Browse by author
Lookup NU author(s): Dr Kevin Wilson,
Dr Faye Williamson,
Dr Joy Allen,
Dr Tom Hellyer,
Dr Clare LendremORCiD
This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).
The development of a new diagnostic test ideally follows a sequence of stages which,amongst other aims, evaluate technical performance. This includes an analyticalvalidity study, a diagnostic accuracy study and an interventional clinical utility study.In this paper, we propose a novel Bayesian approach to sample size determinationfor the diagnostic accuracy study, which takes advantage of information availablefrom the analytical validity stage. We utilise assurance to calculate the required samplesize based on the target width of a posterior probability interval and can chooseto use or disregard the data from the analytical validity study when subsequentlyinferring measures of test accuracy. Sensitivity analyses are performed to assess therobustness of the proposed sample size to the choice of prior, and prior-data conflictis evaluated by comparing the data to the prior predictive distributions. We illustratethe proposed approach using a motivating real-life application involving a diagnostictest for ventilator associated pneumonia. Finally, we compare the properties of theapproach against commonly used alternatives. The results show that, when suitableprior information is available, the assurance-based approach can reduce the requiredsample size when compared to alternative approaches.
Author(s): Wilson KJ, Williamson SF, Allen AJ, Williams CJ, Hellyer TP, Lendrem BC
Publication type: Article
Publication status: Published
Journal: Statistics in Medicine
Print publication date: 10/07/2022
Online publication date: 10/04/2022
Acceptance date: 11/03/2022
Date deposited: 14/03/2022
ISSN (print): 0277-6715
ISSN (electronic): 1097-0258
Publisher: John Wiley & Sons, Inc.
Altmetrics provided by Altmetric