Better understanding of the reasons for loss of housing among persons with mental illness could inform this debate. If these individuals lose their housing because of a distinct set of factors related to symptoms of mental illness, such as disturbed behavior and interpersonal conflict, then the specific vulnerability hypothesis would be supported. However, if the immediate reasons for homelessness among persons with mental illness are mostly similar to those among non-mentally ill homeless persons, then the general vulnerability hypothesis would be supported.
The NSHAPC methods have been described in detail elsewhere (
+15,
+16,
+17,
+18). Briefly, a nationally representative sample of more than 4,200 persons who used homeless services in 52 metropolitan and 24 nonmetropolitan areas in the United States between October 18 and November 14, 1996, was randomly selected and surveyed; the response rate was 96 percent. Services included emergency shelters, transitional housing programs, permanent housing programs for homeless persons, migrant workers' camps, centers for distribution of vouchers for shelter, soup kitchens, food pantries, drop-in centers, mobile food programs, and street outreach programs. Six to eight clients were interviewed in person during each of approximately 700 program visits. Semistructured interviews were conducted by trained Census Bureau field-workers. Each interview lasted about 45 minutes, and participants were paid $10 for their participation. NSHAPC was funded by the U.S. Census Bureau.
This article focuses on 2,974 currently homeless NSHAPC participants. Participants were rated as currently homeless if they reported staying in an emergency shelter, a transitional housing program, a hotel or motel paid for by a shelter voucher, an abandoned building, a place of business, a car or another vehicle, or any other nonresidential space on the day of the survey or during the seven-day period before being interviewed; reported that the last time they had a place of their own for at least 30 days in the same place was more than seven days ago; said that their last period of homelessness ended within the previous seven days; were identified for inclusion in the NSHAPC client survey at an emergency shelter or a transitional housing program, or at a voucher distribution program, but only if there was at least one other indicator of current homelessness; reported getting food from the shelter where they lived within the previous seven days; or reported staying in their own or someone else's place on the day of the interview but said they "could not sleep there for the next month without being asked to leave."
Mental illness. Accurate diagnosis of mental illness in surveys of homeless populations is difficult (
+19,
+20,
+21). Following the suggestion of Susser and colleagues (
+20), two methods for defining mental illness were used in this study: a broad definition based on meeting the NSHAPC criteria for lifetime mental illness, used in previous NSHAPC reports (
+16,
+18), and a narrow definition that requires meeting the NSHAPC criteria for lifetime mental illness and a history of psychiatric hospitalization. NSHAPC criteria include scoring at least .25 on the psychiatric problems domain of the Addiction Severity Index (ASI) (
+22), ever having been treated for psychiatric problems, ever having taken prescribed medications for psychiatric problems, having any history of treatment and at least one ASI psychiatric problem, and having stayed in a psychiatric hospital or a group home for persons with mental illness. Meeting any of these criteria qualifies for the designation of mentally ill in the NSHAPC.
The NSHAPC criteria also include self-report of "a mental health condition" as the most important factor contributing to the individual's continued homelessness. However, because this study examined perceived reasons for continued homelessness among mentally ill persons, this criterion was not used for defining mental illness.
The ASI is a semistructured interview for assessing problems in seven domains—psychiatric, medical, legal, employment, family and social, illicit drug abuse, and alcohol abuse. Ratings on psychiatric problems, drug abuse, and alcohol abuse were used in this study. The psychometric properties of ASI ratings in various populations have been extensively studied and were generally found to be in the acceptable range (
+22,
+23,
+24,
+25,
+26,
+27,
+28,
+29). For example, in a sample of homeless persons, the alpha coefficients of internal consistency for ratings of psychiatric problems, alcohol abuse, and drug abuse were .89, .87, and .70, respectively (
+28). Also, ratings on these domains correlated strongly and specifically with other measures of each domain. For example, for psychiatric problems, the correlation coefficients with the Beck Depression Inventory (
+30) exceeded .50 and with the 90-Item Symptom Checklist (SCL-90) (
+31) exceeded .60.
