When Health Insurance Is Not a Factor: National Comparison of Homeless and Nonhomeless US Veterans Who Use Veterans Affairs Emergency Departments

Abstract

Objectives. We examined the proportion of homeless veterans among users of Veterans Affairs (VA) emergency departments (EDs) and compared sociodemographic and clinical characteristics of homeless and nonhomeless VA emergency department users nationally.

Methods. We used national VA administrative data from fiscal year 2010 for a cross-sectional study comparing homeless (n = 64 091) and nonhomeless (n = 866 621) ED users on sociodemographics, medical and psychiatric diagnoses, and other clinical characteristics.

Results. Homeless veterans had 4 times the odds of using EDs than nonhomeless veterans. Multivariate analyses found few differences between homeless and nonhomeless ED users on the medical conditions examined, but homeless ED users were more likely to have been diagnosed with a drug use disorder (odds ratio [OR] = 4.12; 95% confidence interval [CI] = 3.97, 4.27), alcohol use disorder (OR = 3.67; 95% CI = 3.55, 3.79), or schizophrenia (OR = 3.44; 95% CI = 3.25, 3.64) in the past year.

Conclusions. In a national integrated health care system with no specific requirements for health insurance, the major differences found between homeless and nonhomeless ED users were high rates of psychiatric and substance abuse diagnoses. EDs may be an important location for specialized homeless outreach (or “in” reach) services to address mental health and addictive disorders.

Use of emergency departments (EDs) is of national concern because high rates of ED use are thought by many to indicate poor access to regular health care providers and failure to address preventable illness and injury.¹ Approximately 5% of patients are responsible for a quarter of all ED visits,² and homeless adults are among the most frequent users of ED services.^3–7 Two studies conducted in San Francisco, California, found that 40% of homeless adults used EDs at least once in the past year, a rate 3 times the US norm,³ and that homeless people were nearly 4 times as likely as domiciled people to be frequent ED users, that is, to have visited an ED more than 4 times in the past year.⁷ One large national study of homeless adults found that 32% reported having an ED visit in the past year,⁴ which is much higher than the rate of 13% to 20% for US adults overall.^8,9 Several additional studies have found that homeless adults are disproportionately represented among the most frequent users of EDs.^7,10,11

In addition to using EDs more often than nonhomeless adults, homeless adults who use ED services have been reported to have higher rates of infectious diseases, substance use, and psychiatric illness.^6,10 They have also been reported to be more likely to be ED recidivists, that is, to return to the ED within a short period of time after a prior ED visit.^6,12 Among homeless adults, some identified predictors of ED use have been unstable housing, chronic medical illness, food insecurity, and victimization.^3,4,13,14

An additional factor thought to be associated with ED use is the lack of regular and accessible health care, health insurance, or both. Analyses of national ED utilization data have shown that homeless ED users are more likely to be uninsured than nonhomeless ED users.^5,6 Other studies have shown that homeless patients who use EDs often have no other source of health care,⁴ especially those who are chronically homeless.¹⁵ Thus, homeless patients may be more likely to use EDs because they do not have health insurance for ambulatory care. Moreover, even though more than half of the homeless population is enrolled in public financial assistance programs, these subsidies may not fully address the needs of those with substance use disorders or chronic medical conditions.¹⁰

However, another series of studies have shown that frequent ED users, who are especially likely to be homeless, are more likely to be insured and less likely to be uninsured^2,3,16,17 and, rather than being disengaged from care outside of the ED, frequent ED users have been found in some studies to use outpatient primary and specialty clinic systems quite heavily.^2,18–22 One explanation for these mixed findings may be that some studies are based on health care systems that offer extensive services for the uninsured and homeless, in which insurance may be a marker for higher levels of physical and mental disability rather than an indicator of access or lack of access to ambulatory care.³

The Veterans Affairs (VA) health care system offers a unique opportunity to examine correlates of ED use by homeless individuals. The VA is one of the largest integrated health care networks in the United States and offers veterans equal access to an array of health care services, in which insurance coverage is unrelated to access to services.^23,24 Little examination of homeless patients who use EDs within this type of health care system has occurred. Moreover, the health of homeless veterans is vital to the VA’s established goal of ending homelessness among veterans.²⁵ One study has shown that homeless veterans tend to be frequent users of VA emergency departments,¹¹ and another has shown that homeless veterans are more likely to revisit a VA emergency department within 30 days of a prior VA emergency department visit.¹² However, no studies have been published of how homeless veterans who use EDs differ on sociodemographic and clinical characteristics from other ED users.

