Lo, C., Arora, S., Baig, F., Lawton, M., El Moulden, C., Barber, T. R., Ruffmann, C., Klein, J., Brown, P., Ben-Shlomo, Y., de Vos, M., & Hu, M. T. (2019). Predicting motor, cognitive and functional impairment in Parkinson’s. Annals of Clinical and Translational Neurology, 6(8), 1498-1509. https://doi.org/10.1002/acn3.50853 Publisher's PDF, also known as Version of record License (if available): CC BY Link to published version (if available): 10.1002/acn3.50853 Link to publication record in Explore Bristol Research PDF-document This is the final published version of the article (version of record). It first appeared online via Wiley at https://doi.org/10.1002/acn3.50853 . Please refer to any applicable terms of use of the publisher. University of Bristol - Explore Bristol Research General rights This document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: http://www.bristol.ac.uk/red/research-policy/pure/user-guides/ebr-terms/ RESEARCH ARTICLE Predicting motor, cognitive & functional impairment in Parkinson’s Christine Lo1,2 , Siddharth Arora1,3, Fahd Baig1,2 , Michael A. Lawton4 , Claire El Mouden1,2 , Thomas R. Barber1,2 , Claudio Ruffmann1,2, Johannes C. Klein1 , Peter Brown2,5 , Yoav Ben-Shlomo4 , Maarten de Vos6 & Michele T. Hu1,2 1Oxford Parkinson’s Disease Centre (OPDC), University of Oxford, Oxford, UK 2Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, UK 3Somerville College, University of Oxford, Oxford, UK 4Population Health Sciences, University of Bristol, Bristol, UK 5Medical Research Council Brain Network Dynamics Unit, University of Oxford, Oxford, UK 6Institute of Biomedical Engineering, University of Oxford, Oxford, UK Correspondence Abstract Christine Lo, Level 6, West Wing, John Radcliffe Hospital, Headley Way, Oxford, OX3 Objective: We recently demonstrated that 998 features derived from a simple 9DU. Tel: 01865 226649; 7-minute smartphone test could distinguish between controls, people with E-mail: [email protected] Parkinson’s and people with idiopathic Rapid Eye Movement sleep behavior disorder, with mean sensitivity/specificity values of 84.6-91.9%. Here, we inves- tigate whether the same smartphone features can be used to predict future clin- Funding Information ically relevant outcomes in early Parkinson’s. Methods: A total of 237 This study was funded by the Monument Trust Discovery Award from Parkinson’s UK participants with Parkinson’s (mean (SD) disease duration 3.5 (2.2) years) in (J-1403) and supported by the National the Oxford Discovery cohort performed smartphone tests in clinic and at home. Institute for Health Research (NIHR) Oxford Each test assessed voice, balance, gait, reaction time, dexterity, rest, and postu- Biomedical Research Center (BRC). The views ral tremor. In addition, standard motor, cognitive and functional assessments expressed are those of the authors and not and questionnaires were administered in clinic. Machine learning algorithms necessarily those of the NHS, the NIHR or the were trained to predict the onset of clinical outcomes provided at the next 18- Department of Health. month follow-up visit using baseline smartphone recordings alone. The accu- racy of model predictions was assessed using 10-fold and subject-wise cross val- Received: 26 February 2019; Revised: 26 June 2019; Accepted: 3 July 2019 idation schemes. Results: Baseline smartphone tests predicted the new onset of falls, freezing, postural instability, cognitive impairment, and functional impair- Annals of Clinical and Translational ment at 18 months. For all outcome predictions AUC values were greater than Neurology 2019; 6(8): 1498–1509 0.90 for 10-fold cross validation using all smartphone features. Using only the 30 most salient features, AUC values greater than 0.75 were obtained. Interpre- doi: 10.1002/acn3.50853 tation: We demonstrate the ability to predict key future clinical outcomes using a simple smartphone test. This work has the potential to introduce individual- ized predictions to routine care, helping to target interventions to those most likely to benefit, with the aim of improving their outcome. Introduction to the minimally clinically important 3-point difference on the Movement Disorders Society Unified Parkinson’s Significant heterogeneity in Parkinson’s influences clinical Disease Rating Scale (MDS-UPDRS) that neuroprotective presentation, progression, medication response, and dis- treatment trials are often powered to detect.1–3 While ease complication risk. The Oxford Discovery and Track- such differences are observed in cohort studies, individu- ing Parkinson’s cohorts provide around 2500 community- alized predictions remain challenging. ascertained patients, prospectively followed from early Disability in Parkinson’s is mainly determined by the diagnosis, in whom these phenotypic variations can be onset of postural instability, falls, freezing