Development of a predictive model based on ultrasonographic features of brucella-associated periarthritis of the shoulder.

Purpose: To investigate ultrasonographic characteristics of shoulder involvement in brucellosis and explore correlations between imaging features and shoulder pain for predictive model development. Materials and methods: A total of 111 patients with confirmed brucellosis were enrolled and divided into a shoulder pain group (n=60) and a control group without shoulder symptoms (n=51). All underwent standardized shoulder ultrasound performed by experienced operators blinded to clinical data. Ultrasonographic findings were compared between groups. Correlation analysis and Partial Least Squares (PLS) regression identified key predictors. Predictive models were evaluated using ROC analysis, nomogram visualization, and decision curve analysis (DCA). Results: Compared with the non-shoulder pain group, patients with shoulder pain showed significantly higher frequencies of subacromial-subdeltoid (SASD) bursal effusion (58.3% vs. 5.9%), long head of the biceps tendon (LHBT) sheath effusion (68.3% vs. 15.7%), and rotator cuff hypervascularity (38.3% vs. 11.8%) (all P<0.05). PLS regression identified SASD bursal thickness (≥2.2 mm), SASD effusion, and LHBT sheath effusion as key predictors. The three-variable model demonstrated the highest sensitivity (0.967), while the five-variable model achieved the best overall performance with an area under the curve of 0.957 and specificity of 0.902. Nomogram and decision curve analysis further demonstrated the model's potential clinical utility in identifying shoulder involvement in brucellosis. Conclusion: Key ultrasonographic features are strongly associated with shoulder pain in brucellosis and provide high diagnostic value for early identification of shoulder involvement.