Gender differences in legal involvement among homeless shelter users

Gender differences in legal involvement among homeless shelter users

International Journal of Law and Psychiatry 24 (2001) 583 – 593 Gender differences in legal involvement among homeless shelter users George Tolomicze...

83KB Sizes 3 Downloads 119 Views

International Journal of Law and Psychiatry 24 (2001) 583 – 593

Gender differences in legal involvement among homeless shelter users George Tolomiczenkoa,*, Paula Goeringb a

Assistant Professor, Department of Psychiatry, University of Toronto, Toronto, ON, Canada b Professor, Department of Psychiatry, University of Toronto, Toronto, ON, Canada

1. Introduction As the number of homeless people rises in urban centers such as Toronto (Golden, Currie, Greaves, & Latimer, 1999), different issues among the population are receiving attention. Given that a substantial proportion of unaccompanied adult shelter users are women, services and interventions for homeless persons must address gender-related differences among clients. While gender differences in mental illness and substance abuse have received some attention (Brunette & Drake, 1998; Buckner, Bassuk, & Zima, 1993; Burt & Cohen, 1989; Goering, Wasylenki, St Onge, & Paduchak, 1992), little work has been done focusing on gender differences in legal involvement among homeless adults. The growth of homelessness has fostered a growing consensus that something must be done but, to date, few priorities and strategies have been set. Since legal involvement is an important consideration among this population, an exploration of associated variables, including gender, would help prioritize intervention strategies. Previous studies have consistently underlined the high prevalence of legal involvement among large samples of persons currently homeless (Burt & Cohen, 1989; Rossi & Wright, 1987; Shlay & Rossi, 1992). Legal involvement among homeless persons fits into some formulations as a causal factor (Richman, Convit, & Martell, 1992) and into others as an effect (Caton, Wyatt, Felix, Grunberg, & Dominguez, 1993). In this study, interview data regarding level of legal involvement are presented and analyzed to test bivariate associations and to construct models of potential predictor variables with the outcome of legal involvement. In a recent paper, Phelan and Link (1999) have highlighted the pervasive influence of point-prevalence bias leading to, among other things, overestimation of ‘‘deviant

* Corresponding author. 4th Floor, Health Systems Research and Consulting Unit, CAMH, Clarke Site, 250 College Street, Toronto, ON, Canada M5T 1R8. Tel.: +1-416-535-8501 ext. 4868; fax: +1-416-979-4740. E-mail address: [email protected] (G. Tolomiczenko). 0160-2527/01/$ – see front matter D 2000 Elsevier Science Inc. All rights reserved. PII: S 0 1 6 0 - 2 5 2 7 ( 0 0 ) 0 0 0 6 8 - 6

584

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

lifestyles among homeless people.’’ The current study reduces the effect of this bias by including level of shelter use as a stratification variable in sampling unaccompanied adult shelter users. This way, frequent users, who are more easily found in shelters at any given point in time, are not disproportionately represented and infrequent users are adequately represented in the final sample.

2. Research questions Beyond the chief aim of describing the rates of legal involvement among adult shelter users, these rates are compared with rates of legal involvement in the general population. Gender differences in rates of legal involvement are also presented. Finally, these gender differences are analyzed using background, precursor, and present status variables (both singly and in combination) that are associated with legal involvement.

3. Method 3.1. Sampling strategy A preliminary survey of 561 daytime Drop-in Program users was conducted at 10 different sites, all of which provided food for participants in addition to any other program offerings. The data gathered were used to estimate the percentage who were homeless and had used shelter services during the year prior to the interview date. Among those who were 18 or older, unaccompanied, and homeless (no place to stay for more than seven nights during the past month and little likelihood of obtaining accommodation in the upcoming month), almost all (93.3%) had stayed overnight at a city shelter at least once during the prior year. This finding supported a sampling strategy that capitalized on the existence of a comprehensive database of shelter users for city shelters. The administrative database for 1995 compiled by Metro Community Services (Hostel Services Division) contains sufficient information to determine the characteristics of the entire shelter user population for that year. Since the majority of homeless adults at the daytime congregation sites surveyed reported that they had stayed overnight at one of the shelters, such a population description would reflect the majority of city’s unaccompanied adult homeless population. Sampling proportions were calculated using four criteria, namely, gender, age, type of shelter used, and level of shelter use. Gender proportions were calculated for recruitment from the youth shelters where both men and women are housed, and totals for men’s and women’s shelters were computed using the administrative database. Age was stratified into three ranges: 18–24, 25–40, and 41 and older. Based on the first shelter used during the reference year, proportions for men’s, women’s, and youths’ shelters were computed. No significant differences between individual shelters within a type (average age and total time in shelters during the year) were found. Failure to account for turnover within the homeless population can contribute to the point prevalence bias that Phelan and Link (1999) describe. Simply put, on any given day, longer-

