Level of Agreement With a Multi-Test Approach to the Diagnosis of Diabetic Foot Osteomyelitis

Although bone biopsy has historically been considered the "gold standard" or "standard reference" for the diagnosis of diabetic foot osteomyelitis, some contemporary investigations have provided evidence against this as a single diagnostic test and in support of a combination of clinical, laboratory, and radiographic findings. The objective of this investigation was to measure the level of agreement between several commonly used forms of diagnostic testing for diabetic foot osteomyelitis. A retrospective chart review was performed of 50 consecutive patients admitted to a single tertiary healthcare center with the documented performance of 1) a clinical probe-to-bone test on hospital admission; 2) plain film radiographs prior to any surgical intervention; 3) magnetic resonance imaging prior to any surgical intervention; and an intraoperative excisional bone debridement performed, with samples sent for both 4) histologic analysis and 5) microbiologic analysis. A frequency count of agreement among these 5 tests was performed, and the interobserver (or inter-test) agreement was measured using the kappa statistic. We observed low levels of inter-test agreement between the 5 diagnostic tests (range 42.0%-62.0%), and levels of chance-corrected agreement were well below what would be considered appropriate for a "gold standard" or "standard reference." Levels of the kappa statistic ranged from 0.0 to 0.220, with most inter-test comparisons falling in the "poor agreement" and "slight agreement" interpretation ranges. The highest level of agreement occurred between the plain film radiographs and magnetic resonance imaging (62.0% agreement and kappa statistic of 0.220). Although it is likely that a combination of clinical, radiographic, and laboratory tests provides the best diagnostic approach for diabetic foot osteomyelitis, the data provided herein indicate that the tests themselves might have high intrinsic levels of unreliability and that the specific combination of tests that might be best used remains unclear.