Objectives Knee injury and Osteoarthritis Outcome Score (KOOS)-Child is a modification of the adult KOOS aiming to evaluate knee injury, including ACL deficiency. However, the measurement properties of KOOS-Child have not been assessed in a cohort of children with ACL deficiency. We aimed to study the structure of KOOS-Child using modern test theory models (Rasch analysis and confirmatory factor analysis (CFA)). Methods Data were collected prospectively in a cohort of children with ACL deficiency at three time points: before-and-after ACL surgery, and at 1-year follow-up. For each subscale, structural validity through the fit of a CFA model was evaluated for 153 respondents. Modification indices were examined to find the model of best fit, confirmed using Rasch analysis. Responsiveness was reported for each subscale. Reliability was calculated for each item. Floor and ceiling effects, and Person-item distribution were reported. Results All subscales showed inadequate fit to a unidimensional CFA model. Rasch analysis confirmed these results. Adjusting the subscales improved model fit, although this was still quite poor, except for the quality of life subscale. With one exception, all items demonstrated ceiling effects. Person-item distribution confirmed this. Due to lack of fit, reliability was not reported. All subscales were able to detect change from baseline to 1-year follow-up. Conclusions KOOS-Child exhibits inadequate measurement properties in its current form for children with ACL deficiency. Suggestions to make the subscales fit the models better and improve accuracy of KOOS-Child are presented. However, the large ceiling effects observed may reduce sensitivity and induce type 2 errors.