Defining ranges for certainty ratings of diagnostic accuracy




Poster session 1


Sunday 16 September 2018 - 12:30 to 14:00

All authors in correct order:

Hultcrantz M1, Langendam MW2, Leeflang M2, Lavergne V3, Ansari MT4, Mustafa RA5, Schünemann H6
1 Swedish Agency for Health Technology Assessment and Assessment of Social Services; Karolinska Institutet, Sweden
2 University of Amsterdam, The Netherlands
3 University of Montreal, Canada
4 University of Ottawa, Canada
5 University of Missouri-Kansas City, USA
6 McMaster University, Canada
Presenting author and contact person

Presenting author:

Monica Hultcrantz

Contact person:

Abstract text
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) concept for certainty of evidence represents our confidence that the true effect lies above or below a threshold, or in a specified range. A clarification of how these concepts apply to certainty ratings of diagnostic test accuracy is needed. This is especially important as it relates to frequent lack of direct evidence assessing the effect of tests on patient-important outcomes.

To illustrate possible approaches for defining ranges when assessing the certainty of evidence for diagnostic test accuracy within a systematic review, health technology assessment, or clinical practice guideline.

After initial brainstorming with GRADE project group members, the investigators iteratively refined and clarified the approaches using input from workshops and discussions at GRADE Working Group meetings.

Ranges can be defined both for single test accuracy and for comparative accuracy of multiple tests. For systematics reviews and health technology assessments there are approaches for defining ranges that do not include value judgments about downstream health outcomes. Key challenges arise when clinical meaningful ranges should be set for sensitivity and specificity, in the context of a guideline. At the Colloquium, we will show how these challenges can be addressed using an example from a real-life systematic review illustrating a direct comparison between two test strategies.

The application of the approaches for defining ranges in test accuracy will provide a useful framework when assessing, presenting, and making decisions based on the certainty of evidence for diagnostic test accuracy.

Patient or healthcare consumer involvement:
No involvement at this early stage of methodological development.

Relevance to patients and consumers: 

What is a good test to use for diagnosing a particular condition? This work illustrates how one can define a good test by first addressing what the consequences of receiving a correct as well as a false test result would be on the patient’s health and wellbeing. We hope that this in the long run will make it easier to choose what test to use based on patient values and preferences.