Bland-Altman Analysis Calculator
Compare two measurement methods with mean bias, 95% limits of agreement (LOA), and an interactive Bland-Altman plot — the gold standard for clinical method comparison studies.
About Bland-Altman Analysis
Proposed by Bland & Altman (1986), this method visualises agreement between two measurement methods. For each subject, it plots the difference (A − B) on the Y-axis versus the mean (A + B) / 2 on the X-axis. Three key horizontal lines are drawn: the mean bias and the upper and lower 95% limits of agreement (LOA = bias ± 1.96 × SD of differences).
A new method is considered clinically acceptable when the entire LOA range falls within a pre-specified maximum acceptable difference — a clinical judgement, not a statistical cutoff.
Reporting format: "The new device showed a mean bias of −0.13 mmol/L (95% LOA: −0.70 to 0.43 mmol/L) vs the reference method (Bland & Altman, 1986)."
Frequently Asked Questions
What is Bland-Altman analysis?
Bland-Altman analysis assesses agreement between two measurement methods. It plots the difference between methods against their mean, quantifies mean bias, and defines 95% limits of agreement showing the range within which 95% of differences fall.
Why not use Pearson correlation for method comparison?
Correlation measures whether variables move together, not whether they agree absolutely. Two methods could be highly correlated (r = 0.99) but one consistently give values 20% higher. A Bland-Altman analysis reveals this systematic bias and the spread of individual disagreements.
What are the limits of agreement?
LOA = mean bias ± 1.96 × SD of differences. They represent the range within which 95% of individual pair differences are expected to fall. To judge acceptability, compare the LOA against your pre-specified maximum clinically acceptable difference.
How do I interpret the Bland-Altman plot?
Mean bias near zero = no systematic offset. Random scatter around bias line = only random error. Funnel pattern (spread increases with mean) = proportional bias — consider log transformation. Points outside LOA lines are outlier pairs with greater-than-expected disagreement.
How do I report Bland-Altman results?
Report mean bias, 95% LOA, n, and whether the LOA fall within your pre-specified acceptable limits. Example: "Mean bias −0.13 mmol/L (95% LOA: −0.70 to 0.43 mmol/L, N=12). As LOA fall within the pre-specified ±0.83 mmol/L threshold, the device is clinically acceptable."