EMBalance - validation of a decision support system in the early diagnostic evaluation and management plan formulation of balance disorders in primary care: study protocol of a feasibility randomised controlled trial

Abstract

BackgroundBalance problems are caused by multiple factors and often lead to falls and related fractures, bringing large socio-economic costs. The complexity of balance control mechanisms, the lack of medical expertise, and the absence of specialised equipment contribute to the delayed or incorrect diagnosis and management ofthese patients. Advances in computer science have allowed the development of computer systems that support clinical diagnosis and treatment decisions based on individualised patient data. The aim of the EMBalance decision support system (DSS) is to support doctors facing this clinical challenge, to make a definitive diagnosis and implement an effective management plan. The EMBalance study will determine the accuracy of this supportive tool when used by non-specialist doctors. This study is funded by the European Union’s Seventh Framework Programme.Methods/designEMBalance is a proof-of-concept study designed as a non-commercial, international, multi-centre, single-blind, parallel-group randomised controlled trial to be carried out at four clinical sites in the United Kingdom, Germany, Greece and Belgium. The study is comprised of three stages: internal pilot, phase I (diagnosis) and stage II (management). For this purpose, 200 patients presenting with persistent dizziness (>3 months’ duration) to primary care services will be randomised to either the intervention group (diagnostic assessment with the DSS) or a control group (diagnostic assessment without the DSS). Patients allocated to the intervention group will be assessed by a doctor with the support of the EMBalance DSS, while patients allocated to the control group will receive a visit as per standard practice. Ultimately, all patients’ diagnoses and management plans will be certified by a consultant in neuro-otology.DiscussionEMBalance is the first trial to test the accuracy of a DSS in both the diagnosis of and the management plan for vestibular disorders across the healthcare systems of four different countries. The EMBalance study is the result of a combined effort of engineers and physicians to develop an accurate tool to support non-specialist doctors, with no risk for the patient. This trial will provide reliable information about the benefits of implementing DSSs in primary care while supporting the feasibility of testing the EMBalance algorithms in further research.Trial registrationClinicalTrials.gov NCT02704819. Registered 29 February 2016.Electronic supplementary materialThe online version of this article (doi:10.1186/s13063-016-1568-x) contains supplementary material, which is available to authorized users.

Diagnostic accuracy and usability of the EMBalance decision support system for vestibular disorders in primary care: proof of concept randomised controlled study results

“…The overall concept of the EMBalance project [ 31 ] and the methodology for the EMBalance DSS development [ 27 ] have been previously described. Briefly, the participating medical partners first established and agreed “gold standard” criteria for the diagnosis of balance disorders and treatment guidelines, following nomenclature, classification and recommendations of the Bárány Society ( http://www.baranysociety.nl/ ), i.e.…”

“…The EMBalance project [ 27 , 30 ] aimed to develop and validate a web-based platform used by primary care physicians for the early diagnostic evaluation, and effective management of balance disorders. Herein, we describe the proof of concept clinical evaluation of the EMBalance DSS by means of a study conducted as per the published protocol [ 31 ].…”

Artificial Intelligence Applications in Otology: A State of the Art Review

“…Most recently, a diagnostic system for 12 different disorders by Exarchos et al 28 that could also advise on efficient management is currently being validated in a feasibility randomized controlled trial for diagnosis and management of vestibular disorders across 4 different countries. 29 Therefore, there are numerous machine-learning methods, and most methods, when optimized, fared similarly well. 30…”

“…Most recently, a diagnostic system for 12 different disorders by Exarchos et al 28 that could also advise on efficient management is currently being validated in a feasibility randomized controlled trial for diagnosis and management of vestibular disorders across 4 different countries. 29 Therefore, there are numerous machine-learning methods, and most methods, when optimized, fared similarly well. 30 Beyond expert systems, machine-learning techniques have also been used to power assessment tools that can provide patients with balance indices when training with interactive virtual reality vestibular rehabilitation games with improved vestibular outcomes 31 as well as to identify postural sway patterns typical for balance disorders in a posturography with an overall sensitivity and specificity of 0.93 for postural phobic vertigo, anterior lobe cerebellar atrophy, primary orthostatic tremor, and acute unilateral vestibular neuritis.…”

When vestibular rehabilitation can assist: findings with use of data mining

“…This retrospective dataset, that was built for the development of the EMBalance Decision Support System (DSS) [16], includes data from 1010 patients diagnosed with peripheral and central vestibular disorders (unilateral peripheral dysfunction/failure, psychological disorders, bilateral vestibular failure/dysfunction, cerebellar/pontine lesion, acoustic neuroma, chronic subjective dizziness, persistent postural-perceptual dizziness (PPPD), vestibular neuritis, Menière's disease, migrainous vertigo, vestibular paroxysmia, typical and atypical posterior benign paroxysmal positional vertigo (BPPV), that were collected from the National Hospital for Neurology and Neurosurgery, Queen Square, UK, the 1st Department of Otolaryngology, National Kapodistrian University of Athens (NKUA), Greece, the University of Antwerp, Belgium and the University Clinic of Freiburg, Germany. These data contained more than 160 features (variables), including patients' demographics, medical and disease related history, clinical and laboratory examination results (including auditory tests), treatment (besides medication) and medication (see Table I).…”