[ Main Page of Report | Contents of Report ]
The existing body of literature related to homeless families provides substantial information on the characteristics and service needs of currently homeless mothers and their dependent children but is not robust enough to provide sufficient data with which to develop a typology of homeless families. In order to fill this knowledge gap, this project has employed a step-wise approach to seeking opportunities to collect additional information about homeless families and families at risk of homelessness that could be used in the development of a typology. The first step in the process identified existing major national and multijurisdictional surveys that might yield information through secondary data analysis. A closer analysis of data collected through the Fragile Families study further illuminated additional findings about homeless families and families at risk of homelessness. However, the data were still insufficient to fully inform a typology.
A second step included reviewing ongoing and planned surveys and developing a short battery of housing questions that could be considered for use in future surveys of low-income populations. The third and final step in the process is to identify and develop three separate approaches that Health and Human Services could consider for a future specialized data collection to fill key data gaps with respect to homeless families.
[ Go to Contents ]
Based on previous chapters and the Expert Panel meeting, three options for future research to inform the typology are proposed (see Table 7-1). First, there remains a need to understand the exits and pathways out of homelessness and subsequent residential patterns. A longitudinal, nationally representative study of first-time homeless families requesting shelter would provide critical information on multiple gaps identified.
| Key knowledge gaps | Option 1: National longitudinal study of exit patterns and shelter requests of homeless families using primary data |
Option 2: Longitudinal, cross-regional study of families utilizing homeless shelters (HMIS) |
Option 3: Testing of promising practices to use a "typology" to prevent homelessness and/or expedite exit from homelessness |
|---|---|---|---|
| Geographic diversity | Y | Y | No |
| Families over time, as they move from homelessness into other arrangements | Y | No | No |
| Factors that prevent imminent homelessness | Y | No | Y |
| Dynamics of service use | Y | Y | (Y) |
| Homeless children | Y | No | No |
| Father and father's social networks | No | No | No |
| Key subgroups | |||
| Families that fall back into homelessness despite intervention | Y | Y | Y |
| Families at risk of becoming homeless | No | No | No |
| Moderate needs families | No | (Y) | Y |
| Family separations | Y | No | No |
| Working homeless families | Y | (Y) | (Y) |
| Families in extended family networks | (Y) | No | No |
| Two-parent homeless families | Y | Y | (Y) |
| (Y) — Could potentially fill the gap. | |||
The second option is an analysis of Homeless Management Information System (HMIS) data from a national sample of communities. The analysis of universal items would provide an understanding of the demographic characteristics of families in and across different regions, while the analysis of program-specific data, if available, would permit an examination of the patterns of service use over time and their relationship to outcomes for subgroups of homeless families.
The third option would be targeted to understanding how best to prevent homelessness, with an examination of existing efforts to triage families, such as in Hennepin County's (Minnesota) Homelessness Prevention program in which they use a risk assessment to make decisions on how to prevent homelessness locally. This option, in many respects, would examine "test runs" of typologies in action in different communities.
Expert Panel members all agreed that more than one typology relevant to homeless families would be needed, depending on the purposes for developing the particular typology. After much discussion, four possible goals for a typology were summarized:
[ Go to Contents ]
The Longitudinal Study of Homeless Families is a proposed national longitudinal study of exit patterns and shelter requests of homeless families using primary data. The major research questions could include the following:
Much of the past research involving homeless families has focused on the pathways into homelessness and the characteristics of families who become homeless in comparison to poor families in general. There has not been comparable attention paid to understanding how families exit homelessness and their subsequent residential patterns. During an overall period of lean fiscal times and reduced Section 8 certificates and other forms of public housing, other factors need to be identified that both facilitate families leaving homelessness and block successful exits. Information on both factors should inform intervention efforts, as well as efforts in targeting the limited housing resources to families least able to exit homelessness on their own. Likewise, there is a need to more clearly understand factors that both protect families from, and increase risk for, future homelessness episodes.
Few studies have had a longitudinal perspective that could provide insight into the trajectories families take out of homelessness. Little is known about the types of assistance that families receive and whether they take full advantage of services or benefits for which they may be eligible in order to exit. Research has not been conducted on the extent to which having bad credit, a criminal record, multiple children, and other factors hinder a family's ability to exit a homeless situation, nor has sufficient research been conducted on the factors that influence repeat homelessness among families.
Data collected through a national longitudinal study of homeless families would help with resource allocation; understanding the needs of the population enables resource matching. Basic study design could provide data on the following:
If the study includes multiply homeless families at baseline, there will be greater understanding of repeat homelessness among families. If the study includes a sample of at-risk families, factors that prevent families from becoming homeless will be learned. If the sample is large enough to look at subgroups in regions, contextual factors will be identified that interact with individual factors and family homelessness.
