Reporting of Multivariable Methods in the Medical Literature

BackgroundMultivariable models are frequently used in the medical literature, but many clinicians have limited training in these analytic methods. Our objective was to assess the prevalence of multivariable methods in medical literature, quantify reporting of methodological criteria applicable to mo...

Full description

Saved in:
Bibliographic Details
Published in:Journal of investigative medicine 2008-10, Vol.56 (7), p.954-957
Main Authors: Tetrault, Jeanette M., Sauler, Maor, Wells, Carolyn K., Concato, John
Format: Article
Language:English
Subjects:
Biometrics
Biometry
Data Interpretation, Statistical
Internal medicine
Logistic Models
Medical research
Medicine
Methods
Multivariate Analysis
Proportional Hazards Models
Publishing
Regression analysis
Studies
Survival analysis
Variables
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:BackgroundMultivariable models are frequently used in the medical literature, but many clinicians have limited training in these analytic methods. Our objective was to assess the prevalence of multivariable methods in medical literature, quantify reporting of methodological criteria applicable to most methods, and determine if assumptions specific to logistic regression or proportional hazards analysis were evaluated.MethodsWe examined all original articles in Annals of Internal Medicine, British Medical Journal, Journal of the American Medical Association, Lancet, and New England Journal of Medicine, from January through June 2006. Articles reporting multivariable methods underwent a comprehensive review; reporting of methodological criteria was based on each article's primary analysis.ResultsAmong 452 articles, 272 (60%) used multivariable analysis; logistic regression (89 [33%] of 272) and proportional hazards (76 [28%] of 272) were most prominent. Reporting of methodological criteria, when applicable, ranged from 5% (12/265) for assessing influential observations to 84% (222/265) for description of variable coding. Discussion of interpreting odds ratios occurred in 13% (12/89) of articles reporting logistic regression as the primary method and discussion of the proportional hazards assumption occurred in 21% (16/76) of articles using Cox proportional hazards as the primary method.ConclusionsMore complete reporting of multivariable analysis in the medical literature can improve understanding, interpretation, and perhaps application of these methods.
ISSN:1081-5589
1708-8267
DOI:10.2310/JIM.0b013e31818914ff