Accident rate per million miles: the formula for a NEMT fleet, with preventable and injury rates

Accident rate per million miles is the number of crashes multiplied by 1,000,000 and divided by the miles the fleet drove in the same period. Two things make the number useful or useless: which crashes you count, and which miles you divide by. Contracts and regulators use different definitions, so a fleet that compares its own rate with a published one without checking the definition usually ends up comparing numbers that do not measure the same thing.

This page gives the formula with a worked example, the three crash counts worth tracking, and the published benchmarks with the unit and definition each one uses. The driver-level version of the same measure sits on the driver performance metrics page, and the review that decides what is preventable is in the fleet safety program guide.

How do you calculate it, with an example?

Divide crashes by miles, then scale. This is an example, not a benchmark. A fleet of 8 vans drives 36,000 miles each over twelve months, 288,000 miles in all. It has 3 crashes, 2 of them preventable, and 1 of them involved an injury.

The rate for any count is the count times 1,000,000 divided by 288,000. Dividing that by 10 gives the rate per 100,000 miles, the unit most contracts use.

Per million miles, the same rates are 10.4, 6.9 and 3.5. To convert any published figure, multiply per-100,000 numbers by 10 or divide per-million numbers by 10.

Which crashes should you count?

Count three things separately, because each answers a different question. The full count shows how often the fleet is touched, preventable crashes show the quality of the driving, and injury crashes show severity.

• All crashes. Every contact with a vehicle, a person or an object, including a bumper tap in a lot. Keep this one in your own file; it shows how often the fleet is touched.
• Preventable crashes. The ones your review finds the driver could have avoided. FMCSA’s test in Appendix B to 49 CFR part 385 is whether a driver using normal judgment and foresight could have seen the crash coming and avoided it with steps within the driver’s control. Fault and preventability are separate questions, as the fleet safety program explains.
• Injury crashes. Crashes where someone was killed or needed immediate medical treatment away from the scene. This borrows part of FMCSA’s recordable standard (49 CFR 390.5), which also counts any crash that leaves a vehicle with disabling damage that needs a tow. The standard is written for commercial motor vehicles, defined there as vehicles rated 10,001 pounds or more or built to carry more than 8 passengers including the driver, for pay. Many NEMT vans are below both lines, so the rule does not bind them, but its wording is a clean one to use. The same rule leaves out an occurrence involving only boarding or alighting from a stationary vehicle, so decide how your file treats a rider who falls at the curb and track it as its own event.

Which miles belong in the denominator?

Use total miles driven for your own rate, and the contract’s miles for a contract rate. The Federal Transit Administration divides by vehicle revenue miles, which exclude deadhead, operator training and maintenance testing. A van with a high share of between-trip driving therefore looks worse on revenue miles than on total miles. The same crashes over fewer miles give a larger rate.

For the numerator and denominator, take both from the same twelve months and the same vans. Odometer readings or the GPS totals from trip records work, as long as the method stays the same month to month.

What benchmarks exist, and what does each measure?

No single number applies to every NEMT fleet. The table puts each published figure on one scale, per million miles, so the differences are visible. Read the definition next to the number, because much of the gap between 0.84 and 14 comes from what gets counted.

Where each number comes from:

• FMCSA. The accident factor in a safety rating is the recordable accident rate for the past 12 months. An urban carrier with a rate above 1.7 gets an unsatisfactory accident factor, and other carriers above 1.5 do. FMCSA set those lines by roughly doubling the national averages from its 1994 to 1996 reviews, which were 0.747 for all carriers and 0.839 for carriers within 100 air miles.
• NHTSA. The 1.9 is 6,138,359 police-reported crashes divided by 3,246,817 million vehicle miles traveled, both from NHTSA’s 2023 Quick Facts and covering all vehicles on the road.
• FTA. The National Transit Database listed 939 major safety events in demand-response service in 2024, 926 of them collisions, over 1.117 billion vehicle revenue miles. Only events that meet FTA’s major-event thresholds are in the file.
• Contracts. Access Services in Los Angeles County lists limits of 0.25 preventable incidents and 0.85 preventable collisions per 100,000 miles in its April 2026 board materials, and its report card for the Santa Clarita region shows 0.36 preventable collisions per 100,000 miles for fiscal 2025. Volusia County’s 2026 paratransit scope allows no more than one preventable accident per 100,000 miles, sets a quarterly standard of zero accidents, and deducts $100 per accident unless the accident is shown to be non-preventable. Sumter County, Florida’s coordinated system sets a standard of 1.4 crashes or incidents per 100,000 vehicle miles, and its evaluation recorded 1.84 for fiscal 2025.

The contract numbers sit far above the federal ones, which count only serious events. A contract limit is usually a limit on a broader count of preventable accidents, so read how each contract defines one. A fleet that tracks only injuries and tows will look better than it is against those limits.

Do contracts and regulators set targets you must hit?

Contracts do, and transit agencies set their own. A paratransit contractor inherits the agency’s limits through the contract, as in the Volusia and Access Services examples, and a failure can mean a financial deduction. Transit agencies themselves set annual safety performance targets under FTA’s agency safety plan rule. The 2024 update to the National Public Transportation Safety Plan raised the required measures from seven to 14, including a vehicular collision rate, and FTA calculates these rates on vehicle revenue miles. The ADA paratransit contractor guide covers what else comes with that work.

For brokers, the definition matters more than the target. Ask each one how it defines an at-fault or preventable incident, and track your own rate on the same terms.

How many crashes make the rate reliable?

Not one or two. In the example fleet, a single added crash moves the rate by 0.35 per 100,000 miles. FMCSA builds in the same caution: it uses the recordable rate in a rating only when a carrier has at least two recordable accidents in the previous 12 months. For a fleet this size, report the count and the rate together, use a rolling twelve months, and avoid scoring one driver on a handful of crashes. The driver performance metrics guide suggests a minimum trip count before a rate is used.

Running this in HealthRide

The miles half of the formula comes from trip records, and the crashes half comes from your own crash log. HealthRide records the GPS miles on every trip, and the trip log can be downloaded as a CSV or PDF. That gives you the miles side of the formula without rebuilding it from odometer notes. See reports and trip records.