The Homeless

Historical Context

Multiple historical events have been linked with current conceptualizations of homelessness, including such disparate populations as those created by the 16th-century Elizabethan Poor Law, colonization of the North American continent, and itinerant workers in the late 19th century. For example, Elizabethan Poor Laws were the first attempt to provide service for landless and homeless poverty populations. In the 19th century, discussions of homelessness often focused on itinerant workers, or “hobos.” Historically, homelessness has not necessarily been identified as a “problem.” Wright points out that various descriptions of the homeless, some as recently as the 1950s and 1960s, have romanticized the lives of hobos and migrant workers. However, starting with changes in the population from the time of deinstitutionalization in the 1960s, there is a general consensus that homelessness has emerged as a serious and increasingly important social issue, and that this issue is closely interrelated with substance use and abuse and other psychiatric illness.

It is also important to consider the conduct of research on this population from a historical perspective. Although there are no doubt exceptions, early research on homelessness (for the sake of the current discussion, operationalized as published prior to 1970) was generally ethnographic or even anecdotal in method. Seminal works, such as those by Whyte, Gans, and Liebow, focused on the complex interactions among small groups of urban dwellers. Although more recent reexamination of these works demonstrates the significance of illicit substances in the lives of these “streetcorner” groups, questions of “how many” or prevalence of these disorders were not addressed in these studies.

The 1970s and 1980s witnessed an explosion of research on homelessness, with more than 500 published articles and books listed on the subject in those two decades. Unfortunately, most of this research was also methodologically flawed, presenting population descriptions incorporating a convenience sample, services-limited research consisting of program descriptions, or nonrandomized studies comparing different interventions. It was not until the late 1980s that leaders in the field called for research to move beyond demographic descriptions to conduct more complete and methodologically sophisticated research addressing complex epidemiologic issues.

In the last two decades, numerous methodologically sound cross-sectional studies have concluded that addiction and other psychiatric disorders are disproportionately prevalent in the homeless population. Unfortunately, because of sampling-related issues emanating from varied definitions applied to the problem of homelessness, changes in the population over time, and the lack of an acceptable national sample, our subsequent discussions of homelessness and associated comorbidities represent at best an incomplete snapshot of the problem. Thus answers to the specific questions of how many (e.g., What is the prevalence of psychiatric illness in the homeless population?) vary with these methodological differences, even among studies deemed methodologically adequate for most purposes. Given this situation, we present ranges of likely prevalence estimates rather than provide specific figures of undeterminable validity.

Operationalizing Homelessness

Historically, homeless samples in research studies have often been limited to service-using populations, especially individuals using services directed to homeless populations, such as overnight shelters. General consensus, however, is that this subset captures only a segment of the homeless population that may not be representative of the larger homeless population. Perhaps the most commonly accepted definition of homelessness is that of the 1987 Stewart B. McKinney Homeless Assistance Act, which defines a homeless person as:

(1) an individual who lacks a fixed, regular, and adequate nighttime residence and (2) an individual who has a primary nighttime residence that is (a) a supervised, publicly or privately operated shelter designed to provide temporary living accommodations, (b) an institution that provides a temporary residence for individuals intended to be institutionalized, or (c) a public or private place not designed for or ordinarily used as a regular sleeping accommodation for human beings.

Understanding differences among specific definitions of homelessness requires consideration of a number of factors. These factors include what personal circumstances are considered homeless (e.g., inclusion of individuals doubled up/marginally housed versus only counting individuals literally without housing), how long one must be homeless to be included (one night vs. a longer spell), and whether one self-identifies or is identified by some external criteria as homeless. Currently, operational definitions of homelessness have focused either on individuals who are literally homelessness or those marginally housed. Definitions of literal homelessness include not only those found in shelter settings, following the definition in the McKinney Act, but also individuals sleeping on the streets and in other locations not considered appropriate housing (e.g., subways, abandoned properties). Definitions of literal homelessness vary both in duration and in the types of nonhousing locations included. Inclusion of marginally housed individuals broadens the definition of homelessness to include individuals with precarious housing situations such as those living in single room occupancy buildings and staying with others without paying rent, and has been used to provide broader estimates of the lifetime prevalence of homelessness. Most recently, researchers have identified an additional dimension to measuring homelessness with given definitions that may affect estimates of population prevalence. Eyrich-Garg and colleagues have discussed differences between subjective (self-identified) and objective (identified by others) determinations of homelessness, and have demonstrated significant differences in risk patterns among samples of heavy-drinking women identified with different methods of determining homelessness.

