The President's New Freedom Commission on Mental Health identified as national priorities concerns about the involvement of people with mental disorders in the criminal justice system and about homelessness among those with mental disorders (1). The commission also stressed the need for evidence-based practice in its national agenda for mental health services.
There has been little research on the relationship between homelessness and the criminalization of people with mental disorders. It has been reported that the co-occurrence of homelessness and mental disorder increases the risk of violent criminal activity (2) and that people who are homeless and who have mental disorders and histories of violence tend to not voluntarily use outpatient mental health treatment or emergency shelters (3). Others have expressed concern that people with mental disorders who become homeless are often inappropriately criminalized and that many people who have these problems and who become dangerous should receive services in treatment settings (4,5,6).
The development of informed policies to reduce the involvement of people with mental disorders in the criminal justice system could benefit from data about the current scope of the problem, including the potentially important issue of homelessness. Given the public safety concerns that are often involved in policy debates about these issues, data on rates of violence among people with mental disorders who are incarcerated are also needed. The purpose of the study reported here was to address these issues by describing patterns of incarceration associated with homelessness and mental disorder in the jail system in San Francisco, California.
San Francisco has an estimated adult population of about 664,000 (7). Precise measurement of the number of homeless people in San Francisco is difficult, and estimates have ranged from 5,000 to 14,000 (8,9,10). The Mayor's Office on Housing estimated that on any given night in 2000 there were 12,500 homeless people in San Francisco, of whom 9,375 were individuals and 3,125 were in families with children (8).
This study retrospectively examined data for all episodes of incarceration in the San Francisco County Jail during the first six months of 2000. We assessed rates of homelessness, mental disorder, and violence; relationships between homelessness, mental disorder, and violence; and the relationship between these variables and the number and duration of episodes of incarceration. To our knowledge, this is the first study to evaluate these issues on the basis of the entire population of persons entering the jail system of a large urban area.
The study involved review of computerized administrative databases from the San Francisco County Jail system. The protocol was approved by the committee on human research of the University of California, San Francisco. Because the study involved retrospective analysis of data sets that did not identify individuals, the committee affirmed that informed consent was not necessary.
We studied all people aged 18 years or older who entered the San Francisco County Jail system between January 1 and June 30, 2000. The San Francisco Sheriff's Department operates five main jails as well as an intake and release facility and a ward at the county hospital. We extracted data on criminal charges, demographic characteristics, and use of jail psychiatric services from the agency's information systems. During the six months of the study, about 12,934 people were booked into jail. There were approximately 18,335 episodes of incarceration, because some individuals were booked into jail more than once during the study. The median number of days in jail per episode of incarceration was 4.0, with an interquartile range of three days (25th percentile) to eight days (75th percentile). We measured the variables outlined below.
We obtained diagnostic information for all episodes of incarceration in which inmates received an evaluation by jail psychiatric services during the six months of the study. Upon entry into jail, each detainee undergoes an initial medical screening by a registered nurse, followed by a more detailed medical screening one to two hours later by another nurse. The medical screening includes standard questions about mental health problems, such as suicidal ideation, depression, use of psychiatric medications, and history of psychiatric hospitalizations, as well as observations of abnormal behavior. Inmates whom the nurses identify as having a possible psychiatric problem are referred for evaluation by mental health staff, including psychiatrists, psychologists, social workers, and marriage and family therapists as well as registered trainees in these disciplines who work under the supervision of a licensed mental health professional. Inmates may also be referred for a mental health evaluation later in their incarceration by attorneys, judges, or jail staff or by self-referral. The mental health evaluations include a review of recent and past psychiatric history, substance use history, medical history and use of psychiatric medications, recent psychosocial stresses, a formal mental status examination, review of jail health records and use of the community mental health system, and contact with community health providers if available. Psychiatric diagnoses are based on the signs and symptoms specified in DSM-IV.