We examined the rate of homelessness among veterans who used EDs at VA facilities in fiscal year 2010 and compared homeless ED users with nonhomeless ED users on sociodemographic characteristics, medical and psychiatric diagnoses, use of psychotropic medications, and other clinical characteristics. The results may be informative in determining factors, independent of health insurance status, related to homelessness among ED users. Given the frequent use of EDs by homeless people, targets for further intervention and specialized attention may need to be identified.

METHODS

National VA administrative data from fiscal year 2010 identified 5 531 379 veterans who used Veterans Health Administration services, of whom 930 712 (16.83%) used VA emergency departments. Information on the use of any VA emergency department services and the mean number of VA emergency department visits was available from the administrative data. We used a cross-sectional study design to compare, among ED users, veterans identified as homeless (n = 64 091) with those who were not homeless (n = 866 621) on sociodemographics, medical and psychiatric diagnoses, prescriptions for psychotropic medication, mortality risk indicators, and other clinical characteristics.

Measures

Homeless veterans were defined as veterans who had used specialized VA homeless program services, received a V60.0 International Classification of Diseases, Ninth Revision (ICD-9)²⁶ diagnostic code (indicating lack of housing), or both during the fiscal year.

We based medical and psychiatric diagnoses on ICD-9 diagnostic codes entered by VA clinicians into the medical record. These codes were obtained for any VA visit in fiscal year 2010, as were codes specifically linked to any VA emergency department visit during fiscal year 2010. We obtained types of psychotropic medications from VA administrative pharmacy data.

We summarized comorbidity and severity of medical conditions with the Charlson index.²⁷ The Charlson index is a widely used measure that predicts the 10-year mortality for patients who may have a range of comorbid conditions. Each condition is assigned a score of 1, 2, 3, or 6 depending on the risk of dying associated with the condition, and scores are summed for a total score.

Data Analysis

First, we calculated the percentage of VA emergency department users among homeless VA service users and among nonhomeless VA service users. Using these numbers, we calculated the odds ratio (OR) for ED use among homeless and nonhomeless VA service users. In this analysis, the analytic denominator was all VA emergency department users.

Then we conducted 2 sets of analyses to examine correlates of homeless status among VA emergency department users (so the denominator was no longer all VA service users). In the first set of analyses, we examined homeless status and associations among sociodemographic characteristics, medical diagnoses, psychiatric diagnoses, and psychotropic medication prescriptions of ED users during any VA visit during the study year. In the second set of analyses, we examined the independent relationship of sociodemographics and medical and psychiatric diagnoses specifically associated with any VA emergency department visit during the study year, again with homelessness as the dependent variable. Whereas the first set of analyses addressed the association between general characteristics and homelessness among ED users, the second set of analyses addressed the relationship of specific diagnoses coded during ED visits to homelessness.

For each set of analyses, we made bivariate comparisons between homeless ED and nonhomeless ED users by calculating effect size differences using Cohen’s d for continuous variables and ORs for categorical variables. Given the large sample size, effect size is a more accurate measure of difference than statistical significance.

Then we conducted forward stepwise logistic regressions (variables retained met the criterion of P < .05) with all significant variables identified in bivariate comparisons that had at least a small to moderate effect size (d > 0.30; OR < 0.66; or OR > 1.50). Logistic regression analyses did not include the Charlson index score or psychotropic medications to avoid redundancy with medical and psychiatric diagnoses entered into the model.

