Bland-Altman

Statistical methods for assessing agreement between two methods of clinical measurement

Bland JM, Altman DG Lancet (London, England) 1986 · PMID 2868172 · doi:10.1016/s0140-6736(86)90837-8

Question

How should agreement between two methods of measuring the same quantity be assessed?

Design

Statistical methods paper with worked clinical examples.

Patients

Paired measurements of the same variable by two methods.

Intervention

Plot of the difference between methods against their mean, with limits of agreement

Comparison

Correlation coefficient and regression

Primary outcome

Bias and limits of agreement

Key result

Showed that correlation coefficients are misleading for judging agreement and proposed a simple graphical approach: plot differences against means and report the mean difference (bias) and limits of agreement derived from the spread of the differences.

Limitations

Applies to continuous data; judging whether the limits are acceptable remains a clinical decision.

Impact on practice

The standard approach for comparing measurement methods in orthopaedics (e.g. radiographic angles, implant positioning, new imaging techniques); a frequent exam statistics topic.