Practitioners and social scientists alike realize that not all members of a target population-though they may share some characteristics (for example, low-income unemployed fathers)-have the same service needs. The underlying premise of service-user typologies is that groups of individuals have different constellations of service needs that require a different package and/or intensity of services.
In identifying subgroups, service-user typology research examines variables reflective of service needs. Variables used to create service-user typologies typically include:
· Demographic characteristics reflective of service needs, for example, lack of high school diploma (Cheng et al. 2003; Freedman et al. 2000; McGroder et al. 2003; Rog et al. 1995)
· Past service utilization, such as number of stays in a homeless shelter (Kuhn et al. 1998), frequency of doctor visits (Leopold 1974), and time on welfare (McGroder 2003; Cheng et al. 2003)
· Service needs and challenges, such as unemployment and underemployment (Cheng et al. 2003; Freedman et al. 2000; Rog et al. 1995), physical and mental health (Leopold 1974; McGroder et al. 2003; Rog et al. 1995), family challenges (Leopold 1974), and issues with substance abuse, suicide, and domestic violence (Rog et al. 1995)
· Logistical barriers to service receipt, such as lack of transportation or child care (McGroder et al. 2003)
· Attitudes and beliefs that may affect the likelihood an individual participates in and/or benefits from services, including parenting beliefs (Hagelskamp et al. 2011)
Service-user typologies have used both the additive risk approach and the interactive approach to defining subgroups. For example, Leopold (1974) adopted the additive risk approach to classify poor, urban families in Philadelphia into low- or high-need groups based on their responses to a survey of about seven aspects of family functioning: (1) physical health status; (2) mental health status; (3) home conditions and household practices; (4) economic status and practices; (5) use of community resources; (6) social and community involvement; and (7) family unity and child care practices. Sample families scored between 1 and 45, with lower scores indicating fewer and less serious problems and high scores reflecting several and more serious family problems. The author selected the 27 families with the lowest scores (1 through 10) and labeled them "low need" and selected 27 families with the highest scores (36 through 45) and labeled them "high need." She then compared use of health services provided through a neighborhood community program for the 54 families and found significant differences between the two family types on a number of health care utilization indicators such as completed health examinations, follow-through on referrals, and appointments kept.
Kuhn and colleagues (1998) adopted an interactive approach to verify the existence of three types of homeless shelter users typically defined in the homelessness literature based on anecdotal observations: transitional, episodic, and chronic. As is common with the interactive approach, these researchers used cluster analysis, an exploratory data-driven analytic method that identifies naturally occurring subgroups in a sample by grouping individuals who are most similar to each other (and most different from those in other clusters) based on their configuration of scores on numerous dimensions. Specifically, Kuhn and colleagues used public housing data from New York and Philadelphia to examine the number of times (episodes separated by 30 days) a user stayed in a public shelter and the number of days per episode. The cluster analyses confirmed the existence of the three hypothesized clusters: transitional (fewest episodes and fewest days of homelessness), episodic (most episodes and a moderate number of days of homelessness), and chronic (few episodes but the greatest number of days of homelessness). The Transitionally Homeless accounted for 80 percent of shelter users and tended to be younger, White, and have few mental health, substance abuse, or medical problems. The Episodically Homeless, comprising 10 percent of the sample, were also relatively young but more likely to be non-White and to have mental health, substance abuse, and medical problems. The Chronically Homeless, accounting for half of all shelter days despite the fact that they comprised only 10 percent of shelter users, tended to be older, non-White, and have the most mental health, substance abuse, and medical problems.
A more detailed example illustrating the added insights and potential utility of a cluster-analytic approach to service user typology research can be found in Appendix D.