Validation of a Machine-Learning Clinical Decision Aid for the Differential Diagnosis of Transient Loss of Consciousness - Research findings and validation

Aug 5, 2026

Upload by Miles

The study develops and validates a machine-learning classifier to support differential diagnosis of transient loss of consciousness (TLOC) at first presentation using patient and witness questionnaires. Patients newly presenting with TLOC were prospectively recruited and completed an online questionnaire during home or initial assessment. Two expert raters established TLOC cause after 6-month follow-up, with separate development and validation samples. A patient-only model using nine questionnaire items achieved 80.8% diagnostic identification accuracy, but was not sufficiently accurate for routine care, suggesting that including witness information could improve performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Abstract

Background and Objectives

Methods

Results

Discussion

This is a repository copy of Validation of a machine-learning clinical decision aid for the differential diagnosis of transient loss of consciousness. White Rose Research Online URL for this paper: [https://eprints.whiterose.ac.uk/225923/](https://eprints.whiterose.ac.uk/225923/) Version: Published Version Article: Wardrope, [A. orcid.org/0000-0003-3614-6346](A. orcid.org/0000-0003-3614-6346) , Ferrar, [M. orcid.org/0009-0008-4697-3154](M. orcid.org/0009-0008-4697-3154) , Goodacre, [S. orcid.org/0000-0003-0803-8444 et al](S. orcid.org/0000-0003-0803-8444 et al). (4 more authors) (2025) Validation of a machine-learning clinical decision aid for the differential diagnosis of transient loss of consciousness. Neurology Clinical Practice, 15 (2) . e200448 . ISSN 2163-0402 [https://doi.org/10.1212/cpj.0000000000200448](https://doi.org/10.1212/cpj.0000000000200448) Reuse This article is distributed under the terms of the Creative Commons Attribution (CC BY) licence. This licence allows you to distribute, remix, tweak, and build upon the work, even commercially, as long as you credit the authors for the original work. More information and the full terms of the licence here: [https://creativecommons.org/licenses/](https://creativecommons.org/licenses/) Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [eprints@whiterose.ac.uk](eprints@whiterose.ac.uk) including the URL of the record and the reason for the withdrawal request. [eprints@whiterose.ac.uk](eprints@whiterose.ac.uk)[ ](eprints@whiterose.ac.uk)[https://eprints.whiterose.ac.uk/](https://eprints.whiterose.ac.uk/) Downloaded from [https://www.neurology.org by 2a02:c7c:3f18:9c00:3183:7648:31e7:e2de on 30 April 2025](https://www.neurology.org by 2a02:c7c:3f18:9c00:3183:7648:31e7:e2de on 30 April 2025) RESEARCH ARTICLE OPEN ACCESS Validation of a Machine-Learning Clinical Decision Aid for the Diﬀerential Diagnosis of Transient Loss of Consciousness Alistair Wardrope,1,2 Melloney Ferrar,3 Steve Goodacre,4,5 Daniel Habershon,6 Timothy J. Heaton,7 Stephen J. Howell,1 and Markus Reuber1,2 Neurol Clin Pract. 2025;15:e200448 . doi:10.1212/CPJ.0000000000200448 Correspondence Dr. Wardrope [a.wardrope@sheffield.ac.uk](a.wardrope@sheffield.ac.uk) Abstract Background and Objectives The aim of this study was to develop and validate a machine-learning classiﬁer based on patient and witness questionnaires to support diﬀerential diagnosis of transient loss of consciousness (TLOC) at ﬁrst presentation. Methods We prospectively recruited patients newly presenting with TLOC to an emergency department, an acute medical unit, and a ﬁrst seizure or syncope clinic. We invited participants to complete an online questionnaire, either at home or at time of initial assessment. Two expert raters determined the cause of participants’ TLOC after 6-month follow-up. We used independent development and validation samples to train a random forest classiﬁer to predict diagnosis from participants’ questionnaire responses and validate classiﬁer performance. We compared classiﬁer performance against penalized linear regression and referrer diagnosis. Results We included 178 participants in the ﬁnal analysis, of whom 46 identiﬁed a witness able to complete an additional witness questionnaire. Given low witness recruitment, we developed a classiﬁer based on patient answers only. A classiﬁer trained on 9 items correctly identiﬁed 63 of 78 diagnoses (80.8%) (95% CI 70.0–88.5), an increase over the accuracy of initial assessing clinicians who were only able to diagnose 70.5% correctly. Within this, 96%(87.0%–99.4%) of those expertly rated as having syncope were correctly classiﬁed by the classiﬁer (classiﬁer sensitivity); 40%(20%–63.6%) of those expertly rated after follow-up as having either epilepsy or functional/dissociative seizures were similarly classiﬁed as being nonsyncope (classiﬁerspeciﬁcity)

FAQs

What was the primary aim of the study?

To develop and validate a machine-learning classifier using patient and witness questionnaires to support differential diagnosis of transient loss of consciousness (TLOC) at first presentation.

How were participants and diagnoses handled in the study?

Patients newly presenting with TLOC were prospectively recruited and completed an online questionnaire. Two expert raters determined the cause after 6-month follow-up, and development and validation samples were used for training and testing.

What performance did the final classifier achieve, and was it ready for routine care?

A patient-only model using nine questionnaire items correctly identified 63 of 78 diagnoses (80.8%). Despite comparable performance to standard care, it was insufficiently accurate for incorporation into routine care in its current form.

Aug 5, 2026

Upload by Miles