Prevalence of and Risk Factors Associated With Nonfatal Overdose Among Veterans Who Have Experienced Homelessness

This survey study examines the prevalence of and risk factors associated with nonfatal drug or alcohol overdose among veterans who have experienced homelessness.


Introduction
Drug overdose accounts for approximately 70 000 deaths annually in the United States, 1 and nonfatal overdose accounts for more than 500 000 emergency department visits each year. 2 However, overdose risk varies substantially among individuals and across populations. Individuals who have experienced homelessness, including approximately 4% of Americans, 3 are at increased risk. Overdose is one of the most common causes of death for younger individuals experiencing homelessness, with opioids implicated in most cases. 4  Previous studies on risk factors associated with overdose in individuals experiencing homelessness are limited by several factors. One challenge is that overdose data reported to the National Center on Health Statistics 1,7 do not include housing status. Additionally, studies from single institutions or geographic regions have limited generalizability owing to region-specific variables, including drug supply contamination. 4 To our knowledge, there has not been a nationally representative study of overdose prevalence among individuals experiencing homelessness.
As part of a study to examine the association of primary care service design with care experience among patients experiencing homelessness, we administered a survey that included questions about overdose to veterans who have experienced homelessness. This study describes the prevalence of self-reported overdose involving drugs or alcohol in the preceding 3 years, the reported contribution of various drugs and alcohol to overdose, and the demographic, social, and health characteristics associated with overdose in this population.

Methods
This survey study provides analysis of responses to a national survey of veterans who have experienced homelessness who were receiving primary care at 26 VA medical centers across the United States. The primary aim of the overall study was to compare care experiences across primary care models among veterans who have experienced homelessness. This report represents a secondary aim of that study. This study was approved by the VA's Central Institutional Review Board.
Participants were informed that by answering the survey they were agreeing to participate in the research study, so signed informed consent was not obtained. months. Eligibility also required use of the VA's primary care services at 1 of 26 VA medical centers, including a single active panel assignment and 2 or more visits to a clinic with an administrative code indicating primary care at the same study site in the preceding 24 months. Included VA medical centers were those with the largest H-PACTs. We randomly selected a subsample of eligible veterans from each study site based on initial power calculations and expected survey response rate of 40%, with the goal of a final sample consisting of two-thirds from H-PACTs and one-third from nontailored mainstream primary care clinics. We later excluded participants if they had no contact information from VA or other records or had died prior to survey initiation. To compare survey respondents and nonrespondents, we retrieved demographic and clinical information from VA medical records in the 2 years preceding recruitment.

Survey Measures
In the absence of validated survey questions for experiencing overdose, we devised 2 survey items for this study based on similar questions published in the literature [11][12][13] and on consultation with national experts. Experiencing overdose was based on the question "In the last 3 years, have you had an overdose where you needed to go to the emergency room or get medical care right away?" with a follow-up question asking what substances were involved (eg, alcohol, heroin, fentanyl) (eFigure in the Supplement). Additionally, we devised a similarly worded question to assess whether participants had witnessed someone else overdose ("In the last 3 years, have you seen another person have an overdose where they needed to go to the emergency room or get medical care right away?") The substances involved with witnessed overdoses were not queried.
In addition to self-reported demographic characteristics, including age, race, ethnicity, and current housing status, other measures obtained via survey included self-report of drug or alcohol problems, based on the validated 2-Item Conjoint Scale, 14 which assesses having used alcohol or drugs "more than you meant to" and having "felt you wanted to or needed to cut down" in the past

Statistical Analysis
In the initial step of our analysis, we compared survey respondents and nonrespondents on sociodemographic and clinical variables. We used independent sample t tests to test for group differences on continuous variables and χ 2 tests for differences on categorical variables. Next, we constructed a variable for the propensity to respond using the administrative data that were available for each participant who was invited to participate, which included demographic variables (ie, age, sex, race, and marital status), medical conditions (ie, alcohol use disorder, drug use disorder, posttraumatic stress disorder, psychotic disorders, and number of Elixhauser comorbidities), health care utilization information (ie, number of primary care visits, mental health visits, emergency department visits, hospitalizations, and number of administrative service codes related to homelessness), and primary care type (ie, H-PACT vs mainstream). Responses were weighted by the inverse of response propensity (1 / propensity) so that respondents with lower overall propensity to respond were given greater weight, and vice versa. We then compared individuals who experienced any overdose with those who did not.
Finally, we used multivariable logistic regression models to explore which factors were associated with reporting any overdose, overdose involving drugs, and overdose involving alcohol.
The intent of these models was explanatory and illustrative rather than predictive. Covariates included sociodemographic and clinical variables shown to be associated with overdose in prior studies: age, race, psychological distress, current homelessness, and medical comorbidities. [22][23][24] We also included variables associated with homelessness and adverse health outcomes that we hypothesized would be associated with overdose in individuals experiencing homelessness: severe chronic pain 25,26 and social support. 27,28 The multivariable analyses excluded individuals with missing data on 1 or more study variables. These models controlled for the identity of the VA medical center through application of a random effects term. P values were 2-sided and considered statistically significant at less than .05. Analyses were conducted using SAS statistical software version 9.4 (SAS Institute). Preliminary analyses were conducted in October 2018, and final analyses were conducted in January 2020.

