P16 DEFINING THE PROBABILITY OF INFLICTED BRAIN INJURY BY A META-ANALYSIS: A SYSTEMATIC REVIEW OF PREDICTIVE CLINICAL FEATURES

S Maguire, N Pickard, D Farewell, M Mann, V Tempest, A Kemp. Cardiff University, Cardiff, Wales, UK

Aim: The aim of this systematic review was to define the clinical features that distinguish inflicted from non-inflicted brain injury in children.

Methods: An all-language literature search of 10 databases, conference abstracts and references from 1970 to 2007 was performed. All relevant studies underwent two independent reviews by paediatric, radiology, neurology and ophthalmology specialists, using standardised critical appraisal methodology. Inclusion criteria: primary comparative studies of inflicted and non-inflicted brain injury, in children aged 0–18 years, presenting alive, in which the injury was defined as an intracranial haemorrhage (ICH) visible on neuro-imaging, with a clear inflicted or non-inflicted aetiology confirmed. Exclusion criteria: mixed adult and child data, management or outcome. A meta-analysis was performed combining the presence of ICH with individual clinical features (retinal haemorrhages, rib fractures, long bone fractures, skull fractures, bruises to the head, apnoea or seizures) to calculate their sensitivity, specificity and positive predictive value (PPV) individually. From this, minimal estimates (allowing for underreporting of the feature in the non-inflicted brain injury) and maximal estimates (assuming that any untested individual only has the feature if they have sustained an inflicted brain injury) were derived. When raw data were available (three studies) a multivariate logistic regression analysis was performed, identifying the PPV for multiple clinical features.

Results: Of 305 studies reviewed, 11 were included, representing 1034 children, 625 with inflicted brain injury. Of the …