Untangling the Relationship Between Mental Health and Homelessness Among a Sample of Arrestees

Abstract

Past research has focused on the intertwined relationship between homelessness, mental illness, and criminal justice. Although a well-established correlation between mental illness and homelessness has emerged, a better understanding of how this may be mediated by other prominent risk factors such as substance use or victimization is warranted. The current study uses data obtained from 3,673 recently booked arrestees to examine these relationships. Using structural equation modeling with measured variables, the analyses indicate the relationship between mental health and homelessness to be almost entirely mediated by alcohol use, drug use, and violent victimization. Policy implications are discussed.

Keywords

homelessness mental health substance use victimization incarceration arrestees

Introduction

Homelessness remains a significant public health problem in the United States affecting more than 1.6 million individuals annually and more than 650,000 people being homeless on any given night (U.S. Department of Housing and Urban Development, 2009). A common and important subsection of the homeless population are individuals who are transient and involved in the criminal justice system as evidenced by a robust literature connecting these two groups (e.g., see Desai, Lam, & Rosenheck, 2000; P. J. Fischer, 1988; Greenberg & Rosenheck, 2008; Kushel, Hahn, Evans, Bangsberg, & Moss, 2005; Martell, Rosner, & Harmon, 1995; Snow, Baker, & Anderson, 1989). This well-established relationship not only exists but also appears to be multifaceted and bidirectional (Kushel et al., 2005; also see Travis, Solomon, & Waul, 2001). For instance, Greenberg and Rosenheck (2008) found that jailed individuals had been homeless in the previous year at rates 7.5 to 11 times greater than the general public. Conversely, Greenberg and others noted that homeless offenders routinely have histories of criminal justice involvement that far exceed domiciled individuals (Zaph, Roesch, & Hart, 1996).

The problem, however, is that homelessness and the factors associated with homelessness are not well understood in regard to criminally involved individuals. The risk of homelessness is of particular concern for criminally involved individuals because offenders reentering the community face harrowing obstacles following incarceration. Specific challenges to securing a place to live, gaining employment, and obtaining behavioral health services all can lead to offenders becoming homeless (Kushel et al., 2005).

Although numerous studies exist on the consequences of homelessness for individuals in the criminal justice system, traditionally, they have focused on prevalence rates or statistical models that examine direct relationships between risk factors and homelessness. Given the scope of the dilemma of understanding homelessness and criminal justice involvement, it can be assumed that other risk factors for mental health problems and homelessness are also entwined and intricate. Therefore, the current study proposes a model for understanding the relationships between mental health, substance abuse problems, victimization, and homelessness for criminally involved individuals.

Risk Factors for Homelessness and Criminal Justice Involvement

A growing body of literature has examined many risk factors for homelessness (Booth, Sullivan, Koegel, & Burnam, 2002; Greenberg & Rosenheck, 2008; Susser, Moore, & Link, 1993), creating a range of public policy, behavioral health, and public safety considerations. A primary focus of this body of research has been on the increased rates with which homeless individuals are affected by serious mental health problems (e.g., see Breakey et al., 1989; Fazel, Khosla, Doll, & Geddes, 2008; P. J. Fischer & Breakey, 1991; Gelberg, Linn, & Leake, 1988; Koegel, Burnam & Farr, 1988; Lamb & Lamb, 1990; Sullivan, Burnam, & Koegel, 2000; Susser et al., 1993; Susser, Stuening, & Conover, 1989). Additional research has led academics to understand that mental illness is not only strongly associated with homelessness but that mental health problems are also overrepresented in criminal justice populations (Ditton, 1999; Teplin, 1991; Teplin, Abram, & McClelland, 1996). For example, more than 50% of individuals in jails and prisons suffer from a mental health problem (James & Glaze, 2006), putting the prevalence of mental illness for offenders at far greater levels than the general public (Desai et al., 2000; Feliner & Abramsky, 2003; Guy, Platt, Zwerling, & Bullock, 1985; Teplin, 1990). Multiple studies have also revealed the increased prevalence of homeless individuals with mental illness in the criminal justice system as opposed to the general population (Belcher, 1989; Martell et al., 1995; White, Chafetz, Collins-Bride, & Nickens, 2006). Aside from mental illness being more prevalent in homeless incarcerated individuals (compared with homeless individuals in the community), research has also found that homeless incarcerated individuals were significantly more likely to suffer from a mental illness than non-homeless individuals who were incarcerated (McNiel, Binder, & Robinson, 2005).

Although there is a well-established relationship between mental illness with homelessness and involvement in the criminal justice system, it remains an important area of examination for a number of reasons. First, informed policy regarding persons with mental illness rests on a better understanding of why these individuals are overrepresented in the criminal justice system. Some evidence has argued that the deinstitutionalization movement had a substantial impact on the criminalization of mental illness (Abramson, 1972; see also Torrey, 1997). As a result of deinstitutionalization, the criminalization hypothesis contends that vulnerable populations (such as those with mental illness and homeless) are disproportionately arrested because police officers lack alternative training or solutions to resolve the situation (Lamb & Weinberger, 2001; Teplin & Pruett, 1992). Other scholars have argued that these individuals often become involved in the criminal justice system as a result of minor public disturbances and property crimes that can at least partially be attributed to their lack of housing (Teplin & Pruett, 1992). While there is some evidence supporting the criminalization hypothesis (Teplin, 1984, 1985), other evidence suggests little relationship (Engel & Silver, 2001; Novak & Engel, 2005). For instance, when controlling for criteria such as the influence of drugs and alcohol, non-compliance, or the seriousness of the offense, Engel and Silver (2001) found little support for the criminalization hypothesis, and, in fact, in some instances, persons with mental illness were significantly less likely to be arrested than individuals without mental illness. Regardless, the over representation of the group in the criminal justice system is troublesome because it is often hard to find adequate community treatment (Bachrach, Talbott, & Meyerson, 1987) and individuals face multiple and difficult reentry challenges in the community after being released from incarceration (Kushel et al., 2005; Travis et al., 2001).