Alcohol and drug abuse. Alcohol and drug abuse were rated on the basis of the NSHAPC criteria. Past-month alcohol abuse is ascertained by the NSHAPC when any of the following criteria are met: a score of at least .17 on the ASI alcohol abuse domain, treatment for alcohol abuse during the previous month, getting drunk at least three times a week during the previous month, and having any history of treatment and drinking at least three times a week during the previous month. Past-year and lifetime alcohol abuse were defined similarly by extending the time qualifiers. Persons who reported at least three alcohol-related difficulties during their lifetime from a list of eight difficulties adopted from the Michigan Alcoholism Screening Test (MAST) (
+32) were also categorized as having a lifetime alcohol problem.
Reasons for loss of housing and continued homelessness among currently homeless participants in the National Survey of Homeless Assistance Providers and Clients
Reason for most recent loss of housing
Lost your job or the job ended
Rent increased and you couldn't afford to pay it
Someone who paid the rent or mortgage stopped paying it
Lost welfare or other cash assistance benefit
Not enough money for habitation
You or your children were abused or beaten; violence in the household
Didn't get along with people there
The people you were staying with asked you to leave
The landlord made you leave
Moved out because of a problematic relationship or the end of a relationship with a partner or relative
Had roommate or landlord problems
Went into a hospital or a treatment program
HIV, AIDS, or AIDS-related complex (ARC)
Was pregnant or had just had a baby
Became sick or disabled (other than ARC or AIDS related)
Had a problem with the residence or the area in which the residence is located
Was displaced because the building was condemned, destroyed, or subject to urban renewal
The lease expired or the building was sold
Was looking for work or was forced to relocate to keep your current job
Had problems abiding by the rules of the current residence or program time ran out
Was released, dismissed, or discharged
Moved in with a significant other, a friend, or a relative
Death or illness in the family
Needed a change of scenery or climate
Reason for continued homelessness
Drug or alcohol addiction
Insufficient education, skills, or training
Physical condition or disability
Family or domestic instability
Insufficient services or lack of information about available services
Past-month drug abuse is ascertained by the NSHAPC when any of the following criteria are met: a score of at least .10 on the ASI drug abuse domain, treatment for drug abuse in the previous month, current intravenous drug use, and use of drugs at least three times a week during the previous month. Past-year and lifetime drug abuse were defined similarly by extending the time qualifiers. Individuals who reported having at least three drug-related difficulties during their lifetime from a list of eight that were adopted from the Drug Abuse Screening Test (DAST) (
+33) were also rated as having a lifetime drug problem.
Medical conditions. Ratings of medical conditions were based on self-reports of diabetes, anemia, hypertension, heart disease or stroke, liver disease, arthritis, cancer, and HIV infection or AIDS.
Perceived reasons for loss of housing and homelessness. The study participants were asked about 32 reasons for leaving the last place of their own, such as a house, apartment, room, or other housing where they stayed for at least 30 days. They were then asked to identify which was the main reason. For this study, the responses were categorized by the author into five broad categories, described in the box on this page: financial problems, interpersonal problems, health-related reasons, drug and alcohol abuse, and miscellaneous reasons. Categorization of these items by a second mental health researcher who was blinded to the aims of the study and the author's original categorization produced very similar results (88 percent agreement, kappa=.83, p<.001). Respondents were also asked to select the most important reason for their continued homelessness from a list of ten reasons.
Two sets of analyses were conducted. First, sociodemographic and clinical characteristics of the study participants with mental illness and those without mental illness were compared. Second, the perceived reasons for the most recent loss of housing and continued homelessness were compared among these two groups by using a series of binary logistic regressions in which mental illness was the dependent variable and the reasons for loss of housing and continued homelessness were the independent variables. Each category of reasons for loss of housing was rated dichotomously: 1 if a reason in that category was reported by the participant and 0 if not. Analyses adjusted for sociodemographic and clinical characteristics that had been found to be significantly different between groups in the first set of analyses.
