Table 1.

Summary of identified papers where GPS has been used in patients with dementia disorders. Participant age is presented as mean (SD) where available.

Author (Year)	Study Population	Device(s) and Location Carried	Duration	Sampling Frequency	Study Design	GPS-Derived Outcomes	Data Processing Details	Key Findings
Oswald et al. [38] (2010)	PwD n = 6, MCI n = 6, HC n = 7 (Participants between age 63–80).	SenTra device: GPS receiver, a RF transmitter wristwatch, and a home RF monitoring system [44]. Location: GPS device located in pouch or bag [44].	4 weeks	0.2 Hz	Observational, cross-sectional study	Distance travelled, Walking speed, Distance from home, Daily mobility activity	The GPS data were transmitted via the GPRS protocol to a project server. A valid hour was ≥30 min of valid GPS data; a day was valid only if there were no invalid hours. Full time analysis was carried out on valid days (methodology of processing was not stated).	This study established that the future proposed SenTra project was feasible. However, the SenTra tracking kit placed high cognitive and behavioural demands on participants.
Shoval et al. [33] (2011)	PwD n = 7 (mean age 81.9), MCI n = 21 (mean age 78.3), HC n = 13 (mean age 72.9).	SenTra kit [44]	4 weeks	0.1 Hz	Observational, cross-sectional study	Distance from home, Time OOH	GPS data transmitted via the GPRS protocol to a project server. Using a combination of a GIS and the recorded locations of the participant, the distance from home was calculated. This information was visualized on a ‘spider-web diagram.’	Participants with cognitive impairment travelled shorter distances from home during the day compared with HCs. PwD had a smaller spatial range compared to those with MCI.
Werner et al. [27] (2012)	PwD n = 16, MCI n = 34, HC n = 26, CG n = 66 (Participants aged 63 or older).	SenTra kit [44]	4 weeks	0.1 Hz	Observational, cross-sectional study	Time spent OOH per day, Time spent walking per day, Number of visited nodes, Number of walking tracks per day, Average walking distance, Average walking speed	GPS data transmitted via the GPRS protocol to a project server. From GPS data a node was defined as a stopping point lasting >5 min. A track was the pathway between nodes. Detail was not presented on the processing methods to gather GPS derived outcomes.	The greater the mobility of PwD (with mobility defined through the GPS derived outcomes), the less burden placed on CGs.
Wahl et al. [39] (2013)	MCI n = 76 (mean age 72.9), HC n = 146 (mean age 72.5).	SenTra kit [44]	4 weeks	0.1 Hz	Observational, cross-sectional study	Time spent OOH per day, Number of visited locations	A valid day was when <1 h of missing data was observed. A visited location was defined as a GPS coordinate staying in the same location for >5 min.	The mean number of visited locations was higher in HCs than those with MCI.
Tung et al. [28] (2014)	PwD n = 19 (mean age 70.7), HC n = 33 (mean age 73.7).	GPS receiver on smartphone chipset (Qualcomm RTR6285, Qualcomm Inc., San Diego, CA, USA) Location: Pocket	3–5 days	1 Hz	Observational, cross-sectional study	Life-space area, Distance from home, Time OOH	GPS coordinates were projected to a 2D plane using Matlab R12. Home radius was set to 25 m around home coordinates determined from the participants address and Google Earth. A convex hull, calculated using the standard convex hull operation, was used to determine the area and perimeter measures. The Euclidean distance from the home coordinates was calculated and a distance time series was produced to determine the time spent OOH and distance from home were calculated.	Reduced mobility was observed in PwD compared to HCs, using measurements of the area and perimeter of the convex hull.
Wettstein et al. [40] (2014a)	PwD n = 35 (mean age 74.1), MCI n = 76 (mean age 72.9), HC n = 146 (mean age 72.5)	SenTra kit [44]	4 weeks	0.2 Hz	Observational, cross-sectional study	Walking distance, Walking speed, Walking duration, Time spent OOH, Number of places visited, Number of walking tracks per day	GPS data transmitted via the GPRS protocol to a project server. A valid day had to have OOH behaviour and <1 h of missing GPS data. A visited location was defined as GPS coordinates in the same location for >5 min. A walking track was considered as movement less than 5 km/h.	In PwD, higher walking distance and walking speed were positively correlated with environmental mastery (how capable an individual feels with using environmental resources).
