 174 (63.7)	2567 (60.8)
Medicare, other, n (%)	4013 (7.0)	3530 (7.5)	261 (6.2)
None, n (%)	11 892 (20.8)	10 433 (22.0)	1285 (30.4)
Primary payment source				539.16	<0.01
Self-pay, n (%)	6814 (11.8)	6540 (13.6)	207 (5.2)
Private insurance, n (%)	1739 (3.0)	1656 (3.4)	45 (1.1)
Medicare, n (%)	1336 (2.3)	1034 (2.1)	99 (2.5)
Medicaid, n (%)	37 336 (64.5)	29 815 (61.9)	2423 (61.4)
Other government payments, n (%)	7553 (13.1)	6592 (13.7)	819 (20.7)
No charge (free, charity, special research, and teaching), n (%)	1087 (1.9)	737 (1.5)	162 (4.1)
Other, n (%)	1991 (3.4)	1770 (3.7)	193 (4.9)

Table 3.

Clinical and Treatment-Related Characteristics Based on Housing Status Among 188 238 Treatment Episodes for Medication for Opioid Use Disorder Treatment in the United States From 2016 to 2018.

	Overall	Housed	Experiencing Homelessness	Test Statistic	Adjusted P value
	Overall	F	T	Test Statistic	Adjusted P value
N	188 238	171 115	17 158
Reason for treatment termination				654.23	<0.01
Completed, n (%)	39 526 (20.1)	33 282 (19.5)	2145 (12.5)
Dropped out, n (%)	126 596 (64.5)	110 572 (64.6)	12 670 (73.8)
Terminated by facility, n (%)	30 244 (15.4)	27 261 (15.9)	2343 (13.7)
Length of time in treatment
Did not stay more than 180 days, n (%)	99 045 (50.4)	83 757 (48.9)	10 012 (58.4)
Stayed more than 180 days, n (%)	97 321 (49.6)	87 358 (51.1)	7146 (41.6)
Referral source				305.02	<0.01
Individual (including self), n (%)	136 290 (70.5)	119 597 (70.4)	11 465 (67.6)
Alcohol/drug use care provider, n (%)	18 170 (9.4)	16 133 (9.5)	1900 (11.2)
Other health care provider, n (%)	11 197 (5.8)	9616 (5.7)	1052 (6.2)
School (educational), n (%)	59 (0.0)	49 (0.0)	8 (0.0)
Employer/EAP, n (%)	297 (0.2)	272 (0.2)	20 (0.1)
Other community referral, n (%)	12 390 (6.4)	10 674 (6.3)	1500 (8.8)
Court/criminal justice referral/DUI/DWI, n (%)	14 855 (7.7)	13 550 (8.0)	1023 (6.0)
Prior substance use treatment episodes				57.54	<0.01
No prior treatment episodes, n (%)	47 969 (24.9)	40 515 (24.1)	3646 (21.5)
1 or more prior treatment episodes, n (%)	144 324 (75.1)	127 607 (75.9)	13 317 (78.5)
Primary substance use at admission				1696.92	<0.01
None, n (%)	1147 (0.6)	492 (0.3)	16 (0.1)
Methamphetamine/speed, n (%)	1332 (0.7)	1123 (0.7)	131 (0.8)
Benzodiazepines, n (%)	927 (0.5)	868 (0.5)	45 (0.3)
Inhalants, n (%)	27 (0.0)	21 (0.0)	6 (0.0)
Over the counter medications, n (%)	15 (0.0)	15 (0.0)
Other drugs, n (%)	595 (0.3)	532 (0.3)	48 (0.3)
Alcohol, n (%)	6128 (3.2)	5589 (3.3)	432 (2.5)
Cocaine/crack, n (%)	2919 (1.5)	2543 (1.5)	327 (1.9)
Marijuana/hashish, n (%)	2456 (1.3)	2227 (1.3)	168 (1.0)
Heroin, n (%)	142 527 (73.3)	123 578 (72.3)	14 639 (85.3)
Non-prescription methadone, n (%)	1257 (0.6)	1178 (0.7)	59 (0.3)
Other opiates and synthetics, n (%)	35 010 (18.0)	32 681 (19.1)	1265 (7.4)
Phencyclidine, n (%)	43 (0.0)	34 (0.0)	9 (0.1)
Hallucinogens, n (%)	77 (0.0)	68 (0.0)	7 (0.0)
Frequency of use of the primary substance at admission				305.88	<0.01
No use in the past month, n (%)	36 281 (21.4)	32 746 (22.0)	2384 (15.9)
Some use, n (%)	30 675 (18.1)	27 118 (18.2)	2826 (18.9)
Daily use, n (%)	102 253 (60.4)	88 941 (59.8)	9772 (65.2)
Frequency of use of the primary substance at discharge				109.18	<0.01
No use in the past month, n (%)	54 031 (42.4)	48 444 (43.5)	3955 (38.5)
Some use, n (%)	32 708 (25.7)	28 740 (25.8)	2724 (26.5)
Daily use, n (%)	40 569 (31.9)	34 259 (30.7)	3589 (35.0)
Age at first use of primary substance				167.57	<0.01
18-29, n (%)	103 093 (53.9)	91 523 (54.2)	8727 (51.2)
Over 30, n (%)	39 350 (20.6)	35 005 (20.7)	3265 (19.2)
Under 18, n (%)	48 953 (25.6)	42 424 (25.1)	5050 (29.6)
Current IV drug use reported at admission				600.96	<0.01
Not reported, n (%)	42 838 (30.1)	38 644 (31.0)	2894 (20.9)
Reported, n (%)	99 685 (69.9)	86 068 (69.0)	10 939 (79.1)
Co-occurring mental and substance use disorders				262.93	<0.01
No, n (%)	103 492 (56.6)	91 243 (56.6)	7946 (49.9)
Yes, n (%)	79 394 (43.4)	70 023 (43.4)	7980 (50.1)
Attendance at substance use self-help groups (AA, NA, and other mutual aid) in 30 days prior to admission				0.03	0.8546
No attendance, n (%)	119 164 (78.0)	104 718 (77.8)	10 017 (77.7)
Attendance, n (%)	33 528 (22.0)	29 871 (22.2)	2869 (22.3)
Attendance at substance use self-help groups (AA, NA, and other mutual aid) in 30 days prior to discharge				0.34	0.5827
No attendance, n (%)	118 001 (80.7)	103 979 (80.7)	9819 (80.4)
Attendance, n (%)	28 287 (19.3)	24 940 (19.3)	2388 (19.6)

