A newborn screen designed to predict family risk was examined to: (a) determine whether all families with newborns were screened; (b) evaluate its predictive validity for identifying risk of out-of-home placement, as a proxy for maltreatment; (c) determine which items were most predictive of out-of-home placement. All infants born in Manitoba, Canada from 2000 to 2002 were followed until March 31, 2004 (N = 40,886) by linking four population-based data sets: (a) newborn screening data on biological, psychological, and social risks; (b) population registry data on demographics; (c) hospital discharge data on newborn birth records; (d) data on children entering out-of-home care. Of the study population, 18.4 percent were not screened and 3.0 percent were placed in out-of-home care at least once during the study period. Infants not screened were twice as likely to enter care compared to those screened (4.9 percent vs. 2.5 percent). Infants screening at risk were 15 times more likely to enter care than those screening "not at risk." Sensitivity and specificity of the screen were 77.6 percent and 83.3 percent, respectively. Screening efforts to identify vulnerable families missed a substantial portion of families needing support. The screening tool demonstrated moderate predictive validity for identifying children at risk of entering care in the first years of life. (Published Abstract)