The authors aimed to extract statistically significant associations between non-ocular clinical and diagnostic imaging findings and the presence of significant retinal haemorrhages, and then develop an evidence-based screening algorithm to determine when to consult an ophthalmologist in suspected child abuse cases. Importance was placed on identification of cases on first presentation to the medical system. This was a retrospective comparative effectiveness study reviewing cases over 2014-2020. Two hundred and seventy-four patients met the inclusion criteria of being under 36 months with ophthalmic exam at first presentation. Of these, 81.3% were under 12 months of age. Forensic examination was ordered in 261 and 148 were deemed positive for abuse; six died in hospital. Significant retinal haemorrhages were found in 35. Death or deemed positive for abuse were significantly associated with significant retinal haemorrhage. Significant retinal haemorrhages were associated significantly with seizures, vomiting, Glasgow Coma Score (GCS) <15, bruising and facial / head bruising, abnormal CT / MRI scan, parenchymal haemorrhage, hypoxia / ischaemia / infarct, cerebral oedema, axonal / white matter injury, and subdural and subarachnoid haemorrhage. After multivariate logistic regression model analysis, only abnormal imaging, GCS score and altered mental status had significant correlation with significant retinal haemorrhage. These factors informed the development of the algorithm. The authors acknowledge many confounding variables, and these should be further explored. The screening algorithm now requires validation research.