The potential vulnerability of homeless individuals with mental illness to be victims of crime and violence represents a second consideration for the study of criminal justice system involvement among this population. Individuals with mental illness may have significant difficulty adjusting to being homeless (Greenberg & Rosenheck, 2008; Martell et al., 1995; Solomon & Draine, 1999) and perhaps not as able to successfully engage in critical street survival skills relative to the homeless who do not have mental health problems. These functional limitations, in turn, may lead to greater criminal justice involvement or increased exposure to victimization (e.g., see Dennis & Steadman, 1991; Hiday, Swartz, Swanson, Borum, & Wagner, 1999; Maniglio, 2008). Individuals in transient living conditions were found to have the highest odds ratio of non-violent and violent victimization, surpassing other significant predictors such as education and substance abuse in a recent multivariate analysis of 331 psychiatric patients (Hiday et al., 1999). Other studies have also found that women who are homeless and mentally ill are particularly susceptible to victimization (Goodman et al., 2001; White et al., 2006). This is especially alarming, as research has shown that the mentally ill and homeless populations that are involved in violence seek out help at emergency shelters and psychological treatment facilities less often (Martell et al., 1995). As White et al. (2006) stated, “exacerbations of mental illness and co-occurring homelessness and substance abuse may put the severely mentally ill at risk for both arrest and victimization, further disruptions to lives that are already chaotic” (p. 133).

A third concern in the continued study of the connection between homelessness, mental illness, and criminal justice involvement centers on disentangling the relationship between homelessness and mental health (e.g., see Bassuk, Rubin, & Lauriat, 1984). It is assumed that the relative influences that mental health status, and the associative psychosocial impacts may play into increased risk for homelessness. Some community samples of the homeless have examined whether the link between homelessness and serious mental illness may be mediated by other risk factors. In one example, Draine, Salzer, Culhane, and Hadley (2002) argued that mental illness is not as important an explanatory factor for homelessness when considering the increased social problems such as poverty and unemployment this group faces. Some scholarship has also suggested that the more common risk factor for homelessness in a population of individuals with mental illness may be substance abuse (Caton et al., 1994; Caton et al., 1995). Considering the well-documented high rates of co-morbidity of mental illness and substance abuse (Kessler et al., 1997), developing a greater understanding of the role that substance abuse plays between mental illness, homelessness, and criminal justice involvement is critical as little is known about these factors for offenders.

Like mental illness, research has repeatedly shown substance abuse to be a risk for homelessness as well as potentially exasperating it (e.g., see Booth et al., 2002; Fazel et al. 2008; P. J. Fischer & Breakey, 1991; Greenberg & Rosenheck, 2008; Hartwell, 2003; Koegel et al., 1988; Padgett & Struening, 1992). Furthermore, homeless individuals are more likely to have a co-occurring diagnosis of a mental health and substance abuse disorder (Brunette, Mueser, & Drake, 2004) as are homeless inmates (McNiel et al., 2005). Compounding the relationship between mental illness and substance abuse for co-occurring populations is the question of how important is the relationship between mental illness and criminal justice involvement when controlling for substance abuse. Some scholars have found that the relationship between mental illness and criminal justice involvement disappeared when controlling for comorbid substance abuse (Pandiani, Rosenheck, & Banks, 2003; Quanbeck et al., 2005), a finding supported in another study involving veterans (Erickson, Rosenheck, Trestman, Ford, & Desai, 2008). Studies of this kind highlight the need for scholars to better understand the nature of these relationships between mental health status, homelessness, substance use, and criminal justice involvement.

As discussed, substantial research has examined homelessness and its predictors, as well as risk factors for individuals in the criminal justice system. However, further hypothesis testing of the risk factors examined in the present study are important for multiple reasons. First, increased knowledge surrounding the factors that help sustain homelessness can assist in helping professionals to identify those most at risk for future homelessness who are involved in the criminal justice system. It may be that having a mental illness is not an independent predictor of homelessness without considering additional risk factors. Second, greater understanding of these pathways is of social importance to better inform treatment professionals attempting to provide more informed care to persons with mental illness who are transitioning from incarceration into the community. Treatment that has addiction or victimization as the focal point of the service may be of greater importance in preventing future homelessness for criminal justice involved individuals; perhaps even more so than traditional mental health counseling. Finally, better understanding of the mediating relationships of the predictors for homelessness of criminally involved individuals is important for preventing later cycling in and out of the criminal justice system, thus saving extremely valuable, yet largely limited, resources within the system.

Current Focus

The current study examines the relationship between mental health, alcohol abuse, drug abuse, victimization, and homelessness among an arrestee population. This study tests the relationship between mental health problems and homelessness in a sample of arrestees, questioning whether this relationship is mediated by other factors, such as victimization, alcohol use problems, or illegal substance use problems. Our primary research hypothesis is that having criminal justice involvement and displaying symptoms of mental illness increase the risk of victimization or substance problems, which, in turn, increases the likelihood of one becoming homeless. We also hypothesize that the direct effect of a mental health problem on homelessness is relatively weak compared with the indirect effects of these other predictors (victimization and substance abuse). The current study uses structural equations modeling with measured variables, also known as path modeling, to test this mediating relationship.