Sample. The basic sample would be a random sample of families requesting shelter for the first time. Depending on resources, the sample could include oversamples of families who come from two-parent families, father-only families, and families who are working to allow greater attention to these understudied groups.
The study could be enhanced by the addition of two other cohorts: families who have previously been homeless at least once, and families who are comparably poor, but domiciled and never homeless. This latter group would need to be selected from a separate sampling frame, such as Temporary Assistance for Needy Families (TANF) rolls.
To achieve a nationally representative sample of shelter requests, a stratified, multistage cluster sample would be used. Similar to the design used in the National Survey of Homeless Assistance Providers and Clients (NSHAPC) (Burt, Aron, Douglas, Valente, Lee, and Iwen, 1999), the first stage of the proposed design would include sampling of metropolitan statistical areas (MSAs) and, for non-MSAs, Community Action Agency (CAA) catchment areas. These sampling units would be clustered according to geography, population size, and economic indicators (e.g., per capita income, percent unemployed). Random samples of MSAs and CAAs within each cluster would then be chosen. All homeless shelters within each MSA or CAA would be identified and, if the number is too large, a random sample of these programs would be chosen. If there are specific subgroups that need to be oversampled, such as two-parent families, shelters could be clustered by type of populations served. Depending on resources, either a complete census of families requesting shelters for the first time or a random sample of these families could be sampled.(15, 16)
Time Frame. Families would be contacted to participate in the study at the time of the shelter request and would be followed for at least two years, and up to five years, following the shelter request.
Primary Data Collection. Interviews with the heads of household would be conducted within two weeks of the shelter request; at the time of exit or six months into shelter; and at six- or 12-month intervals subsequent to exit for a period of two to five years. Each interview would include questions on family demographics; family background, including credit history; criminal and legal involvement; residential background (residential follow-back calendar); homeless and shelter background; family separations; service need and use information; current and past trauma, conflict, and violence; and supports available. Data collection would be conducted by local interviewers in each selected community.
Administrative Data. In addition to collecting information through interviews, information could be obtained through the use of administrative databases, particularly the Homeless Management Information System.(17)
Although more in-depth information can be obtained through individual surveys, local HMIS systems can be used to determine the following:
A major advantage of using local HMIS systems is that information can be obtained even for families that cannot be located for a given followup, reducing the amount of missing data. This can be particularly useful in tracking families that return to shelters.
Because the U.S. Department of Housing and Urban Development (HUD) requires only the submission of aggregate HMIS data, however, and has explicitly stated that there will be no Federal effort to track homeless people and their identifying information beyond the local level, access to the local HMIS data will need to be negotiated with each Continuum of Care (CoC) in the targeted sampling areas.
The advantages of a national longitudinal study of homeless families include the ability to:
The likely cost of such a study is greater than other study alternatives. There may be various strategies that could be used to limit costs, such as relying on HMIS data in all communities and including primary data collection in a subset of communities.
[ Go to Contents ]
The Homeless Management Information System (HMIS) is a longitudinal, cross-regional study of families using homeless shelters. Using the HMIS universal data elements, the following questions can be investigated:
Using the program-specific HMIS data elements:
In 2001, Congress directed HUD to provide more detailed information on the extent and nature of homelessness and on the effectiveness of programs funded by the McKinney-Vento Homeless Assistance Act. As a result of this mandate, HUD is requiring each local CoC to develop its own HMIS, a computerized data collection system on homeless individuals and families. As of 2004, there were 444 CoCs operating across the country, with more being established every year. Of these 444 CoCs, 60 percent were already implementing or expanding their HMIS systems, while only one percent were not yet considering any such data collection effort.
By requiring programs and communities to collect demographic, service, and outcome data using standardized data elements, the HMIS system provides a unique opportunity to examine homeless families across programs, providers, and communities. Analyzing HMIS data, particularly from a national sample of CoCs, can help address a number of gaps in what is known about homeless families.
In particular, by showing what services homeless families use and how these services relate to outcomes (such as the length of time a family is homeless, whether they stay out of the homeless system once they leave, and how many exit to more stable housing arrangements), the HMIS data can help allocate appropriate resources to appropriate services. Knowing which families benefit from the various types of services also can inform the development of better treatment matching efforts (e.g., matching families to the appropriate level and intensity of services required).
Using the HMIS universal data elements would help with resource allocation, as these would identify the size and composition of the population to enable resource matching.
