Table 1.

Summary of identified papers where GPS and accelerometer devices have been used in patients with neurodegenerative disorders. Participant age is presented as mean (SD) where available.

Author (Year)	Study Population	Device; Anatomical Site	Duration	Study Design	Reported Sensor Derived Outcome Measures	Data Processing Details	Key Findings
		Wearable GPS based studies
Zhu et al. (2020) [15]	PD, n = 54, age 67.5 (6.3) yr	WIMU-GPS; upper back	14 days (waking hours)	Cross-sectional	Trip frequency, duration outside, life space size	Displacement and LSM data reported using a geospatial statistical approach based on computation of minimum span ellipse fitting all data points for each individual. Surface area of minimum span ellipse used to quantify life space size. Excluded if <6 days (>10 h/d) wear.	Reasonable agreement between GPS data, LSA, and mobility diaries for trip frequency and duration, but not life space size. GPS may help overcome the floor and ceiling effects of the LSA.
		Accelerometer based studies
van Uem et al. (2018) [26]	PD, n = 39; PDD n = 8, age 69 yr (mean of both groups)	DynaPort Minimod; lower back	3 days	Cross-sectional	Time spent in sedentary and active episodes, TEE, number of ”bouts”, mean bout length	Data collected at 100 Hz. Self-report of non-wear and special activities (i.e., sports). Raw data analysed with manufacturer’s proprietary algorithms.	Total energy expenditure predicts PDQ-SI and PDQ-Mobility scores. Sedentary activity predicts PDQ-Mobility.
Pradhan & Kelly (2019) [27]	PD, n = 30, age 68.6 (5.9) y; healthy older adults n = 30, age 67.4 (4.7) yr	Fitbit Charge HR; wrist	14 days	Cross-sectional	Step count, number of bouts, timing of bouts of activity	PA intensity determined by proprietary algorithms that define MET values based on steps/minute and heart rate activity. PA intensity defined as: very active (≥6 METs); fairly active (3–6 METs), lightly active (1–3 METs), and sedentary.	Individuals with PD spent fewer minutes per day in moderate/vigorous activity, more minutes sedentary vs. healthy controls.
Mantri et al. (2019) [28]	PD, n = 30 #, median (IQR) age 70 (69–74) yr	Actigraph GT3X; waist	7 days	Cross-sectional	Step count, MVPA	Raw activity data was analysed in 30 s epochs. PD-specific algorithm [29] used to define steps and activity intensity. MVPA defined as >3.0 METs, or the equivalent of >1.31 m/s brisk walk. Valid wear time ≥ 10 h, excessively high counts excluded.	Median step count correlates with PASE.
Sulzer et al. (2021) [30]	PD, n = 45 #, median (range) age 71 (44–80) yr	DynaPort Minimod; lower back	3 days, data only collected at baseline	Longitudinal	Time spent in physical behaviour, number of behaviour bouts, timing of bouts of activity, step count, TEE	Manufacturer’s proprietary algorithms used to define posture/activity. Lying and sitting combined as SB. Standing, shuffling, and walking periods combined as active behaviour. For valid days (24 h registration, wear time > 80%), behaviours further categorised by time spent in behaviour, no. of behaviour bouts (i.e., >1 s, walking > 3 consecutive steps), mean bout length. TEE and step count calculated.	Longer sedentary bouts predict illness and death.
Moyle et al. (2017) [24]	Dementia, n = 192, age 85.5 (7.7) yr	SenseWear activity-armbands; arm	24 h	Cross-sectional	Step count, TEE, MET, Time spent physically active, lying down, awake, asleep	All data recorded in 60 s epochs. Analysed with manufacturer’s proprietary algorithms based on AI. Valid wear-time defined as ≥21 h.	Individuals in care home engaged in light physical activity, varying dependent on mobility factors, age, and sex.