Abbreviations: EAP, employee assistance program; DUI, driving under the influence; DWI, driving while impaired; HMO, Health Maintenance Organization; AA, alcoholics anonymous; NA, narcotics anonymous.

Baseline Characteristics and Housing Status

All variables except the following were associated with housing status after applying the BH procedure: attendance at substance use self-help groups (AA [“Alcoholics Anonymous”], NA [“Narcotics Anonymous”], other mutual aid) in 30 days prior to admission and discharge. The group experiencing homelessness was more likely to be male, Hispanic, not currently married, without a high school diploma, without employment, not a veteran, and older, and have a recent arrest in the 30 days prior to admission and discharge, have had 1 or more prior treatment episodes, have more frequent substance use at admission and discharge, report first substance use at a younger age, report current intravenous drug use, report co-occurring mental health and substance use disorders, and have no health insurance. There were also significant differences between the primary substance used on admission to treatment, primary payment source for treatment, referral source, region, and racial background between the group of episodes where the patient reported homelessness at treatment admission versus those where the patient was housed.

Housing Status and Reason for Treatment Non-Completion

We examined the relationship between housing status and reason for treatment termination among those treatment episodes that were not coded as “completed.” Housing status was significantly associated with dropout reason (P < .01, df = 1). Compared to those who were housed, individuals experiencing homelessness were significantly less likely to be terminated by the program (15.6% unhoused vs 19.8% housed) and were more likely to experience patient-initiated treatment termination.