RESULTS

Among the 930 712 veterans who used EDs at VAs nationally in fiscal year 2010, 64 091 (6.89%) were homeless and 866 621 (93.11%) were not homeless. By comparison, of all veterans who used any VA services in fiscal year 2010, only 142 695 (2.58%) were homeless, reflecting an OR of 4.25 for ED service use among homeless veterans compared with domiciled veterans. In addition, homeless ED users had a mean of 3.38 (SD = 4.01) ED visits versus a mean of 2.07 (SD = 1.09) for nonhomeless ED users (Cohen’s d = 0.41). Among all VA service users, homeless users had a mean of 1.52 (SD = 3.17) ED visits and nonhomeless users had a mean of 0.33 (SD = 1.10) ED visits.

Bivariate analyses found several medium and large effect size differences between homeless ED users and nonhomeless ED users on sociodemographic characteristics, medical status, psychiatric status, and psychotropic medications during any VA visit in the past year (Table 1). Homeless ED users were younger, had lower incomes, and were more likely to have liver disease and HIV/AIDS but were less likely to have dementia. Notably, homeless ED users were far more likely to have a broad range of psychiatric disorders than nonhomeless ED users, with the strongest effects observed for dual diagnoses and substance use disorders. As a result, homeless ED users were also much more likely to have been prescribed antidepressants, antipsychotics, and lithium.

TABLE 1—

Characteristics of Homeless and NonHomeless ED Users: Veterans Affairs Administrative Data, United States, Fiscal Year 2010

Characteristics	Homeless ED Users (n = 64 091), No. (%) or Mean ±SD	Other ED Users (n = 866 621), No. (%) or Mean ±SD	Effect Size Difference Between Homeless and Other ED Users, d or OR
Sociodemographics
Age, y	51.84 ±10.41	58.90 ±15.73	0.53ab
Male	59 536 (92.89)	799 878 (92.30)	1.09
White	11 482 (17.92)	228 091 (26.32)	0.61
OEF–OIF–OND status	3334 (5.20)	63 633 (7.34)	0.69
Service connection
None	46 353 (72.32)	505 210 (58.30)	1.87
< 50%	7600 (11.86)	205 103 (23.67)	0.85
≥ 50%	10 138 (15.81)	156 308 (18.03)	0.43
Income, $	10 403.03 ±15 997.40	24 531.91 ±43 322.41	0.43a
Medical status
Medical visitsc	13.29 ±13.05	13.00 ±13.49	0.02a
Myocardial infarction	973 (1.52)	23 404 (2.70)	0.56
Congestive heart failure	42 527 (66.35)	564 284 (65.11)	1.06
Peripheral vascular disease	2712 (4.23)	70 238 (8.10)	0.50
Cerebral vascular accident	3191 (4.98)	76 138 (8.79)	0.54
Dementia	306 (0.48)	13 213 (1.52)	0.31
Chronic pulmonary disease	12 954 (20.21)	185 882 (21.45)	0.93
Connective tissue/rheumatic disease	485 (0.76)	12 704 (1.47)	0.51
Peptic ulcer disease	967 (1.51)	12 659 (1.46)	1.03
Liver disease	6283 (9.80)	36 797 (4.25)	2.45
Diabetes without complications	12 027 (18.77)	238 099 (27.47)	0.61
Diabetes with complications	2892 (4.51)	67 727 (7.82)	0.56
Paraplegia and hemiplegia	513 (0.80)	9909 (1.14)	0.70
Renal disease	2581 (4.03)	70 987 (8.19)	0.47
Cancer	3933 (6.14)	106 099 (12.24)	0.47
Moderate/severe liver disease	654 (1.02)	5821 (0.67)	1.52
Metastatic cancer	469 (0.73)	12 896 (1.49)	0.49
HIV/AIDS	1486 (2.32)	6577 (0.76)	3.10
Charlson indexd	2.71 ±2.26	3.80 ±2.89	0.42a
Psychiatric status
Mental health visitse	38.69 ±56.72	4.29 ±17.39	0.82a
Alcohol use disorder	31 800 (49.62)	87 156 (10.06)	8.81
Drug use disorder	34 233 (53.41)	64 681 (7.46)	14.22
Schizophrenia	6753 (10.54)	25 577 (2.95)	3.87
Other psychosis	4984 (7.64)	20 442 (2.35)	3.42
Bipolar disorder	9088 (14.18)	31 990 (3.69)	4.31
Major depressive disorder	14 252 (22.24)	80 826 (9.33)	2.78
Dysthymia	31 680 (49.43)	196 093 (22.63)	3.34
Posttraumatic stress disorder	16 581 (25.87)	128 072 (14.78)	2.01
Other anxiety disorder	15 935 (24.86)	108 010 (12.46)	2.32
Adjustment disorder	11 657 (18.19)	40 714 (4.70)	4.51
Personality disorder	6704 (10.46)	16 119 (1.86)	6.16
Any psychiatric disorder	57 153 (89.17)	438 253 (50.57)	8.05
Dual diagnosis	40 784 (63.63)	102 224 (11.80)	13.08
Psychotropic medications
Psychotropic prescriptions	23.23 ±47.42	9.77 ±29.89	0.34a
Any antidepressants	39 956 (62.34)	336 993 (38.89)	2.60
Any antipsychotics	39 956 (62.34)	336 993 (38.89)	3.62
Any anxiolytics	21 040 (32.83)	103 187 (11.91)	1.50
Any stimulants	20 302 (31.68)	204 316 (23.58)	1.23
Any mood stabilizers	648 (1.01)	7119 (0.82)	1.88
Any lithium	18 929 (29.53)	157 753 (18.20)	3.40
Any opiates	2015 (3.14)	8191 (0.95)	1.12