Varma & Watts (2017) [31]	Alzheimer’s disease, n = 39, age 73.5 (7.9) yr; controls, n = 53, age 73.2 (6.5) yr	Actigraph GT3X+; hip	7 days	Cross-sectional	Intensity, activity type	Aggregation of VM data into 60 s epochs, average VM calculated for minute. Excluded days with <10 h of wear time. Wear/non-wear time classification algorithm used [32]. Intensity defined as 0–149 CPM for SB, 150–2689 CPM for light-intensity, and ≥2690 CPM for MVPA.	Mild AD is associated with lower intensity and complexity of the activity. Not associated with longer sedentary activity.
Kim et al. (2019) [20]	PD, n = 46, age 68 (7.9) yr	ActiGraph GT3X+; wrist and waist	7 days	Cross-sectional	Step count, activity count, time spent in activity type	Activity levels defined as SB (≤100 CPM), light (101–1951 CPM), and moderate (1952–5737 CPM) to vigorous PA (≥5738 CPM). PA intensity determined using cut-off points [33]. Walking speed classed as <1.04 m/s, 1.05 to 1.3 m/s, and > 1.31 m/s. PD-specific algorithm applied to percent time spent in different walking speeds [34].	Higher numbers of steps recorded with wrist accelerometer versus waist accelerometer, however, remained proportionate. Variation in recordings linked to increased tremor and bradykinesia.
Klenk et al. (2016) [35]	NDD (n = 103; n = 34 ataxia, age 58(11.3) yr, n = 15 PSP, age 66.2 (5.5) yr, n = 16 PD, age 72.2 (6.5) yr); n = 18 healthy young adults, age 28.3 (9.6) yr; n = 38 healthy older adults age 70.6 (3.9) yr	ActivPAL3; thigh	7 days	Cross-sectional	Walking duration, step count, cadence, walking bout length, absolute and relative number of walking bouts, number of daily sit-to-stand transfers	Manufacturer’s proprietary algorithms detect upright posture and classified activities as lying or sitting, standing, and walking. Days < 24 h data considered invalid.	Daily walking duration decreased for PD, ataxia and PSP groups versus health controls. Daily walking duration was lower for PD, ataxia and PSP groups compared to healthy controls; walking duration diminished as the diseases progressed.
Porta et al. (2018) [36]	PD, n = 18, age 68 (10.8) yr	ActiGraph GT3X; wrist	3 months	Cross-sectional	Step count, intensity of activity by time period	Data collected using 60 s epochs at 30 Hz, processed in manufacturer’s software. Valid wear-time >16 h/day. Step counts and PA classification based on the cut-off points for wrist-worn accelerometry in older adults [37], and (hip-worn) PD specific algorithms [29,38]. Nero algorithm provides walking speeds cut-points.	Peaks of physical activity in early morning and early evening. Physical activity intensity correlates with quantitative spatiotemporal and kinematic gait parameters.
van Eijk et al. (2019) [39]	Amyotrophic lateral sclerosis, n = 42, age 60 (12) yr	ActiGraph GT9X Link; hip	7 days every 2–3 months	Longitudinal	Activity count, MET	Data captured at 30 Hz, processed using manufacturer’s software in 10 s epochs using low-frequency extension (LFE) algorithm. Non-wear time algorithm applied [32], no activity defined as 150 min in advanced disease. Days < 8 h wear-time excluded. Percentage of time active estimated as the proportion of VM counts > 100 CPM. VM counts translated to METs to summarize average daily MET score.	Decline in physical activity of 0.64% per month. Correlation in decline of activity with ALSFRS daily function score.
Cerff et al. (2017) [40]	PD, n = 48: n = 17 PD-NC age 71 yr, n = 22 PD-MCI age 68 yr, n = 9 PDD age 72 yr	DynaPort Minimod; lower back	3 days	Cross-sectional	Step count, posture, Type, duration, intensity and volume of activity MET	Data collected at 100 Hz and a resolution of 1 milli g -force, manufacturer’s algorithms defined posture/activities. Lying and sitting combined as SB. Standing, shuffling, and walking combined as active behaviour. Days with <24 h, or wear time <80% excluded.	Increased number and longer sedentary bouts for individuals with PD dementia versus PD MCI and PD with cognitive impairment.
