Overall, while the HADS reviewed for this report provide some useful measures of program inputs and process, they do not provide a set of measures of program outcomes or performance (with the possible exception of length of stay) that are readily adaptable to the DHHS homeless-serving programs that are the focus of our overall study. There are, however, some interesting implications that can be drawn from HADS for developing performance measures for DHHS homeless-serving programs and the systems capable of maintaining data that might be collected as part of such systems.
With regard to measures of homelessness several of the systems we reviewed do collect data on duration of episodes of receipt of homeless services (i.e., length of stay in emergency shelters and transitional facilities). Such a measure is particularly helpful in understanding frequency and total duration of homeless individuals receipt of assistance (e.g., duration of each spell of use of emergency shelters). Such data would be particularly helpful in understanding the extent of chronic homelessness and types of individuals most likely to have frequent and lengthy stays in emergency or transitional facilities. This points to the need to collect client-level data on service utilization, which includes dates that services begin and end so that it is possible to examine duration and intensity of services received, as well as multiple patterns of service use (i.e., multiple episodes of shelter use). The HADS also show that it is possible to collect detailed background characteristics on homeless (and other disadvantaged) individuals served, and especially in the case of Hawaiis HADS, to collect data at the time of entry and exit from homeless-serving programs to support pre/post analysis of participant outcomes.
The HADS systems also clearly demonstrate that it is possible to collect data on a core set of data items on homeless individuals receiving services from a substantial number of local homeless and other human services agencies. The systems demonstrate that it is possible to amass such data over a considerable period of time (15 years and longer) for a substantial number of homeless individuals (e.g., hundreds of thousands). In addition, the systems also demonstrate that very large networks of partnering agencies (in excess of 200 agencies) can collaborate on the development and implementation of data systems to track homeless and other types of disadvantaged individuals. Rapid technological advances in recent years particularly the ability to input and retrieve data at remote service locations have facilitated the expansion of such systems and made it possible for a wide variety of human service agencies (in some instances offering substantially different types of services) to share data on the same group of homeless and other disadvantaged individuals. Such sharing of data helps to facilitate inter-agency referrals, can contribute to reduction in duplication of services (e.g., reducing the chance that the same individual may receive utility or food vouchers for the same period from two different local agencies), help agencies to track homeless individuals/families over an extended period, and facilitate reporting of program service levels and results to state and federal agencies. Hence, while demonstrating the feasibility of implementing large systems to collect data on homeless individuals and linking substantial numbers of partnering agencies to collect such data, the HADS that we have reviewed for this study do not suggest a comprehensive list of performance measures that could be applied to DHHS homeless-serving programs. However, if a set of common measures were developed, the implementation experiences of the HADS would be helpful in terms of the lessons suggested for successful implementation of automated systems to maintain such data.
(12) The total of 30,000 homeless individuals includes single homeless adults in emergency shelters and families (composed of both adults and children) in temporary shelters.
(13) Randal Kuhn and Dennis P. Culhane, Applying Cluster Analysis to Test a Typology of Homelessness by Pattern of Shelter Utilization: Results form the Analysis of Administrative Data, American Journal of Community Psychology, Vol. 26, No. 2, 1998.
(14) Using cluster analysis on a sample of 73,263 total homeless individuals, Kuhn and Culhane found that the chronic cluster represented clients with a lone episode to six episodes with stay lengths from 371 to 1095 days over a three-year period; the episodic cluster represented clients with 3 to 14 stays over a three-year period, with stay length ranging from 1 to 895 days; and the transitional cluster included all others in the sample (with fewer spells and/or durations of spell length) that did not fall into the other two clusters.