Regression Confirmatory Analyses

Housing status was statistically significantly related to retention for both logistic regression analyses. When premature treatment termination was the outcome, the housing status coefficient was estimated to be −0.4283 (P < 0.001, 95% CI = [−0.478, −0.378], odds ratio 0.6516) where the negative sign indicates homelessness has a negative effect on retention. For retention at greater than 180 days the housing status coefficient was estimated to be −0.4 (P < 0.001, 95% CI = [−0.434, −0.367], odds ratio = 0.67).

Discussion

To our knowledge, this study is one of the first investigations of differences in U.S. national MOUD treatment outcomes and social determinants of health based specifically on housing status. There were several main findings. First, the patients reporting homelessness at treatment entry into episodes of outpatient MOUD represent a clinically distinct and more socially vulnerable population from those with independent housing. Second, homelessness was significantly and negatively associated with staying in outpatient MOUD treatment for 180 days or longer, a relationship that remained significant after accounting for covariates. Third, homelessness was significantly negatively associated with treatment completion after adjusting for covariates. Finally, treatment episodes where homelessness was reported at treatment entry were more likely to end in a patient-initiated treatment termination and less likely to experience a program-initiated termination compared to those who were housed.

Our findings reveal a number of demographic differences and social and clinical vulnerabilities among people experiencing homelessness that are distinct from those housed at entry into outpatient MOUD treatment. On nearly every social, demographic, and clinical variable, there were significant differences between the group of episodes representing patients experiencing homelessness and those representing patients who were housed, aligning with national data on homelessness. For example, we found the group experiencing homelessness had significantly more males than the housed group. In comparison to their proportion of the general population, males are overrepresented both among the group of people experiencing homelessness in the U.S. and those with substance use disorders.⁴⁶ Similarly, we found a statistically significant association between race and housing status. Looking at Table 1, we show that people experiencing homelessness were empirically more likely to be non-white than the general population, which matches national data on homelessness.⁴⁷ Structural factors such as systemic racism may have led to a disproportionately high rate of homelessness among minoritized people in the United States generally and among individuals pursuing outpatient MOUD.⁴⁸

Our findings highlight the potentially additive or intersectional nature of social determinants of health for people experiencing homelessness and how they might potentially compound to worsen outcomes in MOUD treatment. Examples from the present study include how, compared to episodes among those who were housed, we found the group of episodes among PEH had significantly higher rates of unemployment, no income, receipt of public assistance, lacking health insurance, and arrests in the 30 days prior to admission and discharge. While the impact of SDOH on morbidity and mortality for many medical conditions^49
-51 has been well documented, the literature on the impact of SDOH on MOUD outcomes or interventions is limited.^52
-54 One existing study examining SDOH including income inequality and structural racism demonstrated an association with increased rates of injection drug use and worsened related outcomes such as HIV infection and AIDS-related mortality.⁵⁵ Similarly, social determinants, including housing status and geography, have been shown to be associated with overdose outcomes.⁵⁶ Investigation into interventions that directly address homelessness and other related SDOH is warranted, as are explorations of SDOHs at treatment entry and subsequent treatment planning. Unique to methadone treatment and other outpatient interventions are regulatory requirements of treatment planning at prescribed intervals (eg, 30, 60, 90, and 180 days). This mandatory contact with patients in methadone treatment is unique and could be considered an opportunity to explore and address SDOH.

