Diagnostic performance tests for suspected scaphoid fractures differ with conventional and latent class analysis

Area of Science:

• Orthopaedic diagnostics
• Radiology
• Statistical modeling

Background:

• Diagnosing scaphoid fractures lacks a consensus reference standard, hindering test evaluation.
• Latent class analysis (LCA) is a statistical method for determining diagnostic performance without a gold standard.

Purpose of the Study:

• Compare diagnostic performance of MRI, CT, bone scintigraphy, and physical exam for scaphoid fractures.
• Evaluate LCA's utility in assessing diagnostic accuracy for suspected scaphoid fractures.

Main Methods:

• Utilized data from two studies involving patients with suspected scaphoid fractures.
• Compared diagnostic performance metrics calculated by LCA versus traditional reference standard-based methods.
• Included wrist MRI, CT, bone scintigraphy, and structured physical examination.

Main Results:

• LCA-calculated performance differed from traditional methods for CT and MRI in the first cohort.
• Significant differences observed in MRI and physical examination sensitivity in the second cohort.
• CT sensitivity: 0.78 (LCA) vs. 0.67 (traditional); MRI sensitivity: 0.80 (LCA) vs. 0.67 (traditional).

Conclusions:

• Diagnostic performance characteristics calculated by LCA may differ from traditional reference standard-based calculations.
• LCA shows promise for evaluating diagnostic tests in orthopaedic conditions lacking a consensus reference standard.
• Further research into LCA for diagnostic performance assessment is warranted.