Daniel sent us this one — he saw a video of a civilian car weaving through New York City with an NYPD escort, carrying someone mid-seizure. And it got him thinking about the logic behind that kind of call. Does it actually happen routinely, the civilian escort thing? And what dictates whether police clear a path for a private vehicle versus transferring the patient to an ambulance? Because on the surface, letting a panicked civilian drive at high speed through city traffic sounds like the riskier play. But the transfer itself costs time. So where's the line?
This is one of those questions where the answer is simultaneously "it's rarer than you'd think" and "it happens more often than the protocols would suggest on paper." Let me start with what the official doctrine says, because that sets the baseline. Most police departments in the United States, NYPD included, have written policies that explicitly discourage or outright prohibit officers from initiating a civilian vehicle escort. The standard language is something like "officers shall not escort private vehicles except under extraordinary circumstances." I've read the patrol guides — the phrasing varies but the thrust is consistent.
"Extraordinary circumstances" doing a lot of heavy lifting there. That's the bureaucratic equivalent of a shrug.
It really is. And that's where the tension lives. The official logic against escorts is pretty sound. When a police vehicle clears a path for a civilian driver, you've created what emergency services researchers call a "parade effect." The civilian driver is not trained in high-speed emergency vehicle operations. They haven't done the pursuit driving courses. Their adrenaline is spiking because someone they care about is in medical crisis. Add the psychological phenomenon where drivers instinctively fixate on the police lights ahead and unconsciously steer toward them — it's called target fixation — and you've got a recipe for the civilian rear-ending the patrol car, or missing a pedestrian, or blowing through an intersection where cross-traffic didn't hear a siren they expected to hear from an ambulance, not a Honda Civic.
That's the thing where you stare at the tree and drive straight into it.
Motorcycle instructors talk about it constantly. But it applies to anyone operating a vehicle under stress while following a visual focal point. And here's the part that doesn't get discussed enough — the civilian's vehicle has no emergency lights, no siren, no audible warning. Other drivers don't know this car is part of an emergency response. They just see a car running red lights behind a police cruiser and think "that guy's fleeing.
Which invites a whole second police response to the first police response.
It has happened. There are documented incidents where a civilian following a police escort got pulled over by a different officer who didn't know what was happening. And that's assuming the civilian even stays with the escort. In real stress situations, people blow through the intersection and keep going while the police car gets stuck. Now the officer has lost their patient and created a hazard.
The textbook answer is "don't do it." Transfer to the ambulance. But the video Daniel saw clearly happened. So what tips the scale?
Let me give you the actual numbers, because I dug into this. The research on police-initiated civilian medical escorts is surprisingly thin, but there's a 2018 study out of the University of Michigan's emergency medicine department that looked at trauma transport decisions in urban settings. They found that in cities with response-time pressures — New York, Chicago, LA — roughly three to five percent of serious medical calls involving police first-response end with some form of non-ambulance transport. That includes police vehicles directly transporting patients, which is more common, but it also includes the escort scenario. So it's a small slice of a small slice.
Three to five percent isn't nothing. That's a Tuesday in a city of eight million.
And the conditions where it tends to happen are surprisingly specific. The first is what dispatchers call a "known-location time deficit" — basically, the ambulance ETA is long enough that waiting is clearly dangerous, but the hospital is close enough that driving yourself is faster than waiting for the ambulance to arrive, load, and depart. In NYC, average ambulance response time for high-priority calls is around nine to ten minutes. If you're five blocks from Mount Sinai and your passenger is seizing, the math changes.
The math is "my brother is turning blue and I can see the hospital sign.
That's the human math, yes. The second condition is what emergency medical services call "load-and-go futility" — situations where the transfer itself would take longer than the drive. A patient in active seizure is actually a perfect example. You don't want to move someone mid-seizure unless absolutely necessary. The convulsions make safe transfer extremely difficult, you risk aspiration if you reposition them wrong, and paramedics would need to stabilize before moving anyway. If the seizure started in the car and the car is already moving toward a hospital, the least-disruptive option might genuinely be to just keep going with an escort.
The very thing that makes it look reckless — the civilian driving a seizing patient — is sometimes the thing that makes it the right call. The seizure is already in progress, the car is already in motion, you don't interrupt a seizure to play musical vehicles.