Note. d = Cohen d; ED = emergency department; OEF = Operation Enduring Freedom; OIF = Operation Iraqi Freedom; OND = Operation New Dawn; OR = odds ratio.

^aThe continuous variables are Cohen d values.

^bThe conventional interpretation of Cohen d is 0.2, 0.5, and 0.8, respectively, for a small, medium, and large effect size.

^cMedical visits include outpatient medical and surgical visits.

^dA widely used measure that predicts the 10-year mortality for patients who may have a range of comorbid conditions. Each condition is assigned a score of 1, 2, 3, or 6 depending on the risk of dying associated with the condition, and scores are summed for a total score.

^eMental health visits include psychiatric and substance abuse visits.

Bivariate comparisons of homeless and nonhomeless ED users by diagnoses given specifically during an ED visit in the past year showed similar differences (Table 2). Homeless ED users were much more likely to have ED diagnoses of mental health problems, especially substance use disorders. We found few substantial differences in the ED diagnoses of those medical conditions that we examined, except that homeless ED users were, again, more likely to be diagnosed with liver disease or HIV/AIDS but less likely to be diagnosed with dementia.

TABLE 2—

Homeless Versus Nonhomeless ED Patients With Selected Diagnoses Coded at Any ED Visit: Veterans Affairs Administrative Data, United States, Fiscal Year 2010

Diagnosis at any ED visit	Homeless ED Users, No. (%)	Other ED Users, No. (%)	Effect Size Difference Between Homeless and Other ED Users, OR
Alcohol disorder	8683 (13.55)	17 198 (1.98)	7.74
Drug disorder	7638 (11.92)	10 824 (1.25)	10.70
Schizophrenia	2314 (3.61)	6321 (0.73)	5.10
Other psychosis	1346 (2.10)	6195 (0.71)	2.98
Bipolar disorder	2070 (3.23)	5121 (0.59)	5.61
Major depressive disorder	1467 (2.29)	4348 (0.50)	4.65
Dysthymia	7639 (11.92)	20 557 (2.37)	5.57
Posttraumatic stress disorder	2111 (3.29)	9451 (1.09)	3.09
Other anxiety disorder	4575 (7.14)	18 661 (2.15)	3.49
Adjustment disorder	675 (1.05)	2062 (0.24)	4.46
Personality disorder	671 (1.05)	1204 (0.14)	7.60
Any psychiatric disorder	24 173 (37.72)	93 825 (10.83)	4.99
Any seizure	104 (0.16)	835 (0.10)	1.69
Any insomnia	415 (0.65)	3114 (0.36)	1.81
Any headache	2656 (4.14)	23 841 (2.75)	1.53
Any pain diagnosisa	8227 (12.84)	79 138 (9.13)	1.47
Myocardial infarction	193 (0.30)	5330 (0.62)	0.49
Congestive heart failure	16 858 (26.30)	24 8113 (28.63)	0.89
Peripheral vascular disease	329 (0.51)	6348 (0.73)	0.70
Cerebral vascular accident	662 (1.03)	14 313 (1.65)	0.62
Dementia	35 (0.05)	1750 (0.20)	0.27
Chronic pulmonary disease	4803 (7.49)	64 456 (7.44)	1.01
Connective tissue/rheumatic disease	77 (0.12)	1619 (0.19)	0.64
Peptic ulcer disease	124 (0.19)	1118 (0.13)	1.50
Liver disease	872 (1.36)	5665 (0.65)	2.10
Diabetes without complications	3246 (5.06)	42 893 (4.95)	1.02
Diabetes with complications	300 (0.46)	4347 (0.50)	0.93
Paraplegia and hemiplegia	83 (0.13)	1462 (0.17)	0.77
Renal disease	709 (1.11)	16 680 (1.92)	0.57
Cancer	662 (1.03)	17 195 (1.98)	0.52
Moderate/severe liver disease	153 (0.24)	1341 (0.15)	1.54
Metastatic cancer	56 (0.09)	1791 (0.21)	0.42
HIV/AIDS	382 (0.60)	1529 (0.18)	3.39