To our knowledge, this is the first study to directly compare 180-day treatment retention in outpatient MOUD based on patient housing status. One prior study examining factors related to treatment retention in one year of TEDS-D data identified homelessness as one of several risk factors for reduced length of treatment in outpatient MOUD.²⁵ Our findings build off a recent study that found reduced retention and treatment completion at 90 days in outpatient OUD including MOUD by analyzing a more recent year of TEDS-D data and investigating longer-term treatment engagement, in addition to the expansion of social determinants and patient characteristics discussed above.¹⁸ Our results agree with previous work showing homelessness was an independent risk factor for 12-month treatment discontinuation in low-barrier methadone treatment, where all treatment terminations were patient-initiated³¹ to include different types of MOUD programs, such as those dispensing buprenorphine and naltrexone, and include program-initiated discharges. Prior work has indicated homelessness may be a risk factor for decreased retention in MOUD when including housing status as one potential covariate for treatment retention,^24,27,29,57 but these studies were in smaller samples and did not directly examine retention outcomes, such as length of time in treatment or reason for treatment termination, based on housing status. In the current study, homelessness was significantly and negatively associated with treatment completion and time in treatment in pairwise analysis and in logistic regression after adjusting for covariates, confirming that homelessness may be an independent predictor for MOUD treatment discontinuation nationally. The fact that episodes involving PEH were more likely than those among people who were housed to experience a patient-initiated, rather than program-initiated, premature termination is notable and requires future investigation into reasons PEH may prematurely terminate treatment. Future work into interventions that improve treatment retention among individuals experiencing homelessness is warranted.

Prior studies have examined housing status within the context of the TEDS dataset. One study examining racial and ethnic disparities in treatment effectiveness of outpatient non-MOUD substance use disorder treatment found that homelessness was associated with decreased improvement during treatment using the 2015 to 2017 dataset.⁵⁸ Another study investigating socioeconomic determinants of planned methadone treatment noted that those who were not experiencing homelessness were more likely to receive methadone as part of their treatment⁵⁹ and a later study found that experiencing homelessness was associated with decreased odds of receiving MOUD and also lower odds of waiting greater than 1 week to enter treatment.¹⁶ Similarly a study of the use of methadone within “detoxification” settings found those experiencing homelessness were less likely to have methadone planned for their treatment.¹⁵ The present study’s results build off these studies by demonstrating that, in addition to the disparities that may limit treatment entry into MOUD, those individuals experiencing homelessness that are able to enroll into MOUD have reduced retention length and decreased odds of treatment completion. One important area of future direction is to examine how state and city-specific policies impact disparities based on housing status observed in the present study. Future policies to address addiction among PEH will need to intervene on SDOH such as homelessness to have sustained outcomes.⁶⁰

Limitations

Our study findings should be considered in light of their limitations. Our definition of housing status relied on treatment centers’ determinations; thus, we were unable to provide a specific definition of homelessness beyond what is reported in the dataset. Our study design utilized housing status at treatment enrollment as the independent variable, limiting examination of longitudinal housing trajectories and their influence on MOUD outcomes. In addition, the TEDS database groups all 3 FDA-approved medications for OUD together, limiting evaluation into outcomes based specifically on treatment type (eg, comparison of methadone to buprenorphine to naltrexone). Finally, while we found many statistically significant differences between the 2 groups, it is possible that not all of these findings are clinically significant given the overall small difference in percentages between the 2 groups (eg, veteran experience, referral source, and prior substance use treatment episodes). Future work should examine how these and our other findings may impact patient experience and outcomes in MOUD.

Conclusion

In a national sample of treatment episodes for outpatient MOUD treatment, the episodes in which patients reported homelessness at treatment entry differed significantly in demographic, social, clinical, and treatment-related characteristics, representing a clinically distinct and more socially vulnerable population. Homelessness was significantly and negatively related to treatment retention in MOUD based on 2 separate metrics, length of time in treatment and treatment completion. Strategies to enhance retention in evidence-based treatment for OUD are needed in the midst of a worsening overdose crisis and may need to include systematic strategies for assessing and improving housing status among people experiencing homelessness enrolled in MOUD treatment.