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

585

term shelter users are overrepresented among current shelter residents. To sort subjects into three levels of service utilization, the administrative database supported an alternative to selfreported time in hostels over the prior year. Service users were sorted into one of three groups: those who used a single shelter only once during the year, those who had used a single shelter more than once during the year and those who had used two or more different shelters during the year. This method divided the total population into three comparably sized subgroups with significant between-group differences in average total time spent in shelters during 1995. Furthermore, potential subjects could be asked for more reliably recalled information for the purpose of determining level of shelter use. (‘‘Is this the only shelter you have been at during the past year?’’ If no, then ‘‘high level user.’’ If yes, ‘‘have you been here at some other time during the past year?’’ If yes, then ‘‘middle level user.’’ If no, then ‘‘low level user.’’) Interviewers approached a total of 508 different potential subjects to recruit the final sample stratified as outlined above. Of the 208 subjects who were excluded, the majority (57%) was excluded, for not fitting stratification specifications. This number was elevated due to our effort to make sure that numbers at the halfway point were also proportionately stratified to match the prior year’s population of shelter users. Of the remaining 89, 20 were excluded because they did not meet our criteria for being homeless, 21 refused to participate, 7 were unable to complete an interview in English, and 41 were either unable to focus on the task and/or were threatening toward the interviewer. Subjects approached and not interviewed were given a subway token for their time. Subjects who were interviewed were paid $10 for their time. 3.2. Interviews The project survey was divided into 10 parts, ending with a structured diagnostic interview. Subjects were given the option of taking a break before the diagnostic interview at the halfway point, and interviews averaged less than 2 h in duration. The first nine parts covered sociodemographics, income/employment history, legal issues, social support, homelessness history, adverse life events, physical health, service utilization, and childhood events. The diagnostic section was based principally on the DIS-IV (Robins, Cottler, Bucholz, & Compton, 1995 for PTSD, Depression, Bipolar Disorder, Psychotic Disorders, and Alcohol Disorders) with two sections of the UM-CIDI (Robins et al., 1988 for Childhood Conduct Disorder and Antisocial Personality Disorder; Kessler & Mroczek, 1994 for the short form of the 12-month Substance Use Short form). The latter two sections were substituted, since the format fit better with the people interviewed. Rather than asking for open endorsement of specific antisocial items, the UM-CIDI version has the subject record tally for three diagnostically relevant subsets of diagnostic criteria (behaviors engaged in prior to 18 years of age, for example). We found this immediate level of anonymity better served to collect valid information reliably by minimizing the likelihood of subjects’ withholding information or abandoning the interview. It was also felt that the administration of a full substance-use section was excessively burdensome for this population, given the high rates for these disorders and the difficulty of reconstructing timelines for each substance used in the past. The short form of the modified UM-CIDI was used because of its focus on the past 12 months and its brevity.

586

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

4. Findings 4.1. Sample demographics The majority of the sample (78%) consisted of males (N = 234). The average age was 33.0 years (S.D. = 11.4). The subjects interviewed were predominantly White (73%) with Black (15%) and aboriginal (5%) minority subjects overrepresented, as compared with the general population in Metro Toronto. Only 4% of the study sample was married or in common-law relationships: 72.3% were single, 8.7% were separated, 14.3% were divorced, and 0.7% were widowed. About 41% reported that they had children, and one-third of this group still shared custody of children under 18. Overall, the sample was poorly educated and for the most part had little or no income. About 64.3% had not completed high school, compared to 30.6% for Toronto’s general population (Statistics Canada, 1996). About 37.7% of the study sample had no income. Major sources of income listed by those with some income included welfare (20% of the entire sample), family benefits (11%), and full and part-time work (14%). Long-term residents of the shelter system are generally not entitled to benefits other than a ‘‘personal needs allowance’’ of $3.68/day. This was the only source of income for 5.3% of the sample. Only 3.3% of the sample reported panhandling for money. 4.2. Legal involvement Overall, the level of legal involvement in this sample is very high (Table 1). To compare with Ontario rates, a comparably proportioned random sample [80% male, 20% female, N = 3804, average age = 36.0 years (S.D. = 13.7)] from the Ontario Mental Health Supplement database (Boyle et al., 1996) included 15.4% who said that they had been arrested at least once since the age of 18. Among the younger shelter user sample, almost four times this rate was reported (62.7%, c2 = 405.0, df = 1, P < .001). Table 1 also shows a clear gender effect. While women do show high rates of legal involvement across the categories, the rate among men is consistently much higher. The differences are greatest at the extreme end (having served jail or prison time once or more) and in the frequency of having been held overnight in jail. The gender differences in Table 1 vary from Canada-wide statistics describing charges made against adult men and women (Statistics Canada, 1999). On a year-by-year basis between 1993 and 1997, males had almost Table 1 Level of legal involvement Males Arrested since 18? Held overnight? Convicted? Served jail or prison time once? Served time more than once?

