AIDS and Behavior

, Volume 11, Issue 6, pp 854–863

Housing Status and Associated Differences in HIV Risk Behaviors Among Young Injection Drug Users (IDUs)

Authors

    • Center for Urban Epidemiologic StudiesNew York Academy of Medicine
  • Mary H. Latka
    • Center for Urban Epidemiologic StudiesNew York Academy of Medicine
    • Aurum Institute for Health Research
  • Hanne Thiede
    • HIV/AIDS EpidemiologyPublic Health-Seattle & King County
  • Elizabeth T. Golub
    • Department of EpidemiologyJohns Hopkins Bloomberg School of Public Health
  • Larry Ouellet
    • Division of Epidemiology and Biostatistics, School of Public HealthUniversity of Illinois, Chicago
  • Sharon M. Hudson
    • Health Research Association
  • Farzana Kapadia
    • Center for Urban Epidemiologic StudiesNew York Academy of Medicine
    • Guttmacher Institute
  • Richard S. Garfein
    • Division of International Health and Cross Cultural MedicineUniversity of California San Diego School of Medicine
Original Paper

DOI: 10.1007/s10461-007-9248-1

Cite this article as:
Coady, M.H., Latka, M.H., Thiede, H. et al. AIDS Behav (2007) 11: 854. doi:10.1007/s10461-007-9248-1

Abstract

Using cross-sectional analysis we examined residential status and associated differences in HIV risk behaviors among 3266 young IDUs enrolled in an HIV prevention trial. A three-level outcome (homeless (37%), equivocally housed (17%), housed (46%)) was defined based on responses to two questions assessing subjective and objective criteria for homelessness: “equivocally housed” participants were discordant on these measures. In multivariate analysis, antecedents of homelessness were having lived in an out-of-home placement, been thrown out of the home or in juvenile detention, and experienced childhood abuse; while correlates included receiving income from other and illegal sources, drinking alcohol or using methamphetamine at least daily, using shooting galleries, backloading, and sex work. A subset of these variables was associated with being equivocally housed. HIV risk varies by housing status, with homeless IDUs at highest risk. Programs for IDUs should utilize a more specific definition of residential status to target IDUs needing intervention.

Keywords

Injection drug userHIVHomelessHousing statusRisk behavior

Introduction

Homelessness is a significant problem among injection drug users (IDUs), and homelessness has been found to be associated with human immunodeficiency virus (HIV) (Culhane et al., 2001; Song et al., 2000) and its related risk behaviors (Andia et al., 2001; Lee et al., 2000; Metraux et al., 2004). However, standard definitions of homelessness are lacking and variations in the definition used can impact our interpretation of the data and the way services are provided.

Both subjective (thinking of oneself as homeless) (Song et al., 2000) and objective (sleeping outside or in a shelter) (Andia et al., 2001; Metraux et al., 2004; Reyes et al., 2005) definitions of homelessness are used in the literature. Rarely are they used simultaneously, but when they are, responses can be discordant (Link et al., 1994, 1995; Smereck and Hockman, 1998). One explanation for this is that there are many people who, while not living on the street or in shelters, obtain temporary housing with friends or relatives until being told to leave (Lifson and Halcon, 2001). One’s perception of feeling homeless may be indicative of being in transition towards being on-the-street homeless. It may also indicate circumstances where there is a greater likelihood of engaging in HIV-related risk behaviors (e.g., injecting outside without access to clean water or syringes). Conversely, someone may have re-located from a dangerous street to a safer squat and may not consider him or herself homeless anymore. In both examples, the material conditions of risk may not be substantially different.

