Aims National radiology standards for investigation of suspected physical abuse (PA) in under 1 year olds recommend that a head CT should be performed alongside other investigations. Clinicians express anxiety regarding the radiation dose to the infant brain. This audit looks at adherence to guidelines and whether clinical indicators can be refined.
Methods This audit, in two teaching hospitals, identified infants who underwent investigation for suspected PA (January 2005–2008). Standardised data were extracted from case and radiology records and entered into an anonymised database. Presenting neurological symptoms (seizure, vomiting, apnoea, altered or loss of consciousness and altered behaviour) and neurological signs (focal neurology, bulging fontanelle or enlarging head circumference) plus documented injuries were recorded. Cases were allocated a level of concern of PA; high (unexplained fracture, bruising or head injury), medium (sibling, thermal injury) or low (clear accidental/medical cause).
Results 180 children were identified: 89 (49%) underwent CT head scan (58 high, 20 medium, 11 low risk). 91 had no CT (15 high, 36 medium, 40 low risk), of which 12 had neurological symptoms/signs and 35 had associated injuries suggestive of PA. 31/89 (35%) children who had CT showed intracranial injury (ICI). Table 1 identifies the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for ICI on CT in children with neurological symptoms/signs and in children with neurological symptoms/signs or facial/head bruising.
Conclusions This audit shows clinicians do not wholly follow radiological standards. Clinical decisions to perform a head CT were directly related to the level of perceived risk of PA. Of concern, some high risk cases with neurological symptoms/signs had no CT. Our data shows that should neurological symptoms/signs be used as a predictor of ICI the sensitivity is 81% and specificity 83%. If facial/head bruising is added as an indicator the sensitivity would rise to 100% and specificity falls. We acknowledge an element of selection bias in this data. However, the findings reflect current clinical practice and identify strong clinical indicators for ICI in suspected PA in infancy.