Introduction - The epidemiology of child abuse is poorly documented. We developed an algorithm, using data from the PMSI, to identify hospitalizations resulting from physical abuse among children aged 0 to 5 years. Materials and Methods - We conducted a single-center study to test the algorithm’s validity. The tool was used to identify hospital stays at the Dijon University Hospital that were highly suspicious or suspicious of physical abuse. We examined the child’s injuries and their consistency with the mechanism of injury reported by caregivers to determine whether or not the case constituted abuse. Cross-referencing data from the PMSI system and medical records allowed us to calculate the algorithm’s positive predictive value. Results—Among children aged 0 to 5 years, 54 were included in Group 1 (highly probable abuse) and 102 in Group 2 (suspected abuse). The algorithm’s PPV for Group 1 was 85.2% (95% CI: [75.7–94.7]); it was 50% [40.3–59.7] for Group 2. The younger the children, the higher the PPV. It reached 94.4% [87.0–100] for children in Group 1 aged 1 month to 1 year, and 78.3% [67.9–88.8] for those in Group 2. Discussion-Conclusion - This algorithm is a promising tool for identifying hospital stays due to physical abuse in young children. The identification of suspicious stays (Group 2) still needs refinement. Routine clinical application of the algorithm could improve the diagnosis of abuse.