Smartphone data can reveal early dementia risks during real-world navigation

Study highlights how smartphone-based navigation tasks can serve as an innovative tool for identifying subtle cognitive changes in older adults, potentially offering early insight into dementia risk.

A recent study published in the journal PLOS Digital Health reports that subtle cognitive changes in people with subjective cognitive decline (SCD) can be inferred from smartphone data collected during a wayfinding task.

The importance of diagnosing dementia early

About 58 million individuals are living with dementia worldwide, with 69 million estimated to be in the prodromal stage. The prevalence of dementia is projected to triple by 2050 due to rising life expectancies and population growth in many countries.

Although phase III clinical trials have demonstrated that several drugs can alter the trajectory of dementia, there is currently no cure for this disease. Thus, there remains an urgent need to develop novel diagnostic tools that can assess cognitive functioning in asymptomatic individuals people who may exhibit subtle changes associated with an increased risk of dementia.

Recently, researchers have become increasingly interested in the potential utility of digital cognitive assessment tools, as they have the potential to detect mild cognitive impairment (MCI)-grade episodic memory impairments. These approaches can facilitate the early identification of those at an increased risk of developing dementia, which can provide important insights into when and what interventions should be initiated to ultimately improve patient prognoses.

About the study

Orientation stops as key predictor: The number of orientation stops during wayfinding tasks strongly predicted SCD (subjective cognitive decline) status, offering a practical digital marker for early detection​.

The present study investigated the diagnostic value of movement trajectories and related data during real-world navigation through a smartphone-based wayfinding task. The sample comprised 25 cognitively healthy older adults, 24 younger adults, and 23 SCD patients.

The wayfinding task was performed at the German Center for Neurodegenerative Diseases (DZNE) campus and was assisted by the mobile app “Explore.” All study participants were familiar with the campus area and did not report mobility impairments.

During the task, study participants walked from the DZNE to five points of interest (POIs), covering an 820-meter route. They viewed a map on their smartphone, which showed their current location and pictures and locations of POIs before being told to close the map and subsequently find the POIs.

If needed, study participants were allowed to view the map again, with the number of views recorded. A quick response (QR) code was scanned at the POI to indicate completion and initiate the procedure for the next track.

The relative distances between individual Global Positioning System (GPS) trajectories were measured. Participant subgroups with similar wayfinding patterns were identified using a clustering analysis.

Study findings

The researchers evaluated how well the clusters represented the three participant classes. Most subjects in the first cluster walked directly from one POI to another or showed minor deviations/detours.

Performance variability within SCD group: Although SCD patients generally performed worse than healthy participants, there was notable variability, with some SCD patients demonstrating mid or even high-level performance, underscoring the heterogeneity of this group​​.

The second cluster took a less direct path to POIs, with incorrect turns at certain intersections. Comparatively, the third cluster took paths different from the rest of the sample. The correspondence between participant classes and these clusters was low.

The first cluster comprised 18 younger adults, seven SCD subjects, and 10 healthy older adults. The second cluster included five younger adults, 10 SCD patients, and nine healthy older adults, whereas the third cluster comprised one SCD patient and one healthy older adult.

Five performance measures were also estimated based on user input and GPS data. These data included wayfinding distance, duration, movement speed, number of map views during walking, and number of brief stops while walking.

A latent profile analysis was performed on these measures to identify profiles that were subsequently evaluated for how well they corresponded to participant classes. Study participants in the first profile were high-level performers with less time, distance, quicker movements, and fewer stops and map views while walking. The second and third profiles represented mid and low-level performers, respectively.

Participant classes were well represented by performance profiles. High-level performers were mainly younger adults, two SCD subjects, and five older adults. Most healthy older adults, SCD patients, and five younger adults were mid-level performers. Low-level performance was observed for one healthy older adult and six SCD subjects.

Younger adults significantly differed on all measures, with less time, distance, faster movement, and fewer stops and map views. SCD patients had considerably more stops than healthy older adults, were more likely to view the map while walking, and took more time for task completion.

SCD patients did not walk more distance than healthy older adults, with mean movement speed similar between these two groups. More stops were associated with a significantly greater risk of being an SCD patient.

The number of stops, which varied among SCD patients across tracks, correlated with these individuals' general cognitive and visual memory functioning; however, this association was not statistically significant.

Conclusions

The current study evaluated the potential of using smartphone data of real-world wayfinding to differentiate between SCD patients and healthy older adults. The study findings suggest that behavioral performance indicators contain information on participants’ SCD status and age group.

Healthy younger adults exhibited better overall performance, whereas differences between SCD subjects and healthy older adults were more nuanced. More specifically, differences in the number of stops between SCD patients and healthy older adults were observed. This effect was able to predict SCD status, rendering it a promising digital footprint for cognitive decline associated with dementia.

Journal reference:

• Marquardt, J., Mohan, P., Spiliopoulou, M., et al. (2024). Identifying older adults at risk for dementia based on smartphone data obtained during a wayfinding task in the real world. PLOS Digital Health. doi:10.1371/journal.pdig.0000613