Peer-reviewed Dingus et al., Proceedings of the National Academy of Sciences (PNAS)

Driver crash risk factors and prevalence evaluation using naturalistic driving data

Cited in 4 Likelier entries (4 risks, 0 decisions).

Used in 4 entries

For each citing entry, the verbatim excerpt and Likelier's calculation notes (how the source's number was converted to the lifetime-probability framing) are shown below. Click through to read the full claim ledger.

[1] Eating while driving Risk

Statistic

Overall distraction while driving associated with 2.0x crash risk versus model driving; manual secondary tasks (including eating, reaching, grooming) contribute materially to the 68.3% of crashes in which distraction was a factor

“"The overall risk of distraction while driving was 2.0 times higher than model driving, meaning drivers are at double the risk for more than one-half of their trips when they choose to engage in a distracting activity."”
Calculation notes
```
Dingus 2016 analyzed 3,500+ drivers in the SHRP 2 NDS across six US sites over three years, yielding 905 injurious and property-damage crashes. The 2.0x overall distraction OR is the broadest applicable figure for non-phone manual tasks. The study does not isolate eating specifically in the abstract; the 100-Car NDS is the primary source for the eating-specific OR. Used here to corroborate the general distraction multiplier and to anchor the upper bound of the uncertainty range.
```
Source date: 2016-03-08 · Accessed: 2026-05-04

[2] Hand-held phone call + driving Risk

Statistic: Talking on a handheld cell phone: OR 2.2; texting: OR 6.1; dialing: OR 12.2; overall handheld cell phone interaction: OR 3.6 (all vs model driving in SHRP 2 passenger-car naturalistic sample)

“"The overall risk of interacting with a handheld cell phone is 3.6 times higher than model driving."”

Calculation notes

Dingus 2016 provides the naturalistic-driving complement to McEvoy's case-crossover design. The 2.2x OR for talking on a handheld phone is lower than McEvoy's 4.1 partly because naturalistic studies capture near-crashes differently from hospital attendance databases. The two estimates bracket the plausible per-epoch range (2.2–4.1x); this entry uses 2.2x as a conservative anchor for the exposure-weighted calculation. For context, texting's OR in the same dataset is 6.1 — nearly three times higher than talking on a handheld phone — which is the relevant comparison for drivers who think "just talking" is close to safe.

Source date: 2016-03-08 · Accessed: 2026-05-04

[3] Texting + driving Risk

Statistic

Texting on a handheld cell phone: odds ratio 6.1; handheld cell dialing: 12.2; reaching for a handheld cell phone: 4.8; talking on a handheld cell phone: 2.2; overall handheld cell phone interaction: 3.6 (all relative to model driving)

“"The overall risk of interacting with a handheld cell phone is 3.6 times higher than model driving."”
Calculation notes
```
Dingus 2016 is the canonical peer-reviewed passenger-car number. The 6.1 OR for texting is the per-epoch (six-second window around crashes/near-crashes) risk while the driver is actively texting, not a per-trip or per-year figure. To convert to a lifetime probability we multiply the US per-capita annual car-crash hazard (12.2/100,000, IIHS 2023) by an exposure-weighted factor of ~2.5 for a regular texter, then compound over 59 adult years.
```
Independence note: Dingus 2016 draws from the SHRP 2 Naturalistic Driving Study dataset, which is the primary upstream source for most US naturalistic-driving crash-risk estimates. The VTTI 2009 trucker study below uses a different naturalistic dataset (commercial vehicles), so the two are independent.

Source date: 2016-03-08 · Accessed: 2026-04-11
[4] Video watching + driving Risk

Statistic

Reading/writing on handheld cell phone: OR 9.9; handheld cell dialing: OR 12.2; texting: OR 6.1; browsing: OR 2.7; overall handheld cell interaction: OR 3.6. All relative to model driving in SHRP 2 passenger-car naturalistic sample.

“"Reading or writing on a handheld cell phone (e.g., e-mail, text, browsing) was 9.9 times more likely to result in a crash than model driving."”
Calculation notes
```
Dingus 2016 does not report a specific "watching video" category as a discrete secondary task. The 9.9 OR for reading/writing is the closest analogue: both tasks demand sustained eyes-off-road gaze, typically 5–15 seconds per episode. Video-watching differs from reading/writing in that it holds attention longer and is less likely to be interrupted voluntarily, suggesting a per-epoch OR at or above 9.9. This entry uses ~10 as the working per-epoch estimate, consistent with the reading/writing figure. To convert to a lifetime probability, the US per-capita annual car-crash hazard (12.2/100,000, IIHS 2023) is multiplied by an exposure-weighted factor of ~4x for a regular video-watcher, then compounded over 59 adult years.
```
Independence note: Dingus 2016 draws from the SHRP 2 Naturalistic Driving Study. The Simons-Morton 2014 source below also uses SHRP 2 data, so treat both as drawing from a shared upstream dataset; they are methodologically distinct but not independent samples.

Source date: 2016-03-08 · Accessed: 2026-05-04