Statistically significant logistic regression coefficients associated with any of the reasons for the most recent loss of housing or for continued homelessness would indicate significant differences between the study participants who were mentally ill and those who were not mentally ill on that reason or category of reasons. If individuals with mental illness lose their housing or remain homeless for a distinct set of reasons related to symptoms of mental illness, such as disturbed behavior and interpersonal conflict, the regression coefficients associated with these reasons would be statistically significant. In contrast, if the reasons for loss of housing and continued homelessness are similar between persons who are mentally ill and those who are not, few or none of the regression coefficients would be statistically significant.
Analyses were conducted once by using the broad definition of mental illness and another time by using the narrow definition. The comparison group of persons without mental illness for both analyses comprised the study participants who did not meet the broad criteria for mental illness.
Probability weights provided by the Census Bureau were used in all analyses to make percentages nationally representative. The complex sampling design of the NSHAPC requires adjustment using design elements (strata and primary sampling units). However, because of confidentiality concerns, these data are not made available. Instead, the Census Bureau recommended adjusting standard errors by using an approximation of the design effect (square root of 3) (
+18). All statistical tests reported in this article are thus adjusted. Analyses were conducted by using the logit routine of Stata 8.0 software (
+34) using probability weights provided by the NSHAPC.
The findings of this study are constrained by a number of limitations. First, the data are based on self-report. Errors and biases in causal attribution are common with self-reported data, especially among individuals with severe mental disorders, who may lack insight into their illness and its social impact (
+35). To further explore this possibility, the perceived reasons for the most recent loss of housing were examined among participants who stated that mental illness was the main reason for their continued homelessness. Presumably, these individuals have more insight into the relationship between their illness and homelessness. However, even in this group, 28 percent reported financial problems as the major reason for the most recent loss of housing, 19 percent reported interpersonal problems, 12 percent reported health problems, and 17 percent reported alcohol and drug abuse. Nevertheless, reliance on self-report and lack of objective measures remain the major limitation of this study.
Second, the use of the same correction method for standard errors obtained in the different analyses might have inordinately reduced the power of some of the statistical tests. Thus the results of the statistical analyses should be viewed with caution. Third, the sampling frame of the NSHAPC included institutions that typically serve chronically homeless individuals. Furthermore, the NSHAPC was a prevalence survey. Persons with longer episodes of homelessness are overrepresented in prevalence samples. Thus the findings may not apply to the larger group of persons who experience only transient homelessness (
+36).
Despite these limitations, the NSHAPC data provide useful information about the chronically homeless persons who are the most vulnerable group in the homeless population. The perceived reasons for loss of housing and continued homelessness among persons with mental illness were mostly similar to the reasons reported by those without mental illness. Only a small fraction of the mentally ill individuals reported mental illness as the main reason for their continued homelessness. These findings support the general vulnerability hypothesis.
The findings are consistent with consumer surveys of homeless individuals with mental illness (
+7,
+37) in which the participants attribute their success or failure in obtaining and maintaining housing more to financial factors and housing availability than to mental illness and its treatment. The findings are also consistent with results of recent studies of risk factors of homelessness among persons with mental illness (
+12,
+38). Comparing homeless and nonhomeless individuals who had mental illness and those who did not, Sullivan and colleagues (
+12) concluded that mental illness does not represent a "distinctive pathway to homelessness," but rather an added disadvantage.
However, the findings are at odds with the results of a number of other studies (
+5,
+39) that showed a specific association between symptoms of mental illness and homelessness. These studies may have focused on individuals who had more severe illnesses, for whom mental health problems may be more significant determinants of housing instability (
+40). The divergent results of the studies may also be due to differences in raters' perspectives (
+41). Studies based on providers' assessment (
+5,
+39) often indicate a larger role for psychiatric factors than studies based on self-report (
+4,
+7,
+37).