The aims of this prospective school cohort study were to describe the epidemiology of diagnosed back pain in childhood, classified as either nontraumatic or traumatic back injury, and to estimate the association with physical activity in different settings. Over 2.5 years, 1240 children aged 6–12 years were surveyed weekly using mobile text messages to ask about the presence or absence of back pain. Pain was clinically diagnosed and injuries were classified using the International Classification of Diseases version 10. Physical activity data were obtained from text messages and accelerometers. Of the 315 back injuries diagnosed, 186 injuries were nontraumatic and 129 were traumatic. The incidence rate ratio was 1.5 for a nontraumatic back injury compared with a traumatic injury. The overall estimated back injury incidence rate was 0.20 per 1000 physical activity units (95% confidence interval 0.18–0.23). The back injury incidence rates were higher for sports when exposure per 1000 physical activity units was taken into consideration and especially children horse-riding had a 40 times higher risk of sustaining a traumatic back injury compared to the risk during non-organized leisure time physical activity. However, the reasonably low injury incidence rates support the recommendations of children continuously being physically active.