When to Stop Driving
By Simon Peter Lokomo, MPH — Public Health
Last reviewed: August 2026
The short answer
Older drivers do have higher death rates per mile driven. But the main reason is not that they crash more. It is that they are more likely to die in the crashes they do have, because bodies get more fragile with age. One analysis found fragility accounted for roughly 60 to 95 percent of the excess death rate, while crashing more often explained no more than 30 to 45 percent even among the oldest drivers. That difference matters enormously, because it points to different answers. Stopping driving is also not a free decision: it roughly doubles the risk of depressive symptoms and is linked to declines in physical and social functioning. The goal is not to take the keys as early as possible. It is to get the timing right, and to make what comes after survivable.
If you are having this argument with a parent right now, or dreading starting it, the useful thing to know is that the evidence supports being more careful and less certain than most advice on this subject suggests.
If dementia is the reason you are asking, start with my article on when someone with dementia should stop driving, which covers the specific evidence on driving with cognitive impairment, driving contracts, and the stage-by-stage picture. This article is about older drivers generally: what the crash data actually shows, why stopping carries its own risks, and how fitness to drive is genuinely assessed.
What the crash data actually shows
Two separate things get merged into one worry. The first is how often someone crashes. The second is how likely they are to survive a crash.
A national analysis separated them and found that fragility, meaning susceptibility to injury and death from a given impact, begins rising at around ages 60 to 64 and climbs steadily from there. Fragility accounted for about 60 to 95 percent of the elevated death rate per mile among older drivers, depending on age and sex. Actual over-involvement in crashes did not become marked until age 75, and even then explained no more than 30 to 45 percent of the elevated risk. For drivers under 20, the pattern reverses almost completely: crashing too often explains more than 95 percent of their elevated death rate.
Put plainly, a 20-year-old’s risk comes mostly from their driving. An 80-year-old’s risk comes mostly from their body. The Insurance Institute for Highway Safety reports the same conclusion from more recent data, and adds that drivers 75 and older are roughly four times as likely as middle-aged drivers to die in a side-impact crash and about three times as likely in a frontal one.
There is one more finding worth knowing, because it reframes the whole conversation. In 2024, 70 percent of the people killed in crashes involving a driver 70 or older were the older driver themselves or their passengers. In fatality terms, an older driver is mostly a risk to themselves and to the person in the passenger seat, who is usually their spouse.
What the evidence shows
Higher death rates per mile among older drivers are driven mainly by fragility rather than by crashing more often, with fragility accounting for roughly 60 to 95 percent of the excess. Crash over-involvement does not become pronounced until about age 75 and explains a minority of the risk even then. Most people killed in crashes involving older drivers are the driver and their passengers.
Driving cessation carries measurable costs. A systematic review of 16 studies found it roughly doubled the risk of depressive symptoms, with a pooled odds ratio of 1.91, alongside declines in general health and in physical, social, and cognitive functioning.
Stopping is not a neutral act
Most advice treats giving up driving as pure safety gain with an emotional cost attached. The research does not read that way. A systematic review of 16 studies of drivers aged 55 and over found that driving cessation was associated with declines in general health and in physical, social, and cognitive functioning, and a meta-analysis within it found close to a doubling of the risk of depressive symptoms. A much larger 2024 review, screening more than 6,000 papers and including 42, found the same pattern across depression, anxiety, general mental health, and stress.
The honest caveat is that causation runs both ways. People often stop driving because their health is already declining, so some of what looks like a consequence of stopping is the underlying decline that prompted it. Researchers in this area say so directly. But the association is strong enough, and consistent enough across studies, that stopping should be treated as a real intervention with real risks, not as the obviously safe default.
What follows from that is practical: the decision is not only “should they stop” but “what replaces it.” Losing the car in a place with no alternative is a different event from losing it where family, transit, or a paid driver can fill the gap — see our guide on getting around after a parent stops driving for the options.
Why a cognitive test cannot settle this
Families often hope for a clean threshold, and a memory test looks like one. It is not.
A study of 168 community-dwelling older adults tested how well the Mini-Mental State Examination predicted pass or fail on an actual on-road driving test. The area under the curve was 0.654, where 0.5 is a coin flip and 1.0 is perfect. The authors concluded that the standard cutoff is not sensitive enough to predict on-road performance and that their findings support the existing best practice of not using the MMSE alone for this purpose. A separate study of the Clock Drawing Test alongside the MMSE reached the same conclusion: both have limited utility as indicators of driving problems, and neither is sufficient to base a driving decision on by itself.
This cuts both ways, and that is the point. A poor score does not establish someone is unsafe to drive. A decent score does not establish they are safe.
Does cognitive impairment even increase crash risk?
