Accurate prediction of recovery is essential to determine whether a knee-ankle-foot orthosis (KAFO) is required in the subacute phase of stroke. However, there are currently no reliable methods to predict such recovery. This study aimed to determine whether muscle strength of the affected lower limb (affected side LL strength) in stroke patients in the subacute phase who cannot walk without a KAFO can be used to predict the continuous need for a KAFO, using a hand-held dynamometer. We enrolled patients with severe hemiplegia (n = 51) who were unable to walk without a KAFO for 10 days after stroke onset. They were divided into two groups depending on the continuous need for a KAFO at 1 month after onset; the KAFO and non-KAFO groups. Logistic regression analysis was used to investigate whether the affected side LL strength was a predictor of the continuous need for a KAFO at 1 month after onset. In addition, significant predictors were analyzed using receiver operating characteristic (ROC) curves. The KAFO and non-KAFO groups included 23 (45.10%) and 28 (54.90%) patients, respectively. The affected side LL strength and pusher syndrome severity were identified as predictors of the continuous need for a KAFO. The predictor with the highest predictive ability was the affected side LL strength, with an area under the ROC curve of 0.80 (95% CI, 0.68-0.93). Affected side LL strength may be a highly accurate predictor of the need for a KAFO in the subacute phase of stroke.