Wettstein et al. [41] (2014b)	As per Wettstein et al. [40]	SenTra kit [44]	4 weeks	0.2 Hz	Observational, cross-sectional study	Time spent OOH, Number of places visited	The same data processing method was used as per Wettstein et al. [40]. However, the walking tracks were not processed as this was not a GPS derived outcome for this study.	Behavioural competence was significantly lower in PwD than both MCI and HC. The mean number of activities carried out was also lower in PwD compared with MCI and HCs.
Kaspar et al. [37] (2015)	PwD n = 16, MCI n = 30, HC n = 95 (Participants were in the age range 50–84).	SenTra kit [44]	4 weeks	0.2 Hz	Case Control study	Time spent OOH, Average walking distance, Type of activity, Type of transport	The GPS data were transmitted via the GPRS protocol to a project server. A valid day had <1 h of missing data. Spatial GPS data was interpreted using complex algorithms (specific type not stated), which integrated compound measures, such as acceleration and velocity, alongside geographical background data to distinguish transport modes.	The authors were unable to establish a strong relationship between daily mood and an individual’s mobility.
Wettstein et al. [42] (2015a)	As per Wettstein et al. [40]	SenTra kit [44]	4 weeks	0.2 Hz	Observational, cross-sectional study	Walking distance, Walking speed, Walking duration, Time spent OOH, Number of places visited, Number of walking tracks per day	The same data processing method was used as Wettstein et al. [40]. An addition of this study was the cluster method which used GPS-derived outcomes to identify whether the participants were ‘mobility restricted’, ‘outdoor oriented’ or ‘walkers’.	The mobility patterns in older people were heterogenous. However, it was identified that there was a higher proportion of cognitively impaired individuals in the cluster defined as having restricted mobility.
Wettstein et al. [43] (2015b)	As per Wettstein et al. [40]	SenTra kit [44]	4 weeks	0.2 Hz	Observational, cross-sectional study	Walking distance, Walking speed, Walking duration, Time spent OOH, Number of places visited, Number of walking tracks per day	Data processing method the same as Wettstein et al. [40]	The three cognitive ability groups did not significantly differ in OOH walking indicators (e.g., walking speed). However, OOH mobility indicators (time OOH, number of visited locations) were lowest in PwD.
Harada et al. [45] (2019)	PwD n = 147 (The mean age of the n = 192 baseline participants was 76.3 but the age was not stated for those included in the final study)	Globalsat DG-200 Data Logger Location: Pocket	2 weeks	0.033 Hz	Secondary analysis of a randomised controlled trial	Time spent OOH per day	GPS data was processed in accordance with the GIS system (ArcGIS for Desktop 10.3: Esri Japan Incorporation: Tokyo, Japan). Home radius set to 100 m around the home coordinates; the time spent OOH was determined using this radius. Validity of a day was defined as wear ≥10 h, location started and ended in the home area, no poor connection during the time OOH and, the participant stated they wore the device in their travel diary.	In PwD, a stronger social network was positively correlated with greater time spent OOH. However, no relationship between environmental factors and time spent OOH was observed in PwD.
Thorpe et al. [22] (2019)	PwD n = 6 (Mean age 69.7)	Smartphone (Nexus 5) and a Smartwatch (Sony Smartwatch 3) Location: Pocket (smartphone) and Wrist (smartwatch).	8 weeks	Ranging from 1000 Hz to 0.003 Hz	Longitudinal study	MCP, Action range, Total distance covered outside the home, Time spent OOH, Time spent moving between locations, Number of places visited, Number of trips	The GPS data was filtered in alignment with the upper limit set at 25 m accuracy. The stop and moves were determined from the trajectory with the DBSCAN method applied to determine locations. A stop was defined as GPS coordinates in the same location for >5 min. The MCP was calculated using the R function to determine the smallest convex polygon around the data points. The action range was the geodesic distance from home coordinates and the GPS data [23].	Digital monitoring of mobility and activity has the potential to detect fluctuations in behaviour that the participant might not detect themselves.