Less clearly than you would expect. Driving simulator studies do show that cognitive impairment reduces hazard perception, visual attention, and reaction time. Some studies find dementia associated with more crashes and with adverse driving behaviors like hard braking and sudden acceleration. But other studies have found cognitive decline was not associated with increased crash risk or with poor driving performance at all.
One proposed explanation for the null findings is that people with cognitive decline often notice something is wrong and adjust: driving less, avoiding highways, not driving at night or in rain. That self-regulation may absorb much of the added risk. It also means the person who has already stopped driving after dark may be managing the problem rather than denying it.
I am not arguing that dementia and driving mix safely. Dementia is progressive, and at some point along it driving becomes unsafe for everyone. I am saying that a diagnosis is not itself the moment, that the evidence for exactly when is genuinely mixed, and that anyone telling you the research is settled here has not read it.
| What families often assume | What the evidence supports |
|---|---|
| Older drivers are dangerous to everyone on the road | Most people killed in these crashes are the older driver and their own passengers |
| The elevated death rate means they crash more | Fragility explains roughly 60 to 95 percent of it; crashing more explains a minority |
| A memory test can tell us whether it is time | MMSE predicts on-road pass or fail barely better than chance (AUC 0.654) |
| Stopping is the safe choice with no downside | Cessation roughly doubles the risk of depressive symptoms and tracks with functional decline |
| A dementia diagnosis means the license goes | Four states require clinicians to report it; most do not, and what follows varies |
What an actual driving assessment looks like
The assessment that answers this question is not a memory test. It is a comprehensive driving evaluation, usually done by an occupational therapist who specializes in driver rehabilitation, and it has three parts: a clinical assessment of vision, movement, reaction time, and thinking; an on-road assessment in a real vehicle; and an outcome discussion.
The outcome is not always pass or fail. These evaluations can produce restricted licenses, recommendations for adaptive equipment, or targeted training, which is why they are worth doing before the situation becomes a crisis. A driver who can safely manage familiar local routes in daylight is in a different position from one who cannot drive at all, and only this kind of assessment can tell you which you are dealing with.
Cost varies widely and is often not covered by insurance. One academic medical center lists a clinical fitness-to-drive screening at $200 and states plainly it is not covered. A specialist driver rehabilitation practice lists full evaluations at $450 to $650 depending on travel distance. Sources disagree enough here that you should verify locally and ask directly about coverage rather than trusting a published figure, including mine. A physician referral is often required.
Where the evidence runs out
Whether cognitive impairment actually raises crash risk is unsettled. Simulator studies say yes, some crash studies say yes, and other crash studies find no association at all, possibly because people compensate by driving less and more cautiously once they sense a problem. No screening instrument reliably identifies unsafe drivers on its own.
There is also almost no evidence that mandatory physician reporting laws improve road safety. Researchers searching the literature from 2000 to 2023 found a single relevant study, which found no significant difference in dementia rates among older drivers hospitalized after crashes between states with clinician reporting laws and states without.
How state reporting actually works, and one thing it appears to cause
There is no national rule. As of research conducted for a 2024 study, four states require clinicians to report a dementia diagnosis to the DMV: California, Delaware, Oregon, and Pennsylvania. Fourteen states require the driver to self-report. The remaining 32 states and the District of Columbia have no explicit requirement, though nearly all accept reports from physicians, family members, or law enforcement, usually triggering a medical review or a retest rather than automatic revocation.
Here is the finding that should give everyone pause. That same 2024 study analyzed Medicare data covering 223,036 primary care clinicians and found that clinicians practicing in mandatory-reporting states had a 12.4 percent probability of underdiagnosing dementia, compared with 7.8 percent in self-reporting states and 7.7 percent in states with no mandate. That is roughly 59 percent higher. The likely mechanism is that when a diagnosis automatically threatens someone’s license, patients hide symptoms and clinicians become reluctant to probe.
Meanwhile the authors could find only one study examining whether these laws improve road safety, and it found no significant difference. So the mandates have a documented cost, in missed and delayed dementia diagnoses, and no documented benefit. That is worth knowing if you live in one of those four states and are deciding how candid to be with a doctor, and it is worth knowing that the honest answer to “will telling the doctor cost my father his license” is: in most states, not automatically, and in four states, possibly.
What the evidence does not support
That older drivers are primarily a danger to other people. In 2024, 70 percent of those killed in crashes involving drivers 70 and older were the driver or their passengers.
That the elevated death rate per mile means older people crash more. Fragility accounts for most of it. Crash over-involvement does not become marked until 75 and explains a minority of the risk even then.
That a cognitive screening test can determine fitness to drive. The MMSE predicts on-road performance barely better than chance, and the researchers who tested it explicitly recommend against using it alone for this.
That a dementia diagnosis automatically ends driving. Four states require clinicians to report it. Most states do not, and what happens after a report varies by state.
That mandatory reporting laws make roads safer. One study exists on the question and it found no difference. What those laws are associated with is a substantially higher rate of underdiagnosed dementia.