Because an individual can have more than one psychiatric diagnosis, we describe whether each of the diagnostic categories listed in the results section of this article was present. Because the focus of our study was on severe mental illness, our primary data analyses considered only substance-related disorders that occurred in conjunction with other mental disorders. In addition, because the role of the jail psychiatric service is to evaluate and treat mental illness, the service's screening methods would not be expected to identify inmates with a substance-related disorder that did not co-occur with a major mental disorder.
For the purposes of this study, inmates were considered to be homeless if the arresting officer, who is responsible for recording the home address of each detainee, specified that they were homeless or transient or if they gave as their residence the address of a homeless shelter. In recording the individual's address, the arresting officer considers the detainee's self-reported address and, in circumstances in which there have been previous contacts, may also consider independent knowledge of whether the individual is homeless. We determined whether the residence listed was a homeless shelter by creating a frequency distribution of addresses for all inmates and identifying addresses that were listed as homeless shelters in the San Francisco Homeless Advocacy Resource Manual (9).
Our operational definition of violence followed the approach of Naples and Steadman (10), who classified criminal charges on the basis of concepts developed by the MacArthur Study of Mental Disorder and Violence (11). We defined violence as a charge involving an act of physical aggression against other people, threatening others with a lethal weapon, or a sexual assault. Using this definition, we classified charges for each episode of incarceration as violent or not violent and also categorized them as felonies or misdemeanors, giving priority to the most severe charge if an episode was associated with multiple charges.
We calculated the number of days in jail for each episode of incarceration that began during the first six months of 2000 and also assessed the number of episodes for each inmate.
We extracted descriptive information about rates of homelessness, mental disorder, and violence and evaluated the relationship between these variables by using chi square tests for categorical variables and t tests for continuous variables. We conducted a multiple regression analysis to predict number of days in jail on the basis of these variables, while statistically controlling for demographic characteristics and charge severity. These analyses were based on the episode of incarceration rather than the individual person, because this approach takes into account the fact that some individuals were jailed multiple times and may have had different characteristics—for example, homelessness—depending on the episode.
Our second series of analyses used the individual person as the unit of analysis, to compare the number of episodes of incarceration by persons who were and were not homeless during any incarceration during the study.
Analyses based on episode data
Homelessness, mental disorder, and violence.t1 shows that homeless people accounted for 16 percent of the episodes of incarceration. In 18 percent of the episodes the inmates had a diagnosis of a mental disorder and in 6 percent (1,132 of 18,335) a severe mental disorder (schizophrenia, psychotic disorder not otherwise specified, delusional disorder, major depressive disorder, or bipolar disorder). Charges of violent crimes were present in 16 percent (2,498 of 16,028) of all the episodes of incarceration for which charge data were available.
t2 shows that inmates who were homeless were significantly more likely than those who were not to receive a psychiatric diagnosis and were more likely to be given diagnoses of substance-related disorders, schizophrenia, personality disorders, and psychotic disorder not otherwise specified.
t2 shows that homeless inmates were significantly more likely than inmates who were not homeless to receive a diagnosis of a co-occurring severe mental disorder and substance-related disorder. This finding is further illustrated by analysis of the subset of episodes of incarceration by people with severe mental disorders (N=1,132); 78 percent of those with severe mental disorders who were homeless (172 of 221) also had a diagnosis of a co-occurring substance-related disorder, compared with 69 percent of those with severe mental disorders who were not homeless (629 of 911) (χ2=6.21, df=1, p<.02).
Relationships between violence, mental disorder, and homelessness. As shown in t2, homeless inmates were significantly more likely to be charged with felonies, although when the crimes were categorized according to level of violence, the homeless inmates were less likely to be charged with violent crimes. Inmates who received a psychiatric diagnosis were more likely than those who did not to be charged with felonies (χ2=165.59, df=1, p<.001) (67 percent [1,827 of 2,736] and 53 percent [7,089 of 13,292], respectively) and were more likely to be charged with violent crimes (χ2=259.06, df=1, p<.001) (26 percent [705 of 2,736] and 13 percent [1,793 of 13,292], respectively). Inmates with co-occurring substance-related disorders and severe mental disorders had higher rates of violent charges than those without dual diagnoses, regardless of whether or not they were homeless (χ2=90.30, df=3, p<.001). Of the inmates with dual diagnoses, 26 percent who were homeless (37 of 142) had violent charges, as did 24 percent (120 of 508) who were not homeless. In contrast, of the inmates who did not have a dual diagnosis, 10 percent who were homeless (245 of 2,380) had violent charges, as did 16 percent (2,096 of 12,998) who were not homeless.
