Among the most difficult decisions facing child protective service (CPS) staff members today are (1) assessing the safety of children who are at risk of maltreatment, (2) deciding what types and levels of services may be immediately needed to keep children safe, and (3) determining under what conditions children must be placed in out-of-home care for their protection.
In an effort to increase the quality of caseworker decisions and the consistency with which caseworkers make those decisions and to deal with the dual roles workers have of child protection and family preservation, at least 42 states have adopted risk assessment models [Berkowitz 1991]. The movement toward a risk assessment approach is an apparent effort by states to "encourage a philosophical shift in the focus of investigations from policing activities to family assessment activities; from case substantiation to appropriate planning; and from emphasizing past experiences to examining the future of the child" [Cicchinelli and Keller 1990: 4].
Despite this shift in philosophy, current risk assessment models engender much controversy. Wald and Wolverton  assert that wide-scale implementation of such models is premature because they are neither adequately designed nor validated. Reviews of risk assessment by Doueck et al. , McDonald and Marks , and Pecora  also identify major differences in definitions and purposes of models.
Some risk assessment systems have a component for assessing the safety of maltreated children. Safety assessments are distinguished from risk assessments because they go beyond predicting the potential of maltreatment at some point in the future to suggesting that future maltreatment may be severe and thus harmful to children at a level that may require immediate intervention. Developers of safety evaluation models gear their instruments toward helping workers make decisions about how to keep maltreated children safe at home and when children should be placed in out-of-home care. A comprehensive search of the literature yielded one unpublished review [McDonnell and Heitner 1985] and no published reviews of the similarities and differences in these models. This article seeks to synthesize safety-related research and contrast ten models for evaluating the safety of maltreated children.
According to Webster [1981: 1010], safe means being "freed from harm or risk, unhurt, and secure from threat of danger, harm, or loss"; safety is defined as "the condition of being safe from undergoing or causing hurt, injury, or loss." Thus, for a maltreated child, being safe suggests more than being free from the risk of future maltreatment--it also encompasses freedom from harm or loss. Studies regarding the safety of maltreated children describe how caseworkers make placement decisions or the reasons that maltreated children come into out-of-home care. The authors found no studies that examined the safety of maltreated children from a family preservation framework.
STUDIES EXAMINING PLACEMENT DECISION-MAKING
Previous reviews of placement decision-making research [Kadushin and Martin 1988; Stein et al. 1978; Wells 1988] suggest that researchers have asked caseworkers to rate existing cases in the agency or respond to questions about analogous cases. Taken together, this research suggests that workers use a combination of factors to make placement decisions. Although not conclusive, studies have found relationships between three types of independent variables--caseworker, client, and resource characteristics--and the dependent variable, placement decision.
Caseworker characteristics and perceptions. Several researchers have suggested that characteristics of caseworkers and their perceptions of their clients influence their final judgments and decision-making [Billingsley and Giovannoni 1972; Boehm 1962, 1968; Meyer 1972; Nagi 1977; Phillips et al. 1971; Wolock 1982]. …