Paul et al. (2016) [41]	PD, n = 92, median (IQR) age 67.3 (7.8) yr	StepWatch 3 Step Activity Monitor; ankle	7 days	Cross-sectional	Step count, intensity	Stride counts recorded in 1-min intervals, manufacturer’s software used to transform stride counts into step counts (i.e., step count = stride count × 2) and to calculate amount of step activity, and minutes/day of MVPA (i.e., minutes > 100 steps/minute).	Two days of monitoring determined as adequate period to obtain sufficient activity level estimates.
Elazari et al. (2016) [42]	PD, n = 99 age 64.8 (9.5) yr; healthy older adults, n = 38 age 78.7 (4.4) yr	Dynaport Hybrid; lower back	3 days	Cross-sectional	Type, duration, posture	Data recorded at 100 Hz, analysed using MATLAB. Activities detected automatically based on the local mean of acceleration signals. Transitions between activity segments automatically determined. ML algorithm applied to assess the ability of the entire feature set to discriminate between PD and non-PDs. The algorithm identified the features of PD using SVM in MATLAB.	Machine learning algorithm is able to differentiate walk-to-sit and sit-to-walk transitions which vary according to the severity of PD and between patients with PD and healthy older adults.
Nero et al. (2016) [34]	PD, n = 91, age 73 (6) yr	Actigraph GT3X; hip	7 days	Cross-sectional	Duration, type	Manufacturer’s software used to process raw data in 15 s epochs, episodes of ≥90 min of consecutive 0 s considered non-wear and excluded. Data excluded <4 days data. Normal data band-pass filter option utilized. PA expressed per day in total activity counts (TAC). Brisk walking (>1.0 m/s) in minutes/day calculated [43]. Sedentary time calculated using cut-off points [44].	Motor impairment is negatively associated with PA. Reported an association between PA and both physical function and balance. Motor impairment, body mass index and dyskinesia contribute to the variation in PA levels. PA and brisk walking had principally different associated factors.
Loprinzi et al. (2018) [45]	PD, n = 25, 68.7 yr	ActiGraph GT1M; hip	1–2 weeks	Cross-sectional	Type, intensity	Minimum of 4 days with ≥10 h per day data were included in the analyses. Non-wear defined by a period of a minimum of 60 consecutive minutes of zero activity counts, with the allowance of 1–2 min of activity counts between 0 and 100. Activity CPM of ≥1952 denoted MVPA intensity.	Regardless of motor impairment, age and gender. MVPA improves cognition quantified with MoCA.
Buckley et al. (2020) [46]	Aged residential care residents, n = 257, age 84.5 (7.2) yr, Dementia, n = 34, age 81.53 (7.07) yr	Axivity AX3; lower back	7 days	Cross-sectional	Steps, volume, walking time, bouts per day, pattern or variability of activity behaviours	Programmed to sample acceleration at a frequency of 100 Hz (range ±8 g). Only participants who had at least 3 full days of analysis. Custom automated MATLAB algorithm calculated PA walking outcomes according to a previously established gait model [19,47]	Optimal number of days recording for reliable estimate of activity determined to be between 2–5 days, dependent on required level of dementia care.
van Alphen et al. (2016) [48]	Institutionalised dementia, n = 83, age 83 (7.6) yr Community dwelling dementia, n = 37, age 77.3 (5.6) yr Healthy older adults, n = 26, age 79.5 (5.6) yr	Actiwatch AW-4; wrist	6 days	Cross-sectional data from a longitudinal study	Sedentary behaviour, type, Intensity, amount, duration, patterns	Data collected at 32 Hz using 60 s epochs. Raw data, evaluated without cut-off points. 0 assigned to epochs without acceleration. PA quantified as time spent in specific zones of activity counts with ranges of 100 CPM, <100 CPM classed as SB.	Institutionalised individuals with dementia are more sedentary than community-dwelling individuals with dementia who in turn are sedentary for longer than healthy controls.