Method

Data

The data for the current study come from the Arizona Arrestee Reporting Information Network (AARIN), from the second quarter of 2007 to the fourth quarter of 2008. The AARIN project in Maricopa County was originally established in 1987 under the auspices of the Drug Use Forecasting (DUF) program, and later, the Arrestee Drug Abuse Program (ADAM), both sponsored by the National Institute of Justice (NIJ).¹ In 2007, after NIJ terminated the nationwide program due to funding constraints, a few jurisdictions continued to fund the program through the use of local funds. Maricopa County was one of those sites, with funding provided by the Maricopa County Managers Office. The AARIN program maintained the same methodology as the ADAM project so that trends among recently booked arrestees could continue to be monitored over time. While the AARIN project samples males and females from the adult and juvenile populations, the data used in the present study are restricted to adult arrestees.²

To ensure representative results for the entire population of arrestees in Maricopa County, the AARIN project uses a systematic sampling protocol that includes the collection of data at multiple facilities, with target quotas at each facility used to ensure representative results for the entire population of arrestees in Maricopa County. Data are collected quarterly at all facilities; interviews are conducted during a 2-week period each quarter. During data collection periods, interviews are conducted with arrestees who are randomly selected based on the time they were booked and processed. Consistent with the ADAM sampling strategy, a stock (i.e., arrested during non-data collection hours) and flow (i.e., arrested during data collection hours) selection process is used to ensure a representative sample of arrestees. Arrestees who had been in custody longer than 48 hr were ineligible for participation in AARIN because of time limitations associated with urinalysis testing. Over the study period, nearly 90% of the approached adult arrestees agreed to participate in the study; more than 92% of those who were interviewed also consented to providing a urine specimen. While a total of 3,997 arrestees participated in the survey and provided a urine sample, 162 (4.2%) participants were excluded from the study because of missing data, resulting in a sample of 3,885 for the present study.³

Variables

Below we discuss the variables used and their coding for the present study.

Homelessness

A dichotomous variable was constructed to indicate whether the arrestee was homeless during the month prior to his or her current arrest. Each case was assigned a value of “0,” and considered domiciled, if they reported one of the following living situations: apartment, condominium, hotel, single family house or mobile home, or public housing. A case was assigned a value of “1” if the respondent reported the streets or short-term or emergency shelters as their primary residence.

Mental health problem

A dichotomous measure was created assigning a value of “1” to cases in which respondents indicated that they had been told, treated, medicated, or hospitalized for a mental illness or emotional problem in the past 12 months. All other cases were coded “0,” reflecting no mental health problem. Although not a clinical assessment, this brief, self-reported screening has shown to be a reliable proxy for clinical assessments (Berwick et al., 1991; Spitzer, Kroenke, & Williams, 1999) and has been used in other studies measuring mental health conditions among criminal justice system participants (Ditton, 1999)

Drug use problem

A dichotomous measure was created, assigning a value of “1” to cases in which respondents who indicated that they were dependent on, or felt they needed treatment for, any of the following drugs: marijuana, crack, powder cocaine, crack, heroin, methamphetamine, inhalants, or ecstasy. All other cases were assigned a value of “0” reflecting no drug problem.

Alcohol use problem

Cases were coded “1” in which the respondents reported that they were dependent on, or felt they needed treatment for, alcohol. All others cases were assigned a value of “0” reflecting no alcohol use problem.⁴

Victimization

Cases were coded “1” in which the respondents indicated being a victim of a violent crime in the 30 days immediately preceding their arrest. Types of victimizations included being threatened with a gun, shot at, shot, threatened with a weapon not including a gun, injured with a weapon other than a gun, assaulted without a weapon, and robbed. All other cases were coded “0,” reflecting no history of recent victimization.

Demographic Characteristics

Two control variables were included in the model to guard against spuriousness. First, Gender was included as a dichotomous variable (males = 1, females = 2). Second, a measure for race was included (0 = non-White, 1 = White). Table 1 displays the means for all the variables included in the analysis.

Table 1.

Means and Tetrachoric Correlations (N = 3,673).

		1	2	3	4	5	6	7	8
1	Homelessness	1
2	Mental health problem	.223	1
3	Drug use problem	.242	.336	1
4	Alcohol use problem	.279	.172	.347	1
5	Victimization	.220	.334	.249	.119	1
6	Property crime	.153	.055	.135	.075	.065	1
7	Female	−.008	.237	.086	−.086	.070	.070	1
8	White	.225	.298	.212	−.053	.083	.034	.148	1
M		0.081	0.299	0.317	0.206	0.209	0.205	1.227	0.380

Analytic Strategy

Structural equation modeling with observed variables was estimated in this study, using PRELIS 2.0 and LISREL 8.8. We addressed assumptions of normality in estimating a linear structural equation model. Because all the variables used in the analyses are dichotomous, specifically the key endogenous variable (homelessness), steps were taken to ensure unbiased parameter estimates in the final path models. LISREL 8.8 was used to estimate a path model using diagonally weighted least squares estimation. This first path model examined whether drug use problems, alcohol use problems, and victimization mediate the relationship between mental health problems and homelessness. In the stage of our analysis, we modified the path model based on local fit indices and assessed the direct and indirect effect of having a mental health problem on homelessness, controlling for the other variables in the model.