Footnotes

Acknowledgements

Results from this study were presented as an oral presentation at the Association for Multidisciplinary Education and Research in Substance Use and Addiction (AMERSA) Conference in Boston, Massachusetts on November 10, 2022

Author Contributions

MGG and DTB conceived of the study aims. MGG, DTB, and IDC conceptualized study methodology. IDC conducted the formal analysis, curated the data, and validated the analysis. MGG produced data visualization and oversaw project administration. All authors wrote the manuscript. LMM provided resources for project completion. DTB supervised the project.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This publication was made possible by the James G. Hirsch Endowed Medical Student Research Fellowship and the Yale School of Medicine Medical Student Fellowship (MGG), the National Science Foundation under Award No. 1902440 (IDC), and the National Institutes of Health U01 HL150596-01 and RM1 DA055310 (DTB).

Compliance,Ethical Standards,and Ethical Approval

Study investigators contacted the Human Investigations Committee at the Yale School of Medicine to determine whether submission was appropriate for this study, who determined a study utilizing this publicly available dataset did not require review.

References

Ahmad

FB CJ

Rossen

Sutton

. Provisional Drug Overdose Death Counts. National Center for Health Statistics; 2022.

Yeo

Johnson

Heng

. The public health approach to the worsening opioid crisis in the united states calls for harm reduction strategies to mitigate the harm from opioid addiction and overdose deaths. Mil Med. 2021;187:244-247.

Shover

Falasinnu

Dwyer

, et al. Steep increases in fentanyl-related mortality west of the Mississippi River: Recent evidence from county and state surveillance. Drug Alcohol Depend. 2020;216:108314.

United Nations Office on Drugs and Crime. Opioid Overdose: Preventing and Reducing Opioid Overdose Mortality. United Nations Vienna; 2013.

Fine

Dickins

Adams

, et al. Drug overdose mortality among people experiencing homelessness, 2003 to 2018. JAMA Netw Open. 2022;5(1-2):e2142676.

Baggett

Hwang

O’Connell

, et al. Mortality among homeless adults in Boston: shifts in causes of death over a 15-year period. JAMA Intern Med. 2013;173(3):189-195.

Morrison

. Homelessness as an independent risk factor for mortality: results from a retrospective cohort study. Int J Epidemiol. 2009;38(3):877-883.

Fazel

Khosla

Doll

Geddes

. The prevalence of mental disorders among the homeless in western countries: systematic review and meta-regression analysis. PLoS Med. 2008;5(12):e225.

Nilsson

Laursen

Hjorthoj

Nordentoft

. Risk of homelessness after discharge from psychiatric wards in Denmark: a nationwide register-based cohort study. Acta Psychiatr Scand. 2019;140(5):477-489.

10.

National Academies of Sciences, Engineering, and Medicine. Medications for Opioid Use Disorder Save Lives. The National Academies Press; 2019.

11.

Sordo

Barrio

Bravo

, et al. Mortality risk during and after opioid substitution treatment: systematic review and meta-analysis of cohort studies. BMJ. 2017;357:j1550.

12.

Santo

Jr Clark

Hickman

, et al. Association of opioid agonist treatment with all-cause mortality and specific causes of death among people with opioid dependence: a systematic review and meta-analysis. JAMA Psychiatry. 2021;78(9):979-993.

13.

Wakeman

Larochelle

Ameli

, et al. Comparative effectiveness of different treatment pathways for opioid use disorder. JAMA Netw Open. 2020;3(2):e1920622.

14.

Hall

Walters

Gould

Lim

. Housing versus treatment first for supportive housing participants with substance use disorders: a comparison of housing and public service use outcomes. Subst Abus. 2020;41(1-2):70-76.

15.

Dunn

Huhn

Strain

. Differential adoption of opioid agonist treatments in detoxification and outpatient settings. J Subst Abuse Treat. 2019;107:24-28.

16.

Yang

Roman-Urrestarazu

Brayne

. Differences in receipt of opioid agonist treatment and time to enter treatment for opioid use disorder among specialty addiction programs in the United States, 2014-17. PLoS ONE. 2019;14(12):e0226349.

17.

Han

Doran

Krawczyk

. National trends in substance use treatment admissions for opioid use disorder among adults experiencing homelessness. J Subst Abuse Treat. 2021;132:108504.

18.

