This morning we have a fairly short post about a narrow (but very important) angle of a story that I think we've failed to do justice to.
By now most of you know the basics: 16 year-old Ma'Khia Bryant, kitchen knife in hand, lunged towards another young woman whose back was tight to a car hood. A just-arrived officer, his warnings to Ma'Khia having gone unreplied to, stopped her by shooting her four times in the back of her torso, ultimately causing her death.
Here's the angle I’m interested in (which amazingly it seems no media outlets have demonstrated any curiosity about): what was the likely fatality risk to the potential stabbing victim had the officer not intervened to the extreme that he did?
I suspect that journalists failed to raise this question because they imagined it difficult to answer. But difficult here really just meant “maybe two hours of reading plus a few emails with a researcher”, because it happens that said data both exists and has already been parsed in ways that get us pretty close. Journalism just didn’t look for it.
(For the time-starved, the TLDR here is "a 0.5% fatality risk per stab seems like a good rule of thumb, though we should invest into narrower research that can give us a more precise answer". The story here elsewise is about diving into that data and comparing it to the perceptions informing our current policies and conversations.)
We reward corrections. See something wrong, misleading, or unfair? Say something. It helps. New readers can also learn more about our mission here.
Story Note: Ma'Khia was Black, and the officer white. This matters. But I’m ignoring it for this particular writeup because one way we can have better conversations about inequitable policing is by first aligning on what standard reasonable conduct looks like. Once we have that baseline it's a lot easier to calculate the degree to which it's being diverged from in cases involving certain minorities, etc.
I. Expert Risk Perception
Let's start with the media-quoted experts and how they framed the risks of officer inaction.
Here’s Mike Woods, the local interim police chief:
If an officer is “faced with someone employing deadly force, deadly force can be the response,” Woods told reporters.
Here’s Charles Ramsey, a former Philadelphia police commissioner:
“[If] you don't shoot the one young lady — she stabs the other one and she dies...”
Here’s Philip Stinson, “a Bowling Green State University professor who studies the fatal use of force by law-enforcement officers”:
“It's a terribly tragic situation, and my heart goes out to the girl and her family and friends,” he said. “But from looking at the video, it appears to me that a reasonable police officer would have had a reasonable apprehension of an imminent threat of serious bodily injury or death being imposed…”
Here’s James Scanlon, “a retired Columbus Division of Police SWAT officer who spent 33 years with the division, has since trained officers, and served as an expert witness at trials in use-of-force cases”:
“An officer is justified in using deadly force if his life or the life of someone else is at risk,” Scanlon said Wednesday. “Few would argue that there weren't at least two lives there that were at serious risk.”
In this case, Scanlon said, Reardon wasn't trying to protect himself, “but to save the life of someone he doesn't even know. ... It's a shame that no one has recognized that that officer, in all likelihood, saved one or more lives.”
So the consensus here is that Tionna Bonner’s life was at serious risk, and that when we add in the uncontroversial notion that law enforcement is often permitted to take a life to save a life, it’s clear that Officer Reardon did the only sensible thing.
But just how serious a peril was Tionna in?
II. The Importance of Good Estimation
Let’s make three pragmatic concessions here:
The elapsed time from Reardon leaving his car to shooting Ma’Khia was just 10 or 11 seconds. He didn’t have opportunity to weigh context or any signals about intensity or intent — or for anything but a snap decision.
While we can point out how other countries have police forces that aren’t nearly as reliant on guns, we need to ground ourselves in what’s reasonable for US cops based on US cultural norms.
While Tionna was putting up some defense, let’s assume that we can’t count on it having been effective enough to fully block Ma’Khia’s attack.
But even if we concede that a drawn gun is the reasonable baseline here, we can still imagine that four bullets to Ma’Khia’s torso was not the only plausible remedy. Either of a warning shot or a single bullet were plausible median options. But of course we’d then have to allow for the possibility that Ma’Khia would have been able to complete one full stab before either waking up to what she was doing or shifting focus to the reality of having been shot. What would this have meant for Tionna?
Say the risk of her dying from a single stab wound was 20%. I’d imagine few communities would want their police officers to take that kind of a gamble.
My intuition though was that there would be some pretty wild variance in terms of guessing at the risk was here, and that most would heavily overestimate it. So I did a (super unscientific) Twitter poll to get a rough sense:
(Note that Twitter limits you to four options, which makes these findings even less scientific. I suspect that many would have gone quite a bit higher than 20%, as at least one respondent explicitly said they would have.)
Anyway, the ranges here are a problem. We can’t expect officers to guess well, nor to do math amid the fog of police action. If we want consistent behavior from them, we need to equip them with simple rules of thumb that they can build of a sort of muscle memory around during training drills.
So what is the right rule of thumb here?
III. The Data We Have
While neither of these papers intended to answer our specific question, each gives us useful data to work backwards from:
Study 1 found that roughly 2% of patients with a stab wound who ended up at a level I/II US trauma center died (almost equally split between those who died after admission and those who arrived dead).
Study 2 found that this number (using similar but not quite identical data) was roughly 3.6% if we limit ourselves to those stabbed in the torso.
While these rates are already low-ish as is, it's important to factor that fatalities are almost certainly skewed towards those stabbed multiple times, where the rule of thumb we’re searching for is fatality risk per stab. And from those numbers it seems reasonable to imagine a figure in the neighborhood of 0.5%.