Non-normal Data

Non-normal data are common in criminology and criminal justice research. Researchers are often interested in what factors lead to a particular outcome such has homicide, drug use, conviction, incarceration, and gang membership. These variables are often measured dichotomously. The problem with non-normally distributed data, however, is that the usual method of estimating a structural equation model is maximum likelihood, which assumes that variables are normally distributed. Estimating parameter estimates for exogenous dichotomous variables is not a problem using maximum likelihood estimation as the coefficients can be interpreted in a manner similar to the interpretation of ordinary least squares (OLS) regression. However, dichotomous or categorical endogenous variables present a problem when using maximum likelihood estimation. Given that the current study uses dichotomously measured variables, which are non-normally distributed, steps were taken to ensure that the final parameter estimates were not biased.

One technique for analyzing non-normal data is to use a tetrachoric correlation matrix. This involves the hypothesized existence of continuous latent variables underlying each dichotomous variable (X*) and analyzing these rather than the binary variables. Furthermore, for each continuous latent variable, there is a latent threshold parameter (v; Muthén, 1993). The latent threshold is essentially the cut point at which a respondent goes from 0 → 1 on the measured dichotomous variable. Thus, we assume X = 1 if X* > v and X = 0 if X* < v. The tetrachoric correlation requires a large sample size (Chen & Popovich, 2002), which we have in the current study (N = 3,673), and gives us the correlation coefficient for the latent, normally distributed, continuous variables. The tetrachoric correlation matrix can be found in Table 1.

The second step in dealing with non-normal data is to estimate an asymptotic covariance matrix. In an asymptotic covariance matrix, the diagonal elements represent the expected variance of a parameter estimated with repeated sampling. Thus, our unbiased standard errors can be derived by taking the square root of the diagonal elements in the asymptotic covariance matrix (Skrondal & Rabe-Hesketh, 2005). The asymptotic covariance matrix was estimated in PRELIS 2.0 and was used to calculate the significance of parameter estimates in the current study. To analyze the tetrachoric correlations and the diagonal elements of the asymptotic covariance matrix, diagonally weighted least squares estimation was used in LISREL 8.8 for the following path models.

Model Identification

The first step in estimating a structural equation model is to examine whether the model is identified. A model is identified when a unique solution is possible for the parameters to be estimated. One necessary condition for identification is the “t-rule.”

t \leq \frac{(p + q) (p + q + 1)}{2},

where p = number of Y variables, q = number of X variables, t = number of parameters to be estimated.

The t-rule assesses whether the model is over-parameterized, that is, whether the model attempts to estimate more parameters than unique pieces of information in the matrix to be analyzed. The current study analyzes a total of eight variables, and thus provides 36 pieces of non-redundant information.⁵ There are 26 parameter estimates in the first model and 27 in the second, indicating that the t-rule is satisfied for the current analysis (Bollen, 1989). While there is no foolproof way to ensure identification (Bollen, 1989), given the above test, we had no reason to believe that either of the models presented were not identified before estimation.

Results

Univariate and Bivariate Relationships

Table 1 displays the tetrachoric correlations and means for all the variables used in the analysis. Just more than 8% of the arrestee population indicated that they were homeless during the 30 days prior to their arrest. About 30% reported having a mental health problem in the previous 12 months. More arrestees reported having a drug use problem (31.7%) than alcohol use problem (20.6%). About 21% of the respondents indicated that they had been victimized in the past 30 days, and the same percentage were arrested for a property crime as their current most serious charge. Approximately 23% of the respondents were female and 38% were White. Also of interest, homelessness and mental health problems were correlated at r = .223; thus, there is a medium-size bivariate relationship between these two variables.

Mediation Model

Figure 1 displays the proposed model to be tested. Specifically, the path model tests whether the relationship between having a mental health problem and being homeless is mediated by the presence of a drug use problem, alcohol use problem, and/or victimization. Figure 1 displays only the relationships among the endogenous variables. The two exogenous variables (gender and race) were included in the model to guard against spuriousness; only the significant paths were included in the final model. See the appendix for the standardized coefficients between the two exogenous variables and all the endogenous variables (for the final model). Global fit indices of the model presented in Figure 1 indicate marginally good fit. The chi-square corrected for non-normality⁶ was 176.56 (p < .001) with 10 degrees of freedom (see Table 2 for all fit indices). A significant chi-square suggests that the model does not fit the data perfectly. The chi-square test is very conservative, however, and other fit indices should be considered when evaluating the global fit of a path model (Bollen, 1989). For instance, the root mean square error of approximation (RMSEA) was .070 (95% confidence interval [CI] = [.062, .079]), which is slightly above the desired level of .05, which indicates good fit. The standardized root mean square residual (SRMSR), which has a desired value of below .05, was .042. And finally, the comparative fit index (CFI) was .95, which is right at the desired level. Thus, the global fit indices indicate reasonable fit.

Table 2.

Original and Modified Fit Indices for Path Model.

	df	χ²	RMSEA	RMSEA CI	SRMSR	CFI
Original	10	176.56**	0.070	[.062, .079]	0.042	0.950
Modified^a	9	54.70**	0.044	[.035, .053]	0.023	0.980

Note. RMSEA = root mean square error of approximation; CI = confidence interval; SRMSR = standardized root mean square residual; CFI = comparative fit index.

Allowed the disturbances of alcohol dependence and drug dependence to correlate.

p < .05. **p < .01.

An examination of the local fit indices indicated that the largest modification index was 288.76 in the psi matrix between drug use problem and alcohol use problem. This is an indication that the disturbances between the two variables should be allowed to correlate. It is reasonable to assume that there is a common source of alcohol and drug use problem that lies outside of the given model. Thus, allowing the two disturbances to correlate will account for this and likely improve the fit of the model.