Duration of incarceration.t3 summarizes the results of a multiple regression analysis predicting number of days in jail (12). In this model, homeless inmates had a length of stay that was 4.5 days longer on average than nonhomeless inmates. Similarly, inmates with a diagnosis of co-occurring severe mental disorders and substance-related disorders stayed significantly longer than other inmates, on average 16.9 days longer than inmates without these conditions. Inmates who had been charged with felonies had longer incarcerations than those who had not been charged with felonies, as did inmates charged with violent crimes compared with nonviolent crimes. In addition, inmates who were male and nonwhite tended to stay longer than inmates without these characteristics. The findings suggest that when demographic characteristics and severity of criminal charges are controlled for, inmates who are homeless and who have co-occurring severe mental disorders and substance-related disorders are incarcerated for substantially longer than their counterparts who do not have these characteristics.
Analyses based on the individual person
The preceding analyses were based on the episode of incarceration. Similar results were found in parallel analyses based on the individual person (who may have had multiple episodes of incarceration) as the unit of analysis (t1). For example, 30 percent of the individuals who were homeless during any incarceration during the study period (716 of 2,402) also had a diagnosis of a mental disorder (although not necessarily during the same episode of incarceration), compared with 19 percent of the inmates who were not homeless (2,011 of 10,532) (χ2=134.30, df=1, p<.001). Homeless inmates with a co-occurring severe mental disorder and a substance-related disorder were more likely to have multiple episodes of incarceration than inmates without these characteristics (χ2=837.70, df=3, p<.001). Of the homeless inmates, 52 percent with a dual diagnosis (96 of 184) and 44 percent without a dual diagnosis (976 of 2,218) were incarcerated multiple times. Of the inmates who were not homeless, 30 percent of those with a dual diagnosis (147 of 487) and 17 percent of those without a dual diagnosis (1,744 of 10,045) were incarcerated on multiple occasions.
The results of this study provide support for the current widespread professional concern about the large numbers of people with mental disorders who are in the criminal justice system. Approximately 18 percent of the episodes of incarceration in the San Francisco County Jail system over the six months of this study involved people who were identified as having a mental disorder. In addition, the results showed that one of every six episodes of incarceration involved a person who was homeless and that homeless inmates had elevated rates of mental disorders. The findings suggest that homelessness needs to be considered in efforts to reduce the criminalization of people with mental disorders. Although it may be that San Francisco has higher rates of homelessness than some other locations, many urban areas currently face similar problems, and the results of this study affirm that the issue is of considerable scope. Below we discuss the implications of the results for research and policy.
Co-occurring homelessness and substance use and mental disorder
The finding that homeless people with mental disorders who are in jail have a higher rate of co-occurring substance-related disorders than those who are not homeless is consistent with research in other settings (13,14). The increased comorbidity of substance-related disorders and severe mental disorders among homeless people who enter the criminal justice system represents a challenge for intervention, given evidence that such comorbidity predicts decreased treatment adherence (15). Because our sample relied on a clinical intake screening to identify mental health problems, it likely provides a conservative estimate of rates of mental disorder among incarcerated people who are homeless. It is possible that other methods, such as structured diagnostic interviews or longitudinal follow-up, would identify even higher rates (14).