Using the program-specific HMIS data elements would help provide data on the following:
An advantage of using HMIS data is that the information is already being collected in a number of communities around the country. One problem with the use of such administrative data, however, is that the only information available is that which is already being collected. Although HUD is encouraging CoCs to collect a wide range of information on everyone receiving homeless services, only a smaller set of items is required to be collected on every person. As a result, the knowledge gaps that an analysis of HMIS systems might address will depend upon the comprehensiveness of data collection in the specific HMIS systems examined.
The universal HMIS data elements required to be collected on everyone are as follows:
If only these basic, universal data elements are available, an analysis of HMIS databases from CoCs around the country could provide the following:
More detailed, program-specific data elements are also collected as part of the HMIS. This information must be collected on all individuals and families participating in various HUD-funded programs, including the Supportive Housing Program, Shelter Plus Care, and Housing Opportunities for Persons with AIDS (HOPWA). CoCs are encouraged to collect this information on everyone tracked in the HMIS, but since this is not mandated, the extent to which this information is available would need to be determined on a case-by-case basis. These program-specific and outcome data elements include the following:
If this more detailed information on family characteristics, service use, and outcomes can be obtained, then a study of HMIS databases could also provide the following:
Finally, it might be possible in a number of communities to link HMIS data with information from other government databases, such as public assistance or public housing data. This would provide even more information about each family that could be used both descriptively and to better understand what characteristics and services are related to exiting and staying out of homelessness.
Sample. As already noted, by a congressional mandate, HUD is requiring local communities to develop a computerized data collection system. Since 2001, HUD has been working with local jurisdictions to develop and implement the HMIS. Individual CoCs will soon be required to submit information to HUD electronically based on Federal HMIS guidelines published in July 2004. These guidelines outline a set of universal elements that every CoC will be required to collect on all persons receiving homeless services, more detailed information that needs to be collected on everyone receiving services through McKinney-Vento-funded programs, along with a set of additional, recommended data elements.
Individual CoCs will be required to annually submit only aggregate information to HUD, however. As noted earlier, HUD has made it clear that "the HMIS initiative will include no Federal effort to track homeless people and their identifying information beyond the local level." As a result, the Federal guidelines state that "any research on the nature and patterns of homelessness that uses client-level HMIS data will take place only on the basis of specific agreements between researchers and the entity that administers the HMIS."(18) Since it would not be feasible, nor necessary, for a study to coordinate with more than 400 CoCs operating across the nation, a sample of CoCs would need to be created.
To identify CoCs to approach for being in an HMIS study, a stratified, multistage cluster sample would need to be used. The CoCs would first be clustered on the basis of geography (e.g., programs in the South or Northeast), as well as possibly by community size (total population), and estimated size of the homeless population (based on prior research). One important set of criteria would also likely be the extent to which the HMIS is operational in a community, including the number of homeless service providers participating in the HMIS effort and the extent to which detailed information is being collected on everyone in the homeless assistance system. Once various clusters of CoCs have been established based on this sort of criteria, communities could be randomly selected to provide a comprehensive national sample of CoCs and, by extension, homeless families.
This sort of multistage cluster sampling procedure has already been used to select communities involved in the first Annual Homeless Assessment Report (AHAR). Although the AHAR will eventually include information from all CoCs, a sample of 80 communities was selected to provide information for the first annual report. Of these 80 communities, 18 were chosen because they have the largest homeless populations (e.g., New York City, Chicago, Los Angeles). The remaining communities were randomly selected after clustering them by their population size and region. The result is a nationally representative sample of communities.
After a sample of CoCs has been selected, each agency administering the HMIS that agreed to participate in the study would provide client-level data to be analyzed. The data submitted could include retrospective data on people and families already served, as well as periodic updates to enable researchers to track families over time.
Time Frame. The HMIS is designed to track people and families over time and record their history within the homeless service system. As a result, it would be possible to examine families from the beginning of each community's HMIS system. In order to compare results across HMIS systems, however, a common starting point would need to be established. When to set that starting point would be a function of the implementation histories of the HMIS systems in the selected communities.
Another data collection factor that would need to be taken into account, either in selecting communities or determining the starting point for data collection, is the extent of HMIS coverage. In order to be confident in the results obtained from any analyses, the Federal Government recommends that the HMIS cover at least 75 percent of the emergency and transitional housing beds in the community. Since it may have taken each CoC some time to begin collecting information on 75 percent or more of the homeless beds, the date when information can be reliably obtained from an HMIS is, therefore, likely to be later than the date when data collection initially started.