Modified Mediation Model

The modification made to the original model was to allow the disturbances from alcohol use problem and drug use problem to correlate (i.e., element 3,4 in the psi matrix). Global fit indices indicate that the modified model was a better fit (see Table 2). While the model chi-square was still significant, and the model does not fit the data perfectly, other indicators suggest that the model was a good fit. For example, the RMSEA was .045 (95% CI = [.035, .055]), the CFI was .98, and the SRMSR was .022, all of which indicate that the model fits the data well.

Figure 2 displays the standardized path coefficients for the modified model. All the paths from the two exogenous variables and the endogenous variables have been excluded for model clarity, additionally; all disturbance terms have been excluded from the model in light of parsimony.⁷

Figure 2 shows that the direct effect of mental health on homelessness is small and non-significant, indicating that the relationship between the two variables is entirely mediated by alcohol use problem, drug use problem, and victimization. Table 3 displays the squared multiple correlations for the endogenous variables, as well as the standardized direct and indirect effects of the endogenous variables on homelessness. Having a mental health problem has a standardized indirect effect on homelessness of .27, while the direct effect is only .01. Also of importance is the large direct effect alcohol use problem has on homelessness (β* = .39), while drug use problem has a much smaller direct effect (β* = .06). Because these are standardized effects, they can be compared directly that the presence of an alcohol use problem has a 6.5 times greater direct effect on homelessness than the presence of a drug use problem. About 63% of the variation in homelessness was explained by the final path model (see Table 3).

Table 3.

Squared Multiple Correlations and Standardized Direct and Indirect Effects on Homelessness.

	r ²	Direct	Indirect
Homelessness	.63	—	—
Mental health problem	.28	0.01	0.27
Drug use problem	.30	0.06	0.00
Alcohol use problem	.26	0.39	0.00
Victimization	.32	0.28	0.15
Property crime	.43	—	—

Discussion

The present study examined the relationship between mental health problems and homelessness within a sample of recently booked arrestees. Research has consistently recognized mental health problems to be associated with homelessness, although the current study found the relationship between mental health and homelessness to be entirely mediated by alcohol use, drug use, and violent victimization. Notably, these findings suggest that a strong direct effect between mental health problems and victimization is present. Although accounting for the largest direct effect, victimization does not provide as strong of a direct effect on homelessness as alcohol does. More individuals endorsed drug problems over those of alcohol although the direct effect for alcohol is almost 5 times greater than drug use on homelessness. It may be that increased access to alcohol, in contrast to illegal substances, is one reason that alcohol plays a larger role in mediating mental health problems and homelessness in this population. In addition, the fact that alcohol is normally inexpensive compared with illegal substances may also play a role in its availability and subsequent increased direct effect on homelessness. The results also reveal a strong effect leading individuals from mental health problems to victimization and, from there on, to drug and alcohol problems. A closer analysis of this relationship might indicate the functionality of “self medication,” wherein the offender who has been the target of victimization abuses alcohol or drugs to modulate experiences of trauma, depression, and anxiety conditions, all common results of victimization. Assuming that violent victimization is quite traumatic for many respondents, using drugs and alcohol might be a fundamental coping mechanism for responding to this trauma.

Before discussing the implications of this study, it is important to mention three primary limitations. First, these are cross-sectional data and while structural equation modeling can test a causal model with cross-sectional data, caution must be exercised in interpretation of the findings. The current study did not assess how individuals became homeless; instead, it examined the factors associated with an arrestee’s current homelessness status. The goal of the current study was to untangle the relationship between these variables as they currently exist. Second, the findings should not be generalized to the general homeless population, as this was a sample of arrestees. Although a large percentage of homeless individuals have interaction with the criminal justice system, it certainly is not the case that all homeless individuals get arrested. This sample of homeless individuals is likely a subset that is most at risk for deleterious outcomes and more chronically homeless on average. While the findings from the current study should not be generalized to all homeless individuals, the examination among arrestees is important given that research has found that homelessness is associated with non-violent crime. S. N. Fischer, Shinn, Shrout, and Tsemberis (2008) went as far as to state, “a strong argument can be made that homelessness encourages non-violent crime” (p. 262). Finally, and as previously noted, the self-report assessment of mental health problems used in this study is not a clinical assessment resulting in a diagnosis or a formal screening to indicate Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV; American Psychiatric Association, 1994) disorder. Despite these limitations, this study offers important insights into the relationship between mental health, substance use, victimization, and homelessness, specifically for those involved in the criminal justice system.

These findings suggest three significant policy implications. The first policy consideration comes from the finding that the association between mental health problems and homelessness can potentially be interrupted if more attention is paid to the issues stemming from poor mental health, specifically substance abuse and victimization. By treating these two risk factors for persons with mental illness, we may in fact also be treating homelessness. Comprehensive programs that focus on problems for criminally involved homeless individuals beyond finding a home and a job are keys to keeping these individuals off the street. For instance, wide-ranging and long-lasting addiction treatment as described by Hartwell (2003), or an increased range of community support services and continuum of care (Center for Mental Health Services, Center for Substance Abuse Treatment, 1998), could be better used.

Second, the findings from the current study suggest that drug problems and especially alcohol problems are important predictors in the path between mental health problems and homelessness. While it is true that those who are homeless need housing, it is equally critical that these individuals are provided access to evidence-supported mental health and substance abuse treatment. As Drake, Osher, and Wallach (1991) pointed out “the distinction between providing appropriate living environments and mental health treatments emerges throughout” (p. 1149). The instrumental functions that material needs such as housing and food provide should not overshadow the importance of emotional and behavioral health for those at risk individuals who are entangled within the criminal justice system. Developing and implementing evidence-supported programs aimed specifically at treating co-occurring mental health and substance disorders in arrested homeless individuals may be paramount in assisting them in securing and maintaining stable housing and employment after incarceration.

