# Radial Head Fractures: Is the Mason Classification Still Effective Today? A Large-Sample Validation of Intra- and Inter-Observer Reliability

## TL;DR

This study evaluates the reliability of the Mason–Johnston classification for radial head fractures using a large sample of X-rays and multiple observers.

## Contribution

The study provides a large-sample validation of the Mason–Johnston classification's reliability and reproducibility in diagnosing radial head fractures.

## Key findings

- Inter-observer agreement for the Mason–Johnston classification was moderate, with Fleiss’ κ values of 0.49 and 0.50.
- Intra-observer agreement showed moderate to substantial levels of concordance.
- Reliability was higher for Type IV fractures compared to Types II–III.

## Abstract

Introduction: Various classifications of radial head fractures have been reported in the literature, most of them are based solely on conventional radiographic criteria. The Mason–Johnston classification, currently the most widely used system worldwide, is affected by the limitations of conventional radiographs. The aim of our study is to confirm or refute the low reliability and reproducibility of the Mason–Johnston classification. Materials and Methods: The study collected elbow X-rays showing radial head fractures from 2011 to 2021. Images were evaluated by eight orthopedic surgeons and one radiologist consultant from different hospitals for classification. The first phase assessed inter-observer agreement, comparing classifications among participants. After four months, the same images were randomly reordered and then reclassified to evaluate intra-observer agreement. A total of 90 elbow X-rays from 50 women and 40 men were analyzed. Inter- and intra-observer agreement was assessed using Fleiss’ kappa, Krippendorff alpha, and Cohen’s kappa. Results: Overall inter-observer agreement by unweighted Fleiss’ κ was moderate in both sessions (κ = 0.49 and κ = 0.50), with overall pairwise percent agreement 63% and prevalence- and bias-adjusted κ (PABAK, k = 4) ≈ 0.50. As an ordinal sensitivity analysis, Krippendorff’s α (ordinal) was 0.726 and 0.744, indicating substantial agreement. Type-specific reliability was moderate for Types II–III and higher for Type IV. Unweighted Cohen’s kappa coefficients were calculated to assess intra-observer agreement, demonstrating moderate to substantial levels of concordance. Conclusions: The Mason–Johnston classification shows moderate inter-observer reliability, especially for Types II–III, and moderate to substantial intra-observer agreement.