Road Safety Behavioural Science · Randomised Controlled Trial
Seatbelt Reminders or Sensor-Triggered Alerts? Protocol for a Four-Arm Factorial Randomised Controlled Trial of the ADR 68/01 Seatbelt Requirements and Per-Seat Buckle Sensing on Buses
Australian International Technology Solutions Company
2 August 2026
Summary in plain language
There are two ways to get bus passengers to wear a seatbelt. You can tell them, which is what Australia’s new rule does with labels, a lit sign and one recorded announcement. Or you can put sensors in every seat and make a noise at the seat of anyone sitting there unbuckled, which is what a seatbelt warning system does. Nobody has tested these two against each other, and nobody has measured either of them against a bus that has neither.
This trial does all of that at once. Buses are randomly assigned to one of four set-ups: a standard bus, a bus with the ADR 68/01 seatbelt requirements only, a bus with a seatbelt warning system only, or a bus with both. Every bus carries the same sensors, so every arm is measured with the same instrument. Then we count how many passengers are actually buckled.
Four arms, 40 vehicles, randomised by vehicle, six months from baseline to follow-up: month 1 records seatbelt use silently in every vehicle, months 2 to 5 apply the four assigned conditions, and month 6 returns every vehicle to silent recording under its baseline conditions. The four arms are set out in Table 4, the eight comparisons they produce in Table 5, and the results that would contradict the hypotheses or fail the design criteria in Table 10.
Abstract
Background. Australian bus passengers wear seatbelts at rates reported between 7 and 45 per cent.3 Car occupants in the same country wear them at about 97 per cent.4 Two approaches aim to increase bus passenger seatbelt use. The ADR 68/01 seatbelt requirements, which apply to new bus models from 1 November 2026, provide reminders: a label on each seatback, an illuminated cabin sign, and one recorded announcement after the doors close.26 Those three measures, and nothing else in that broad rule, are what this paper means by the phrase, and Section 1 defines it. A seatbelt warning system responds to an unfastened belt: sensors in every seat detect who is seated and unbuckled, and an alert sounds at that seat until the belt is fastened. Neither has been measured against the other, and neither has been measured against a bus with neither intervention.
Objective. To determine which of these two methods raises passenger seatbelt use, by how much, whether the two work better in combination than either does alone, and whether seatbelt use remains higher after the trial reminders and alerts are withdrawn.
Method. A four-arm factorial randomised controlled trial in belt-fitted Australian buses and coaches, randomised by vehicle. Two things are varied independently: the ADR 68/01 seatbelt requirements (fitted or not fitted) and a per-seat buckle-sensing warning system (enabled or disabled). That produces four arms: a standard bus, a bus with the ADR 68/01 seatbelt requirements only, a bus with a seatbelt warning system only, and a bus with both. Every vehicle in every arm is fitted at enrolment with per-seat occupancy and buckle sensors that log silently, so the standard-bus arm is measured with exactly the same instrument as the treated arms. Ten vehicles are enrolled per arm, 40 in total, from at least six operators, stratified by service type, operator and baseline wearing rate. Month 1 establishes baseline seatbelt use through silent logging in all vehicles. At the start of month 2, vehicles are randomised to the four arms, which remain in their assigned conditions through month 5. In month 6, trial-added reminders and alerts are withdrawn and every vehicle returns to silent logging under its baseline conditions.
Outcome measures. The primary outcome is the proportion of occupied seats with the belt latched five minutes after the last door closes, taken per trip from the sensor log. Secondary outcomes cover belt use across the whole trip, time to buckle, false alerts, belt defeat, passenger acceptability, driver overrides, hardware reliability, and changes during months 2 to 5 and persistence of seatbelt use after withdrawal in month 6.
Criteria. Success and failure are defined before enrolment. Ten criteria for results that would contradict the trial’s hypotheses or fail its design requirements are fixed in advance and reported whether or not they are met. They include the threshold below which the warning system is judged to provide no meaningful additional benefit in an already compliant vehicle, the threshold below which the ADR 68/01 seatbelt requirements are judged ineffective on their own, the threshold above which they are judged to provide a benefit beyond sensing alone, and a false-alert rate above which the system fails its own design criteria whatever its measured effect.
Keywords: seatbelt reminder systems; buckle sensing; factorial randomised controlled trial; cluster randomisation; negative reinforcement; present bias; alarm habituation; bus passenger safety; ADR 68/01; ADR 4/05