Diagnostic accuracy of an abbreviated MRI protocol for detecting radiographically occult hip and pelvis fractures in the elderly.

Abstract

To determine the diagnostic accuracy of an abbreviated, two-sequence MRI protocol using limited pulse sequences for the detection of radiographically occult hip and pelvis fractures in the elderly compared to the complete MRI examination. One hundred and eleven consecutive emergency department patients age 65 or older who had undergone MRI to evaluate for clinically suspected hip fracture after negative radiographs were included in the study. The large field-of-view coronal T1 and STIR sequences were isolated from the complete six-sequence MRI protocol and reviewed independently in a blinded fashion by two musculoskeletal fellowship-trained radiologists who recorded presence or absence of fractures of the proximal femora or pelvis, fracture type, and presence or absence of soft tissue injury. Test accuracy was calculated with 95% confidence intervals and accuracy of fracture classification for the abbreviated protocol was compared to that made on the basis of the full exam. For proximal femoral fractures, the abbreviated protocol had a pooled sensitivity and specificity for the two readers of 100 and 97%, respectively. For pelvic fractures, sensitivity was 92% and specificity was 98%. The kappa coefficient for fracture classification was 0.90 for reader 1 and 0.88 for reader 2, indicating excellent agreement for both readers in fracture classification compared to the classification made based on the complete MRI protocol. An abbreviated MRI protocol that includes only coronal T1 and STIR sequences maintains high sensitivity and specificity for hip and pelvis fracture detection and fracture classification.