Among IDUs, HIV prevalence estimates differ by housing status and tend to increase as housing becomes more unstable (Reyes et al., 2005; Robertson et al., 2004; Smereck and Hockman, 1998; Zolopa et al., 1994). One study reported an HIV prevalence of 28% among on-the-street homeless IDUs, 14% among transitionally housed IDUs (living with friends or family but considering themselves homeless) and 11% among housed IDUs (Reyes et al., 2005). Unstable housing is also associated with increased risk of HIV infection due to the behaviors that take place in these environments (Aidala et al., 2005; Corneil et al., 2006). It is likely that drug users living on the street or in unstable housing have limited resources for maintaining safer injection-related behaviors and may find it necessary to engage in survival activities that are fraught with HIV risk (Andia et al., 2001; Bourgois et al., 1997). Risk behaviors that are associated with unstable housing status include sharing injection paraphernalia (Reyes et al., 2005; Robertson et al., 2004) shooting gallery use (Celentano et al., 2001; Deren et al., 2003), engaging in paid sex (Andia et al., 2001), and inconsistent condom use with casual sex partners (Sears et al., 2001).

Early lifetime instabilities, such as abuse and neglect, are also associated with future poor housing status and substance use (Nyamathi et al., 1999; Zlotnick et al., 2004). One study of methadone treatment drop-outs found that those who reported having multiple sex partners were more likely to have experienced childhood physical and sexual abuse, and those who were neglected in childhood were more likely to be HIV positive (Kang et al., 2002). Among HIV positive men, a history of childhood sexual abuse predicted subsequent early initiation of injection drug use (Holmes, 1997).

In this analysis, we evaluated whether two definitions of self-reported homelessness were associated with varying levels of HIV risk behavior among a large sample of young IDUs from five cities in the United States. One question assessed subjective, or self-perceived homelessness (thinking of oneself as homeless), while the other assessed a more objective definition (sleeping outside or in a shelter for more than a week in the past month). We hypothesized that HIV risk behaviors would be higher among the equivocally housed and higher still among those who unequivocally reported themselves as homeless. Evaluating the risk behaviors associated with these two definitions of housing status will increase our understanding of the housing situations of young IDUs in order to develop and provide more effective interventions and services.

Methods

The data for this analysis derive from the baseline data collected as part of the Collaborative Injection Drug Users Study III/Drug Users Intervention Trial (CIDUS III/DU-IT), a randomized, controlled behavioral intervention trial with longitudinal follow-up addressing the primary prevention needs of HIV- and HCV-seronegative injection drug users. The purpose of the study was to evaluate the efficacy of a small-group six-session behavioral intervention to reduce the unsafe injection and sexual behaviors to limit the acquisition of HIV and HCV among enrolled IDUs. The study methods have been described elsewhere (Garfein et al., 2005), but briefly, between May 2002 and January 2004, IDUs in Baltimore, Chicago, Los Angeles, New York City and Seattle were recruited through street outreach, flyers and participant referrals to take part in baseline screening for trial enrollment.

The purpose of the baseline component was to assess participants for eligibility for participating in the intervention trial. At the baseline visit, participants completed a behavioral assessment using audio computer assisted self interview (A-CASI) technology. The baseline interview assessed sociodemographic characteristics, drug use behaviors, and sexual practices. Questions referred to behaviors in the last three or six months, or lifetime. After interview participants received pre-test counseling for HIV and HCV antibody testing from trained counselors, blood was drawn and tested. Subsequently, HIV and HCV-seronegative participants were invited to take part in the trial. The study was approved by the institutional review boards of the Centers for Disease Control and Prevention and the individual study sites.

Participants

Participants were eligible to take part in the baseline assessment visit if they: (1) were between the ages of 15 and 30 years; (2) injected drugs in the six months prior to screening; (3) spoke English and (4) agreed to undergo a self-administered assessment of their risk behaviors and have their blood drawn and tested for HIV and HCV. Additionally participants had to agree to provide contact information and have no plans to move from the study area in the year following the baseline visit. The sample used in this analysis included all baseline participants, regardless of their serostatus.