von Rosen et al. (2021) [49]	PD, n = 301, age 71.4 (6.4) yr	Actigraph GT3X+; hip	7 days	Cross-sectional	Duration, type, intensity, sedentary behaviour	Manufacturer’s software used to process raw data, collected using 60 s epochs. Non-wear defined as ≥60 min of 0 counts, allowing for 2 min of counts between 0 and 100. Data included from participants with at least 1 day with ≥10 h wear time. Epochs classified as intensity levels using cut-off points for older adults: SB (<100 CPM), LIPA (100–1040 CPM), and MVPA (≥1041 CPM) [50].	The group categorised as being ”sedentary” as opposed to “light-mover” or “steady-mover” performed worse on tests of balance, mobility, and had an increased likelihood of falling.
Moyle et al. (2018) [51]	Dementia, n = 455, age 85.3 (1.0) yr	SenseWear Professional 8 armband; arm	24 h at baseline, during two single days in weeks 5, 10 and 15	Longitudinal/RCT	Step count, duration, MET Time spent lying down, awake, and asleep	Manufacturer’s proprietary software and algorithms used. Data recorded in 60 s epochs. Valid wear time for analysis defined as ≥21 h.	A robotic seal intervention given to individuals cared for in a dementia care home led to a reduction in step count both during the daytime and at night and a reduction in daytime PA.
Murphy et al. (2017) [52]	Dementia, n = 20, age 78.7 (11.8) yr	Sensewear Armband; arm	7 days	Cross-sectional	TEE, step count, sedentary behaviour, duration	Limited data processing information provided.	TEE positively correlated with weight and negatively correlated with duration of sleep.
Leavy et al. (2021) [53]	PD, n = 89, age 71.0 (6.0) yr	Actigraph GT3x; hip	7 days	Cross-sectional	Intensity, step count, duration, sedentary behaviour	Manufacturer’s software used to process raw data, collected in 60 s epochs. Non-wear defined as ≥60 min of 0 counts, allowing for 2 min of counts between 0 and 100. Data included if ≥4 days with ≥9 h wear-time. Epochs classified as intensity levels using cut-off points: SB (<100 CPM), LIPA (100–1040 CPM), and MVPA (≥1041 CPM) [54,55].	Data collected during COVID-19 pandemic. 67% reported a pandemic-related reduction in exercise habits. Being female, aged over 70 years and older, and mobility problems associated with being less physically active.
Kok et al. (2017) [56]	Dementia, regular special care unit n = 48, age 82.9 (8.3) yr Small-scale homelike SCU, n = 67, age 83.3 (6.3) yr	Actiwatch; wrist	7 days	Longitudinal study, data collected at 3-month intervals	Duration, sleep activity	Limited data processing information provided. Incomplete recordings were excluded from analysis. Valid data was collected 7 × 24 h for each measurement. In case missing or invalid data device worn for one more week.	Patients with dementia moved from regular Special Care Unit (SCU) to a small scaled homelike SCU. No effect on activity levels between groups following intervention.
Hartman et al. (2018) [57]	Dementia n = 45, age 79.6 (5.9) yr Control, n = 49, age 80.0 (7.7) yr	Phillips Acti-watch 2; wrist	7 days	Cross-sectional	Sedentary behaviour, intensity, duration	Manufacturer’s software used to export data, sleep intervals manually set and excluded by custom software in MATLAB. Measures obtained every 30 s, wrist accelerations translated to counts to estimate PA/SB. Data included if ≥6 valid days (>10 h). Counts per epoch converted to CPM. Cut-off points of 145, 145–274, 274–597, and >597 CPM used for SB and very light, light-to-moderate, and MVPA. Prolonged SB defined as 30 min with <145 CPM without 1 min of >145 CPM.	People with dementia demonstrated increased sedentary periods and decreased time in light-moderate and moderate-vigorous activity as compared with healthy controls.