of gait, and studied.1 We used data-driven approaches to identify fast dementia.4,5 The time to reach these disease milestones and slow motor progressor subtypes, with differences akin varies considerably, leading to increased outcome 1498 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. C. Lo et al. Smartphone Predictions in Parkinson’s variation and requiring larger sample sizes to demonstrate for device details).18 Smartphone tests assess: (1) Voice potential treatment effects.6 A number of models have (participants hold the phone to their ear, take a deep aimed to predict these clinically relevant outcomes. A 3- breath, and say “aaah” at a comfortable and steady tone step falls prediction tool by Paul et al. attached the great- and level, for as long as possible); (2) Balance and (3) est weighting to whether individuals reported falling at Gait (with the phone in a trouser pocket or arm band, baseline, yet limited numbers prevented the prediction of participants stand still and then walk a distance of 20 the onset of falls in those without falls at baseline, an out- yards before turning and walking back); (4) Dexterity come of greater interest to the treating clinician.7,8 (participants tap alternately between two buttons on the Ehgoetz et al. recently reported a logistic regression screen at a comfortable rate); (5) Non-cued reaction time model utilizing the Hospital Anxiety and Depression Scale (participants press on a button as it appears on the (HADS) and the Freezing of Gait (FOG) questionnaire screen, keeping their finger down whilst it is there and total score to predict the onset of future freezing, which lifting their finger off as it disappears); (6) Rest and (7) requires external validation.5 Velseboer et al. described a Postural tremor (participants hold the smartphone in the similar model utilizing age, the numbers of animals hand most affected by tremor if they have tremor, or named in a verbal fluency task and the UPDRS axial score their dominant hand if they do not, while their hand is at to predict a composite adverse outcome but it was not rest or held outstretched in front of them). possible to distinguish between death, dementia or postu- The seven smartphone tasks take 6–7 min in total to ral instability.9 Models to date have relied on combina- perform. All seven tasks constitute one smartphone tions of different clinical questionnaires and assessments, recording. Incomplete recordings, where all seven tasks requiring time and skill to administer, in order to make were not performed within a 15-min time period, were specific predictions; to the best of our knowledge no sin- excluded from analysis. gle test has been able to predict multiple future clinical Clinical data collected at in-person 18-monthly longitu- outcomes. dinal clinic visits were matched to smartphone recordings A multi-device study is being planned, that uses smart- performed in clinic and at home within 3 months of the phones to capture questionnaire data and to store tremor clinic visit; henceforth referred to as a time window. data recorded by smartwatches, alongside tablet-based Smartphone recordings contributed at different time win- assessments, with the aim of differentiating Parkinson’s dows, related to different longitudinal clinic visits, were from Essential Tremor.10 However so far, studies using treated independently for the purpose of analysis. Smart- smartphones alone have focused on equipping the clinician; phone recordings analyzed were collected between 8 in distinguishing individuals with and without Parkin- August 2014 and 7 November 2017. son’s11,12 and working to derive smartphone scales with Future outcomes were defined according to the results which to measure disease severity.13,14 Our aim was to use of clinical assessments and questionnaires performed at smartphone data to predict outcomes of direct clinical rele- the next 18-month clinic visit as detailed in (Table 1) and vance to people with Parkinson’s and clinicians, alike. included the new onset of (1) falls (>1 self-reported fall in the preceding 6 months), (2) freezing (a freezing fre- Methods quency of at least “about once a month” on the FOG questionnaire19), (3) Postural instability (Hoehn and The Oxford Parkinson’s Disease (PD) Centre Discovery Yahr20 stage ≥ 3), (4) Cognitive impairment21 (a Mon- study15 is a longitudinal cohort study that recruits partici- treal Cognitive Assessment (MoCA)22 score < 26), (5) pants with early Parkinson’s who fulfil the United King- Difficulty doing hobbies (MDS-UPDRS23 part II item 2.8 dom PD Brain Bank criteria for probable PD.16 score ≥ 3 indicating major difficulty or an inability to do Continued participation depends upon participants being activities of enjoyment) and (6) the self-reported need for ascribed a probability of Parkinson’s of at least 90% by future help at home.