Measures

For this analysis we created a three-level dependent variable (homeless, equivocally housed, housed). This residential status variable was based on a subjective question (“In the past six months, have you thought of yourself as homeless?”) and an objective question (“In the past six months, have you slept in a car, abandoned building, public park, shelter, squatting place, or other non-dwelling for more than seven nights in a row?”) asked during the baseline assessment visit. The subjective question was intended to capture one’s perception of feeling homeless. The objective question was intended to capture material conditions that might impact an individual’s ability to practice safer sexual and injection behaviors, both of which in turn could impact HIV and HCV infection risk. For the residential status variable, participants who responded affirmatively to both questions were coded as “homeless” and those with negative responses to both questions were coded as “housed”. Those with discordant response were collapsed into one group and coded as “equivocally housed”.

We examined independent variables in the following domains for their associations with housing status: socio-demographic characteristics, HIV and HCV antibody status, early lifetime instabilities, and current HIV-related risk behaviors, including substance use and sexual behaviors. Socio-demographic variables included age, gender, race/ethnicity, education, source of income and the number of years injecting. HIV antibody detection (standard ELISA screening and western blot confirmation), and HCV antibody detection (EIA 2.0 or 3.0, Ortho Diagnostic Systems, Inc., Raritan, NJ) were performed using blood collected at the baseline visit. Participant samples that were reactive on an initial HCV EIA were retested in duplicate and if either or both of the repeat samples were reactive, the participant was considered positive for HCV.

Early lifetime instabilities included a history of out-of-home placements, being thrown out of the home before age 18, being put in juvenile detention, and experiencing childhood physical or sexual abuse. These events were selected because homeless persons have high rates of family discord and abuse, which may lead to a subsequently troubled life (Tyler and Cauce, 2002). As 2% of the sample was between the ages of 15 and 17, it was possible that instabilities could overlap in time with residential status. However, we assessed differences in results when this age group was both included in and excluded from the analysis. Finding no differences, the 15–17 year olds were retained. For these early lifetime variables alone we were able to establish temporality and determine whether earlier life circumstances predicted current housing status.

In this cross-sectional analysis, current HIV-related risk behaviors were those that occurred in the three months prior to interview. Substance use was defined either as at least daily use or less than daily use, and included the use of alcohol, non-injection crack and speedball (injecting heroin and cocaine together). Crack and cocaine use are often associated with elevated HIV risk (Bogart et al., 2005; McCoy et al., 2004; Somlai et al., 2003), hence we examined two ways in which crack/cocaine could be administered. Other substance use variables included: methamphetamine use (non-injection and injection alone, and injection with heroin, ≥ daily or less), frequency of injection (≥ daily or less), use of shooting galleries (settings where others are injecting and where resources for injection are provided: yes or no), receptive syringe sharing, paraphernalia sharing (sharing of cookers, cotton filters or water) and backloading, defined in this paper as using a shared syringe to measure and divide drugs. Sexual behaviors were assessed using the following measures: sex work (defined as trading sex for money or drugs; yes or no), number of sex partners (none, 1–2, ≥3) and condom use (never, inconsistent, consistent) during vaginal and/or anal sex with steady or casual heterosexual and same-sex male partners. We stratified by relationship status of the partner, since condom use often varies by this dimension (Seage, et al., 2002).

Data analyses

Housing status was associated with city, which could lead to potential confounding of associations between housing status and other variables of interest. Therefore, associations were adjusted for city using generalized Mantel-Haenszel tests. Since the dependent variable contains three levels, multinomial logistic regression analysis was performed to identify factors independently associated with housing status. Independent variables found to be significant in bivariate analysis (P < 0.10) were considered candidates in the multivariate model. Since the assumption of proportionality of odds across the response categories was not met, the multinomial logit model for unordered categories was used. This approach was used to test our hypothesis that involvement in HIV risk behaviors would increase as housing status become more unstable. Significant variables were retained in the final multivariate analysis. Fit for the model was judged by the likelihood ratio test.

