A Randomized Field Trial of Smartphone-based Feedback to Encourage Safe Driving

This is the mean of the four behavior scores for the intervention period (and, in a follow-up analysis of effect sustainability, the post-intervention period). Scores can range from 0 -100, 100 being the safest driving score.

A proprietary CMT score based on participant phone use-especially handheld phone use-while driving on a scale of 0-100, 100 being no phone use while driving (safest).

A proprietary CMT score based on the amount of time the participant drove over the speed limit on a scale from 0 -100, 100 being no incidences of driving over the speed limit.

A proprietary CMT score based on the frequency of a participant's hard brakes on a scale of 0-100, 100 meaning no hard brakes.

A proprietary CMT score based on the frequency of a participant's fast accelerations on a scale of 0-100. 100, being no fast accelerations.

This is a composite outcome that measures the proportion of total trip time in which the driver is engaged in handheld phone call use or non-call handheld use (e.g. texting, swiping, and typing), as measured by the Way to Drive app. Several studies (e.g. Klauer, NEJM, 2014) have demonstrated the association between handheld phone use (e.g., reaching for phone, typing, swiping, dialing) and increased crash risk. This outcome is also known as the active phone use percentage. Passive phone use (e.g. phone is streaming GPS navigation directions or music without any typing, swiping, or holding of the phone) is not included in this outcome.

This is a composite measure of the underlying driving metrics for the four behaviors of interest. For each behavior a z-score will be computed for each participant and take the mean of these four z-scores to get their overall riskiness. This will be done for both the intervention period and the post-intervention period.

If a participant texts "stop" or "bye" to stop receiving our intervention text messages, this could indicate unacceptability of push notification interventions. This will be measured as a binary score, where participants that text "stop" or "bye" will be marked as 1, and those that do not will be marked as 0.

If a participant asks to be unenrolled, this is an even stronger indicator of unacceptability. This will be measured as a binary score, where 1 means the participant was unenrolled before the end of the intervention period, and 0 means they were not.

This will be derived for each arm based on a 1-item question on a scale of 0 -10, 0 bring not at all likely to recommend the program, and 10 being extremely likely.

This will be a 4-item acceptability scale administered in the exit survey. For each participant the mean of the 4 items will be calculated for an overall acceptability score ranging from 1-5 (5 is the best).

Intervention arm participants will be asked to rate the helpfulness of each of the 4 intervention components (SMS feedback, weekly dashboard, SMS tips, $100 incentive), each on a scale of 1 (not at all) to 5 (extremely), with an option to indicate they did not see or were not aware of the intervention component.

Number of hard braking evens per 100 miles of driving

Based on duration of speeding events detected by the app.

Number of fast acceleration events per 100 miles of driving.