Duration of incarceration
Inmates who were homeless and who had co-occurring severe mental disorders and substance-related disorders were held in jail for substantially longer than demographically similar inmates who had been charged with similar crimes. Possible reasons for the finding that homeless inmates were incarcerated longer are that they have difficulty making bail, they are less likely to be released on their own recognizance, they often have bench warrants for failure to appear in court, and they often are arrested multiple times for a similar crime. In addition, inmates with mental disorders may be incarcerated for longer for reasons related to the evaluation and treatment of their psychiatric problems, such as court-ordered treatment, waiting to be stabilized on medications before entering community treatment programs, being held in a safety cell because of behavior that is dangerous to themselves or others, waiting for openings in community treatment programs, or undergoing evaluation for competence to stand trial. More generally, the finding that people who were homeless and had severe mental disorders with co-occurring substance-related disorders stay longer in jail than others charged with similar crimes suggests that the jail system is de facto assuming responsibility for problems that have traditionally been subsumed by other systems, such as community mental health, substance abuse treatment, and social welfare.
The proportion of jail inmates in our study who were identified by psychiatric services staff as having severe mental illness (6 percent) is comparable to the findings in other jurisdictions (16,17,18,19). Such research has generally found that 6 percent to 15 percent of inmates in county jails have severe mental illness. This similarity between our sample and others supports the generalizability of our findings concerning rates of severe mental disorder. However, because psychiatric services were delivered during only about 18 percent of the episodes of incarceration, it is likely that the true prevalence of substance abuse among persons without severe mental illness was underestimated in our sample. Researchers who have used other methods have reported that more than half of inmates without severe mental illness meet diagnostic criteria for substance-related disorders (17,18,19).
Implications for public safety
Homelessness, when considered in isolation, was not associated with elevated rates of violence in our sample. However, it must be recognized that, because entry into jail is frequently predicated on behavioral disturbance, the comparison group of nonhomeless persons in jail also has an elevated rate of behavioral dysfunction compared with the general population. Although only a small proportion of the general adult population of San Francisco is homeless (about 1 percent [9,375 of 664,000]) (7,8), the proportion of people incarcerated on the basis of charges of violent crimes who were homeless in our study (about 11 percent [282 of 2,498]) suggests that failure to address the service needs of this population may have a modest adverse impact on public safety. In our sample, inmates with dual diagnoses were more likely than those without co-occurring disorders to be homeless and to be charged with violent crimes than other inmates. Taken together, the findings are consistent with those of a previous study in New York City that found that homeless inmates with mental disorders had elevated rates of violence compared with the general population (2) and with the results of previous research showing that co-occurring substance abuse and severe mental disorder are associated with elevated rates of violence (20).
Our measurement of violence on the basis of criminal charges has limitations, because charges are not the same as convictions, charges may not have been sensitive to violence that went unreported or led to psychiatric hospitalization, and the public nature of violence by people who are homeless may have increased its detection by police. However, criminal charges do represent the behavior alleged as a basis for incarceration.
Our findings support the widespread concern about the large numbers of people with mental disorders who are in jail (5,6). In the six months of this study, psychiatric services were delivered in a similar number of episodes of entry to the San Francisco county Jail (N=3,234) as were delivered in all episodes of entry to the San Francisco County Hospital's psychiatric emergency service (N=3,202) in an earlier six-month interval (21). Various explanations have been proposed for the large number of people with mental disorders who are being placed in jail, such as deinstitutionalization, civil commitment statutes that emphasize the right to refuse treatment, limited availability of community-based mental health services and affordable housing, and police decision-making practices (18,22,23).
Regardless of the cause, the high rate of mental disorders among jail detainees represents an important problem. To the extent that the crimes committed by these individuals are related to their untreated mental or substance-related disorders, this pattern represents a failure of the current health care delivery system. These individuals are detained in criminal justice systems that have been designed more for incapacitation and punishment than for treatment; although psychiatric services are provided in jail, the setting is suboptimal for mental health treatment. In addition, the increased rate of multiple episodes of incarceration shown by this group suggests that many individuals with mental disorders are not being stabilized under the current system but, rather, are cycling between the street and jail.