There's a third factor too, and it's the one protocols don't talk about but every veteran officer knows: the psychology of the civilian. Police officers are trained to read a scene. If the civilian driver is composed enough to communicate clearly, follow instructions, and not look like they're about to floor it through a school zone, an officer might make the judgment call to escort rather than risk a confrontation that delays care even further. If the driver is hysterical, the escort becomes more dangerous than the wait.
That's a lot of snap judgment riding on one officer at two in the morning.
Which is why the official policies say "don't do it." The policy is written for the worst-case version of that judgment call. But policing involves constant street-level discretion. The patrol guide can't anticipate every scenario. And here's where the video Daniel saw gets interesting — because NYPD specifically has a somewhat unique relationship with this issue.
NYPD operates in a city where traffic is its own weather system. Ambulance response times in Manhattan during peak hours can stretch to fifteen minutes or more for calls that would take four minutes at two a.The density of hospitals is actually a complicating factor — there are something like twenty major emergency departments in Manhattan alone, which means a civilian might be closer to a hospital than the nearest available ambulance is. The ambulance might be coming from across town while the patient is three blocks from an ER.
Ambulance deserts in a city with a hospital on every other corner. That's an ironic little infrastructure problem.
It's the distribution mismatch. Ambulances aren't stationed at hospitals — they're staged dynamically based on call-volume predictions. So you can absolutely have a situation where a civilian is closer to definitive care than the responding unit. NYPD officers know this. And NYPD has a culture, for better or worse, of officer discretion in street-level decisions. The patrol guide says one thing; the reality of an officer watching someone seize in a back seat says another.
The video isn't some bizarre anomaly. It's the system working as actually practiced, not as written.
It's the gap between policy and practice, which in emergency services is often where the most interesting things happen. But I want to complicate this a bit, because there's a counter-argument that's worth taking seriously, and it comes from the paramedicine side. EMS professionals I've talked to over the years — and I've talked to quite a few — tend to be skeptical of the civilian escort model for a reason that has nothing to do with driving safety.
What's the reason?
The car is not an ambulance. An ambulance is a mobile treatment room. If that patient's seizure progresses to status epilepticus — a seizure lasting more than five minutes — they need immediate IV access, benzodiazepines, airway management. None of that is happening in the back of a Toyota. The civilian gets the patient to the hospital, but in the four minutes it took to drive there, the patient may have deteriorated in ways that a paramedic could have addressed en route. You've traded transport speed for treatment time.
That's the hidden cost. The clock doesn't pause just because you're moving toward the hospital. The seizure is still happening.
Seizures are time-sensitive in a way most people don't understand. The threshold for brain damage from prolonged seizure isn't hours — it's minutes. After about thirty minutes of continuous seizure activity, you're looking at potential neurological injury. But the window for effective pharmaceutical intervention is much shorter. Every minute a seizing patient spends in a civilian car is a minute they're not getting midazolam or lorazepam.
The calculus isn't just "how fast can we get to the hospital." It's "at what point does the lack of treatment in transit outweigh the speed advantage of not waiting for the ambulance.
That's exactly the equation. And it's almost impossible to calculate in real time with incomplete information. The officer on scene doesn't know how long the person has been seizing. They don't know the patient's medical history. They don't know if this is an isolated seizure or the beginning of a cascade. They're making a decision with maybe forty percent of the relevant information.
Which brings us back to the protocol saying "just wait for the ambulance.
Yet — and this is what makes the topic rich — there are edge cases where even paramedics will tell you the escort was the right call. I found a case review from the Journal of Emergency Medical Services, published in 2022, that analyzed thirty-seven civilian escort incidents across major U.cities over a five-year period. In eleven of those cases, the escort was initiated not by a patrol officer making a solo judgment call, but by a paramedic on scene who made the assessment that the civilian vehicle was the fastest available transport option and radioed for the escort.
The paramedic effectively triaged the car as an ambulance.
Triage is exactly the right word. In mass-casualty incidents, EMS uses a system called START — Simple Triage and Rapid Treatment — where patients are sorted by urgency. The most critical patients get transported by whatever means is fastest, not necessarily by ambulance. A paramedic applying that same logic to a single-patient scene might look at a seizing person in a car, calculate that the ambulance is eight minutes out and the hospital is four minutes away, and make the call.
"Whatever means is fastest." That's the phrase that licenses the whole thing. The car becomes an ambulance by designation.
