User:StreamunrealGlow/Bland–Altman plot
 | This is the sandbox page where you will draft your initial Wikipedia contribution.
If you're starting a new article, you can develop it here until it's ready to go live. If you're working on improvements to an existing article, copy only one section at a time of the article to this sandbox to work on, and be sure to use an edit summary linking to the article you copied from. Do not copy over the entire article. You can find additional instructions here. Remember to save your work regularly using the "Publish page" button. (It just means 'save'; it will still be in the sandbox.) You can add bold formatting to your additions to differentiate them from existing content. |
Article Draft
[edit]Interpretation
[edit]Interpretation of a Bland-Altman plot is contingent on the construction of the plot and the data at hand. Variations of the originally-defined plot have been introduced throughout the years and each should be interpreted accordingly.
The original plot displays a scatter plot of differences between a reference system to a trusted golden-standard system. The limits of agreement thereby stand as a confidence interval for which most of the difference lies between systems. The mean difference represents a general bias between the two systems: a positive average difference indicates that the reference system produces generally larger values than the gold-standard, and a negative average difference indicating that the reference system produces lower-values for the reference system[1]. A traditional Bland-Altman plot that produces a mean difference close to 0 indicates agreement between systems, though the limits of agreement indicates how varied the differences are across multiple values. Thresholds for appropriate limits of agreement should be defined a-priori for whatever the use case, as there is no general rule of thumb for appropriate thresholds[2].
The normality of differences is a prerequisite for the construction of limits of agreement since they're based on standard-deviation values. In the event that the distribution of differences are not normal, then the limits of agreement must be constructed such that they're not contingent on normal distributions. Bland and Altman's follow up paper on the topic indicates that percentiles of differences may be used in place for traditional limits of agreement[3]. Interpretation of these modified limits of agreement are identical to traditional implementations; the limits of agreement represent the majority of differences between systems.
Tests for heteroscedasticity are beneficial for assessing whether a proportional bias exist[4]. To give a definitive envelope for the proportional bias, tests such as the Breusch–Pagan test or the White test can provide statistically significant indicators of the relationship[5].
In the event that the magnitude of the data has an effect on the differences, several visual and construction options can be used to present the plot.
For example, a regression may be beneficial for illustrating how the differences between systems grows relative to the magnitude of the data. A slope closer to 1 indicates what is known as a 'proportional bias' given that the difference grows proportional to the magnitude of the data.If the data happens to be heteroscedastic, then the evaluation of the relationship. between systems is said to be dependent on the magnitude of the data in a different manner than a simple proportional bias. In such a case, a higher magnitude indicates a higher variance in the difference between systems, which has implications for the systems' impact on the data that are particular to the systems[6].
Alternatively, the plot may be presented in terms of proportional percentages between systems as opposed to actual differences. In such a case, each individual point of the scatter plot represents the proportion of differences between the two systems[7].
If for some reason the scale of the differences follows an exponential growth, then it may beneficial to plot the differences in a logarithmic manner. The relationship between the two systems then illustrates a multiplicative expression as opposed to a linear relationship[6].
Applications
[edit]Since the Bland-Altman plot assesses bias between systems, the plot has general prominence in a broad field of applicability. The plot has gained prominence in fields such as optometry, nutrition, radiology, surgery, medicine, veterinary, and other sciences[8][9][10].
The visualization and quantified output is useful wherever a comparison between two systems is needed. The interpretation of a plot can therefore represent agreement as well as relative or system bias between systems. This is particularly useful for evaluating medical instruments that record continuous output, especially by comparing novel tools to established tools that have already established as credible tools[8].
It is standard practice to preemptively define appropriate thresholds for limits of agreement on a case-by-case basis on the premise that agreement between systems should be specific for it's use case and take into consideration any nuance.
References
[edit]- ^ Martin Bland, J.; Altman, DouglasG. (1986-02-08). "STATISTICAL METHODS FOR ASSESSING AGREEMENT BETWEEN TWO METHODS OF CLINICAL MEASUREMENT". The Lancet. Originally published as Volume 1, Issue 8476. 327 (8476): 307–310. doi:10.1016/S0140-6736(86)90837-8. ISSN 0140-6736.
- ^ Giavarina, Davide (2015-06-15). "Understanding Bland Altman analysis". Biochemia Medica. 25 (2): 141–151. doi:10.11613/BM.2015.015. ISSN 1330-0962.
- ^ Bland, J Martin; Altman, Douglas G (1999-04-01). "Measuring agreement in method comparison studies". Statistical Methods in Medical Research. 8 (2): 135–160. doi:10.1177/096228029900800204. ISSN 0962-2802.
- ^ Chhapola, Viswas; Kanwal, Sandeep Kumar; Brar, Rekha (2015-05-01). "Reporting standards for Bland–Altman agreement analysis in laboratory research: a cross-sectional survey of current practice". Annals of Clinical Biochemistry. 52 (3): 382–386. doi:10.1177/0004563214553438. ISSN 0004-5632.
- ^ "Conditional Heteroscedasticity and Unit Root Tests", Unit Root Tests in Time Series Volume 2, Palgrave Macmillan, ISBN 978-1-137-00331-7, retrieved 2025-04-12
- ^ a b Ludbrook, John (2010). "Confidence in Altman–Bland plots: A critical review of the method of differences". Clinical and Experimental Pharmacology and Physiology. 37 (2): 143–149. doi:10.1111/j.1440-1681.2009.05288.x. ISSN 1440-1681.
- ^ Giavarina, Davide (2015-06-05). "Understanding Bland Altman analysis". Biochemia Medica. 25 (2): 141–151. doi:10.11613/BM.2015.015.
- ^ a b Zaki, Rafdzah; Bulgiba, Awang; Ismail, Roshidi; Ismail, Noor Azina (2012-05-25). "Statistical Methods Used to Test for Agreement of Medical Instruments Measuring Continuous Variables in Method Comparison Studies: A Systematic Review". PLOS ONE. 7 (5): e37908. doi:10.1371/journal.pone.0037908. ISSN 1932-6203. PMC 3360667. PMID 22662248.
{{cite journal}}: CS1 maint: unflagged free DOI (link) - ^ Carkeet, Andrew (2020-01). "A Review of the Use of Confidence Intervals for Bland-Altman Limits of Agreement in Optometry and Vision Science". Optometry and Vision Science. 97 (1): 3. doi:10.1097/OPX.0000000000001465. ISSN 1538-9235.
{{cite journal}}: Check date values in:|date=(help) - ^ Moore, A Russell (2024). "A review of Bland–Altman difference plot analysis in the veterinary clinical pathology laboratory". Veterinary Clinical Pathology. 53 (S1): 75–85. doi:10.1111/vcp.13293. ISSN 1939-165X.