Scoliosis screening is important for timely initiation of brace treatment to mitigate curve progression in skeletally immature children. Scoliosis screening programs frequently include the protocol of referring children screened positive with Scoliometer and Moiré Topography for confirmatory standard radiography. Despite being highly sensitive (88%) for detecting those who require specialist referral, the screening program was found to have more than 50% false positive rate that leads to unnecessary radiation exposure. Radiation-free ultrasound has been reported to be reliable for quantitative assessment of scoliosis curves. The aim of this prospective diagnostic accuracy study was to determine the accuracy of ultrasound in determining the referral status for children initially screened positive for scoliosis. 442 schoolchildren with a mean Cobb angle of 14.0 ± 6.6° were recruited. Using x-ray as the gold standard, the sensitivity and specificity of ultrasound in predicting the correct referral status were 92.3% and 51.6% respectively. ROC curve analysis revealed an area under curve of 0.735 for ultrasound alone and 0.832 for ultrasound plus scoliometer measurement. The finding provided strong evidences on the accuracy of ultrasound in determining the referral status that could result in more than 50% reduction of unnecessary radiation exposure for children undergoing scoliosis screening.