Skip to content
Omnikalc

Bland-Altman calculator: how well do two measurement methods agree?

Enter the mean and standard deviation of the paired differences and the sample size to get the bias, the limits of agreement and their confidence intervals.

Updated on

Educational tool, not medical advice: this page does not diagnose and does not replace a doctor, nurse or pharmacist. It has not been reviewed by a healthcare professional. Editorial and medical policy

Your inputs

Value out of range

Value out of range

Value out of range

Value out of range

Results

Upper limit of agreement

9.04

Agreement acceptable: the limits and their confidence bounds stay within the acceptable difference
Lower limit of agreement
-6.64
95 % confidence interval of the bias: lower
-0.08
95 % confidence interval of the bias: upper
2.48
Lower limit: outer 95 % bound
-8.86
Upper limit: outer 95 % bound
11.26

This calculator is for general information and education. It is not medical advice and does not replace the advice of your doctor or a registered dietitian.

A Bland-Altman analysis tells how well two measurement methods agree. The Bland-Altman calculator takes the mean difference (bias), the standard deviation of the differences, the number of pairs and the largest acceptable difference, and gives the limits of agreement, the confidence interval of the bias and the outer confidence bounds of each limit.

How the calculation works

  • Measure the same subjects with two methods and take the difference for each pair. The bias is the mean of the differences, and s is their standard deviation.
  • The limits of agreement are bias ± 1.96 × s. About 95 percent of the differences between the two methods should fall between them.
  • The confidence interval of the bias is bias ± t × s ÷ √n, with t the Student value for n − 1 degrees of freedom.
  • The limits themselves have a standard error of s × √(3 ÷ n), and the outer bounds are each limit moved outward by t times this error.

Example

Two blood pressure monitors are compared on 40 patients. The mean difference is 1.2 mmHg and the standard deviation of the differences is 4.0. The limits of agreement are −6.64 to 9.04. The bias has a 95 % interval of −0.08 to 2.48, and the outer bounds of the limits are −8.86 and 11.26. If differences up to ±12 are acceptable, all fit, so the agreement is acceptable.

Why it matters

When a new device or test is compared with an established one, correlation alone is misleading. The limits of agreement show the size of the differences between the two methods, and the confidence bounds show how reliable those limits are for the sample size.

Good to know

The method assumes that the differences are roughly normal and that the bias and spread are constant across the measurement range. The confidence intervals use approximations that are good for moderate samples, and the limits are wide when the sample is small. Repeated measures on the same subject need a method that accounts for them.

The Cohen’s kappa calculator measures agreement for categories, and the standard error calculator gives the uncertainty of a mean.

Frequently asked questions

What is a Bland-Altman analysis?

It is a way of comparing two ways of measuring the same quantity, such as two thermometers or a new test against a standard one. Instead of correlation, which only shows whether the methods move together, it looks at the differences and shows how far apart the two methods can be.

Why not use the correlation coefficient?

Two methods can be highly correlated yet disagree by a large amount, for example if one always reads ten units higher. Correlation measures association, not agreement. The limits of agreement give the range of differences you can expect between the methods in practice.

How do I decide whether the agreement is acceptable?

The statistics cannot decide it. You choose beforehand a difference that would not matter in practice, based on clinical or technical judgment. If the limits of agreement, ideally including their confidence bounds, fall inside that range, the methods can be used interchangeably.

What assumptions does the method make?

The differences should be roughly normally distributed and their spread should not change with the size of the measurement. Plot the differences against the mean of the two methods to check. If the spread grows with size, work with percentage differences or logarithms.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.