That stopping is automatically the safe option. It roughly doubles depression risk and tracks with broader functional decline. It is a real intervention with real costs, which is an argument for planning it well, not for avoiding it.
When to get help
If you have specific observations, get a comprehensive driving evaluation rather than arguing about it. Ask their doctor for a referral to an occupational therapist certified in driver rehabilitation. An independent on-road assessment moves the conversation off your judgment versus theirs.
If something has changed suddenly, treat that as medical rather than as a driving question. New confusion, a sudden change in vision, a fall, a medication change, or a near-miss that seems out of character can all have treatable causes. My guides to delirium and to medications that increase fall risk cover the sorts of things that are worth ruling out.
Before anything else, look at the car itself. Older drivers tend to drive older vehicles with fewer safety features, and given that fragility is the dominant risk, moving someone into a newer car with side airbags and modern crash protection addresses the actual mechanism of harm. It is the most underrated intervention in this whole area.
If they are unsafe and will not stop, nearly every state accepts a request for driver re-examination from a family member, and some allow it confidentially. Check your state DMV for the specific process. This is a serious step and worth taking after an evaluation rather than instead of one.
Key takeaways
- Higher death rates among older drivers come mainly from fragility, not from crashing more often. Fragility explains roughly 60 to 95 percent of the excess.
- Most people killed in crashes involving older drivers are the driver and their own passengers.
- Driving cessation roughly doubles the risk of depressive symptoms and is associated with declines in physical, social, and cognitive functioning. It is not a cost-free decision.
- Cognitive screening tests cannot determine fitness to drive. The MMSE predicts on-road performance barely better than chance.
- Whether cognitive impairment actually increases crash risk is genuinely contested, partly because people compensate once they notice a problem.
- A comprehensive driving evaluation by an occupational therapist, including an on-road assessment, is the tool that actually answers the question. Expect roughly $200 to $650 and often no insurance coverage.
- Four states require clinicians to report dementia to the DMV. Those states show substantially more underdiagnosed dementia, and the single study on whether such laws improve safety found no difference.
- Because fragility is the dominant risk, putting an older driver in a newer, safer car is one of the highest-value and least-discussed interventions available.
Frequently asked questions
My mother was diagnosed with early Alzheimer’s. Does she have to stop driving now?
Not necessarily, and not automatically in most states. Early-stage dementia does not by itself establish that someone cannot drive safely, which is why some state DMVs retest rather than revoke. What it does establish is that the situation will change, so this is the moment for a comprehensive driving evaluation and for planning ahead, while she can participate in the decision.
Will telling the doctor cost my father his license?
In most states, no, not automatically. Four states require clinicians to report a dementia diagnosis to the DMV, and even there the usual result is a review or retest rather than immediate revocation. It is worth knowing your state’s rule before the appointment. It is also worth knowing that avoiding the conversation has its own costs, since undiagnosed dementia means untreated symptoms and no planning time.
He passed a memory test at the doctor. Does that mean he is safe to drive?
No. That is the same limitation in reverse. Screening tests do not predict on-road performance well in either direction. If you have specific observations, an on-road evaluation is the thing that answers them.
What are the actual warning signs worth acting on?
Concrete, observable changes rather than age, and the signal is change from their own baseline. My article on driving with dementia has a full breakdown of what to look for. The point I would add here is that noticing a warning sign tells you to arrange an on-road evaluation, not that you have established anything. Screening on your own observations has the same problem as screening with a memory test.
Is there a point at which everyone should stop, just based on age?
No, and the data argues against setting one. Crash involvement varies enormously between individuals of the same age, which is why an assessment of the person outperforms any rule about the birth year.
She has already stopped driving at night and on highways. Is that denial or good judgment?
Most likely good judgment. Self-regulation of this kind is one proposed reason some studies find no increased crash risk with cognitive decline. Someone narrowing their driving in response to their own limits is doing roughly what a driving evaluation would recommend.
How do I bring it up without a fight?
The specific thing this article adds is the framing. Leading with fragility works better than leading with competence, because it is true and it is not an accusation: the risk is mostly to him and to whoever is in his passenger seat, and it is about how bodies handle impacts, not about whether he is a good driver. For fuller conversation approaches, including written driving agreements, see my article on driving with dementia. Whichever framing you use, pair it with a concrete plan for getting places rather than a request to hand over keys.
This article is for general information and is not medical or legal advice. Driving laws, reporting requirements, and DMV procedures vary considerably by state and change over time. Discuss specific concerns with your parent’s doctor and check your state DMV for current rules.
How this site is funded: I do not currently earn a commission on anything mentioned here. If that changes I will say so on every page it applies to, and it will not change an assessment. Several articles on this site advise readers not to buy popular products, and that stays true whatever the funding.
