Objective To establish the pathological grades of Henoch-Schonlein purpura nephritis(HSPN) in children with diagnostic prediction models by stepwise Fisher discriminant in children. Methods Based on the investigation of 28 clinical indicators from 144 cases with HSPN came from Children′s Hospital of Chongqing Medical University, the sensitive indicators were found and stepwise Fisher discriminant model was established and its accuracy in predicting the pathological classification of HSPN was tested. Results There were 5 laboratory indicators and clinical manifestations with different pathological grades of HSPN.In children with pathological grade Ⅱ, Ⅲ and Ⅳ, 5 indicators were screened (P<0.05) and stepwise Fisher discriminant models were established.And the correct rate of comprehensive diagnosis was (61.371±8.740)% in 100 random sampling diagnostic simulations; in children with pathological grade Ⅲa and Ⅲb, 5 indicators were also screened (P<0.05) and stepwise Fisher discriminant models were established.And the correct rate of comprehensive diagnosis was (68.015±5.736)% in 100 random sampling diagnostic simulations. Conclusions The stepwise Fisher discriminant models established in this research have a better diagnostic accuracy in forecasting for pathological grade of HSPN, and have a certain guiding value on early treatment and prognosis evaluation of children with newly diagnosed HSPN. Key words: Stepwise Fisher discriminant model; Henoch-Schonlein purpura nephritis; Pathology; Child