Objective: To construct a predictive model for radiation dermatitis induced by radiation therapy for esophageal cancer based on logistic regression analysis and to test its accuracy. Methods: A total of 152 patients receiving radiotherapy for esophageal cancer in Department of Radiation Oncology of Affiliated Hospital of Jiangnan University from January 2017 to March 2022 were selected and randomly divided into a modeling group (n=90), including dermatitis and no dermatitis, and a validation group (n=62). Univariate and multivariate logistic regression analysis were used to screen clinical factors. The predictive score model was constructed according to the risk factor weight, the risk of radiation dermatitis was divided into three grades: low risk, medium risk, and high risk. The risk factor weights of different risk levels in the modeling group and the validation group were recorded. The receiver operating characteristic (ROC) curve was used to analyze the accuracy of the modeling group and validation group. Results: In the modeling group, grade 0 radiation dermatitis occurred in 21 patients (23.33%), grade 1 in 54 patients (60.00%), grade 2 in 11 patients (12.22%), and grade 3 in 4 patients (4.44%). Logistic regression analysis showed that tumor diameter, surgical complications, radiation dose, serum albumin level, combined diabetes, body mass index (BMI), and the use of compensation membrane (Bolus) were the important risk factors for radiation dermatitis (all P<0.05). The number of low-risk, medium-risk and high-risk radiation dermatitis cases in the modeling group were 3, 13, and 53, respectively, and that in the validation group were 1, 11, and 35, respectively. The area under the curve (AUC) of ROC was 0.968 and 0.953 respectively. Hosmer-Lemeshow test showed good goodness of fit (χ²=87.148, P=0.059; χ²=72.433, P=0.058, respectively). Conclusion: The main clinical risk factors for radiation dermatitis induced by radiotherapy for esophageal cancer are tumor diameter, surgical complications, radiation dose, serum albumin level, comorbidity with diabetes, BMI, and the use of Bolus. The constructed prediction model can quantitatively assess the risk. It has a high degree of agreement with the actual situation and has high clinical application value.