Finally, for this population of recently booked arrestees, these results suggest that homelessness is frequently associated with being arrested for a property offense. One possible explanation for this finding is that those who were homeless were often engaging in crime as a consequence of being homeless as other studies have found (P. J. Fischer, 1988; Snow et al., 1989; Teplin & Pruett, 1992). Past research has noted the increasingly important roles that law-enforcement and correctional agencies play in this process (P. J. Fischer & Breakey, 1991; Freudenberg, 2001; Kushel et al., 2005; McNiel et al., 2005). If the criminal justice system is more proactive in screening arrestees for mental health problems, and subsequently directing them to appropriate services, the path from mental illness to homelessness can potentially be interrupted. Initial contact with the criminal justice system might serve as a warning signal, wherein early risk factors can be identified and dealt with before individuals become ensnared in the complicated cycle of repeated homelessness and criminal justice involvement.

Overall, this study provides evidence that the relationship between mental health and homelessness, in a sample of arrested persons, is almost entirely meditated by drug use, alcohol use, and violent victimization. Future research should continue to examine the relationships between positive predictors for homelessness to assist in expanding our understanding of how to assist persons with mental illness and homeless individuals in their transition from incarceration to the community, as well as keeping them from cycling in and out of the criminal justice system. Such a focus has the potential for identifying strategies that result in cost savings, as the homeless population uses proportionately more emergency services, which cost government agencies a substantial amount of money (Folsom et al., 2005). If researchers can identify ways in which pathways can be interrupted, this will result in a more efficient social service system, as well as foster quality of life improvements for many homeless individuals by providing more targeted programming to meet their individual needs.

Footnotes

Appendix

Standardized Path Coefficients From X to Y

	Sex	White
Homelessness	0.22*	0.13*
Mental health problem	0.38*	0.27*
Drug use problem	—	0.14*
Alcohol use problem	0.10*	—
Victimization	0.24*	—
Property crime	—	—

p < .05.

Authors’ Note

The opinions expressed here are those of the authors and are not necessarily those of Maricopa County.

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: The research was funded, in part, by the Maricopa County, Arizona Managers office.

Notes

Author Biographies

Andrew M. Fox is an assistant professor in the Department of Criminal Justice and Criminology at the University of Missouri–Kansas City. His research interests include social network analysis, gangs, mental health, and communities. He received his PhD from Arizona State University.

Philip Mulvey is an assistant professor at Illinois State University in the Department of Criminal Justice Sciences. His research interests focus on the interface between the mental health and crime justice systems, the social control of marginalized populations, and criminal justice policy. He received his PhD from Arizona State University.

Charles M. Katz is Watts Family director of the Center for Violence Prevention and Community Safety and is a professor in the School of Criminology and Criminal Justice at Arizona State University. He received his PhD from the University of Nebraska at Omaha in 1997. His research interest include gangs and gang control, police organizational theory, and crime control policy.

Michael S. Shafer is a professor in the School of Social Work and director of the center for Applied Behavioral Health Policy at Arizona State University. He received his PhD from Virginia Commonwealth University in 1988. His research interests are concentrated in the implementation of evidence-supported approaches to the treatment of mental illness and substance use disorders across diverse government sectors, including child welfare, criminal justice, and health care systems.

References

Abramson

(1972). The criminalization of mentally disordered behavior: Possible side effects of a new mental health law. Hospital & Community Psychiatry, 23, 101-105.

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

Bachrach

L. L.

Talbott

J. A.

Meyerson

A. T.

(1987). The chronic psychiatric patient as a “difficult” patient: A conceptual analysis. New Directions for Mental Health Services, 33, 35-50.

Bassuk

E. L.

Rubin

Lauriat

(1984). Is homelessness a mental health problem? American Journal of Psychiatry, 141, 1546-1550.

Belcher

J. R.

(1989). On becoming homeless: A study of chronically mentally ill persons. Journal of Community Psychology, 17, 173-185.

Berwick

D. M.

Murphy

J. M.

Goldman

P. A.

Ware

J. E.

Barksy

A. J.

Weinstein

M. C.

(1991). Performance of a five-item mental health screening test. Medical Care, 29, 169-176.

Bollen

K. A.

(1989). Structural equations with latent variables. New York, NY: John Wiley.

Booth

B. M.

Sullivan

Koegel

Burnam

(2002). Vulnerability factors for homelessness associated with substance dependence in a community sample of homeless adults. American Journal of Drug and Alcohol Abuse, 28, 429-452.

Breakey

W. R.

Fischer

P. J.

Kramer

Nestadt

Romanoski

A. J.

Ross

. . .Stine

O. C.

(1989). Health and mental health problems of homeless men and women in Baltimore. Journal of American Medical Association, 262, 1352-1357.

10.

Brunette

M. F.

Mueser

K. T.

Drake

R. E.

(2004). A review of research on residential programs for people with severe mental illness and co-occurring substance use disorders. Drug and Alcohol Review, 23, 471-481.

11.

Caton

C. L. M.

Shrout

P. E.

Dominguez

Eagle

P. F.

Opler

L. A.

Cournos

(1995). Risk factors for homelessness among women with schizophrenia. American Journal of Public Health, 85, 1153-1156.

12.

Caton

C. L. M.

Shrout

P. E.

Engle

P. F.

Opler

L. A.

Felix

Dominguez

(1994). Risk factors for homelessness among schizophrenic men: A case-control study. American Journal of Public Health, 84, 265-270.

13.

