Standardizing ordinal subadult age indicators: Testing for observer agreement and consistency across modalities

The Subadult Virtual Anthropology Database (SVAD): An Accessible Repository of Contemporary Subadult Reference Data

“…A recent study showed no differences in long bone lengths and dental development between the deceased (medical examiner's office context) and living (hospital/clinical context) individuals from the U.S. and South African samples [10]. Although these findings should be verified on the other SVAD samples, they suggest there is no bias in skeletal and dental indicators that is related to the type of collaborating institution, the context of origin of the individuals (forensic or clinical), or the type of medical image (CT scan or radiograph), and that these indicators could be compared across samples for anthropological research [28].…”

“…Stacks of two-dimensional images, such as CT scans and MRIs, have the advantages of radiography (i.e., the ability to see beyond the external surfaces), but the additional benefit of alleviating orientation, position ambiguity, magnification, and superimposition issues that are inherent to two-dimensional radiograph imaging. The possibility for visualizing internal structures and high image resolution have proven time and again that the rendered image series [21][22][23][24][25] and the specific derivatives, such as segmented bone surfaces and the skeletal and dental variables collected directly from the image series or the virtual reconstructions, are reliable and accurate [26][27][28][29][30][31].…”

“…Intra-and inter-observer agreement rates for ordinal data (Table 5) were assessed using quadratic weighted Cohen's Kappa. The original presentation of the agreement rates [28] used linear weighted Cohen's Kappa; the authors have since recognized that quadratic coefficients are more appropriate. Essentially, the quadratic approach put more weight on a difference in ordinal scores of two or more increments (e.g., a difference between a score of 1 and a score of 3 for dental development) than a difference in ordinal scores of one increment (e.g., a difference between a score of 1 and a score of 2), which limits the risk of overestimating the resulting Cohen's kappa values [59].…”

“…All TEM, %TEM and Cohen's Kappa values for each indicator are available in the SVAD data collection protocol stored in the SVAD Zenodo Community [49]. [28]; the original publication employed linear weighted Cohen's Kappa.…”

“…Along with fellow forensic and biological anthropology colleagues [28,32,33,[60][61][62][63][64][65][66], we recognize the need to standardize and "optimize" data collection and ease the access to samples, medical images, and skeletal derivatives, and the resulting anthropological methods. The SVAD is substantial in size and diversity and the derivatives are freely available via the SVAD Zenodo Community for fellow researchers to use for their own research questions.…”