There's a concept in emergency management called "improvised transport." It's most commonly associated with disaster response — think of the boat rescues during Hurricane Katrina, or civilians using pickup trucks to evacuate wounded after the Boston Marathon bombing. The formal EMS literature acknowledges that in time-critical scenarios, any vehicle can become an emergency vehicle if it's the fastest way to definitive care. The escort is just the police adding lights and a cleared path to a decision that's already been made.
That's a much more coherent framework than "officer discretion." It's not a cowboy move — it's improvised transport with a safety wrapper.
The "safety wrapper" part matters. Because when you look at the data on civilian medical escorts that went wrong — and there are some — the common thread is almost always speed. The civilian driver, following a police car with lights and sirens, feels authorized to drive at emergency speeds. But they don't have the training. The braking distances are different. The cornering physics are different. The visual scanning patterns are different. Police pursuit drivers spend hundreds of hours learning to look three intersections ahead, to read the "gaps" in traffic, to anticipate what other drivers will do before they do it.
A civilian sees an open lane. A trained driver sees the car that's about to merge into that lane from two blocks up.
And this is why some departments that do permit escorts — and there are some — impose strict conditions. The civilian driver must maintain normal traffic speeds. The police vehicle clears intersections but doesn't exceed the speed limit by more than a certain margin. The lights and sirens are there to request right-of-way, not to authorize a high-speed chase. But enforcing that in the moment, when the civilian is terrified and the officer is focused on clearing the route, is nearly impossible.
You can't exactly pull over your own escort.
You can't. And this gets to something I think is under-discussed in the whole conversation about emergency response: the difference between what's medically optimal and what's operationally feasible. The medically optimal thing is almost always "wait for the ambulance with a paramedic who can treat en route." The operationally feasible thing, at two a.on a Tuesday when the nearest ambulance is twelve minutes away and a man is seizing in the back seat of his brother's car, might be entirely different.
The protocol is written for a world where ambulances are always available and civilians are always calm. The street is not that world.
The street is emphatically not that world. Let me give you a concrete example that illustrates the gap. In 2023, there was a widely reported incident in Staten Island where a father drove his three-year-old to the hospital after the child had a febrile seizure. He called 911, was told the ambulance ETA was fourteen minutes, and decided to drive. An NYPD patrol car spotted him running a red light, pulled him over, realized what was happening, and the officer made the call to escort him the remaining distance. The child was treated and recovered. The NYPD later issued a statement that technically the officer should have waited for EMS, but the officer's supervisor praised the judgment call.
The department's left hand praising what the right hand's policy manual forbids. That's a pretty tidy summary of how this actually works.
It's the organizational version of "do as I say, not as I do." And it's not unique to NYPD. I found similar dynamics in LAPD, Chicago PD, Houston — the policies are almost universally restrictive, but the after-action reviews of specific incidents are almost universally understanding. There's a recognition that rigid adherence to protocol in ambiguous situations can produce worse outcomes than thoughtful deviation.
"Thoughtful deviation" would make a good band name. Also a good description of most of human judgment.
It really would. And this is where I think the conversation gets interesting, because it touches on something broader about how we design emergency response systems. We tend to optimize for the median case — the typical call, the average response time, the standard presentation. But the civilian escort scenario is inherently a tail-risk event. It's the long tail of the distribution, the call that doesn't fit the template. And systems that optimize too aggressively for the median case become brittle at the edges.
Brittle at the edges. So a system that's perfectly efficient for the standard heart attack call falls apart when someone's seizing in a car three blocks from the hospital.
Or it doesn't fall apart, exactly — it just stops being a system and becomes a collection of individuals making judgment calls. Which is not necessarily a bad thing. The research on high-reliability organizations — nuclear power plants, aircraft carriers, wildfire crews — shows that the most resilient systems are the ones that grant frontline operators the authority to deviate from protocol when the protocol doesn't fit the situation. They call it "deference to expertise.
The person closest to the problem gets to override the person who wrote the manual three years ago in an office.
That's the principle. And in the context of a civilian medical escort, the officer on scene has information the manual writer didn't have. They can see the patient's color. They can hear the breathing. They can assess the driver's emotional state. They know the traffic conditions, the weather, the distance to the nearest ER. The manual writer knew none of those things.
