This dissertation focuses on the problem of assessing the risk of harm to children from child maltreatment involving overt acts or the omission of appropriate care on the part of their caregivers. Data collected in 1994 from three Canadian child protection agencies in Winnipeg, Manitoba; Thunder Bay, Ontario; and Ottawa, Ontario; and rater-scored, closed case files from Winnipeg, Manitoba; are analyzed to construct significant case profiles. Logistic regression analysis is used to specify which variables are relevant to predicting the severity of the current maltreatment incident from information concerning the demographic characteristics of the adult perpetrator and the child victim, the adults' maltreatment patterns, and variables which describe attitudinal and behavioral characteristics of these adults. Given the limitations which exist concerning the quality of these data, this study is not conclusive but may be useful as a means of specifying guidelines for risk assessment practices within child protection services. Parsimony, and the use of data which can be accurately assessed without the participation of the clients, are of particular importance. The results of this work may also be relevant to the problems of designing further research in this subject area.