Detection of Risky Driving Behaviors in the Naturalistic Environment in Healthy Older Adults and Mild Alzheimer’s Disease
DOI: 10.3390/geriatrics3020013
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This study addresses the challenge of analyzing large volumes of naturalistic driving data to detect risky behaviors in older adults, particularly those with mild Alzheimer’s disease (AD). While naturalistic driving assessments offer high ecological validity, manual review of continuous video is time-intensive and impractical for large-scale application. The authors aimed to validate a semi-automated, computerized method that flags specific high-risk driving epochs—such as rapid stops, lane deviations, turns, and intersections—for manual review, thereby reducing the data burden while maintaining sensitivity to cognitive impairment. The study utilized archival naturalistic driving data from 44 participants with mild AD and 16 age-matched healthy controls (HC). All participants passed a standardized Rhode Island Road Test (RIRT) before having cameras installed in their vehicles for a two-week recording period. The automated system used GPS, speed, and front-view video to detect specific events: turns were identified by direction changes, intersections by proximity to a geographic database, rapid stops by deceleration thresholds, and lane changes by tracking lane markers. Flagged events were manually reviewed and rated using a modified DriveCam® scoring system (Mockingbird), which categorized errors into eight types (e.g., distractions, poor awareness, fundamentals) and assigned severity scores. These error scores were then used to train a logistic model tree classifier to predict diagnostic group membership. Results indicated that the automated event-based method effectively distinguished between groups. AD participants exhibited higher error scores than HC participants when controlling for mileage driven. Specifically, AD drivers made more severe errors related to driving fundamentals, such as lane maintenance and failing to look far enough ahead. In contrast, HC drivers made more errors related to distraction and risky behaviors, such as failing to keep an exit route. The logistic model trained on the top eight error types achieved 91.7% overall accuracy, 97.7% sensitivity, and 75.0% specificity in predicting AD diagnosis, outperforming both the standardized road test and composite ratings of continuous video review. Furthermore, the automated error scores correlated significantly with road test performance and measures of global cognition (MMSE) and maze navigation, validating the method’s clinical relevance. The findings demonstrate that focusing on high-risk driving epochs is a sensitive and cost-effective data reduction strategy for analyzing naturalistic driving data. This approach allows for the efficient detection of cognitive impairment-associated driving behaviors, offering a scalable alternative to full manual video review. The study highlights distinct error profiles between AD and healthy drivers, suggesting that automated monitoring could serve as a valuable tool for assessing driving safety in aging populations and those with early-stage dementia.
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-25 |
| archive | success | openalex | — | — | 5 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-25 |
| chunk | success | chunk | — | — | 1 | 2026-06-25 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-25 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-25 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Methodological Resource: dataset resource, validation psychometrics
- Theoretical Contribution: computational model