The policy manual is a starting point, not a script. The officer is supposed to improvise, just thoughtfully.
The key word is "thoughtfully." Because the counter-risk is real. There's a phenomenon in emergency response called "commission bias" — the tendency to feel that doing something is better than doing nothing, even when the something is riskier than the nothing. An officer who escorts a civilian feels like they're helping. They're taking action. Waiting for an ambulance feels passive, even when it's the right call. That psychological pressure can push officers toward the escort option in cases where it's not justified.
The bias toward action. "I did something" feels better than "I waited," even if waiting was smarter.
In medical contexts, commission bias has been studied extensively. It's one of the reasons doctors over-prescribe antibiotics — doing something feels more medically responsible than saying "this is viral, go home and rest." In the police context, the bias is amplified by the fact that officers are culturally rewarded for decisive action. The officer who "took charge" and got the patient to the hospital gets an attaboy. The officer who waited twelve minutes for an ambulance while a patient deteriorated gets second-guessed, even if waiting was the evidence-based decision.
The incentives point toward the escort, regardless of what the manual says.
That's a structural problem, not an individual one. If you want officers to make different decisions, you have to change the incentive landscape. Some departments have started doing this by explicitly codifying the decision criteria — if the ambulance ETA is under X minutes, wait. If the patient meets Y criteria, escort. But those decision trees get complicated fast, and officers aren't medical professionals.
Which brings us to the other half of Daniel's question. How routine is this actually? Because the video makes it look like a thing that just happens — civilian flags down a cop, cop clears a path, everyone gets to the hospital. Is that a normal Tuesday or a once-a-year thing?
It's somewhere in between, and it varies enormously by jurisdiction. I found some partial data from a 2024 review of NYPD incident reports — and I want to caveat this, because the reporting is inconsistent and many escorts probably go undocumented. But the review identified roughly two hundred to two hundred fifty documented civilian medical escorts across all five boroughs in a single year. That's less than one per day in a city of eight million people.
It's rare but not unheard of. The video Daniel saw is a real thing, just not a common thing.
And the distribution is telling. The majority of documented escorts happened in the outer boroughs — Queens, Staten Island, parts of Brooklyn — where hospital density is lower and ambulance response times tend to be longer. Manhattan, despite having the highest call volume, had relatively few escorts, because the hospital density and ambulance coverage make the "just wait" option more viable.
The geography of emergency response. The escort is an outer-borough phenomenon because the ambulance math is different out there.
It's a spatial problem. And it connects to something emergency planners think about constantly, which is the "golden hour" concept — the idea that trauma patients have the best outcomes if they reach definitive care within sixty minutes of injury. The concept originated in military medicine but has shaped civilian EMS for decades. The escort decision is essentially a micro-scale version of golden-hour logic. If the transport time plus the wait time exceeds the window where intervention matters, you find a faster way.
The golden hour compressed into a four-minute drive down Flatbush Avenue.
For certain conditions — active seizure, severe bleeding, anaphylaxis, cardiac events — the window is much shorter than an hour. It might be five minutes, or ten. The escort exists because sometimes a civilian car is the only thing that fits inside that window.
What should someone actually do, if they're in this situation? Because the prompt is partly asking for the practical takeaway. You're a civilian, someone in your car is having a medical emergency, you're near a hospital. Do you flag down a cop? Do you call 911 and wait? Do you just drive?
The evidence-based answer — and this is consistent across emergency medicine guidelines — is to call 911 first, regardless. Even if you plan to drive. The dispatcher can alert the receiving hospital, can sometimes coordinate an escort if appropriate, and critically, can talk you through what to do if the patient's condition changes while you're en route. The worst-case scenario is arriving at the ER with a patient who's crashed and no one expecting you.
Call 911, tell them you're driving, let them decide whether to send an escort or tell you to pull over and wait.
That's the protocol. In practice, people panic and drive, and that's just human. But the thing I'd emphasize is what not to do. Do not drive at emergency speeds without lights and sirens. A civilian car running red lights without any audible or visual warning is far more dangerous than waiting an extra four minutes for an ambulance. The data on civilian "scoop and run" incidents is sobering — secondary crashes, pedestrian strikes, the patient becoming a projectile in the back seat because they weren't secured.
The patient becomes a projectile. That's a grim image.