Results

Between May 2002 and January 2004, 3285 participants completed the baseline assessment, of which 3266 had complete data on the housing questions and were included in this analysis. In the previous 6 months, 50% of participants felt homeless (subjective measure of homelessness), while 41% reported sleeping in a car, abandoned building, public park, shelter, squatting place, or other non-dwelling for more than one week (objective measure) (data not shown). Participants with affirmative responses to both questions assessing the subjective and objective measures of homelessness were coded as “homeless” and those with negative responses were coded as “housed”. Those with discordant responses were coded as either “subjective homeless only” (14%) or “objective homeless only” (3%), indicating an affirmative response to one question only. The two discordant groups tended to exhibit intermediate and similar characteristics and thus were combined as “equivocally housed” for the final analysis. The three-level residential status variable was created based on the results from bivariate associations between the four categories of housing status and independent variables, adjusting for site (Fig. 1). One risk factor from each category of risk (sociodemographics; early lifetime instabilities; substance use; injection drug use; sexual behaviors) is shown in Fig. 1 to be illustrative, not comprehensive.
https://static-content.springer.com/image/art%3A10.1007%2Fs10461-007-9248-1/MediaObjects/10461_2007_9248_Fig1_HTML.gif
Fig. 1

Bivariate associations between sociodemographic and HIV risk behavior variables and residential status among IDUs aged 15–30 in 5 U.S. cities (n = 3266)

From the subjective and objective measures, more than one third (37%) of participants in this sample were classified as being homeless (positive response to both measures), 17% were equivocally housed (discordant response to both measures) and 47% were housed (negative response to both measures) (Table 1). The mean age of participants was 24 years, 69% were male, 57% were White, 61% had at least a high school education, and 40% held a full or part-time job (Table 1). In bivariate analysis, source of most income, all of the early lifetime instabilities and all of the risk behaviors except for the condom use variables varied significantly by housing status (Table 1). Sources of income varied by housing status, with a tendency of paid employment to increase as housing became more stable. Homeless and equivocally housed were more likely to report illegal means of supporting themselves and homeless persons were more likely to report income from “other” sources. There was no difference in HIV or HCV serostatus by housing status, but all measures of early lifetime instabilities increased as housing status was more unstable.
Table 1

Sociodemographic characteristics, adverse early life circumstances, and HIV-related injection and sexual risk behaviors by residential status among IDUs aged 15–30 in 5 U.S. cities (n = 3266a)

Variable

Total

Homeless

Equivocally Housed

Housed

Chi-Square

3266

1200

546

1520

n (%)

36.7%

16.7%

46.5%

Sociodemographics

Site

    

Baltimore

971 (29.7)

15.7

35.8

38.6

 

Chicago

795 (24.3)

 9.3

19.0

38.1

 

Los Angeles

536 (16.4)

27.8

18.1

 6.9

 

New York

380 (11.6)

16.8

12.1

 7.4

 

Seattle

584 (17.9)

30.4

15.0

 9.0

 

Age—mean (SD)

23.9 (3.6)

23.6 (3.7)

24.1 (3.5)

23.9 (3.6)

 17.0

15 df

Male

2264 (69.3)

71.8

65.5

68.8

5.33

Race

    

6.29

6 df

Non-Hispanic White

2078 (63.6)

57.0

61.9

70.7

 

African American

246 (7.5)

 7.5

 9.0

 7.2

 

Hispanic

541 (16.6)

20.2

16.0

14.2

 

Other

373 (11.4)

15.4

13.1

 8.0

 

≥12th grade or GED

1986 (60.8)

58.6

57.9

63.8

5.19

Source of most income

    

68.8***

4 df

Job

1301 (39.8)

27.6

41.7

49.3

 

Other sources

1246 (38.2)

45.5

32.5

34.8

 

Illegal sources

701 (21.5)

26.9

25.7

15.8

 

Number of years injecting

     

mean (SD)

5.0 (3.7)

 5.6 (3.9)

 5.0 (3.8)

 4.5 (3.4)

3.2

Positive HIV status

89 (2.7)

 3.0

 3.4

 2.4

2.11

Positive HCV status

1099 (33.6)

36.8

36.0

32.0

4.42

Early Lifetime Instabilities

Lived in out-of-home placements

735 (22.5)

35.1

24.0

12.1

 46.9***

Thrown out of house

2067 (63.3)

