30 The use of information theory to evaluate clinical testing regimes. (23rd February 2023)

• Record Type: Journal Article
• Authors: Hussein, Mhelai; Bowyer, Stuart A; Booth, John; Briggs, Lydia; Bryant, William A; Key, Daniel; Shah, Mohsin; Sebire, Neil J
• Is Part Of: Archives of disease in childhood. Volume 108(2023)Supplement 1
• Page Start: A11
• Page End: A11
• Publication Date: 2023-02-23
• Subjects: Children -- Diseases -- Periodicals; Infants -- Diseases -- Periodicals; 618.920005
• DOI: 10.1136/archdischild-2023-gosh.30
• Languages: English
• ISSNs: 0003-9888

Abstract

Abstract : Data Research, Innovation and Virtual Environments Unit (DRIVE), NIHR Great Ormond Street Hospital BRC Diagnostic testing has an important role in modern medical practice. Despite the benefit that laboratory tests provide, over-testing presents issues. At a systematic level, inappropriate testing wastes finite health resources at a time where health systems are under unprecedented financial strain. Conversely, under-testing can result in important clinical information being missed. As such it is important to identify the most appropriate testing regimes to ensure tests are only conducted as and when necessary. Conventional metrics of diagnostic test performance do not quantitatively answer how much diagnostic uncertainty is reduced by conducting the test. As the purpose of diagnostic testing is to reduce uncertainty in diagnosis, traditional metrics lack crucial clinical information. Information Theory (IT) provides a means of quantifying information gain. We studied how the principles of IT can be applied to identify redundancies arising from clinical testing regimes in the current creatinine testing protocol at the Great Ormond Street Hospital (GOSH). Mutual information (MI) was applied to quantify redundant information as per the current testing protocol. Creatinine test results were obtained from routinely acquired electronic patient records collected by GOSH. The dataset was analysed in Aridhia Digital Research Environment software using Python. Data was first cleaned, and MI was subsequently calculated. Based on 68, 589 creatinine samples taken across 671 patients, our results showed that between 28 to 60 days post-transplant, the current testing protocol at GOSH (testing every three days) provides similar amounts of clinical information (normalised MI score of 0.610) as testing every four and five days (MI scores of 0.591 and 0.602 respectively). This indicates the possibility that testing frequency could potentially be reduced with little loss of clinical information. With further study, this method could be applied to a range of laboratory tests to define testing regimes that optimise diagnostic information whilst minimising health system demand.