Center for Mental Health Services, Center for Substance Abuse Treatment. (1998). Refocusing upstream: Interim status report of the Center for Mental Health Services and Substance Abuse Treatment, collaborative program to prevent homelessness. Washington, DC: Author.

14.

Chen

P. Y.

Popovich

P. M.

(2002). Correlation: Parametric and nonparametric measures. Thousand Oaks, CA: Sage.

15.

Dennis

D. L.

Steadman

H. J.

(1991). The criminal justice system and severely mentally ill persons: An overview. Washington, DC: U.S. Department of Health and Human Services.

16.

Desai

R. A.

Lam

Rosenheck

R. A.

(2000). Childhood risk factors for criminal justice involvement in a sample of homeless people with serious mental illness. Journal of Nervous and Mental Disease, 188, 324-332.

17.

Ditton

(1999). Mental health and treatment of inmates and probationer. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

18.

Draine

Salzer

M. S.

Culhane

D. P.

Hadley

T. R.

(2002). Role of social disadvantage in crime, joblessness, and homelessness among persons with serious mental illness. Psychiatric Services, 53, 565-573.

19.

Drake

R. E.

Osher

F. C.

Wallach

M. A.

(1991). Homelessness and dual diagnosis. American Psychologist, 46, 1149-1158.

20.

Engel

R. S.

Silver

(2001). Policing mentally disordered suspects: A reexamination of the criminalization hypothesis. Criminology, 39, 225-252.

21.

Erickson

S. K.

Rosenheck

R. A.

Trestman

R. L.

Ford

J. D.

Desai

R. A.

(2008). Risk of incarceration between cohorts of veterans with and without mental illness discharged from inpatient units. Psychiatric Services, 59, 178-183.

22.

Fazel

Khosla

Doll

Geddes

(2008). The prevalence of mental disorders among the homeless in western countries: Systematic review and meta-regression analysis. PLoS Medicine, 5, 1670-1681.

23.

Feliner

Abramsky

(2003). Ill-equipped: U.S. prisons and offenders with mental illness. New York, NY: Human Rights Watch.

24.

Fischer

P. J.

(1988). Criminal activity among the homeless: A study of arrests in Baltimore. Hospital & Community Psychiatry, 39, 46-51.

25.

Fischer

P. J.

Breakey

W. R.

(1991). The epidemiology of alcohol, drug, and mental disorders among homeless persons. American Psychologist, 46, 1115-1128.

26.

Fischer

S. N.

Shinn

Shrout

Tsemberis

(2008). Homelessness, mental illness, and criminal activity: Examining patterns over time. American Journal of Community Psychology, 42, 251-265.

27.

Folsom

D. P.

Hawthorne

Lindamer

Gilmer

Bailey

Golshan

. . . Jeste

D. V.

(2005). Prevalence and risk factors for homelessness and utilization of mental health services among 10,340 patients with serious mental illness in a large public mental health system. American Journal of Psychiatry, 162, 370-376.

28.

Freudenberg

(2001). Jails, prisons, and health of urban population a review of the impact of the correctional system on community health. Journal of Urban Health, 78, 214-235.

29.

Gelberg

Linn

Leake

(1988). Mental health, alcohol and drug abuse, and criminal history among homeless adults. American Journal of Psychiatry, 145, 191-196.

30.

Goodman

L. A.

Salyers

M. P.

Mueser

K. T.

Rosenberg

S. D.

Swartz

Essock

S. M.

. . .Swanson

(2001). Recent victimization in women and men with severe mental illness: Prevalence and correlates. Journal of Traumatic Stress, 14, 615-632.

31.

Greenberg

G. A.

Rosenheck

R. A.

(2008). Homelessness in the state and federal prison population. Criminal Behavior and Mental Health, 18, 88-103.

32.

Guy

Platt

J. J.

Zwerling

Bullock

(1985). Mental health status of prisoners in an urban jail. Criminal Justice and Behavior, 12, 29-53.

33.

Hartwell

(2003). Deviance over the life course: The case of homeless substance abusers. Substance Use & Misuse, 38, 475-502.

34.

Hiday

V. A.

Swartz

M. S.

Swanson

J. W.

Borum

Wagner

H. R.

(1999). Criminal victimization of persons with severe mental illness. Psychiatric Services, 50, 62-68.

35.

James

D. J.

Glaze

L. E.

(2006). Mental health problems of prison and jail inmates. Washington, DC: U.S. Department of Justice, Office of Justice Programs.

36.

Katz

C. M.

Webb

V. J.

Gartin

P. R.

Marshall

C. E.

(1997). The validity of self-reported marijuana and cocaine se. Journal of Criminal Justice, 25, 31-42.

37.

Kessler

R. C.

Crum

R. M.

Warner

L. A.

Nelson

C. B.

Schulenberg

Anthony

J. C.

(1997). Lifetime co-occurrence of DSM-III-R Alcohol Abuse and Dependence with other psychiatric disorders in the national comorbidity survey. Archives of General Psychiatry, 54, 313-321.

38.

Koegel

Burnam

Farr

R. K.

(1988). The prevalence of specific disorders among homeless individuals in the inner city of Los Angeles. Archives of General Psychiatry, 45, 1085-1092.

39.

Kushel

M. B.

Hahn

J. A.

Evans

J. L.

Bangsberg

D. R.

Moss

A. R.

(2005). Revolving doors: Imprisonment among the homeless and marginally housed population. American Journal of Public Health, 95, 1747-1752.

40.

Lamb

H. R.

Lamb

D. M.

(1990). Factors contributing to homelessness among the chronically and severely mentally ill. Hospital & Community Psychiatry, 41, 301-305.

41.