[EDIT 05/28/21: One thing I should have clearer about here is that trauma centers are not synonymous with ERs. While they’re often co-located, patients are generally only triaged into a trauma center when there’s a danger to life or limb. So that 2% total fatality figure isn’t being watered down by scrapes or non-serious wounds.]
Of course, there are caveats here:
Average fatality risk is low because stab wounds in a struggle are widely distributed across body areas, some of which are far less vulnerable. But in constructing a rule for police intervention, it’s not unreasonable to round up to a more conservative figure like 1%. It’s just more philosophy than science.
These datasets cover assaults, accidents, and self-inflicted wounds. Those two studies didn’t calculate fatalities by category (though future studies could).
Both papers are missing some stabbing victims who died without ever reaching a hospital. (One older UK study found that 80% of fatalities happened pre-hospital, where the US studies found closer to 50%. While we wouldn’t expect those numbers to match, we can assume the real US number is a bit higher.)
Only 85% of Americans live within an hour of "the protective halo of a trauma center", which means we're probably undercounting deaths a bit there too (as the study data came from ~700 participating trauma centers that presumably offer slightly higher standard of relevant care vs. normal hospitals).
(For more on these caveats, see the comments section. I’ve pasted a few excerpts from my email exchange with Dr. Ahmad Zeineddin, the lead author of the first study.)
Anyway, I perceive two truths here:
While it isn’t the job of academics to decide how police departments should act on data, sufficient data does exist in the National Trauma Data Bank to somewhat decisively quantify fatality rates per number of stab wounds from assaults. We’d just need to commission a more specific study.
While there are too many variables (wound location, weapon size/shape, force, intent, etc.) to ever allow a precise rule, we can probably get close enough to give communities the means of deciding on broad policies.
All we need to do is mix the right researchers with a bit of grant money.
(Note that these community conversations would also need to weigh the risks that officers will sometimes miss, and that some on-target bullets will either get deflected or pass through their targets and strike others. While it's reasonable to imagine that these risks are quite low, they're certainly non-zero.)
IV. The Moral / Where to Go From Here
This isn't about whether Officer Reardon's actions violated the Graham v. Connor reasonableness standard or any local regulations. I certainly don't see a crime here. I see a 23 year-old who followed his training in neutralizing a threat. (The justice system is still reviewing, so we could theoretically be in for a surprise based on details not captured on the film, though that’s pretty unlikely at this point.)
But even if he met the standard of what was expected of him, the question is whether that standard reflected what Ma’Khia’s community would have wanted had they had a better sense of the data. If they believed that the probability of Tionna’s fatality from a single blow was into significant double-digits, we can perhaps imagine them lamenting Ma’Khia’s death while also acknowledging the tragic necessity of it. But what if that number was 0.5%? Or even 1%? Compared to the double-digit likelihood of Ma’Khia dying from those four bullets (plus whatever risk of one of those bullets striking Tionna), we could expect a very different conversation overall.
Now obviously different communities would still reach varied conclusions even with clear data in hand. But at least those choices would be informed. The entire basis of productive democracy is people with competing preferences engaging in good-faith debates around the stable center of good data. The data part is essential.
But we won’t get this data unless someone funds a study or two, and the chances of that happening would go way, way up if journalism were to shift from deferentially quoting overconfident and data-starved experts to raising awareness of the information gap and pointing out how trivially it could be closed.
Interesting analysis! I think the question that you didn't fully address here is whether the well-being of the aggressor should be given significant weight relative to the victims. My personal opinion is that the well-being of the victims should take full precedence over the aggressors and the correct course of action is to disable the aggressor as quickly as possible (making sure victims are safe) before providing medical assistance to the aggressor.
Excerpts from my emails with Dr. Zeineddin.
----
Me:
Your paper came up in my research on baseline mortality from stab wounds, and was very helpful. Though one thing that wasn't clear was whether it was capturing 100% of the dead (in that the NTDB's data is predicated on ICD codes, which seem conditional on hospital admittance). To your knowledge, do the DOA-inclusive numbers in table 3 capture those who aren't sent to the hospital with hope of rescue/resuscitation?
My naive sense was that even if someone was clearly dead they'd still be sent to the hospital so that an ME could look them over. But wasn't 100% clear on that (or if MEs used the same codes), and I don't know anyone who works in trauma to ask. So thought you might know offhand.
----
Dr. Zeineddin:
Other than the voluntary nature of data contribution you also highlight another [data] limitation in your question. Do all patients with a stab (or [gun shot wound]) present to a hospital or trauma center even if they are clearly dead at the scene? The answer is no. They could present to a hospital that is not a trauma center where they would not be captured in these databases or they could be pronounced at the scene as you were asking.
The laws regarding the ability of EMS personnel to pronounce death at the scene differ by jurisdiction so there is no clear percentage of how many victims are pronounced at the scene and how many are brought in to the hospital and pronounced there. *NTDB should be able to capture the latter but not the former.* Another factor that complicates this issue is that EMS in jurisdictions where they are able to pronounce at the scene by law, might elect to bring the victim in to the hospital due to safety concerns at the scene or as to not escalate the tense situation at the scene by declaring the death of the victim.
----
(As an aside, he also noted that "with the improvement in trauma care, the fatality rate of stab wounds or even gunshot wounds to a single body region have improved significantly over the 14 years studied." So there's that too.)