Friesen

Young

. Experiencing homelessness as a risk factor for negative treatment outcomes among individuals receiving outpatient treatment for opioid use disorder in the United States. Univ Tor Med J. 2021;98:56-62.

19.

Henwood

Lahey

Rhoades

Winetrobe

Wenzel

. Examining the health status of homeless adults entering permanent supportive housing. J Public Health. 2018;40(2):415-418.

20.

Stafford

Wood

. Tackling health disparities for people who are homeless? Start with social determinants. Int J Environ Res Public Health. 2017;14(12).

21.

Catalyst

NEJM

. Social determinants of health (SDOH). NEJM Catalyst. 2017;3(6). https://catalyst.nejm.org/doi/full/10.1056/CAT.17.0312

22.

Daniel

Bornstein

Kane

. Addressing social determinants to improve patient care and promote health equity: an American college of physicians position paper. Ann Intern Med. 2018;168(8):577-578.

23.

Agbese

Leslie

Manhapra

Rosenheck

. Early discontinuation of buprenorphine therapy for opioid use disorder among privately insured adults. Psychiatr Serv. 2020;71(8):779-788.

24.

O’Connor

Cousins

Durand

Barry

Boland

. Retention of patients in opioid substitution treatment: a systematic review. PLoS ONE. 2020;15(5):e0232086.

25.

Krawczyk

Williams

Saloner

Cerdá

. Who stays in medication treatment for opioid use disorder? A national study of outpatient specialty treatment settings. J Subst Abuse Treat. 2021;126:108329.

26.

Williams

Nunes

Bisaga

Levin

Olfson

. Development of a Cascade of Care for responding to the opioid epidemic. Am J Drug Alcohol Abuse. 2019;45(1-2):1-10.

27.

Klimas

Nosova

Socias

, et al. Factors associated with discontinuation of methadone maintenance therapy (MMT) among persons who use alcohol in Vancouver, Canada. Drug Alcohol Depend. 2018;186:182-186.

28.

Lundgren

Sullivan

Maina

Schilling

. Client factors associated with length of stay in methadone treatment among heroin users who inject drugs: quantitative analysis of state-level substance abuse treatment utilization data. J Addict Med. 2007;1(1-2):26-32.

29.

Huissoud

Rousson

Dubois-Arber

. Methadone treatments in a Swiss Region, 2001–2008: a registry-based analysis. BMC Psychiatry. 2012;12(1-2):238.

30.

Damian

Mendelson

Agus

. Predictors of buprenorphine treatment success of opioid dependence in two Baltimore City grassroots recovery programs. Addict Behav. 2017;73:129-132.

31.

Gaeta Gazzola

Carmichael

Madden

, et al. A cohort study examining the relationship among housing status, patient characteristics, and retention among individuals enrolled in low-barrier-to-treatment-access methadone maintenance treatment. J Subst Abuse Treat. 2022;138:108753.

32.

Perreault

Rousseau

Lauzon

Mercier

Tremblay

Héroux

M-C

. Determinants of retention in a canadian low-threshold methadone maintenance program. J Maint Addict 2008;3(2-4):37-51.

33.

Alford

LaBelle

Richardson

, et al. Treating homeless opioid dependent patients with buprenorphine in an office-based setting. J Gen Intern Med. 2007;22(2):171-176.

34.

Hersh

Little

Gleghorn

. Integrating buprenorphine treatment into a public healthcare system: The San Francisco department of public health’s office-based buprenorphine pilot program. J Psychoact Drugs. 2011;43(2):136-145.

35.

Hood

Banta-Green

Duchin

, et al. Engaging an unstably housed population with low-barrier buprenorphine treatment at a syringe services program: Lessons learned from Seattle, Washington. Subst Abuse. 2020;41(3):356-364.

36.

Dunphy

Peterson

Zhang

Jones

. Do out-of-pocket costs influence retention and adherence to medications for opioid use disorder? Drug Alcohol Depend. 2021;225:108784.

37.

McClellan

Moriya

Simon

. Users of retail medications for opioid use disorders faced high out-of-pocket prescription spending in 2011-2017. J Subst Abuse Treat. 2022;132:108645.