Fleiner et al. (2016) [58]	Dementia, n = 45, age 45 (7) yr	uSense sensor device; lower back	3 days	Cross-sectional as part of an RCT	Posture, intensity, type, sedentary behaviour, duration	Data sampled at 100 Hz, limited data processing information provided.	Overall low level of activity and wide variety of activity patterns observed. Patients spent 45% of their time lying, 41% sedentary while sitting or standing, 7% active while sitting, and 7% walking.
Garshol et al. (2020) [43]	Dementia, farm-based n = 29, age 74.0 (7.2) yr Regular day care, n = 39, age 83.4 (8.1) yr	Actisleep+ and Actigraph; wrist	7 days	Cross-sectional	Intensity, sedentary activity, duration, step count	Manufacturer’s software used to report PA, limited data processing information provided.	Higher levels of moderate activity were observed in individuals attending farm-based day care activities versus standard care, whilst similar time spent on sedentary or light activity for both groups.
Adams et al. (2017) [59]	PD, n = 16, age 68 (8.7) yr HD (n = 15), age 55.0 (10.7) yr Prodromal HD, n = 5, age 38 (8.6) yr Controls without a movement disorder n = 20, age 58 (16.2) yr	BioStampRC; placed on chest, limb, thigh and forearm	2 days	Cross-sectional	Posture, duration, type	Data collected at 31.25 Hz. Chest and thigh sensors identified posture. Further analysis differentiated walking and standing. Walking durations identified with normalized autocorrelation-based analysis [60]. Walking split into epochs within upright posture intervals, lags of 0.3–1.2 s considered as plausible range of step durations.	Individuals with PD spend less time lying down than those with Huntington’s disease but a similar amount of time to healthy controls.
Feng et al. (2020) [61]	Case control study: Idiopathic REM sleep behaviour disorder (iRBD), n = 88, age 69.8 (7.7) yr Non-RBD controls, n = 44, age 70.1 (10) yr Clinically diagnosed alpha-synucleinopathies *, n = 44, age 70.7 (8.8) yr Prospective nested case-control study: non-convertors **, n = 66, age 70.9 (7.5) yr; convertors, n = 22, age 72.1 (7.6) yr	Actiwatch Spectrum Plus, wrist	7 days	Case-control study and a prospective nested case-control study	Pattern, duration, sleep-activity, type (rest-activity patterns)	Manufacturer’s software used to report data by 60 s epochs. Data excluded as invalid/missing data if exceeded 33% of total for day/interval.	Actigraphy-derived napping (percentage of time, episodes, and duration) occurred more in those with PD and DLB, than iRBD and, in turn, controls with a similar pattern of decreasing activity across the three groups. More napping and lower activity levels were seen in the iRBD group who had a clinically diagnosed alpha-synucleinopathies (at ~2 years) as compared with those that had not “converted”.

n, number; PD, Parkinson’s disease; AD, Alzheimer’s disease; ALS, amyotrophic lateral sclerosis; DLB, dementia with Lewy bodies; IMU, inertial measurement unit; iRBD, idiopathic rapid eye movement sleep behaviour disorder; GPS, global positioning system; PA, physical activity; LSA, life space assessment; PDQ, Parkinson’s Disease Questionnaire; TEE, total energy expenditure; ALSFRS-R, the ALS Functional Rating Scale; VM, vector magnitude; MVPA, moderate-to-vigorous physical activity; PASE, Physical Activity Scale for the Elderly; PSP, progressive supranuclear palsy; MET, metabolic equivalent of task estimates; MoCA, Montreal Cognitive Assessment; MCI, mild cognitive impairment; REM, rapid eye movement; SB, sedentary behaviour; CPM, counts per minute; SVM, support vector machine; AI, artificial intelligence; IQR, interquartile range; * alpha-synucleinopathies are the specific spectrum of neurodegenerative conditions that include PD; ** converters are individuals with REM sleep behaviour disorder who go on to develop PD, a known risk factor [62]; #, sensor data from subset.