A Review of the Use of Confidence Intervals for Bland-Altman Limits of Agreement in Optometry and Vision Science

Abstract

Significance: Confidence intervals are still seldom reported for Bland-Altman 95% limits of agreement. When they are reported, 50% of articles use approximate methods and 50% use exact methods.

Purpose: Bland-Altman limits of agreement can be unreliable estimates, especially for small sample sizes. However, authors seldom use confidence intervals for limits of agreement. This article reviews their use in Optometry and Vision Science.

Methods: A keyword search for "Bland," "Altman," "Bland-Altman," "LoA," and "limits of agreement" was conducted on the Optometry and Vision Science website within a time range from January 2016 to December 2018.

Results: Fifty articles were reported or were judged to use Bland-Altman analysis; sample sizes ranged from 3 to 2072. Eight of these article reported confidence limits for limits of agreement, four of which used exact methods and four used Bland and Altman's approximate method.

Conclusions: Use of confidence intervals for limits of agreement has increased in Optometry and Vision Science but is far from universal. To assist researchers in calculating exact confidence limits for Bland-Altman limits of agreement, spreadsheets are included for performing the calculations and generating Bland-Altman plots with the confidence intervals included.