74.0

70.2

52.5

104.6***

Juvenile detention

1009 (30.9)

41.7

32.4

21.9

 29.4***

Childhood physical abuse

1310 (40.1)

55.0

46.9

29.6

 79.8***

Childhood sexual abuse

830 (25.4)

34.9

31.0

18.3

 32.2***

Current Risk Behaviors (past 3 months)

≥Daily alcohol use

330 (10.1)

15.9

 9.9

5.6

 35.2***

≥Daily non-injection crack use

478 (14.6)

18.6

17.3

10.9

 42.2***

≥Daily speedball use

489 (15.0)

17.5

18.4

11.8

 27.2***

≥Daily methamphetamine use

385 (11.8)

20.4

12.3

4.8

26.6***

≥Daily injection

1496 (45.8)

48.9

43.7

44.1

 10.6**

Shooting gallery use

1283 (39.3)

54.4

47.4

26.6

184.3***

Receptive syringe sharing

1721 (53.7)

57.8

58.1

48.9

 46.1***

Injection paraphernalia sharing

2460 (76.2)

80.7

78.1

71.9

 35.9***

Backloading

1164 (39.8)

48.5

44.1

31.8

 85.6***

Sex work

614 (18.8)

26.1

25.1

11.3

 68.3***

Number of sex partners

    

 24.6***

4 df

0

433 (13.3)

13.5

13.2

14.9

 

1–2

1520 (46.5)

42.5

46.3

55.9

 

3+

1126 (34.5)

44.1

40.6

29.3

 

Condom use

     

Heterosexuals—Steady partners

    

9.3

6 df

Consistent

291 (8.9)

 7.4

12.1

10.3

 

Inconsistent

828 (25.4)

28.5

25.3

25.0

 

Never

1317 (40.3)

40.2

41.7

43.2

 

Heterosexuals—Casual partners

    

 11.5

6 df

Consistent

402 (12.3)

12.5

12.3

13.9

 

Inconsistent

435 (13.3)

17.9

13.1

11.7

 

Never

457 (14.0)

15.7

15.1

14.3

 

Homosexual men—Steady partners

    

5.49

6 df

Consistent

55 (37.2)

36.1

44.6

32.6

 

Inconsistent

22 (14.9)

20.8

 3.1

15.2

 

Never

71 (47.9)

44.1

53.3

53.1

 

Homosexual men—Casual partners

    

0.33

6 df

Consistent

57 (39.9)

38.3

43.1

40.0

 

Inconsistent

35 (24.5)

25.5

25.6

23.2

 

Never

51 (35.7)

37.2

32.9

37.4

 

a 19 missing on homeless outcome variables

P < 0.05b

** P < 0.01

*** P < 0.001

b P -values adjusted for city and calculated using generalized mantel-haenszel tests or fisher’s exact test

All injection-related risk behaviors and most sexual risk behaviors varied by housing status. Compared with stably-housed persons, the homeless and equivocally housed were more likely to drink alcohol and use crack/cocaine and methamphetamine at least daily, and the homeless were most likely to inject at least daily. Shooting gallery use and HIV injection risk practices (receptive syringe sharing, paraphernalia sharing and backloading) increased as housing became less stable. The homeless and equivocally housed were more likely to have recently engaged in sex work and have three or more sex partners. Condom use among either heterosexual or homosexual male partners did not vary significantly by housing status.

In multivariate analysis, seven independent variables remained associated with being equivocally housed, while eleven correlates remained associated with being homeless (Table 2). Two early lifetime instability variables and five risk behavior variables were associated with being both homeless and equivocally housed, while two additional early lifetime and socio-demographic variables were associated with being homeless. The two socio-demographic variables that were associated with housing status were: females were less likely to be homeless, and participants reporting “other” or illegal sources of income were twice as likely to be homeless. Early lifetime instabilities independently associated with being homeless included living in an out-of-home placement, and having been in juvenile detention. Having been thrown out of the house before age 18 and being physically abused as a child were associated with being homeless and equivocally housed (Table 2).
Table 2