Females

Entire sample

N

%Yes

N

%Yes

N

%Yes

170 165 150 114 89

72.6 70.5 64.1 48.7 38.0

18 13 17 8 6

27.3 19.7 25.8 12.1 9.1

188 178 167 122 95

62.7 59.3 55.7 40.7 31.7

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

587

five times the number of charges (all offenses) against them than did women. In contrast, the ratio in this sample of homeless adults for arrests was less than 3:1. 4.3. Variables associated with legal involvement Variables potentially associated with legal involvement run the gamut from family background and precursor variables (parental socioeconomic status (SES), housing problems, and physical or sexual abuse) to present status variables (current diagnosis of an affective or psychotic disorder, antisocial personality disorder, current alcohol abuse, and current substance abuse). These are summarized at a descriptive level separately by gender in Table 2. Significant gender differences are found in all three blocks of conceptually and temporally related variables. Some of these variables may be indirect predictors of legal involvement by either increasing the likelihood of homelessness episodes (having been homeless as a child runaway, for example, may influence adult coping strategies) or the duration of homeless episodes (someone with a serious substance abuse problem likely will have more difficulties finding a landlord who will, over time, accommodate him or her, for example). Other variables are more directly related to legal involvement (e.g. antisocial personality traits reflecting behavior outside of social norms). Table 2 Potential predictor variables by sex Block and variables

Men (N = 234)

Women (N = 66)

Background variables Age (average; years)* Ethnicity (minority; %)* Education (mean)a Parental SES (mean)a

33.7 23.9 3.5 2.8

32.3 39.4 3.7 2.9

Precursor variables Homeless during childhood (%)** Conduct disorder (%)*** Family poverty (%)a Other family troubles (%)a Physical abuse (%)* Sexual abuse (%)** Distressing life events (mean)a

46.6 61.1 35.5 63.7 37.6 15.8 3.3

27.3 28.8 42.4 74.2 51.5 48.5 4.5

Present status variables Social support (mean)* Substance abuse/dependence (%)a Affective disorder (%)* Antisocial personality disorder (%)*** Psychotic disorder (%)***

2.0 58.1 46.6 35.0 2.6

2.3 28.8 60.6 12.1 13.6

a

Sex difference: N.S. * Sex difference: P < .05. ** Sex difference: P < .01. *** Sex difference: P < .001.

