Multitasking additional-to-driving: Prevalence, structure, and associated risk in SHRP2 naturalistic driving data

Bálint, András; Flannagan, Carol; Leslie, Andrew; Klauer, Sheila G.; Guo, Feng; Dozza, Marco · 2020 · openalex

DOI: 10.1016/j.aap.2020.105455

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 investigates the prevalence, structure, and crash risk associated with multitasking additional-to-driving (MAD), defined as engaging in at least two secondary tasks while driving. Addressing a gap in literature where few studies specifically analyze multiple distractions in real-world settings, the authors utilize data from the Second Strategic Highway Research Program Naturalistic Driving Study (SHRP2 NDS). The dataset comprises 3,546 drivers and over 37,000 events, including control driving segments, crashes of varying severity, and near-crashes. The research aims to quantify MAD prevalence across different event types, compute odds ratios (ORs) for crash risk, and visualize the co-occurrence structure of secondary tasks. The methodology involves analyzing secondary task engagement coded from video recordings. For control segments, tasks were coded within random 6-second epochs; for safety-critical events, tasks were coded from 5 seconds prior to the precipitating event until conflict resolution. The study defines MAD as the presence of two or more secondary tasks and Single Task Additional-to-Driving (SAD) as exactly one task. To assess robustness, a sensitivity analysis was conducted using a generalized set of 14 task groups (G14) to determine if results varied based on specific coding definitions. Risk was quantified using odds ratios comparing MAD and SAD against driving with no secondary tasks. Additionally, novel graph-based visualizations were developed to map task prevalence and co-occurrence frequencies, using statistical tests to identify significant deviations from independent occurrence. Results indicate that MAD prevalence increases with event severity: it occurs in 11% of control driving segments, 22% of crashes and near-crashes, 26% of Level 1–3 crashes, and 39% of rear-end striking crashes. Using the generalized G14 definitions yielded slightly lower but consistent prevalence rates. The odds ratios for MAD compared to no-task driving were 2.38 for crashes and near-crashes, 3.72 for Level 1–3 crashes, and 8.48 for rear-end striking crashes. Engagement in three or more tasks further elevated risk, with an OR of 16.53 for rear-end striking crashes. Visual analysis revealed that task co-occurrence structures differ significantly between normal driving and safety-critical events; for instance, passenger interaction and internal eye glances co-occur significantly more in crashes than in control segments. The study concludes that multitasking substantially elevates crash risk, particularly for rear-end collisions, confirming that performance reductions observed in simulator studies manifest in real-world crashes. The findings remain consistent regardless of whether specific SHRP2 codes or general task groups are used, suggesting the risk associated with MAD is robust to definition variations. The developed visualization tools provide a means to map complex task interactions, highlighting that the number and type of concurrent tasks vary distinctly across driving conditions. These results underscore the critical safety implications of multiple driver distractions and support the need for interventions targeting multitasking behaviors.

Key finding

Multitasking additional-to-driving significantly increases the risk of safety-critical events, with the highest odds ratio observed for rear-end striking crashes.

Methodology

naturalistic

Sample size: 3546

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 2026-06-10; verification: verified.

Topics

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

• temporal
• naturalistic crash near crash
• visual manual
• distraction detection algorithms
• incidence prevalence
• manual

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).

• Empirical Findings: observational prevalence, crash risk outcomes
• Methodological Resource: dataset resource