Objectives: To describe the outcome of referral to the statutory authorities for infants under 2 years with non-accidental head injury (NAHI), and to establish whether the authorities held sufficient information to develop a risk profile for these cases. Methods: Retrospective review of cases admitted to hospital in Auckland, New Zealand from 1988 to 1998. Records from the hospital admission, child protective services and Police were reviewed, up to 19 years from diagnosis. Results: Of 39 infants, 33 survived to leave hospital. Documentation of risk factors was erratic, and sometimes incongruent between agencies. Inter-agency case conferences took place in 17/39 (44%). The Department of Child, Youth and Family Services (CYF) used an informal family agreement to secure safety in 15/33 survivors (45%). Family Group Conferences occurred in 17/33 (52%). Nine of 33 were placed permanently outside the home (27%), two (6%) with unrelated caregivers. Charges were laid in 18/39 cases (46%). Fifteen cases came to trial, with 14 convictions (36%). Of the survivors, 44% were later renotified to CYF. There was no obvious relationship between type of intervention and re-notification. Conclusions: Ensuring the safety of an infant with NAHI, and identifying and taking appropriate action with regard to the offender, are complex tasks. In New Zealand, data collection is often incomplete and inter-agency practice and collaboration is variable. Although the rate of prosecution was relatively high by international standards, many children were later notified again for further concerns of abuse or neglect, suggesting that our interventions have been only partially successful. Practice implications: This paper suggests that all infants admitted to hospital with non-accidental head injury should become part of a prospective inter-agency research study, using a standardised data collection instrument. This should include the systematic collection of all data known or suspected to be associated with risk of child abuse, and incorporate long-term prospective follow-up, regardless of child protective or legal outcomes. Without large numbers followed prospectively and according to sound methodology, it is difficult to prove which forms of intervention are better than others at reducing the risk of further abuse. (Contains 2 tables.)