588

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

The first of three separate blocks consisted of background variables: age in years, ethnicity (coded dichotomously as White or nonwhite), education (ordinal scale from 1 to 8), and parental SES (Hollingshead’s (1975) four-factor index using parents’ ordinal scores from 1 to 5 reflecting highest level of employment and education achieved). These variables are conceptually linked in that they are fixed descriptors that cannot be targeted for early intervention. The second block consisted of other possible precursor variables: whether or not the subject spent a week or more homeless as a child, whether criteria for childhood conduct disorder were met as reported in the diagnostic interview, family poverty, other family troubles, physical and/or sexual abuse, and number of distressing life events. Indicator variables for family poverty, other family troubles, and abuse were derived from a scale for assessing childhood risk factors for homelessness developed by Koegel et al. (1995). Family poverty was scored if any of the following situations were reported by the subject as true for themselves between ages 6 and 18: family received welfare or charity, did not have enough to eat, or could not afford lights or heat; primary wage earner was unemployed most or all of the time; or had trouble affording housing to the point of relying on others for subsidized housing. Other family troubles was scored if at least one of the following was present between the ages of 6 and 18: adult members of household had problems with alcohol or drugs; a household member was disabled for a month or more; a parent spent at least several days in jail; or physical or sexual abuse had occurred within the household (not necessarily victimizing the subject). Physical and/or sexual abuse was scored if the subject himself or herself had been abused. The number of distressing life events was the sum from a list of 15 events (fired from job, having been robbed, and excluding events associated with having been arrested to avoid collinearity with legal involvement) the subject reported as distressing and that had occurred within the year prior to his or her current homeless episode. The third and final block consisted of present status variables: level of social support (ordinal scale scored from 1 to 5 indicating how many people subject felt he or she could confide in), antisocial personality disorder (currently meeting criteria or not), current substance abuse/dependence symptoms (drug and or alcohol disorder(s)), and recent symptoms of an affective or psychotic disorder. 4.3.1. Bivariate associations Table 3 lists the three groups of potential predictor variables and the correlation of each with progressively greater levels of legal involvement. There are differences between the sexes in the pattern of relationships. For some predictors, for example, sexual abuse, number of distressing life events, and current substance use, there are stronger associations at several levels of legal involvement among women than among men. For age, however, the correlations are in the opposite direction. Older men are more likely than younger men to have legal involvement at all levels, while there is a trend in the opposite direction for women. Current problems with substance abuse are consistently associated with legal involvement at all levels for both men and women. Interestingly, women show associations between low levels of legal involvement (arrested or convicted) and childhood conduct disorder (this could, however, reflect the limitations of the subsample size). The parallel associations hold, for men, at higher levels of legal

Arrested M Background variables Age Ethnicity Education Parental SES Precursor variables Homeless during childhood Conduct disorder Family poverty Other family troubles Physical abuse Sexual abuse Distressing life events Present status variables Social support Substance abuse or dependence Affective disorder Psychotic disorder Antisocial personality

F

Held overnight

Convicted

M

M

F

Served time F

M

Served more than once F

M

F

.25*** .06 .12 .01

.20 .08 .22 .04

.33*** .08 .07 .08

.04 .17 .16 .17

.32** * .06 .12 .07

.04 .19 .21 .15

.27* .01 .10 .05

.15 .10 .30* .12

.17** .04 .18** .09

.15 .04 .22 .08

.02

.16

.02

.12

.02

.11

.07

.31*

.10

.29*

.10 .07 .14* .12 .14* .12

.33** .03 .21 .19 .36** .25 *

.14* .09 .14* .13* .15* .13*

.12 .11 .21 .18 .28* .21

.09 .05 .14* .14* .18** .10

.25* .13 .27* .23 .40** .28*

.22** .01 .11 .09 .19** .04

.19 .15 .12 .08 .19 .11

.25*** .10 .08 .08 .14* .08

.16 .15 .07 .02 .23 .11

.13 .18**

.00 .36**

.08 .29**

.01 .36**

.13* .18**

.14 .31*

.09 .26***

.20 .58***

.11 .27***

.15 .50***

.10 .04 .17**

.01 .05 .19

.13* .01 .16*

.01 .14 .18

.14* .01 .12

.03 .07 .21

.09 .10 .25***

.01 .12 .14

.06 .07 .26***

.08 .02 .12

All other values indicate N.S. For males: N = 234, and for females: N = 66. * P < .05. ** P < .01. *** P < .001.

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

Table 3 Bivariate correlations between legal involvement and potential predictors

589

590

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

involvement (held overnight and/or served time once or more). For men, antisocial personality is significantly associated with legal involvement at each level except for convictions, while no significant associations are found among the women. Finally, the strongest associations are found for both men and women between high levels of legal involvement and current substance abuse or dependence. 4.3.2. Multiple predictor models Two logistic regressions were run. The variables listed in Table 2 were entered to predict a minimal level of legal involvement as a dichotomous outcome (having been arrested at least once since the age of 18). Because of the clear gender differences across both the independent and dependant variables and in patterns of bivariate associations, analyses were conducted separately for men and women. Hierarchical entry was used retaining significant predictors within a particular block before variables from the subsequent block were entered using a stepwise-entry procedure. The final list of significant predictors is presented in Table 4. Among the larger group of men (N = 216, after listwise deletion), at least one variable from each block of potential predictors was included in the final model. In this group, the observed rate of having been arrested at least once was 71.3% (103 of the 216 had served time at least once). Being older, less educated, meeting criteria for childhood conduct disorder, a greater number of distressing life events during the year prior to the current episode of homelessness, and a pattern of current alcohol or substance abuse are associated with an arrest history. Overall, the model correctly classified 74.5% of sample (161 of 216) with a smaller false positive rate than false negative rates (23.2% falsely positive and 38.7% falsely negative). Among the women, only 56 subjects were included in the analysis after listwise deletion of cases with any missing data. The smaller subsample of women necessarily restricts the power of the multivariate analysis. Only 14 (25%) reported having been arrested as adults. Being younger and meeting criteria for childhood conduct disorder are significantly associated with an arrest history. Conduct disorder, however, was significantly intercorrelated with other Table 4 Predictors of a minimum of one arrest since age 18 Men

Women

Significant predictors

b (S.E.)