Note. ED = emergency department; OR = odds ratio.

^aPain diagnoses included musculoskeletal pain, diabetic or other neuropathic pain, other chronic or acute pain, and generalized pain.

Among all ED users, multivariate analyses showed that controlling for other factors, homeless ED users had more than twice the odds of having a diagnosis of drug use disorder, schizophrenia, or adjustment disorder in the past year than nonhomeless ED users, with the largest effect observed for drug use disorder (Table 3).

TABLE 3—

Forward Stepwise Logistic Regression of the Association of Sociodemographics and Clinical Characteristics During Fiscal Year 2010 With Homeless Status Among ED Users: Veterans Affairs Administrative Data, United States, Fiscal Year 2010

Independent Variable	Homeless vs Nonhomeless Users, OR (95% CI)
Age	0.95 (0.94, 0.95)
Service connection
None (Ref)	1.00
≥ 50%	0.41 (0.37, 0.47)
< 50%	0.67 (0.59, 0.75)
Income	1.00 (1.00, 1.00)
Dementia	1.26 (1.05, 1.51)
Cancer	0.83 (0.71, 0.97)
No. mental health visitsa	1.01 (1.01, 1.01)
Alcohol use disorder	1.80 (1.55, 2.08)
Drug use disorder	2.81 (2.39, 3.30)
Schizophrenia	2.40 (2.04, 2.83)
Major depression	1.13 (1.01, 1.27)
Dysthymia	1.30 (1.19, 1.43)
Posttraumatic stress disorder	1.26 (1.12, 1.42)
Adjustment disorder	2.76 (2.45, 3.12)
Personality disorder	1.63 (1.36, 1.95)
Dual diagnosis	1.62 (1.33, 1.97)

Note. CI = confidence interval; ED = emergency department; OR = odds ratio.

^aMental health visits include psychiatric and substance abuse visits.

Multivariate analyses further showed that among ED users, the largest effects indicated that homeless ED users were more likely to have an ED diagnosis of alcohol use disorder, drug use disorder, schizophrenia, bipolar disorder, dysthymia, or HIV/AIDS than nonhomeless ED users (Table 4). Notably, homeless ED users had more than 4 times the odds of being diagnosed with a drug use disorder at an ED visit than did nonhomeless ED users and 3 times the odds of being diagnosed with an alcohol use disorder or schizophrenia at an ED visit.