Lamb

H. R.

Weinberger

L. E.

(2001). Deinstitutionalization: Promise and problems. San Francisco, CA: Jossey-Boss.

42.

Maniglio

(2008). Severe mental illness and criminal victimization: A systematic review. Acta Psychiatrica Scandinavica, 119, 180-191.

43.

Martell

D. A.

Rosner

Harmon

R. B.

(1995). Base-rate estimates of criminal behavior by homeless mentally ill persons in New York City. Psychiatric Services, 46, 596-601.

44.

McElrath

Dunham

Cromwell

(1995). Validity of self-reported cocaine and opiate use among arrestees in five cities. Journal of Criminal Justice, 23, 531-540.

45.

McNiel

D. E.

Binder

R. L.

Robinson

J. C.

(2005). Incarceration associated with homelessness, mental disorder, and co-occurring substance abuse. Psychiatric Services, 56, 840-846.

46.

Muthén

(1993). Goodness of fit with categorical and other non-normal variables. In Bollen

K. A.

Long

J. S.

(Eds.), Testing structural equation models (pp. 205-243). Newbury Park, CA: Sage.

47.

Novak

K. J.

Engel

R. S.

(2005). Disentangling the influence of suspects’ demeanor and mental disorder on arrest. Policing: An International Journal of Police Strategies & Management, 28, 493-512.

48.

Padgett

Struening

E. L.

(1992). Victimization and traumatic injuries among the homeless: Association with alcohol, drug, and mental problems. American Journal of Orthopsychiatry, 62, 525-534.

49.

Pandiani

J. A.

Rosenheck

Banks

S. M.

(2003). Elevated risk of arrest for Veteran’s administration behavioral health service recipients in four Florida counties. Law and Human Behavior, 27, 289-298.

50.

Quanbeck

C. D.

Stone

D. C.

McDermott

B. E.

Boone

Scott

C. L.

Frye

M. A.

(2005). Relationship between criminal arrest and community treatment history among patients with bipolar disorder. Psychiatric Services, 56, 847-852.

51.

Skrondal

Rabe-Hesketh

(2005). Structural equation modeling: Categorical variables. In Everitt

Howell

(Eds.), Encyclopedia of Statistics in Behavioral Science (Volume 4, pp. 1905-1910). New York: Wiley.

52.

Snow

D. A.

Baker

S. G.

Anderson

(1989). Criminality and homeless men: An empirical assessment. Social Problems, 36, 532-549.

53.

Solomon

Draine

(1999). Using clinical and criminal involvement factors to explain homelessness among clients of a psychiatric probation and parole service. Psychiatric Quarterly, 70, 75-87.

54.

Spitzer

R. L.

Kroenke

Williams

J. B.

(1999). Validation and utility of a self-report version of PRIME-MD: The PHQ primary care study. Journal of American Medical Association, 282, 1737-1744.

55.

Sullivan

Burnam

Koegel

(2000). Pathways to homelessness among the mentally ill. Social Psychiatry and Psychiatric Epidemiology, 35, 444-450.

56.

Susser

Moore

Link

(1993). Risk factors for homelessness. American Journal of Epidemiology, 15, 546-556.

57.

Susser

Stuening

E. L.

Conover

(1989). Psychiatric problems in homeless men: Lifetime psychosis, substance use, and current distress in new arrivals at New York City shelters. Archives of General Psychiatry, 46, 845-850.

58.

Teplin

L. A.

(1984). Criminalizing mental disorder: The comparative arrest rate of the mentally ill. American Psychologist, 39, 794-803.

59.

Teplin

L. A.

(1985). The criminality of the mentally ill: A dangerous misconception. American Journal of Psychiatry, 142, 593-599.

60.

Teplin

L. A.

(1990). The prevalence of severe mental disorder among male urban jail detainees: Comparison with the epidemiologic catchment area program. American Journal of Public Health, 80, 663-669.

61.

Teplin

L. A.

(1991). Police handling of the mentally ill: Styles, strategies, and implications. In Steadman

H. J.

(Ed.), Effectively addressing the mental health needs of jail detainees (pp. 10-14). Seattle, WA: National Coalition for the Mentally Ill in the Criminal Justice System.

62.

Teplin

L. A.

Abram

K. M.

McClelland

G. M.

(1996). Prevalence of psychiatric disorders among incarcerated women. Archives of General Psychiatry., 53, 505-512.

63.

Teplin

L. A.

Pruett

N. S.

(1992). Police as streetcorner psychiatrist: Managing the mentally ill. International Journal of Law and Psychiatry, 15, 139-156.

64.

Torrey

E. F.

(1997). Out of the shadows: Confronting America’s mental illness crisis. New York, NY: John Wiley.

65.

Travis

Solomon

A. L.

Waul

(2001). From prison to home: Dimensions and consequences of prisoner reentry. Washington, DC: Urban Institute Press.

66.

U.S. Department of Housing and Urban Development. (2009). The 2008 annual homeless assessment report to congress. Washington, DC: Office of Community Planning and Development.

67.

Webb

V. J.

Katz

C. M.

Decker

S. H.

(2006). Assessing the validity of self-reports by gang members: Results from the Arrestee Drug Abuse Monitoring Program. Crime & Delinquency, 52, 232-252.

68.

White

M. C.

Chafetz

Collins-Bride

Nickens

(2006). History of arrest, incarceration and victimization in community-based severely mentally ill. Journal of Community Health, 31, 123-135.

69.

Zaph

P. A.

Roesch

Hart

S. D.

(1996). An examination of the relationship of homelessness to mental disorder, criminal behavior, and health care in a pretrial jail population. Canadian Journal of Psychiatry, 41, 435-440.