The primary objectives of this study were to determine how frequently knee magnetic resonance imaging (MRI) changes (1) diagnosis, (2) diagnostic confidence, and (3) management. A secondary objective was to correlate these changes with specific patient/physician characteristics and develop a prediction model using these characteristics. Six orthopedic specialists prospectively completed surveys when ordering knee MRI (n=93). Pre-MRI surveys recorded history, symptoms, signs, diagnosis, diagnostic confidence, and planned management. Post-MRI surveys recorded diagnosis, confidence, and planned management. Changes in diagnosis, management, and diagnostic confidence were correlated with patient/physician characteristics using chi-square and logistic regression tests. A multiple variable model was created with the most significant variables from the univariate analysis, and a c-index was used for cross-validation. Magnetic resonance imaging changed diagnosis in 29.3% and management in 25.3% of cases. Confidence in diagnoses after MRI increased, on average, by 10.6%. Change in diagnosis was significantly correlated with lateral joint line pain (P=.012) and tenderness (P=.006). The 3 most significant predictors for change in management were ligament pathology (P=.017), medial-sided pain/tenderness (P=.051), and age (P=.133). A 3-variable model using these predictors was significantly better than chance alone at predicting management changes (c-index: model=0.766; cross-validation=0.661). Magnetic resonance imaging frequently changed diagnosis and management and improved diagnostic confidence in a large minority of patients with internal derangement of the knee, even after evaluation by subspecialized physicians. A statistical model using specific patient characteristics can be created to predict when MRI will change management.
Copyright 2014, SLACK Incorporated.