As discussed in the preceding text, the many definitions of homelessness emerging from variations on an array of elements comprising this concept are destined to yield inconsistent sample characteristics and prevalence estimates. Because there is no unified definition of homelessness, there can be no single gold standard for determining the status of homelessness of individuals, and, therefore, it should be understood that in the remainder of this chapter, the relevant research to be reviewed necessarily consists of work derived from samples based on a variety of nonuniform definitions of homelessness from different perspectives. Although we are careful to identify both definitions of homelessness and the resulting types of samples included in specific studies (and to present critiques of current estimates in part based on this limitation), readers are encouraged to pay attention to these issues and remain cognizant of how these choices can subtly or even dramatically influence estimates of homelessness prevalence and observed characteristics of the population being studied.

Population Size Estimates

A number of methodological and conceptual issues must be considered in answering the question, how many homeless people are there? Similar to the complexities described around defining homelessness discussed in the previous section, issues requiring explication in interpreting estimations of population size include considerations of the sampling source and measurement methods (e.g., agency-based versus epidemiologic samples, neighborhood vs. urban area vs. national samples, point vs. recent or lifetime prevalence estimates vs. incidence).

Early prevalence estimates of homeless populations consisted of cross-sectional point prevalence estimates projected from samples counted at one or more overnight shelters. In one of the more thorough studies of this type, Burt and Cohen estimated that there were 194,000 adult users of homeless shelters and soup kitchens in cities of 100,000 or more in a given week in 1987. Although basing their estimate on national shelter numbers represented a methodological improvement on previous estimates, because their estimate excluded multiple other sources of homeless people (e.g., soup kitchens, unsheltered locations), it was generally considered a substantial underestimate. Other commonly discussed population estimates (e.g., census enumeration) attempted to determine the size of homeless populations on a given night using single enumeration methods. However, we agree with an assertion made by Burt and Cohen and endorsed by many others that these single-night estimations are also likely to miss substantial proportions of the literally homeless population, and thus represent significant undercounts. Populations underrepresented include the literal homeless (particularly those sleeping in hidden locations, such as in abandoned buildings) and those housed for single nights or for short spells. For these reasons, we will not further consider single-night estimates here.

Perhaps the best of the prevalence estimates emerge from the seminal work on homelessness of Burt et al. Using data from the 1996 National Survey of Homeless Assistance Providers and Clients (a survey of a variety of providers for two, 1-month periods) and extrapolating from previous estimations, they were able to arrive at reasonable estimates of how many service-using individuals were homeless on a given day or week, and estimating the total number of homeless individuals (both accessing and not accessing services) for the same periods. Readers wishing to understand more about how these estimates were reached are invited to explore the details of the various methods and estimates provided in this work.

In examining the various estimates, the best defensible figures of homeless service users who were homeless at the time of receipt of services were approximately between 440,000 and 840,000 in a given week and between 260,000 and 460,000 on a given night (including adults and children) in the National Survey of Homeless Assistance Providers and Clients. Using their methods for estimating the proportion of individuals not using services, Burt and colleagues argued that between 1.4 and 2.1 million adults were homeless in a given year. This number is not out of line with other estimates for approximately the same time period. More recently, using multiple enumeration strategies, the “Homelessness Counts” report gave a higher estimation of around 750,000 on a given night. It is important to note that in estimating lifetime prevalence, a telephone household survey found that 6.5% of adults had experienced a spell of literal homelessness at some time in their lives, and that 3% had been homeless within the past year, numbers far greater than any of the previous estimates.

An ongoing debate in the homelessness arena is the accuracy of these population estimates over time and their applicability to current population size and generalizability across locations. In terms of current population estimates, a relative consensus holds that the size of the homeless population increased in the 1980s and that the population size has remained stable or grown since. However, as the National Alliance to End Homelessness has pointed out, consistent enumerations are lacking beyond the flawed census attempts in 1990 and 2000, and, therefore, any discussions around changing size of the population are more speculative than factual. Thus the estimates presented here, while representing best available evidence, cannot be considered precise or even necessarily accurate. In terms of generalizability of findings across locations, Culhane and colleagues have used administrative records from homeless service providers to attempt to examine population size across multiple jurisdictions. Their results, although representing the state of the art, point out once again the difficulties in estimating population size, as they find rates ranging from 0.1% to 2.1% in different cities of the overall population on a yearly basis using approaches similar to those applied in administrative records data collection.