TABLE 4—

Forward Stepwise Logistic Regression of the Association of Sociodemographics and Diagnoses Made During ED Visits With Homeless Status Among ED Users: Veterans Affairs Administrative Data, United States, Fiscal Year 2010

Independent Variables	Homeless vs. Nonhomeless Users, OR (95% CI)
Age	0.98 (0.98, 0.98)
Service connection
None (Ref)	1.00
≥ 50%	0.48 (0.47, 0.50)
< 50%	0.61 (0.60, 0.63)
Income	1.00 (1.00, 1.00)
Diagnosis at ED visit
Alcohol use disorder	3.67 (3.55, 3.79)
Drug use disorder	4.12 (3.97, 4.27)
Schizophrenia	3.44 (3.25, 3.64)
Other psychosis	1.61 (1.50, 1.73)
Bipolar disorder	2.33 (2.19, 2.48)
Major depressive disorder	1.98 (1.84, 2.12)
Dysthymia	2.32 (2.24, 2.40)
Posttraumatic stress disorder	1.21 (1.14, 1.28)
Other anxiety disorder	1.43 (1.37, 1.49)
Adjustment disorder	1.99 (1.80, 2.20)
Personality disorder	1.70 (1.51, 1.91)
Any headache	1.26 (1.20, 1.32)
Myocardial infarction	0.64 (0.55, 0.74)
Cerebrovascular accident	0.80 (0.74, 0.87)
Dementia	0.37 (0.26, 0.53)
Liver disease	1.25 (1.15, 1.35)
Renal disease	0.81 (0.75, 0.88)
Cancer	0.71 (0.66, 0.77)
Metastatic cancer	0.68 (0.51, 0.89)
HIV/AIDS	2.07 (1.83, 2.35)

Note. CI = confidence interval; ED = emergency department; OR = odds ratio.

Given the significantly higher rate of psychiatric diagnoses among homeless ED users, we conducted supplementary analyses only on VA service users with psychiatric diagnoses (including substance use disorders) to determine whether homeless veterans with psychiatric diagnoses were also more likely to use ED services than nonhomeless veterans with psychiatric diagnoses. Of the 114 422 homeless VA service users with psychiatric diagnoses, 49.95% were ED users compared with 24.73% of the 1 771 964 nonhomeless VA service users with psychiatric diagnoses who were ED users, resulting in an OR of 3.04.

DISCUSSION

Homeless veterans had 4 times the odds of using VA emergency departments than nonhomeless veterans, which is comparable to or higher than the ORs calculated from several non-VA studies.^3,4,7 Thus, despite being part of a system that allows ready access to outpatient care without any insurance barrier, homeless patients in the VA system continue to use the ED at high rates. This finding suggests that factors other than lack of health insurance are driving ED use among homeless veterans in contrast with previous studies of the general homeless population that have highlighted the lack of health insurance as a reason for frequent ED use.^4–6 This finding may also have broader implications for the general homeless adult population (nonveterans), suggesting that efforts to provide the homeless with health insurance coverage (e.g., expansion of Medicaid) may increase their access to health services but may not substantially reduce their ED use. Homeless ED users in the VA system used more ED services, on average, than nonhomeless ED users, consistent with research conducted in other settings^2–6; it may also be worth noting that most VA service users who used EDs used them more than once a year.

The most striking observation in this study was that homeless VA emergency department users were substantially more likely than nonhomeless ED users to have a variety of mental health diagnoses, especially severe mental illness, dual diagnosis, and substance use disorders. Among all VA emergency department users, after adjusting for relevant sociodemographics, clinical characteristics, and other psychiatric diagnoses, homeless ED users had twice the odds of being diagnosed with a drug use disorder or schizophrenia than other ED users at any VA visit in the past year and 3 times the odds of being diagnosed with a substance use disorder or schizophrenia specifically at an ED visit. Thus, not surprisingly, homeless ED users also had dramatically more mental health visits than nonhomeless ED users and were more likely to be prescribed a variety of psychotropic medications. Although homeless ED users were less likely to be VA service connected than other ED users, homeless veterans are a priority group within the VA, and so homeless veterans likely receive increased attention and access to specialized VA services compared with other VA service users, regardless of their service connection. So even though homeless veterans are a national priority and many are connected to outpatient mental health care, they are still frequent users of ED services, suggesting the need for more comprehensive outpatient mental health services tailored to this population.