Wald Sig.

Background variables Age Education

0.071 (0.018) 0.215 (0.129)

0.001 0.096a

Precursor variables Conduct disorder No. of distressing life events

0.434 (0.254) 0.166 (0.079)

0.088a 0.036

Present status variables Substance abuse/dependence Summary statistics

0.699 (0.339) 0.039 c2 = 33.7, df = 5, P < .001

a

b (S.E.) 0.055 (0.035)

1.44 (0.529)

Wald Sig. 0.114a

0.006

c2 = 14.266, df = 2, P < .001

These variables are retained based on their within-block contribution and have significance levels at P > .05 after variables in subsequent block(s) are entered into the model.

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

591

variables in the second block, namely, sexual abuse, physical abuse, family trouble, and number of distressing life events. Any one of these other variables, therefore, would do almost as well in the model with age in predicting legal involvement. Using childhood conduct disorder and age, the model correctly classified 76.8% with a higher false positive rate (45.5%) than false negative rate (17.8%).

5. Discussion 5.1. Prevalence of legal involvement Clearly, legal involvement is an important issue among homeless unaccompanied shelter users. Rates of legal involvement are significantly higher among this group than they are for the general population. While most homeless adults are men, it is important to consider women who, representing roughly 20% of adult homeless persons, display different patterns of legal involvement. The disproportionately lower percentage of women held overnight, for example, could reflect a systematic gender bias that makes it more likely for men to be held in jail overnight under similar sets of conditions. A lack of availability of appropriate facilities for women may be a contributing factor. 5.2. Predictors of legal involvement Aside from current substance abuse, the variables most strongly associated with legal involvement at all levels tend to reflect the lasting traits and long-term impact of childhood poverty (education), safety (sexual abuse), and guidance (childhood conduct disorder). Analyses failed to show significant associations between major psychiatric disorders (affective or psychotic DSM-IV Axis I diagnoses) and legal involvement, though overall, the group showed high rates of current symptoms. Given these conditions, diversion programs designed to detect persons with significant mental health issues may help to alleviate only a portion of the burden on the court system while protecting this group from adverse conditions and outcomes associated with jailing such offenders (Porporino & Motiuk, 1994). Among men, a comparable number of offenders, however, displays Axis II traits of antisocial personality that begins in childhood as conduct disorder (for women, this link is significant with conduct disorder only, and there are insignificant negative correlations between antisocial personality disorder and three levels of legal involvement). The positive association between age and level of legal involvement among men probably reflects the increasing likelihood of an instance of legal involvement over an increasing span of time among a group of people for whom the commission of minor illegal acts can be a matter of survival. Given the significant correlation between age and number of times homeless among men (r = .18, P = .008, N.S. among women), the older men have generally been without the financial and social resources to find and maintain housing longer. It is likely that the negative association between age and legal involvement among women reflects a cohort effect. The younger women reported more current problems with substance abuse, which in turn is associated with higher levels of legal involvement [dividing the females into

592

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

two groups based on substance abuse, those reporting no abuse or dependence symptoms in the past year (N = 19) averaged 26.8 years of age, and the 47 who did not averaged 34.5 years old (t = 2.18, df = 64, P = .03)]. 5.3. Implications Some of the critical issues in serving people who are homeless and mentally ill as outlined by Rowe, Hoge, and Fisk (1996) are relevant for those who are homeless and have significant involvement with the law. Why serve this group? A population-based prevention approach would seek to stem the rising tide of people arrested and jailed while without housing. A cost–benefit perspective would have to be broad enough to see the direct and indirect impacts of this group on a community’s productive output. How do we serve this group? The gender differences outlined in this article suggest interventions applicable to specific target populations. Our findings indicate, for instance, that specialized substance-abuse treatment facilities for younger women would be a good place to start. Nontraditional treatment approaches, as Rowe et al. describe them, would also help to extend services to include more homeless persons. Interagency and interdisciplinary collaborations would need to be fostered to unite and coordinate the efforts of legal, medical, and social welfare workers in working with this population. Finally, the last two critical issues described by Rowe et al., that is, addressing racial issues and including formerly homeless persons among staff, should also be considered when planning interventions and formulating policy. The high rates of legal involvement also must inform efforts to find housing solutions. If, as a society, we value the provision of basic needs for all, the commitment in terms of social support programs and structural components (affordable housing stock) will have to grow, given the latest Statistics Canada estimates of the rising numbers of children who live below poverty lines. To date, however, mounting evidence strongly suggests that Canada’s federal and provincial levels of government are not willing to direct resources toward supporting this societal value, and that, within communities across Canada, we have learned to tolerate levels of unmet need that are far too high.

