E-bike injuries among US children surged in 2025: national estimates from emergency department reports, 2019–2025, using machine learning to read injury narratives

Abstract

Background: E-bikes and e-scooters are now common on US streets, but tracking their injuries is hard. In 2024 the US Consumer Product Safety Commission gave e-bikes their own product code; before then, e-bikes shared a code with mopeds. National estimates for 2025, and adjusted comparisons of children's e-bike and bicycle injuries, are lacking. Methods: I used all injuries in the National Electronic Injury Surveillance System, a national sample of US emergency departments, for 2019–2025. I labelled each injury by device from its product codes and written narrative. For unclear records in the old shared code, I trained a text classifier on 2024–2025 records, when e-bikes and mopeds had separate codes. I estimated national visits with survey-based 95% CIs and compared hospitalisation with regular bicycles using weighted logistic regression. Results: The classifier separated e-bikes from mopeds with an area under the curve of 0.995, and my 2022–2024 e-bike estimates were within 4% of official figures. E-bike visits rose from 27 700 in 2022 to 101 300 in 2025 (95% CI 69 500 to 133 100). Children's visits rose sevenfold, from 3600 to 25 600 (16 700 to 34 400): one in four e-bike injuries. Children injured on e-bikes had higher odds of hospitalisation than children injured on bicycles (adjusted OR 1.68, 1.25 to 2.26); adults did not. Conclusions: E-bike injuries among US children are rising fast and are more severe than bicycle injuries. Age, speed and helmet rules for young riders deserve urgent attention. Preprint; submitted to Injury Prevention and not yet peer reviewed. Code: https://github.com/subash28kumar/ebike-injuries-neiss