Multinomial Logistic Regression Analysis Identifying Correlates of Homelessness and Equivocally Housed among IDUs aged 15–30 in 5 U.S. cities

 

Homeless vs. Housed

Equivocally housed vs. Housed

(n = 1200)

(n = 546)

Adjusted Odds ratio

95% CI (P-values)

Adjusted Odds ratio

95% CI (P-values)

Sociodemographics

Gender

    Male

1.00

Reference

  

    Female

0.75*

0.61–0.94

1.17

0.91–1.49

Income

    Jobs

1.00

Reference

  

    Other sources

2.02***

1.63–2.50

0.96

0.74–1.25

    Illegal sources

1.93***

1.48–2.50

1.31

0.97–1.77

Early Lifetime Instabilities

Lived in out-of-home placements

    No

1.00

Reference

  

    Yes

1.81***

1.40–2.34

1.37

1.00–1.89

Thrown out of house

    No

1.00

Reference

  

    Yes

1.78***

1.46–2.19

1.57***

1.23–1.99

Juvenile detention

    No

1.00

Reference

  

    Yes

1.34*

1.07–1.68

1.19

0.91–1.56

Childhood physical abuse

    No

1.00

Reference

  

    Yes

1.83***

1.50–2.23

1.58***

1.24–2.00

Current Risk Behaviors (past 3 months)

Alcohol use

    Never/ < Daily

1.00

Reference

  

    ≥Daily

2.36***

1.69–3.29

1.83**

1.22–2.73

Methamphetamine use

    Never/ < Daily

1.00

Reference

  

    ≥ Daily

3.16***

2.25–4.43

2.16***

1.44–3.24

Shooting gallery use

    No

1.00

Reference

  

    Yes

2.54***

2.08–3.09

1.91***

1.51–2.42

Backloading

    No

1.00

Reference

  

    Yes

1.55***

1.28–1.89

1.37**

1.08–1.72

Sex work

    No

1.00

Reference

  

    Yes

1.66***

1.29–2.15

1.66***

1.23–2.23

P < 0.05

** P < 0.01

*** P < 0.001

All of the substance use-related variables were associated with being homeless and equivocally housed, including alcohol use, methamphetamine use, shooting gallery use and backloading. Sex work was the only sexual risk behavior independently associated with being homeless and equivocally housed.

Discussion

We observed a high prevalence of unstable housing in this sample. In this relatively young group of IDUs more than half were classified as homeless by at least one measure. Furthermore, we found that these two simple housing status questions identified IDUs with different levels of behavioral risk for HIV. Although homeless and equivocally housed participants are at increased risk for participating in HIV risk behaviors compared to housed participants, the equivocally housed exhibited fewer HIV risk behaviors compared to the homeless. This finding indicates that, in terms of accurately defining homelessness for the purposes of determining associated HIV risk, it is important to assess both self-perceived homelessness and specific material conditions (living on the street or shelter), because it is the combination of these measures which is most informative.

Our definition of homelessness, which combined both subjective and objective definitions, may be useful for other researchers. We found it was efficient and simple to assess, and it easily distinguished trends in HIV risk. This approach also identified a sub-group–the equivocally housed–who may not readily identify as “homeless” but experience some elements of this status associated with increases in HIV risk behaviors. In comparison, a study by Andia and colleagues used a detailed definition of housing status and assigned subjects to five residential status categories (living in parent’s home, living in own home, living in other’s home, living in temporary housing, and homeless or living in streets/shelters) and found that high risk behaviors were more likely among homeless IDUs (Andia et al., 2001). Similar to our definition, this assessment described HIV-associated risks, and identified sub-groups of the most severely homeless people in need of urgent attention. Although this five-level definition was informative, a briefer assessment might be more useful and efficient in some research or service provision settings where time is limited.