References Boyle, M. H., Offord, D. R., Campbell, D., Catlin, G., Goering, P., Lin, E., & Racine, Y. A. (1996). Mental health supplement to the Ontario health survey: methodology. Canadian Journal of Psychiatry, 41, 549 – 558. Brunette, M., & Drake, R. E. (1998). Gender differences in homeless persons with schizophrenia and substance abuse. Community Mental Health Journal, 34, 627 – 642. Buckner, J. C., Bassuk, E. L., & Zima, B. T. (1993). Mental health issues affecting homeless women: implications for intervention. American Journal of Orthopsychiatry, 63, 385 – 399 (Review). Burt, M. R., & Cohen, B. E. (1989). Differences among homeless single women, women with children, and single men. Social Problems, 36, 508 – 524. Caton, C. L., Wyatt, R. J., Felix, A., Grunberg, J., & Dominguez, B. (1993). Follow-up of chronically homeless mentally ill men. American Journal of Psychiatry, 150, 1639 – 1642. Goering, P. N., Wasylenki, D. A., St Onge, M., & Paduchak, D. (1992). Gender differences among clients of a case management program for the homeless. Hospital and Community Psychiatry, 43, 160 – 165. Golden, A., Currie, W. H., Greaves, E., & Latimer, J. E. (1999). Taking responsibility for homelessness: an action

G. Tolomiczenko, P. Goering / International Journal of Law and Psychiatry 24 (2001) 583–593

593

plan for Toronto. Report of the Mayor’s homelessness action task force. Toronto, Ontario, Canada: City of Toronto (ON). Hollingshead, A. B. (1975). Four factor index of socioeconomic status. New Haven, CT: Department of Sociology, Yale University. Kessler, R. C., & Mroczek, D. (1994). UM-CIDI Mental health screening scales. Ann Arbor, MI: Institute for Social Research, University of Michigan. Koegel, P., Melamid, E., & Burnam, A. (1995). Childhood factors for homelessness among homeless adults. American Journal of Public Health, 85 (12), 1642 – 1649. Phelan, J. C., & Link, B. G. (1999). Who are ‘‘the homeless’’? Reconsidering the stability and composition of the homeless population. American Journal of Public Health, 89, 1334 – 1338. Porporino, F. J., & Motiuk, L. L. (1994). The prison careers of offenders with mental disorders. Corrections Service Canada, Report R-33. Ottawa, ON: Federal Government Press. Richman, B. J., Convit, A., & Martell, D. A. (1992). Homelessness and the mentally ill offender. Journal of Forensic Sciences, 37, 932 – 937. Robins, L., Cottler, L., Bucholz, K., & Compton, W. (1995). Diagnostic interview schedule for DSM-IV (DIS-IV). St. Louis, MO: University of Washington, Department of Psychiatry. Robins, L. N., Wing, J., Wittchen, H. U., Helzer, J. E., Babor, T. F., Burke, J., Farmer, A., Jablenski, A., Pickens, R., & Regier, D. A. (1988). The composite international diagnostic interview: an epidemiologic instrument suitable for use in conjunction with different diagnostic systems and in different cultures. Archives of General Psychiatry, 45, 1069 – 1077. Rossi, P. H., & Wright, J. D. (1987). The determinants of homelessness. Health Affairs, 6, 19 – 32. Rowe, M., Hoge, M. A., & Fisk, D. (1996). Critical issues in serving people who are homeless and mentally ill. Administration and Policy in Mental Health, 23, 555 – 565. Shlay, A. B., & Rossi, P. H. (1992). Social science research and contemporary studies of homelessness. Annual Review of Sociology, 18, 129 – 160. Statistics Canada (1996). http://ceps.statcan.ca/english/profil.