G67 Exploring the potential impact of a clinical prediction tool for paediatric abusive head trauma

Abstract

Aims The predicting abusive head trauma (PredAHT) clinical prediction tool has been derived and validated. Based on the presence or absence of six clinical features (long-bone fractures/ rib fractures/retinal haemorrhage/head or neck bruising/apnoea/ seizures) this electronic calculator estimates the probability of abusive head trauma (AHT) in young children with intracranial injury. We aimed to explore the impact of this tool on clinicians’ own probability estimates of AHT. Methods A study of six clinical vignettes was conducted in two teaching and two district general hospitals with 29 clinicians: community(15), general(9) and emergency(2) paediatricians, radiologists(2) and one neurosurgical nurse. One vignette was designed as probable AHT, another was designed with a history and clinical features representing accidental head injury, and four were more indeterminate, designed to introduce uncertainty into the clinical decision. Clinicians were asked to estimate the probability of AHT in each vignette; they were then presented with the probability estimate from the PredAHT calculation and asked whether this altered their initial probability estimate. The impact of the tool in each vignette was analysed using linear modelling and linear mixed effects modelling. Interrater reliability of clinicians’ probability estimates was assessed with Intraclass correlation (ICC) based on single rating, absolute agreement, two-way random effects models. Results The tool significantly influenced clinicians’ probability estimates in all six vignettes. The greatest impact was demonstrated in a vignette with few concerning features in the history, but several concerning clinical features (tool score high). The least impact was demonstrated in a vignette with a history incompatible with the child’s motor development, but no additional clinical features (tool score low). The tool had the greatest impact for clinicians with the least paediatric experience. Agreement between clinicians was improved by the PredAHT; interrater reliability ranged from ‘poor’ to ‘good’ prior to using the tool (ICC.593, 95% CI.347 .899) and ‘poor’ to ‘excellent’ (ICC.661, 95% CI.417 .922) after using the tool. Conclusions PredAHT a numerical, electronic, evidence-based clinical prediction tool encourages clinicians to critically consider their probability estimates of AHT in light of the calculated score, and reduces variability in their estimates. The impact of the tool should be tested in clinical practice.