Among this cohort of young IDUs, early traumatic or unstable life events remained associated with later homelessness. These were the only measures that we could prospectively evaluate with current housing status. Childhood physical abuse and being thrown out of the house were particularly predictive, in that they were associated with being both homeless and equivocally housed. These findings are consistent with other reports, which have identified these same risk factors for homelessness (Heffron et al., 1997; Herman et al., 1997; Stein et al., 2002). Given the young ages of these IDUs, it is not clear how or whether early initiation of substance use is related to these childhood life events, but it is possible that these destabilizing life events–particularly physical abuse–may have provoked early substance use behaviors in these participants (Belenko et al., 2005; Kang et al., 2002).

Being homeless or equivocally housed is clearly associated with a number of HIV-related risk behaviors, but did not translate into elevated HIV or HCV prevalence. This finding may be due to issues related to temporality. HIV or HCV infections might have occurred many months, or even years, prior to the study period. In this population of IDUs current housing status may not be predictive of past housing status, which would have been the time when HIV/HCV infections most likely occurred. This finding may also be due to limitations of the cross-sectional study design or, given the young age of participants that much of the high-risk behavior occurred with people who were not yet infected thus reducing the likelihood of exposure. For HCV, the lack of significance may also be due in part to the high transmissibility of HCV in settings with moderate prevalence. Consistent with the literature, we found that frequent alcohol use, methamphetamine use, backloading, shooting gallery use and sex work were associated with being both homeless and equivocally housed (Andia et al., 2001; Deren et al., 2003; Diaz et al., 2001; Gleghorn et al., 1998; Koester et al., 2005; Metraux et al., 2004; O’Toole et al., 2004; Semple et al., 2004). What was unique about this study was that we found these same associations to be true for younger IDUs as well.

We observed that the absence of key elements related to sustaining one’s livelihood were associated with housing status–specifically, those who did not have gainful employment and who engaged in sex work were most likely to be homeless or equivocally housed. While the causes and consequences of homeless and marginally-housed people are myriad and may involve structural as well as behavioral factors, our findings suggest that such IDUs are at elevated risk for engaging in HIV risk behaviors because of the risky nature of the behaviors used to support themselves (Essien et al., 2004), and because stable housing provides the means to more safely inject (Latkin et al., 1994). Being homeless could also lead to unsafe injection practices due to difficulties in storing and having ready access to clean injection equipment and water and safe injection space (Luciano et al., 2000). Homelessness, and the attendant inability or unwillingness to carry around one’s own injection equipment (Case et al., 1998; Vlahov and Junge, 1998), may result in the need to frequent shooting galleries, which provide access to often unsterile injection equipment. Injecting within such settings has long been documented as unsafe because of the likelihood of sharing syringes and injection paraphernalia and of doing so across social networks (Case et al., 1998; Fuller i, 2003; Marmor et al., 1987; Reyes et al., 1996; Thorpe et al., 2002). Unstable housing could also necessitate injecting in outdoor settings or abandoned buildings, which is associated with HIV seroconversion (Friedman et al., 1995).

We found homelessness to be twice as common in Los Angeles and Seattle compared to other study sites. Differences in the prevalence of homelessness by site could be explained by a number of factors, including differences in the underlying population, the social structure, the climate, or available resources for homeless persons. Differences in the underlying population may be widespread, in part because the Chicago sample included a large suburban element, virtually absent in the other cities. The social structure in Baltimore is oftentimes characterized by large households with extended families that many might consider primary. While someone in Baltimore, living with a member of his extended family, might not consider himself to be homeless, a participant in one of the other cities might. The climate on the west coast is more tolerable for a homeless person than in the other cities, which enables homeless people to live on the street or in makeshift shelters rather than temporarily with friends or family. And finally, there is often the perception that there are more readily available resources for homeless persons in the western cities, which could lead to an increased population or which could lead to their knowledge of the study; hence we may have recruited a cross-section of homeless youth in Los Angeles and Seattle, but a distinct, more stable subgroup in the other cities. More research is needed to determine these differences in homelessness across sites.

