Multiple pitfalls can occur with the conduct and analysis of a study of diagnostic tests, resulting in biased accuracy. Our conceptual model includes three stages: patient selection, interpretation of the index test and disease verification. In part 2, we focus on (1) Interpretation bias (or workup bias): where the classification of an indeterminate index test result can bias the accuracy of a test or how lack of blinding can bias a subjective test result, and (2) Disease verification bias: where the index test result is incorporated into the gold standard or when the gold standard is applied only to a select population as the gold standard is an invasive test. In an example with age-adjusted D-dimer for pulmonary embolism, differential verification bias was a limitation due to the use of two gold standards—CT for a high-risk population and follow-up for symptoms in a low-risk population. However, there are circumstances when certain choices in study design are unavoidable, and result in biased test characteristics. In this case, the informed reader will better judge the quality of a study by recognising the potential biases and limitations by being methodical in their approach to understanding the methods, and in turn, better apply studies of diagnostic tests into their clinical practice.
- research, methods
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Contributors BK, MKH and RW generated the content together. BK wrote the review article, KH edited the article and drafted tables, and RW edited the article and drafted figures and table. All authors have read and approved the manuscript.
Funding BK is supported by NHLBI K08 (grant # 1K08HL140105-01).
Competing interests BK is the site investigator for Ortho-Clinical Diagnostics.
Provenance and peer review Not commissioned; externally peer reviewed.
Correction notice This article has been corrected since it was published Online First. The title of the companion article was updated to Recognising Bias in Studies of Diagnostic Tests Part 1: Patient Selection in table 1 and in the references.
Patient consent for publication Not required.