NEMT Route Optimization: How It Works and How to Judge the Results
NEMT route optimization is building every driver's day so all trips are covered on time with the fewest empty miles. Unlike a navigation app, it decides which driver takes which trip and in what order, while respecting pickup windows, appointment times, vehicle equipment, seats, driver shifts, and ride-time limits. It works best when the whole fleet is planned at once and re-planned as the day changes.
On this page
Three different jobs: scheduling, routing, navigation
People use “routing” for three separate jobs. NEMT route optimization is the middle one.
| Job | The question it answers | Example |
|---|---|---|
| Scheduling | When should this trip happen? | Pickup at 7:40 AM for a 9:00 AM appointment |
| Route optimization | Which vehicle takes which trips, in what order? | Van 3 takes these nine legs, in this sequence |
| Navigation | How does the driver get from this stop to the next one? | Turn-by-turn directions to the clinic |
A navigation app answers the third question well and the first two not at all, which is where Google Maps falls short for NEMT. Route optimization takes a day of scheduled trips and a fleet of drivers, and produces each driver’s run.
The rules an NEMT plan has to respect
A plan that saves miles but breaks a contract rule is not a good plan. These are the constraints an NEMT route must satisfy, with current examples:
- Pickup windows. Virginia’s fee-for-service program tells members that drivers arrive no earlier than 15 minutes before and no later than 15 minutes after the scheduled pickup time.
- Appointment times. The Medi Trans manual in Louisiana asks for riders to arrive at least 15 minutes before the appointment and never more than two hours early. See the on-time performance guide for how arrivals are scored.
- No early boarding. Indiana’s broker criteria require that zero trips make a rider board before the scheduled pickup, and CareOregon bars pickups before the scheduled time or window.
- Vehicle equipment. Wheelchair riders need a lift or ramp, stretcher riders need a stretcher vehicle, and oxygen needs have to be matched. These are hard rules, not preferences.
- Seats. Escorts and extra riders count toward capacity.
- Ride-time limits on shared rides. Where multi-loading is allowed, the Medi Trans manual does not let any rider stay aboard more than 45 minutes longer than the direct trip.
- Driver shifts. Start times, end times, breaks, and where each driver starts the day.
- Boarding time. Securing a wheelchair or loading a stretcher takes minutes that a map does not show.
- Your own decisions. Locked assignments, rider and driver preferences, and gaps held for will-calls.
One rule works the other way. The order you run trips in does not change what you can bill. CareOregon’s manual only allows providers to bill each trip using the most cost-effective and reasonably direct route from origin to destination, so under that rule a chained run cannot add billable miles.
Why it is hard to do by hand
Operations researchers call this the dial-a-ride problem. A 2011 study in Computers and Operations Research, built on the Austrian Red Cross’s patient transport operation, notes that transporting patients or elderly people is usually modeled this way. It also models the mix every NEMT dispatcher knows: trips booked in advance, trips that appear during the day, and return trips that may or may not be requested once the rider is at the hospital.
The difficulty is the number of possible plans. One van with 10 trips, done one rider at a time, can run them in 3,628,800 different orders. Add a second van and every trip also has a choice of vehicle, which multiplies the count again. A 10-van day with 90 trips has more possible plans than anyone could check. A skilled dispatcher gets a good plan by instinct and experience. An optimizer can search far more options against every rule at once, and the number of options it has to beat grows with every van you add.
Night-before planning and same-day re-planning
Route optimization happens twice a day in a well-run operation.
- The night before. Most trips are known by then. Virginia’s fee-for-service program requires 5 business days’ notice for routine rides, and MTM in Wisconsin requires most requests at least two business days ahead. Plan the whole fleet at once, review it, and send manifests.
- During the day. Same-day urgent trips, hospital discharges, will-calls, cancellations, and driver call-outs change the picture. Virginia’s handbook says same-day urgent requests may take up to three hours for transportation to arrive. A good re-plan moves only what needs to move and leaves trips already underway alone.
