Estimating the probability of abusive head trauma: a pooled analysis

Area of Science:

• Pediatrics
• Forensic Medicine
• Child Abuse Research

Background:

• Distinguishing abusive head trauma (AHT) from nonabusive head trauma is critical for child protection.
• Clinical features alone can be ambiguous in diagnosing AHT.

Purpose of the Study:

• To identify combinations of clinical features that effectively differentiate AHT from nonabusive head trauma in young children.
• To establish the diagnostic value of specific clinical presentations in suspected AHT cases.

Main Methods:

• An aggregate analysis of individual patient data from six comparative studies involving children under three years old with intracranial injury.
• Utilized multiple imputation for combined clinical features, employing a bespoke hot-deck imputation strategy to manage missing data.
• Focused on features such as apnea, retinal hemorrhage, rib/skull/long-bone fractures, seizures, and head/neck bruising.

Main Results:

• Excluding non-significant variables (gender, age, skull fractures), the positive predictive value (PPV) for AHT ranged from 4% to 97% based on feature combinations.
• Apnea showed a significant association with AHT (OR: 6.89).
• The presence of rib fracture or retinal hemorrhage with any other feature, or any three or more significant features, yielded an odds ratio (OR) >100, indicating a PPV >85% for AHT.

Conclusions:

• Specific combinations of clinical findings can reliably estimate the probability of abusive head trauma (AHT).
• The developed model shows promise for refining diagnostic tools to aid clinical decision-making in suspected AHT cases.
• Further validation in large-scale prospective studies with expanded datasets is recommended.