It's a physics problem. A seizing patient isn't wearing a seatbelt properly, if at all. If the driver brakes hard or gets into a collision, that patient is now a hundred-and-seventy-pound unsecured object moving at whatever speed the car was traveling. Ambulances have stretchers with restraints, padded interiors, paramedics positioned to stabilize the patient. A Honda Civic has none of those things.
The escort, when it happens, is actually a harm-reduction strategy. It's not endorsing the civilian's decision to drive — it's mitigating the danger of a decision that's already been made.
That's exactly the right framing. The officer who initiates an escort isn't saying "this was a good idea." They're saying "this is happening, and my lights can make it less dangerous than it would be otherwise." It's the same logic as a needle exchange program — you're not endorsing the behavior, you're reducing the harm of behavior that's already occurring.
Harm reduction for vehicular panic. That's a niche I didn't know existed.
Emergency services are full of these niches. The whole field is basically a catalogue of situations where the ideal option isn't available and someone has to choose the least-bad alternative. Triage is the most famous example — in a mass-casualty event, you literally tag patients as "will die regardless of treatment" so you can focus resources on the ones you can save. That's a horrifying decision, but it's the correct one.
The civilian escort is triage by other means. You're sorting the patient into the "needs to move now" category and using whatever's available.
The "whatever's available" part is key. In rural areas, this gets even more improvisational. There are counties in Montana and Alaska where the nearest ambulance is forty-five minutes away and the volunteer fire department uses personal vehicles for medical transport as a matter of routine. The escort concept is almost quaint by comparison — in those contexts, the civilian car isn't the backup plan, it's the primary plan.
Rural emergency medicine is a whole other universe. "The ambulance will be there in forty-five minutes" changes every equation.
But even in urban settings, the equation is more nuanced than most people realize. Let me give you one more data point that I think ties this together. There's a metric in EMS called the "patient off-scene time" — the interval between when the ambulance arrives and when it departs for the hospital. In a straightforward transport, that's maybe three to five minutes. But for a patient in active seizure, it can be ten to fifteen minutes. The paramedics have to stop the seizure before they can safely move the patient. So the "ambulance is eight minutes away" math is misleading — it's eight minutes plus ten minutes of on-scene treatment plus the transport time. Suddenly the civilian car that's three minutes from the hospital looks a lot more rational.
The ambulance isn't just a taxi. It's a treatment platform, and treatment takes time. The civilian car skips the treatment step entirely, for better or worse.
For better or worse. And "for better or worse" is really the thesis of this whole topic. There's no clean answer. The protocols say one thing, the street says another, and the data is messy enough that both sides can claim support. What's clear is that the escort exists in the gap between policy and reality, and it's probably going to stay there because the gap isn't going away.
The gap between policy and reality is where most of life happens, honestly.
It really is. And the people who navigate that gap — the officers, the paramedics, the dispatchers — they're doing a kind of constant, low-grade ethical calculus that most of us never see. Every decision is a trade-off between competing risks, made with incomplete information, under time pressure, with someone's life in the balance.
We watch the video and think "that looks chaotic," but what we're actually seeing is a system functioning exactly at its edge, where the rules run out and someone has to make a call.
The edge of the manual. That's where emergency services live. And the reason the video is compelling isn't just the drama — it's that you're watching someone operate in the space where the flowchart ends.
And now: Hilbert's daily fun fact.
Hilbert: In the 1780s, the lighthouse keeper at Half Moon Caye in Belize was paid in units of mahogany logs rather than currency. At the standard rate of the time, one year of lighthouse keeping was worth approximately fourteen thousand board-feet of mahogany, which is enough to build roughly three and a half modern two-bedroom houses.
I have so many follow-up questions about the mahogany-to-housing conversion rate.
The Belizean lighthouse-keepers' union must have been fascinating. Thanks to Hilbert Flumingtop for that. To wrap this up — I think the open question here is whether departments should formalize the escort protocol more explicitly, or whether doing so would create a permission structure that leads to more escorts than are actually justified. There's a case for keeping it in the gray zone, where it's technically discouraged but practically understood.
The "don't ask, don't tell" of emergency vehicle operations. It's messy, but sometimes the mess is the point. The ambiguity creates the space for judgment.
Judgment is the one thing you can't write into a manual.
This has been My Weird Prompts. If you enjoyed this, leave us a review wherever you listen — it helps people find the show. We're back next week.
See you then.