Brokers do their own version upstream. Indiana’s criteria for its broker call for an auto-routing system that assigns at least 90% of qualifying trips to a transportation provider within 48 hours of the request. The broker decides which provider gets a trip. Your optimization decides which of your drivers runs it and when.
How to tell whether it is working
Judge a plan by the numbers it produces over weeks, not by how it looks on the map. Compare four weeks before and four weeks after, on a similar mix of trips:
| Measure | Hypothetical before | Hypothetical after |
|---|---|---|
| Trips per vehicle hour | 1.05 | 1.22 |
| Deadhead share | 44% | 37% |
| On-time pickups | 91% | 95% |
| Trips unassigned after planning | 6 a day | 0 |
| Dispatcher time to build tomorrow | 2.5 hours | 30 minutes |
The values are illustrative. The rule is not: productivity should rise without on-time performance falling. A plan that packs more trips into each van but makes riders late is borrowing against your broker scorecard. The NEMT KPIs guide has formulas for each measure, and the deadhead guide covers empty miles in detail.
Planning by hand until you have a tool
If you plan manually today, a fixed order of work gets you most of the way:
- Place standing orders first. They repeat, so they anchor the week.
- Place wheelchair and stretcher trips next, on the vehicles that can carry them.
- Group the rest by zone and time.
- Chain each drop-off to a nearby next pickup.
- Add boarding time to every leg based on level of service.
- Leave gaps for will-calls near the facilities that generate them.
- Read each van’s day start to finish and look for impossible gaps or back-to-back legs across town.
The dispatch workflow guide puts this planning step in the context of the whole day.
What to look for in a route optimizer
- It plans the whole fleet at once, not one van at a time.
- It treats equipment and level of service as hard rules.
- It honors appointment deadlines, not only pickup windows.
- It counts escorts toward seats.
- It starts same-day re-plans from where each driver actually is.
- It leaves locked assignments alone.
- It shows the plan before applying it and explains any trip it could not place.
- It uses travel times that reflect the time of day.
The buyer’s checklist covers how to test an optimizer on one of your own days before you commit.
Planning the day with Ryder Go
Ryder Go plans and assigns the whole day in one click. It matches wheelchair, stretcher, and oxygen needs to each vehicle’s equipment, honors appointment times, and keeps locked assignments exactly where you put them. Every plan arrives as a proposal for your dispatcher to apply or reject, and anything it could not place is listed with a clear reason. See how the route optimizer works.
Frequently asked questions
- Is Google Maps enough for NEMT routing?
- For directions, yes. For planning, no. A map app finds the fastest way from one stop to the next. It does not decide which of your drivers should take which trip, whether a wheelchair rider is on a van with a lift, or whether a pickup order still gets everyone to their appointments on time.
- Does route optimization replace the dispatcher?
- No. It replaces the hours of building the plan by hand. The dispatcher still reviews the plan, handles exceptions the software cannot know about, such as a rider who needs the same driver every time, and manages the calls, changes, and broker relationships that fill the rest of the day.
- How much can route optimization save?
- It depends on how tight your routes are today. Measure deadhead share, trips per vehicle hour, and on-time performance for four weeks before and four weeks after, on a similar mix of trips. Those numbers from your own fleet are the only savings figure worth trusting.
- Should I optimize the night before or during the day?
- Both. Build the full plan the night before, when most trips are known. Then re-plan during the day for cancellations, same-day adds, will-calls, and driver call-outs, touching only what changed and leaving trips already underway alone.
- What is the dial-a-ride problem?
- It is the name operations researchers use for planning door-to-door rides where each rider has a pickup, a drop-off, and time limits. A 2011 study of the Austrian Red Cross's patient transport service notes that transporting patients or elderly people is usually modeled as a dial-a-ride problem. In other words, it is the textbook name for the NEMT planning problem.
- Can route optimization handle shared rides?
- Some tools can, where your contracts allow shared rides. The plan then has to respect ride-time limits for every rider on board. For example, the Medi Trans manual in Louisiana does not allow any rider in a multi-load vehicle to ride more than 45 minutes longer than the direct trip.