Classification of Drivers with HIV-Associated Neurocognitive Disorders using Virtual Driving Test Performance Data

Grethlein, David; Kandadai, Venk; Dampier, Will · 2023 · Crossref

DOI: 10.32473/flairs.36.133381

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 identifying drivers with neurocognitive impairment (NCI), specifically HIV-associated neurocognitive disorders (HAND), using data from a virtual driving test (VDT). Traditional screening methods for HAND are often resource-intensive, lack ecological validity, and have limited sensitivity for mild cases. The authors aim to determine if VDT performance data can reliably classify NCI status, potentially offering a more accessible tool for identifying individuals who require specialist referral to mitigate dangerous driving behaviors. The researchers analyzed data from 81 participants: 62 people with HIV (PWH), of whom 35 had confirmed HAND, and 19 HIV-negative controls, of whom 7 had non-HAND NCI (e.g., Parkinson’s, Alzheimer’s). All participants underwent a comprehensive neuropsychological assessment to confirm NCI status prior to the VDT. The VDT platform recorded 10 Hz multivariate time series data, from which 2,601 scalar features were extracted. The study employed a two-stage process: first, feature selection using Kruskal-Wallis ranking to identify the top 100 features most indicative of impairment; second, classification using homogeneous ensembles of five classifiers, including Support Vector Machines (SVM), Multi-Layer Perceptrons (MLP), Random Forests, and others. The experiments were conducted across three populations: PWH with HAND, HIV-negative controls with non-HAND NCI, and a combined group. The results demonstrated varying levels of classification accuracy depending on the population. In the PWH population, an SVM ensemble with radial basis function kernels and class balancing achieved 69.4% accuracy and a risk ratio of 2.09 for detecting HAND. In the HIV-negative control group, an MLP ensemble with 20% dropout achieved 84.2% accuracy and a risk ratio of 8.25 for detecting non-HAND NCI. For the combined population, an MLP ensemble with 0% dropout classified general NCI with 63.0% accuracy and a risk ratio of 1.67. Feature analysis revealed that HAND was most associated with deficits in attention, working memory, motor function, and executive function. In contrast, non-HAND NCI was linked to processing speed, executive function, and visuospatial memory, manifesting as erratic gaze and braking behaviors. Combined NCI was characterized by higher driving error accumulation and poor vehicle following distances. The study concludes that VDT data can distinguish drivers with NCI from those without, with higher reliability in specific subgroups than in the combined population. The findings suggest that different cognitive domains underlie driving impairments in HAND versus other neurocognitive conditions. However, the authors note limitations, including the small sample size, which may lead to overfitting, and the binary nature of the NCI classification, which does not account for severity. Despite these constraints, the work supports the potential of virtual driving simulations as a scalable, ecologically valid tool for screening neurocognitive impairments in driving contexts.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-09; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• mci dementia driving
• cognitive impairment
• cognitive capacity variation

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: validation psychometrics, tool software
• Theoretical Contribution: computational model