38.

Cochran

. Some methods for strengthening the common χ² tests. Biometrics. 1954;10(4):417-451.

39.

Armitage

. Tests for linear trends in proportions and frequencies. Biometrics. 1955;11(3):375-386.

40.

Benjamini

Hochberg

. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B. 1995;57(1-2):289-300.

41.

Seabold

Perktold

. Statsmodels: Econometric and statistical modeling with python. Proceedings of the 9th Python in Science Conference, Vol. 57, Austin, TX; 2010:61.

42.

Hunter

. Matplotlib: a 2D graphics environment. Comput Sci Eng. 2007;9(3):90-95.

43.

Walt

SVD

Colbert

Varoquaux

. The NumPy array: a structure for efficient numerical computation. Comput Sci Eng. 2011;13(2):22-30.

44.

McKinney

. Pandas: a foundational Python library for data analysis and statistics. Python High Performance Sci Comput. 2011;14(9):1-9.

45.

Pedregosa

Varoquaux

Gramfort

, et al. Scikit-learn: machine learning in Python. J Mach Learn Res. 2011;12:2825-2830.

46.

Nilsson

Nordentoft

Hjorthoj

. Individual-level predictors for becoming homeless and exiting homelessness: a systematic review and meta-analysis. J Urban Health. 2019;96(5):741-750.

47.

Henry

Watt

Mahathey

Ouellette

Sitler

Associates A. 2019 Annual Homeless Assessment Report to Congress Part 1: Point in Time Estimates of Homelessness. U.S. Department of Housing and Urban Development; 2020.

48.

Jones

. Does race matter in addressing homelessness? A review of the literature. World Med Health Policy. 2016;8(2):139-156.

49.

Havranek

Mujahid

Barr

, et al. Social determinants of risk and outcomes for cardiovascular disease. Circulation. 2015;132(9):873-898.

50.

Schneiderman

McIntosh

Antoni

. Psychosocial risk and management of physical diseases. J Behav Med. 2019;42(1-2):16-33.

51.

Liu

M-Y

Khan

. Association between psychosocial stress and hypertension: a systematic review and meta-analysis. Neurol Res. 2017;39(6):573-580.

52.

Galea

Vlahov

. Social determinants and the health of drug users: socioeconomic status, homelessness, and incarceration. Public Health Rep. 2002;117(Suppl 1):S135.

53.

Park

Rouhani

Beletsky

LEO

Vincent

Saloner

Sherman

. Situating the continuum of overdose risk in the social determinants of health: a new conceptual framework. Milbank Q. 2020;98(3):700-746.

54.

Dasgupta

Beletsky

Ciccarone

. Opioid crisis: no easy fix to its social and economic determinants. Am J Public Health. 2018;108(2):182-186.

55.

Friedman

Tempalski

Brady

, et al. Income inequality, drug-related arrests, and the health of people who inject drugs: reflections on seventeen years of research. Int J Drug Policy. 2016;32:11-16.

56.

Barocas

Wang

Marshall

BDL

, et al. Sociodemographic factors and social determinants associated with toxicology confirmed polysubstance opioid-related deaths. Drug Alcohol Depend. 2019;200:59-63.

57.

Kerr

Hayashi

, et al. Factors associated with methadone maintenance therapy discontinuation among people who inject drugs. J Subst Abuse Treat. 2018;94:41-46.

58.

Sahker

Pro

Sakata

Furukawa

. Substance use improvement depends on race/ethnicity: outpatient treatment disparities observed in a large US national sample. Drug Alcohol Depend. 2020;213:108087.

59.

Rivers

Dobalian

Oyana

Bae

. Socioeconomic determinants of planned methadone treatment. Am J Health Behav. 2006;30(5):451-459.

60.

Miller-Archie

Walters

Singh

Lim

. Impact of supportive housing on substance use-related health care utilization among homeless persons who are active substance users. Ann Epidemiol. 2019;32:1-6.e1.