Aim This study aimed to investigate the incidence of relapse and FR/SDNS in Chinese children with SSNS and to develop clinical prediction models for relapse and FR/SDNS. Methods This retrospective cohort study involved 339 newly onset SSNS patients between 2006 and 2016. The incidence of relapse and FR/SDNS were estimated using the Kaplan–Meier method. Prediction models were constructed based on Cox proportional‐hazards regression. Results The median follow‐up time was 8.7 years. The cumulative incidence of relapse at 1‐, 2‐, and 5‐year was 51.0%, 62.5%, and 66.6%. The cumulative incidence of FR/SDNS at 1‐, 2‐, and 5‐year was 18.4%, 29.0%, and 32.9%. The final prediction model for first relapse included four variables (serum albumin, triglycerides, IgM, and time to first remission). The model's discriminative ability was low (Harrell's C index = 0.62). The final prediction model for FR/SDNS included four variables (serum albumin, lipoprotein(a), time to first remission, and time to first relapse). The discrimination and calibration of the prediction model for FR/SDNS were acceptable (Harrell's C index = 0.73, Brier score at 1‐ and 2‐year were 0.11 and 0.17). Conclusion The first relapse and FR/SDNS mainly occurred in the first 2 years after initial SSNS onset. The prediction model for relapse developed using common clinical parameters performed poorly, while the prediction model for FR/SDNS might be useful. Summary at a Glance The incidence of relapse in Chinese children with SSNS might not be as high as previously reported in cohort studies or randomized controlled trials of other countries. Besides, this study developed a valuable prediction model for FR/SDNS in SSNS children using common clinical and laboratory variables.