One of the strengths of this analysis was its sample of young IDUs. Many studies have focused on older IDUs who exhibit relatively high levels of HIV infection. Young IDUs often have not yet contracted HIV, thus there is still time to intervene to prevent infection. A second strength was the large sample size drawn from various settings within multiple cities. While we observed differences in homelessness by site, key risk factors remained after controlling for site, indicating that these findings may be generalizable to a wide number of settings and cities. Third, the use of ACASI may have decreased self-report bias in assessing stigmatized drug and sexual behaviors (Macalino et al., 2002; Metzger et al., 2000; Perlis et al., 2004).

Limitations included the use of cross-sectional data, particularly with respect to risk behavior and HIV and HCV prevalence; however, we were able to establish temporal associations between adverse early lifetime instabilities and current housing status. Another limitation was the definition used to identify those who were equivocally housed. One possible explanation for the associations seen with HIV risk behaviors and early lifetime instabilities is misclassification; however this is likely to be non-differential, and would therefore lead to a reduction of the association towards the null. The strong associations seen when comparing homeless versus housed IDUs suggest that the definition of equivocally housed did not introduce such a bias.

We have created a simple and brief measure of housing status which incorporates both subjective and objective criteria. We found that it was able to readily identify and distinguish HIV-related substance use and sexual risk behaviors (drinking alcohol, using a shooting gallery, injection risk and sex work) and early lifetime instabilities (having lived in out-of-home placements, been thrown out of the home, and experienced childhood abuse) by housing status. For service providers working among drug-using populations where the prevalence of homelessness is widespread, the standard, oftentimes narrow definitions of homelessness may be limited in their utility to help identify those who need intervention the most. A more accurate one-time assessment could be an important tool for prioritizing potential risk and service delivery amongst IDUs, a large portion of who are homeless or on the verge of becoming homeless.

Acknowledgements

The DUIT Study Group includes the following people: Steffanie Strathdee, Elizabeth Golub, Marie Bailey-Kloch, Karen Yen-Hobelman (Baltimore); Lawrence Ouellet, Susan Bailey, Joyce Fitzgerald (Chicago); Sharon Hudson, Peter Kerndt, Karla Wagner (Los Angeles); Mary Latka, David Vlahov, Farzana Kapadia (New York); Holly Hagan, Hanne Thiede, Nadine Snyder (Seattle); Richard Garfein, David Purcell, Paige Ingram, Andrea Swartzendruber (CDC). The authors also acknowledge the following people for their contributions to this research: Yvette Bowser, Peter O’Driscoll, Janet Reeves, Marcella Sapun (Baltimore); Angus Atkins-Trimnell, Mary Bonilla, David Cosey, Jaime Delgado, Julio Garcia, Michelle Giles, Erin Kubalanza, Michael Phillips, Edward Snulligan (Chicago); Marrisa Axelrod, Elizabeth Faber, Lawrence Fernandez Jr., Christian Geannette, Roberto Rojas (Los Angeles); Ebele Benjamin, Sebastian Bonner, Micaela Coady, Joanna Cruz, Sandra DelVecchio, Dirk Jackson, Gregory Malave, Joan Monserrate, Danielle Ompad, Clarisse Miller O’Shea, Yingfeng Wu, Manny Yonko (New York); Jennifer V. Campbell, Stanley Brown, Rong Lee, Susan Nelson, Jef St. De Lore, Carrie Shriver, Jeanette Frazier, Jean Pass, Paul Swenson (Seattle); Yuko Mizuno, Janet Moore, Ann O’Leary, Vincent Raimondi, Scott Santibanez, RobertoValverde (CDC); IanWilliams,Wendi Kuhnert, Himal Dhotre, Leigh Farrington, (CDC Division of Viral Hepatitis); Suzette Bartley, Dollene Hemmerlein (CDC Serum Bank Branch). This study was funded in its entirety by a cooperative agreement from the Centers for Disease Control and Prevention. U64/CCU317662; U64/CCU517656; U64/CCU917655; U64 CCU217659; U64/CCU071615.

Copyright information

© Springer Science+Business Media, LLC 2007