Exploring Test-Retest Reliability and Longitudinal Stability of Digital Biomarkers for Parkinson Disease in the m-Power Data Set: Cohort Study

Abstract

Background Digital biomarkers (DB), as captured using sensors embedded in modern smart devices, are a promising technology for home-based sign and symptom monitoring in Parkinson disease (PD). Objective Despite extensive application in recent studies, test-retest reliability and longitudinal stability of DB have not been well addressed in this context. We utilized the large-scale m-Power data set to establish the test-retest reliability and longitudinal stability of gait, balance, voice, and tapping tasks in an unsupervised and self-administered daily life setting in patients with PD and healthy controls (HC). Methods Intraclass correlation coefficients were computed to estimate the test-retest reliability of features that also differentiate between patients with PD and healthy volunteers. In addition, we tested for longitudinal stability of DB measures in PD and HC, as well as for their sensitivity to PD medication effects. Results Among the features differing between PD and HC, only a few tapping and voice features had good to excellent test-retest reliabilities and medium to large effect sizes. All other features performed poorly in this respect. Only a few features were sensitive to medication effects. The longitudinal analyses revealed significant alterations over time across a variety of features and in particular for the tapping task. Conclusions These results indicate the need for further development of more standardized, sensitive, and reliable DB for application in self-administered remote studies in patients with PD. Motivational, learning, and other confounders may cause variations in performance that need to be considered in DB longitudinal applications.

Things to Consider When Automatically Detecting Parkinson’s Disease Using the Phonation of Sustained Vowels: Analysis of Methodological Issues

“…This is the largest corpus of its kind to date which is publicly available, with 5876 unique participants contributing a total of 65,022 recordings. Previous studies have used the mPower corpus not only to create and test PD detection models [12][13][14]25,27,30,36,37], but also to design real-time PD diagnosis tools and applications [26], classify voice impairment level [38], measure longitudinal reliability and stability of these metrics [20], quantify and improve diagnosis techniques on signals recorded in noisy environments [39,40], and even screen for symptoms of depression reported by PwPD [41].…”

“…This is usually more adequate when employing acoustic analysis or features proposed to be used in the stable part of the phonation, such as complexity or frequency and amplitude perturbation features [18]. Therefore, some studies take corpora distributed with the full recording and shorten them to as short as one second [14,19,20]. Nevertheless, the onset and offset contain articulatory information and might add extra differentiation between classes, especially when employing certain features such as Mel-Frequency Cepstral Coefficients (MFCC) to characterize them.…”

“…Then, we discarded all participants below the age of 30 to create our reference subset of usable participants S R (Table 5). These participants were discarded as PwPD below 30 years old are atypical, might require a special analysis and, at the same time, most studies found on the literature do not consider them [20,25,30,37]. Within S R , the age distribution curves between PwPD and HC subjects are starkly different in both mean and shape, as shown in Figure 2.…”

A smartphone-based tapping task as a marker of medication response in Parkinson’s disease: a proof of concept study

“…As such, the BRAIN test may be less sensitive to subtle motor fluctuations evoked by the small sequential movements needed for the Cloud-UPDRS tapping task (Akram et al 2022 ). Other studies using repetitive or two-finger paradigms required tapping on the spot or on different keys, included no distance component (Akram et al 2022 ; Arora et al 2015 ; Goñi et al 2021 , 2020 ; Lipsmeier et al 2018 , 2021 ; Omberg et al 2021 ; Sahandi Far et al 2021 ; Simonet et al 2021 ; Surangsrirat et al 2022 ). Here, we found, similar to a recent study (Thijssen et al 2022 ), that the average frequency and the inter-tap distance improved with levodopa, with a few exceptions.…”

“…In this study, we aimed to investigate the validity of a finger tapping task on a smartphone (Stamate et al 2018 ) as a marker of patients’ medication status at home. We selected finger tapping because it provides an objective measure of hand bradykinesia (Alberts et al 2021 ; Lipsmeier et al 2021 ; Omberg et al 2021 ; Pal et al 2001 ; Surangsrirat et al 2022 ), has some diagnostic value (Akram et al 2022 ; Goñi et al 2021 , 2020 ; Lee et al 2016a , b ; Lipsmeier et al 2018 ; Noyce et al 2014 ; Omberg et al 2021 ; Sahandi Far et al 2021 ; Simonet et al 2021 ; Trager et al 2020 ) and can be captured easily and safely in an unsupervised setting. As well, several studies on repetitive and alternating tapping using a keyboard, a tablet, or other forms of motion analysis found differences in ON–OFF states (Akram et al 2022 ; Bologna et al 2018 ; Hasan et al 2019 ; Simonet et al 2021 ; Thijssen et al 2022 ; Wissel et al 2017 ).…”