Initial worsening of symptoms after the start of corticosteroid administration is a major concern in the treatment of myasthenia gravis (MG). However, the risk factors or specific patient backgrounds related to this issue have not been fully understood. We aimed to determine the risk factors and developed a scoring system for predicting initial worsening in generalized MG.We enrolled 62 generalized MG patients with anti-acetylcholine receptor antibody. Initial worsening was defined as an increment of three points in the Quantitative MG score within 2 weeks after the start of steroid treatment. A multivariate logistic regression model was used to determine the risk factors, and predictive scores were assigned. Bootstrap resampling was applied to evaluate the risk score model's internal validity.Steroid-induced initial worsening occurred in 26% of MG patients and was correlated with thymoma-associated or early-onset MG (p = 0.018), initial prednisolone doses ≥40 mg/day (p = 0.029), and upper limb weakness (p = 0.039). Stepwise multivariate logistic regression identified these three clinical factors for predicting initial worsening in MG. A predictive score of 0-3 points had a bootstrapping area under the curve of 0.770 (0.625-0.878).Our scoring system based on three clinical characteristics can predict the likelihood of steroid-induced initial worsening in MG.