The differences observed between homeless and nonhomeless ED users may be similar to those found between homeless VA users and nonhomeless VA users overall²⁸ and confirms previous studies that have found high rates of substance use and psychiatric illness in homeless ED populations^6,10 as well as epidemiological studies of the general population that have shown higher risk for past homelessness among people with psychiatric and substance use disorders.^29,30 The VA has been called a “national safety net” because VA service users are generally more disabled and sicker than other health care service users.³¹ Thus, our findings are worth underscoring because we did not simply compare homeless veterans with nonhomeless veterans; rather, we compared homeless veterans with a fairly disabled, sick population of treatment-seeking nonhomeless veterans and yet the differences in psychiatric and substance abuse problems remained. Moreover, not only did we examine these diagnostic codes during any VA visit, but we examined codes made specifically during VA emergency department visits that showed similar patterns.

These findings are important because non-VA emergency departments have nationally recognized that mental health–related ED visits constitute a relatively small but increasing component of care.^32,33 Additionally, when we examined rates of ED use among VA service users with only psychiatric diagnoses (including substance use disorders), we found that homeless VA service users still had 3 times the odds of using ED services than nonhomeless VA service users, supporting and enhancing our findings of frequent ED use among homeless veterans.

It is interesting to note that 1 randomized controlled trial found that compassionate care, or simply having volunteers talk with and listen to homeless patients during their ED visits, can reduce the frequency of their use of these services.³⁴ More attention to the general social isolation of homeless patients as well as to their mental health problems more specifically may be needed. Additionally, advocates have suggested that more comprehensive ED discharge planning to facilitate access of homeless patients to supported housing and other psychiatric rehabilitation programs⁶ is needed, and several studies have shown that supportive housing and care coordination can reduce future ED use.^35–37

Homeless ED users were younger and had slightly greater indication of mortality risk than nonhomeless ED users, as indicated by their age-adjusted Charlson index scores. We found few differences, however, in the specific medical conditions examined, except that homeless ED users were more likely to have liver disease and HIV/AIDS than nonhomeless ED users. In the VA’s equal-access health care system,^23,24 homeless and nonhomeless veterans may use EDs for the same medical reasons because health insurance is less of a barrier to medical care and ED users may be regular users of primary care and other medical services.^2,18–22 The higher rates of liver disease and HIV/AIDS found among homeless ED users compared with nonhomeless ED users is consistent with the high rates of these conditions found in homeless populations.^38–42 Liver disease and HCV among homeless individuals has been found to be related to both injection and noninjection drug use, to chronic alcohol use, and to being male.⁴¹ HIV/AIDS has also been found to be related to injection and noninjection drug use as well as to being non-White, having male homosexual contact, and being sexually victimized.^38,39 These findings suggest that substance abuse among homeless individuals may be related to their increased risk of these medical conditions and that these conditions deserve particular attention among homeless populations, including homeless veterans.

Limitations

Several study limitations require comment. We defined homelessness as “ever homeless” during the 1-year period of the study and whether diagnoses were documented before or after the homelessness occurred could not be determined. VA administrative data are only as accurate as clinical records allow, and diligence of clinician documentation may affect our results. Many medical conditions, including some common ED diagnoses (e.g., injuries) were not available in our data set, and complete data on race/ethnicity were not available. Some patients received more than 1 diagnosis during an ED visit, and which diagnosis was the primary presenting complaint for the visit is unknown. The high rate of psychiatric illness found among homeless ED users may reflect other aspects of social disconnection that could be addressed by social services rather than by medical care,⁴³ but such data are not captured in health system records. Given the large sample size and high statistical power to detect differences, we focused on effect sizes in our analyses. The research design was cross-sectional and based on medical and psychiatric diagnoses over a 1-year period, so causality and directionality of associations found are more suggestive than conclusive. Strengths of the study include a large national sample, assessment of mortality risk and specific medical conditions, and regression models that focused on the independent effect of relevant variables.

Conclusions

In the VA health care system, one of the largest integrated health care systems in the United States and one in which its users have access to a range of health care services with no specific requirements for health insurance, the greatest differences between homeless ED users and nonhomeless ED users were in the prevalence of psychiatric and substance abuse problems. Thus, EDs may be an important location for specialized homeless outreach or “in” reach services, in which homeless ED users can be referred to comprehensive social services and to mental health and substance abuse assistance. Given concern about the use of EDs among homeless populations, greater attention to providing specialized services for homeless ED users, including social services and psychiatric rehabilitation, may reduce their use of EDs.