Bayat et al. [46] (2021)	PwD n = 7, HC n = 8 (All participants were ≥65 or older).	SafeTracks Prime Mobile GPS Device Location: Pocket	8 weeks	0.017 Hz	Case control study	Number of destinations, Sequence of destinations, Time spent at each destination	4 of the 8 weeks of captured GPS data were extracted. Home location of each participant was determined using DBSCAN algorithm. The trajectory segmentation method [47] extracted the locations visited by each participant. Extracted destinations were clustered and each destination was assigned a cluster ID [47]. Different entropy methods (random, heterogeneous spatial and spatiotemporal) and algorithms were used to assess randomness of individuals mobility.	There was lower spatial and temporal randomness in mobility patterns in PwD compared to HC. Therefore, across the collected data there was a 5% chance, on average, that a PwD would choose a location at random but an 8% chance in HC.
Chung et al. [48] (2021)	PwD n = 1, CG n = 1 (PwD 64, CG 62)	Garmin™ Vivoactive HR Location: wrist watch	1 week	Not stated	Case study	Total distance moved, Movement speed, Convex hull area, Total wear time, Location (home vs. other), Total time OOH, Total time at home, Heart rate	The GPS data were extracted in TCX and CSV formats. The participant wore the device longer than the intended 7-day study period therefore generating 9 days of complete GPS data. GPS track plots used to describe locations visited with total distance moved and speed of movement determined for each track. LSM visualized by plotting and calculating the convex hull of GPS points using mapview package (CITE). Home radius was set as ≤1000 ft around home coordinates.	The participant engaged in OOH activities every day from late morning until the evening. The travel diary correlated with the GPS-derived outcomes and provided additional information on the type of activity the participant carried out.
Liddle et al. [49] (2021)	PwD n = 3, MCI n = 15 (Participants mean age 86.7)	Smartphone based GPS system Location: Pocket	Required 105 to 240 h of GPS data.	Not stated	Longitudinal observational study	Life space area, Time at home, Maximum distance from home, Trips OOH, Time left at home	Custom algorithms (not stated) were used to create metrics. The locations extracted from the GPS data were plotted to visualize the life space area and the shape and perimeter of the life space area were analysed. The home area was defined as 500 m from the home location and the time spent OOH was when the participant left the home radius and did not return for a period > 5 min.	The authors found no relationship between life space and cognition. However, an association with life space and driving status was found with non-drivers having a lower life space compared with drivers.
Sturge et al. [50] (2021)	PwD n = 2, MCI n = 5 (Participants were aged 59–93).	QStarz BT—1000X	2 weeks	Not stated	Observational, cross-sectional study	Visited locations, Distance from home, Life Space Area	GPS data extracted and processed in Microsoft Excel then imported into V-Analytics to store the participants locations and trips over the study period and for time-space movement analysis. Activities were created if GPS location points were connected within an 80 m radius for >5 min. GPS locations exceeding this radius were considered as a distinct trip. Activities were imported into ArcMap 10.5.1. to visualize participants’ spatial movement with activities then defined into routine activity space (<7.5 km of the home coordinates) and occasional activity space (>7.5 km).	Cognitively impaired individuals still engaged in activities beyond their neighbourhood area.
Bayat et al. [51] (2022)	PwD n = 7, HC n = 8 (All participants were ≥ 65).	SafeTracks Prime Mobile GPS device Location: Pocket, purse or bag	4 weeks	Not stated	Case control study	Maximum distance from home, Radius of gyration, Life space area, Number of destinations, Number of unique destinations, Time at home, Time OOH, Time on foot, Time in vehicle, Trip time period, Total number of trips, Outdoor activity duration, Types of activities	Data processing as described by Bayat et al. was used in this study [46]. A distance-based probabilistic model based on Google places, API, was used to retrieve information about visited locations of the participants to define their OOH activities (i.e., shopping, leisure, medical services).	PwD undertook more medical-related and fewer sport-related activities compared to HCs. PwD spent less time walking than cognitively intact individuals.

GPS = global positioning system, GIS = geographic information system, GPRS = general packet radio service, RF = radio frequency, GSM = global system for mobile communications, DBSCAN = density-based spatial clustering of applications with noise, PwD = people with dementia, CG = caregivers, MCI = mild cognitive impairment, HC = healthy control, LSM = life space mobility, OOH = out of home, API = application programme interface.