The study constructed a prognostic model for knee osteoarthritis (OA) based on baseline characteristics, imaging findings, and clinical indicators to assess patients' risk of adverse outcomes. The retrospective analysis included 345 patients divided into a training set (n=241) and a validation set (n=104) in a ratio of 7:3. Independent risk factors for sustained clinical deterioration were body mass index (BMI), total bone marrow lesion volume (TBLV), tibiofemoral angle (TFA), WOMAC functional subscore, serum hs-CRP and urinary uCTX-II (all p<0.05), while medial joint space width (mJSW) was a protective factor (p<0.05). Among machine learning models, Random Forest (RF) had the highest AUC of 0.910, which was better than SVM (0.885) and GBM (0.824). The RF model was selected as optimal for predicting the risk of treatment failure. Key predictors were BMI, uCTX-II and serum hs-CRP. There were no differences in baseline characteristics between groups (p>0.05).