Homeless Management Information System. One advantage of using an administrative database such as the HMIS is that information is being collected on an ongoing basis. Therefore, instead of collecting data through repeated waves of interviews, as is typically done in a survey effort, HMIS data can be collapsed into any time frame desired, such as annually, quarterly, or monthly. There is less flexibility in the extent of information available on each family, or family member, from the HMIS system, however. The universal data elements, listed earlier, are the only variables that will be available on everyone in every community implementing an HMIS. Although this is not a very extensive amount of information, even these data can be used to help address some of the major research questions:
More detailed, program-specific data elements can also be collected as part of the HMIS. This information must be collected on everyone involved in various HUD-funded programs, including the Supportive Housing Program, Shelter Plus Care, and HOPWA. The CoCs are encouraged to collect this information on everyone tracked in the HMIS system but, since this is not mandated, the extent to which the information is available would need to be determined on a case-by-case basis. The availability of this more detailed information, also listed earlier, would make it possible to expand the descriptive information available on each family and to create more refined subgroups of families (e.g., families experiencing domestic violence or substance abuse). It would also be possible to examine the services that families received and explore the relationship between services and basic outcomes, such as length of time in the homeless system and whether the family unit, or individual family members, fall back into homelessness over time.
Finally, there are a handful of data elements that are not required for anyone in the HMIS system but that CoCs are encouraged to collect: employment, education, health, pregnancy, more detailed veteran's data, and information on children's education participation. If this level of information is available on most people in the HMIS systems examined, then it would be possible to examine even more closely the relationships among family characteristics, services received, and various types of outcomes, such as finding a job or keeping children enrolled in school.
Other Administrative Data. Another important feature of the HMIS system is that information is collected that can be linked with other databases. Individual CoCs, for example, have been able to link their HMIS records with databases from the following:
If these linkages could be established for CoCs involved in a national study, they would provide an opportunity to examine even more about each family. Public assistance records, for example, can help show how many families were receiving services before they became homeless, how many obtained services after becoming homeless, whether public assistance came before or after exiting the homeless system, and whether receipt of public assistance is related to whether a family falls back into the homeless system.
There are a number of advantages to this option:
There are also limitations to this option.
Extent of Coverage of Providers Within a Community. Not all homeless service providers necessarily need to participate in the HMIS, and it may take a while for some CoCs to get the participation of most, if not all, providers. To the extent that the HMIS system does not cover all homeless providers, it may miss some homeless families. In particular, there may be biases in the information available because of the lack of participation by certain types of providers. Many domestic violence shelters, for example, have expressed concerns regarding security and client privacy within the HMIS.
Extent of Coverage of Families. The HMIS is limited to providing information on families that receive services from homeless service providers. While it is likely to include most, if not all, families who live in shelters, the HMIS could miss families living in motels, living on the streets, or those who are doubled-up.
Variation in Data Quality. The Federal guidelines provide sites with a great deal of flexibility in how data are collected, including interviews with clients, interviews with staff, review of staff notes, and the like. In addition, many complex variables, such as disability or mental health status, are only grossly measured (Yes/No) and may or may not be based on solid, clinical information. The data also provide little indication of the level of services needed. Finally, the degree to which complete information is available on every person would need to be assessed on a case-by-case basis.
Data May not be Readily Available. As noted earlier, any study that relies on HMIS data would need to negotiate with each individual CoC for access to client-level data. Obtaining approval from multiple CoCs could well be a very cumbersome process and there is no guarantee that any selected CoC will agree to participate in a study. Providing adequate time and resources to establish a good working relationship with any selected CoC is thus likely to be an important aspect of any study involving HMIS records. Furthermore, there is likely going to be a tradeoff in the number of CoCs from which data can be obtained and the depth of information that can be collected. The most detailed studies, those that take advantage of both rich HMIS databases and the ability to link to other databases, can probably be conducted in only a handful of sites at one time, limiting the national representativeness of the study. Conversely, studies that try to use the large number of CoCs operating or developing will likely need to be satisfied with using only the more basic, universal data elements.
[ Go to Contents ]
This option tests promising practices to use a typology to prevent homelessness and/or expedite exit from homelessness. The following questions can be investigated:
One goal for a typology of homeless families would be to identify families' risks for homelessness and barriers to housing in order to address the issues prior to entering shelter so that the incidence of homelessness among families could be reduced. In particular, a prevention-oriented typology would provide the ability to rank families according to levels of risk of homelessness and the probability of a quick exit. Such a typology would allow for distinguishing families in desperate need from those with moderate needs.