A variety of new legal mechanisms have the potential to help address the needs of this population. Mental health courts, which endeavor to mobilize treatment resources for offenders whose criminal behavior is due to untreated mental illness, are rapidly being developed across the country (24). The promise of mental health courts has engendered optimism among stakeholder groups who are concerned with criminalization of people with mental disorders, although research is only beginning to assess what components of these courts are effective (25,26). Some evidence indicates that outpatient civil commitment can reduce the criminalization of persons with mental disorders (27,28). Previous commentary has expressed concern that, without sufficient allocation of resources, such legal interventions may have limited impact (29).
The findings of our study are relevant to these concerns in that a substantial number of people with mental disorders who are high users of jail were found to be homeless. Such individuals have many problems over and above nonadherence to community psychiatric treatment, such as poverty, co-occurring substance-related problems, impoverished social networks, histories of victimization, and comorbid medical problems (30). Solutions to these complex problems will likely require allocation of substantial resources for intervening across service delivery systems, in addition to leveraging adherence with mental health treatment.
Previous research suggests that allocating resources for integrated service systems, assertive community treatment, dual diagnosis programs, community outreach, and supportive housing with on-site substance abuse services, medication, and social services can be helpful with people who are homeless and have mental disorders (31,32,33). Although such interventions may not necessarily reduce overall expenditures (34), they may yield other societal benefits, such as reduction in suffering, encouragement of recovery among people with mental disorders who are homeless, enhancement in community quality of life, and modest improvements in public safety.
A limitation of this study is its reliance on retrospective review of administrative databases. This data source may not have been sensitive to dimensions of homelessness, mental disorder, and violence that might have been measured by other methods, such as interviews with the inmates or interviews with collateral informants such as family members and service providers (35). It is possible that other methods would identify even higher rates of homelessness and mental disorder in the jail population. On the other hand, the instances of homelessness, mental illness, and violence that we identified with this method likely do represent relatively clear-cut instances that would have prompted the line staff to record them.
Our results suggest that a substantial proportion of people who were detained in jail in this study's urban setting were homeless and that such individuals had elevated rates of co-occurring severe mental disorders and substance-related disorders. Development of public policy to decrease the involvement of people with mental disorders in the criminal justice system needs to consider addressing issues associated with homelessness and substance-related problems in addition to facilitating engagement in community psychiatric treatment.
This study was supported in part by a grant from the Psychiatric Foundation of Northern California. The authors thank Jean Mariani, Richard Juster, Ph.D., Judy Lam, Ph.D., Shane Johnston, Patrick Owens, William Roth, and Maurice Hearne.
Dr. McNiel and Dr. Binder are affiliated with the department of psychiatry of the University of California, San Francisco. Ms. Robinson is with San Francisco Jail Psychiatric Services. Address correspondence to Dr. McNiel at Langley Porter Psychiatric Institute, University of California, 401 Parnassus Avenue, San Francisco, California 94143-0984 (e-mail, email@example.com). This paper was presented in part at the International, Interdisciplinary Conference on Psychology and Law, held July 9 to 12, 2003, in Edinburgh, Scotland.
Characteristics of inmates of the San Francisco County Jail at the level of the individual and by episodes of incarceration
Characteristics associated with episodes of incarceration of San Francisco County Jail inmates who were and were not homeless (N=18,335)
Summary of random-effects multiple regression analysis predicting number of days in jail on the basis of homelessness, demographic characteristics, diagnosis, and criminal chargesa
aBecause some individuals had more than one episode of incarceration, a random-effects regression model for unbalanced data was used (12), which accommodates the repeated-measures nature of these data. Error: ei,t=ui + vi,t. Covariance parameter estimate: individual (u)=726.51, residual (v)=1,265.56. Model fit: -2 residual log likelihood=158,964.7. Null model (no random effects) likelihood ratio test: χ2=103.98, df=1, p<.001. The analysis was based on the 83 percent of episodes of incarceration (15,306 of 18,335) for which complete data were available for all the variables listed in the table. Of the 17 percent of episodes with missing values (3,029 of 18,335), most had missing criminal charge data. Criminal charge data were missing for 14 percent of episodes for homeless inmates (416 of 2,938) and 12 percent of episodes for nonhomeless inmates (1,891 of 15,397).