Unadjusted Associations With Experiencing Nonfatal Overdose
Among 5694 veterans included in the overdose analyses, the mean (SD) age was 56. 4

Substances Involved With Nonfatal Overdose
Alcohol was the most common substance involved in a recent overdose, reported by 192 veterans (3.7%) ( Table 2). Overdose involving any drug was reported by 228 veterans (4.6%). Specifically,

Adjusted Associations With Experiencing Any Overdose, Overdose Involving Alcohol, and Overdose Involving Drugs
Results of the multivariable models for experiencing any overdose, overdose involving alcohol, and overdose involving drugs are presented in Table 3  , and veterans who reported alcohol problems were less likely than veterans who did not report alcohol problems to report an overdose involving drugs (OR, 0.74 [95% CI, 0.59-0.91]).

Discussion
In this large survey study of veterans who have experienced homelessness, 7.4% reported an overdose requiring urgent medical attention in the past 3 years, and 16.2% had witnessed someone else overdose. Witnessing overdose was significantly associated with personal experience of overdose. Alcohol was the predominant substance reported in overdose, although 1.7% of veterans reported overdose involving opioids.
One potentially actionable finding from this study is that alcohol was the most common substance reported with overdose, nearly as common as all drugs combined and more than 2-fold as common as opioids. Previous studies have shown that alcohol, cannabis, and cocaine are the most widely used substances among individuals experiencing homelessness. 29,30 Alcohol overdose, or acute intoxication, is a frequent cause for emergency department visits. 31 Although alcohol rarely appears as the sole toxin in fatal overdoses, 32 heavy alcohol use has a substantial negative effect on long-term morbidity and mortality. 33 Alcohol use disorder is also treatable with approved medications and interventions, such as motivational interviewing and cognitive behavioral therapy. 34 Despite this, data show that most individuals with alcohol use disorder do not receive treatment. 35 The second most common overdose cause was opioids. Opioid overdose in this sample (3-year prevalence, 1.7%) was higher than for the general population 2 and comparable with an analysis of patients with Medicaid insurance who had received at least 1 opioid prescription, in which 0.5% of patients per year required medical services for overdose involving opioids. 36 A 2018 study using Medicaid data 37 found that individuals who survived nonfatal opioid overdose were at 24-fold higher risk of death compared with the general population, with drug overdose, respiratory illness, HIV, and suicide all contributing to mortality. For health systems that serve individuals experiencing homelessness, a patient's report of recent overdose signals elevated medical risk and may even merit standardized screening or offer of additional services.
Additionally, maximizing access to the opioid overdose reversal agent naloxone is advisable. 38 However, since pharmacies are required to dispense naloxone to a named individual rather than a facility, this can be difficult to accomplish for homeless shelters. Finally, opioid overdose could also be prevented through medication therapy for opioid use disorder, 39 yet focused efforts to assure delivery of treatment for opioid use disorder in this population are uncommon. 40 The findings of this study may suggest additional targets for clinical and policy intervention. The association of witnessing overdose with experiencing overdose suggests that efforts at risk mitigation should take into account patients' social networks. Although relevant to substance use, 41 few studies explore social networks among veterans who have experienced homelessness, a unique population whose relationships may be shaped by history of military service and unpredictability of housing. Peer-based programs have been applied widely in VA medical centers, 42 including with veterans who have experienced homelessness. 43 Peer-based strategies that focus on homelessness may be an opportunity for tailored assistance, potentially tapping into the experience and wisdom of veterans who have experienced homelessness in reaching their peers.
Given the prevalence of high emotional distress in individuals who have experienced overdose, enhanced mental health services could mitigate some risk for individuals residing on the streets (also known as sleeping rough), in homeless shelters, or newly in housing. This could be advanced through

Limitations
This study has some limitations. First, the cross-sectional nature of the survey prevents causal inference. Current homelessness, for example, was associated with overdose, but it is possible that having an overdose or serious addiction contributed to loss of residence. Second, the response rate was only 40.2%, which would be considered suboptimal under some circumstances. However, this response rate is comparable to that obtained in an evaluation of national VA primary care (47%), 51 and roughly 2-fold that obtained by VA's primary care survey of veterans who have experienced homelessness. 6 Furthermore, since we had access to rich data from VA records to compare respondents and nonrespondents, we were able to weight our analyses by the propensity to respond, which at least somewhat reduces the risk of nonresponse bias. It is also important to highlight that there have been few recent efforts to survey homeless-experienced populations on a national basis, 52,53 so this study likely represents the best available data to study these questions in this population. Third, generalizability may be limited because our sample consisted only of veterans who accessed VA services and who were reachable through mail or telephone. In general, the VA updates a mailing address with every visit, and a 2014 report 54 found that 89% of people experiencing homelessness who use VA services had mobile telephones. We speculate that veterans who have experienced homelessness and who are lacking both a mailing address and a telephone number could be at higher risk of overdose than individuals included in this study. Fourth, our study relies on self-reported survey data, and the validity of self-report of overdose is uncertain. However, many factors associated with overdose risk in this study are consonant with prior studies that used medical records, including substance use disorders, 22-24 race, 1,55 psychological disorders, 23,56 and having witnessed others overdose. 57 Fifth, the survey did not include an option for methamphetamine, which could have contributed to the other substance category, and merits specific queries in future research.

Conclusions
This study, the largest survey of veterans who have experienced homelessness to date, to our knowledge, found that recent nonfatal overdose was a relatively common issue among veterans who have experienced homelessness, occurring in 7.4% of survey respondents, with alcohol being involved more often than opioids. Improving access to mental health and addiction treatment for veterans who are experiencing homeless or who are recently housed and targeting health and social services for those at increased risk could enhance the safety of this population.