There are two concerns with trying to identify families at risk of homelessness on a broad scale, however. First, it is likely that an identification strategy that has fewer "false positives" will be based on a complex risk profile, rather than on one or two factors. As an example, Shinn and colleagues in New York City developed a model including 20 predictors to distinguish new applicants for shelter from the public assistance caseload in 1988 and correctly identified 66 percent of shelter entrants, while targeting 10 percent of the public assistance caseload (Shinn, Weitzman, Stojanovic, Knickman, Jiminez, Duchon, James, and Krantz, 1998). Second, the incidence of homelessness, even among poor families, is still too small to make widespread screening and prevention efficient. Resources targeted to an at-risk population are likely to be spent on more families that would never become homeless, than to reach those families whose homelessness could have been prevented.
A more efficient method for identifying families at risk of homelessness and in need of prevention services might be to use a risk assessment strategy to triage families who present at the shelter door for the first time. Several communities around the country are implementing systems that are using various levels of information to try to determine who can be diverted from the shelter; perhaps with some level of resources, who can be referred elsewhere; and who may require shelter services.
In Hennepin County, Minnesota, homeless service providers have developed a classification system for treatment matching of shelter usage by assessing needs and triaging families in real time. Classification is used at a very practical level and provides a method for service providers to use when making decisions about who receives shelter. In particular, Hennepin County operates the Rapid Exit Program, an innovative program that facilitates rapid rehousing by relying on early identification and resolution of a family's or individual's "housing barriers" and provides the assistance necessary to facilitate their return to permanent housing.
A study of Hennepin County or similar systems would, in effect, provide an opportunity to validate the utility of home-grown typologies.
A basic study of a prevention practice would provide information on the following:
Basic Study Design. The basic study design would be an evaluation of one or more existing best practices at the county or state level where homeless service providers are using an empirical approach to determine need for preventive services. The goal would be to determine how effectively and appropriately the system matches services to needs. Rather than impose a classification system upon communities, this project would seek to find existing or developing systems that could be assessed and tracked over time, using HMIS or other administrative data in addition to primary data.
The first step would be to either issue a call for proposals to systems implementing such programs or to fund a scan of states and communities to identify these initiatives in place. Based on this first step, one or more best practices could be selected for examination.
The major evaluation question would be to determine how effectively the system prevents future homelessness for those diverted at the front door of the system. The study would involve examining the characteristics of each family, the resources and services available and accessed, and the residential arrangements following triaging. The outcome studied would be incidence of homelessness and the length of the homeless episode for each subgroup of families having various constellations of needs and receiving specific levels of service.
The basic study design would be descriptive, tracking families over time with respect to the interventions received and changes in family stability (including both residential stability and family composition). The HMIS data could be used if program-specific elements are included.
Alternate Study Designs. In systems where more than one preventive approach is being used, a randomized study might be possible in which families receiving the same assessment ratings would randomly receive different levels of preventive service. An alternative comparative approach would involve assessing and tracking families in a comparable community where the best practice triaging approach did not exist. Data would be compared over time on homelessness rates and service use.
Sample. The sample would be families who request shelter, are at imminent risk of homelessness, and have not been homeless in the past.
Timeframe. Families would be followed for at least two years, and up to five years, following the shelter request. Data even in the first 12 months may provide an indication of the effectiveness of the triaging in preventing at least the initial onset of homelessness.
Administrative Data. Ideally, administrative data could be accessed through the HMIS system that would provide information on the family background and demographics, service needs, past and ongoing service use, family composition and stability, and family residential arrangements.
Primary Data. Primary data collected through baseline interviews with the families could be used to supplement administrative data if needed. Followup interviews also could be included if administrative data are lacking on key domains such as family stability, residential arrangements, and service use.
The advantage to this option is the ability to examine the effectiveness of typologies in place. Limitations to this proposed approach include:
[ Go to Contents ]
12. This is relevant if the study involves a cohort of multiply homeless families in addition to first-time homeless.
13. This can be investigated only if the study is national with sufficient local samples or a set of local studies.
14. This is relevant only if the study includes a comparable sample of poor families who are at risk of homelessness.
15. In some communities, the sample would be selected from a central screening center rather than from individual shelters.
16. The sample, depending on interest, could be expanded to include all families requesting shelter, not just first-timers.
17. If the study is designed to use HMIS data, then it may make sense to use local Continuum of Care (CoC) as the primary sampling unit. CoC's could further be clustered by geography, location (e.g., rural/urban), and whether they have an operating HMIS system in order to select a final sample.
18. From Federal Register, July 30, 2004, Homeless Management Information Systems (HMIS): Data and Technical Standards of Final Notice, Docket No. FR 4848-N-02.
Top of Page | Contents
Main Page of Report | Contents
of Report
Home Pages:
Human Services Policy
(HSP)
Assistant Secretary for Planning and Evaluation
(ASPE)
U.S. Department of Health and Human Services
(HHS)