Although discussions of overall population size over time have been inconsistent at best and lacking at worst, some persuasive evidence points to recent changes in the composition of those who are homeless. North and colleagues, using three comparable representative samples each examined a decade apart within a single urban environment, noted significant increases in substance use and mood disorders among homeless cohorts over time. Their findings suggest that the homeless population may be changing, and that some of the differences found across studies are likely attributable to changing characteristics of the population rather than simply variation created by use of different sampling strategies and study of different environments. Furthermore, they argued that observed changes in the population over time may represent unintended consequences of changes in national policy.

Chronic Versus Short Term

Efforts to understand the composition of the homeless population require examination of linked issues of duration of homelessness and number of spells of homelessness that have long received considerable attention in the homelessness literature and have focused efforts to help this population toward specific subgroups with distinct characteristics. Currently, much of the federal policy is aimed at addressing the chronic homeless population. Classifications of homelessness generally break the population into some variation of three not-always-distinct groups: crisis/first-episode, episodic, and chronic. Estimates of proportions for the chronic subgroups vary from almost half falling into the chronic category to less than one-fourth and as low as 10%. Caton and colleagues examined predictors of remaining homelessness over 18 months in a cohort of newly homeless individuals, finding that shorter duration of homeless spells was associated with employment, no history of substance treatment or incarceration, and younger age. These observed differences indicate that these subgroups are distinct, with the additional implication that they may have differing treatment needs.

A Critique

Careful readers will note that much of this discussion of the homelessness population has included repeated cautions about the role of methodological issues in shaping the findings, including the definition of homelessness, sampling methods (e.g., service-using vs. non–service-using samples), and evolution of the population over time, to name only a few. Although numerous articles, books, and governmental reports have debated each of these issues separately and together, a broad conclusion from this literature is that it collectively yields only a vague understanding of the size and composition of the homeless population. We echo numerous other writers in noting the frustrations and complexities of integrating a large, methodologically flawed body of information that has been unable to describe this multifaceted population coherently or precisely. Much more could be written, including similar discussions of proportions of the population falling into various demographic subgroupings, but all would be marred by this same general critique. Given the focus of this chapter on substance use disorders and associated psychiatric and medical risk factors, we now move away from this general discussion of the homeless population to the central task of examining substance use disorders and comorbidities.

Association Between Substance Use Disorders and Other Risk Factors and Homelessness

Before launching a discussion of rates of risk factors in the homeless population, it is important to address the relationships of substance use disorders and other risk factors associated with homelessness. Generally, an implicit assumption in the popular literature holds that the disproportionate findings of these risk factors in the homeless population indicate that substance use and abuse/dependence and other mental illness cause these individuals to become homeless. However, evidence on onset of homelessness and psychiatric disorders has called into question this assumption.

Research on the causal nature of psychiatric disorders on homelessness has, in fact, concluded that the association between these factors is not simply unidirectional. O’Toole and colleagues found evidence for changes in alcohol and drug abuse patterns after first onset of homelessness, including escalating use for some individuals and diminished use among others. North and colleagues compared the relative timing of onsets of substance use disorders and other psychiatric disorders with first episode of homelessness and found that the proportion of homeless individuals with onset of their illnesses prior to the onset of their first episode of homelessness was similar to the proportion of a national community sample with onset of illness before an age comparable to that of the homeless sample’s age at first homelessness. Earlier assumptions of direct unidirectional causality from psychopathology, to homelessness have largely been abandoned by the experts who now argue that there are also multiple indirect effects related to having a psychiatric disorder that may not only increase individual risk for entering homelessness but also create barriers to exiting homelessness.

Prevalence Rates

When we use the term “substance use disorder,” we are referring to substance abuse or substance dependence as defined in the Diagnostic and Statistical Manual of Mental Disorders (DSM; editions III, III-R, IV, IV-TR, or 5 depending on when the research was conducted). Meeting diagnostic criteria for either alcohol use disorder or for any specific drug use disorder (e.g., cocaine use disorder, cannabis use disorder, opioid use disorder) qualifies one for a diagnosis of substance use disorder.

There is a general scientific consensus that the prevalence of substance use disorder is disproportionately high in the homeless population. According to epidemiologic studies, the lifetime prevalence of substance use disorder is estimated to be in excess of two-thirds of homeless people. Women range between 31% and 63%, and men between 71% and 75%, Systematic shelter-selected samples (e.g., 30) yield similar lifetime prevalence rates. The current (12-month) prevalence of substance use disorder is estimated to be somewhere between 38% and 52%.

One study found that substance use disorders accounted for most of all lifetime psychiatric disorders among a representative sample of homeless people. A lifetime psychiatric diagnosis was detected in 88% of men and 69% of women, and a lifetime substance use disorder was identified in 84% of men and 58% of women.