India currently has an infant mortality rate (IMR) of 73 and aims to reach 60 per 1000 live births by 2000 A.D. The "at risk" approach which has been traditionally used for Maternal and Child Health services could help to reduce costs. The main objective of the study was to identity socio-demographic "risk factors" at family level for underfive deaths and assess the validity and efficiency of a risk index scale for this purpose. A computerised database on about 71,000 individuals in 28 villages in Ballabgarh Block exists since 1987. All the underfive deaths in the study area during the period 1991-95 were compared with age and sex matched controls on socio-demographic variables. All variables which were found significant at 10% level were taken in for logistic regression. The variables found significant were used to construct a ten point scale. This scaling system was applied to all the families with an underfive child during the two year period 1996-97. Validity and efficiency of this approach was calculated. A total of 849 cases and their age and sex matched controls were studied. The variables which were significantly associated with risk of underfive deaths were: not received measles vaccine (2.19; 1.58-3.04), history of sibling death (2.03; 1.19-3.45), maternal illiteracy (1.86; 1.23-2.81), never used a contraceptive (1.59; 1.17-2.14), having more than 4 children (1.46; 1.04-2.05). About 40% of houses were labelled as high risk. The sensitivity and specificity were around 60%. The improvement in efficiency by the risk approach was 33%. Risk approach is helpful in identifying families who are at greater risk of having underfive deaths. It results in a modest increase in the efficiency of services.