New study: State medical examiners wrongly excused homicides committed by Maryland police officers
After the Derek Chauvin trial, the state ordered an audit of autopsies done on people who had died in police custody. The results are disturbing.

When former Minneapolis police officer Derek Chauvin was tried for the murder of George Floyd, former Maryland Chief Medical Examiner David Fowler testified for the defense. Fowler told the jury that Chauvin’s actions did not cause Floyd’s death. He disagreed with Minnesota medical examiner Andrew Baker’s classification of Floyd’s death as a homicide.
That alarmed a lot of medical professionals, activists, and state officials. Fowler’s office had been involved in a number of controversial in-custody deaths in Maryland. Other medical professionals had questioned his office’s classification of those deaths as natural, accidental, or undetermined. Such classifications typically close off any further investigation into police culpability. They also tend to shut down discussion about whether such deaths could have been prevented with different policies or better training.
After the Chauvin trial, Maryland officials ordered an audit of how Fowler’s office classified in-custody deaths. The results of that audit were published in May. They were alarming.
The audit designers first found 87 deaths in police custody in which police had used restraint. Then assembled a team of medical examiners to review the autopsy reports in those cases and come up with a manner of death determination — whether the death was a homicide, accident, natural death, or undetermined, meaning there was insufficient evidence to make a determination. The audit used a process called sequential unmasking to expose the reviewers to only the evidence they needed to draw their conclusions, while shielding them from irrelevant but potentially biasing information, such as the race of the deceased, or whether the deceased had a criminal record.
Of the 87 deaths the auditors examined, they found 36 in which the Maryland Office of Chief Medical Examiner (OCME) should have classified the manner of death as a homicide, but instead classified it as something else. The audit also found that OCME pathologists were less likely to classify deaths in police custody as a homicide when the decedent was black. The auditors, who again were initially blinded to the race of the decedent, had no statistically significant racial disparity in their manner of death determinations.
As I’ve reported here previously, law enforcement interests and police-adjacent companies like Axon have expended a wealth of resources trying to dispel the idea of positional asphyxiation — that police are killing people by tasing, beating, or putting their weight on suspects while they’re restrained, particularly in the prone position. They’ve pushed dubious concepts like “excited delirium” or “sickle-cell trait” to explain away why otherwise healthy people die at the hands of law enforcement.
Some factions of the medical examiner community have been complicit in pushing these ideas. During a contentious listserv discussion after Floyd’s death, for example, a former state medical examiner in Milwaukee — who also consulted for Chauvin’s defense — wrote, “Is there anyone in our profession that has not, at one point or another, quipped about ‘spinning the wheel of death’ and picking one?”
Yet when studies have demonstrated cognitive bias in how autopsies are conducted, many in the medical examiner community have lashed out with anger and defensiveness, and refused to implement procedures that could help filter out some of that bias.
The Maryland study is just the latest piece of evidence that police are routinely restraining people in ways that are causing premature, unnecessary death — and that by refusing to shield themselves from bias, medical examiners are ensuring that those deaths continue.
I spoke with William Thompson, professor emeritus of law, criminology, and psychological science at the University of California-Irvine, and one of the co-authors of the Maryland study.
I’ve edited our conversation for length and clarity.
Can you talk a little about how this report came about?
It’s fallout from the George Floyd case. When Derek Chauvin was tried for Floyd’s murder, the former chief medical examiner in Maryland, David Fowler, testified for Chauvin defense. He claimed that the manner of death in the case should have been labeled “undetermined” rather than “homicide.” He offered various theories as to why he thought it was not clear that this was a homicide, even though we all saw Chauvin kneel on Floyd’s back for an unconscionably long period of time.
So, after Fowler gave that testimony, a large group of medical professionals wrote an open letter complaining about Fowler’s testimony and calling for an investigation. It was ultimately signed by over 500 medical professionals. Interestingly, it did not include a lot of active medical examiners. But it did include people who have expertise in forensic and other areas of pathology, pulmonology, cardiology, and various other specialties. They sent the letter to the governor and the attorney general of Maryland.
The letter demanded an investigation into Fowler’s office while he was the state’s chief medical examiner. That office does all of the autopsies in Maryland, including every death in police custody. And Maryland had had a number of controversial cases of its own involving death in police custody — Anton Black, Freddie Gray, Tyrone West. Just a lot of cases where someone ended up dead after encounters with the police.
In some of those cases, the medical examiner's office found the manner of death to be “undetermined” rather than homicide. The medical examiner's determination about whether a death was a homicide is really important in two ways. First, it’s the predicate for further investigation into whether the death was a crime, and whether or not someone should be prosecuted for it. A homicide finding doesn't necessarily mean that someone is criminally culpable, but it might. It flags the case for prosecutors to continue investigating.
If a medical examiner determines that a death was a homicide, it may or may not be a crime. But anything short of homicide usually ends any further criminal investigation. Is that correct?
That’s correct. What we usually see is that they’ll say the manner of death is “undetermined.” And when they do that, it means they believe there isn’t enough evidence to say if it was a homicide. And when they do that, it’s very, very unlikely that any criminal investigation will follow.
So that’s one part of it. But the other reason manner of death is important is for public health. A homicide determination also has really important public health ramifications, because it flags the case as a case where somebody died due to the actions of another human being. It means that, but for the actions of another human being, this person would still be alive. And if we see that people are dying because of the actions of other people, it raises the obvious question of whether the actions that caused this person to die were really necessary. Could this death have been prevented? Could we prevent similar deaths in the future just by changing policies or behaviors?
But if a death is classified as “undetermined,” these questions don’t get asked. If the medical examiner says “undetermined” or “accidental,” that's pretty much the end of it.
So with deaths in police custody, only a homicide determination leads to difficult questions like, did the police really have to do what they did here? Was there some other, less harmful way this could have been handled?
So for both of these reasons, it’s really important that medical examiners be willing to make the homicide determination when appropriate — and that they have a relatively low threshold for doing so. You don’t need proof beyond a reasonable doubt to classify a death as a homicide. But there’s good evidence that medical examiners have a higher threshold when they’re looking at deaths in police custody.
Why do you think that is?
One problem is that the standards medical examiners apply when making these determinations are too vague. They’re put out by an organization called the National Association of Medical Examiners. NAME has been criticized for being hidebound, conservative, and oriented toward law enforcement. We looked at the standards closely, and on some level I think they get it right. They apply a “but for” test for any kind of unnatural death — “but for” the actions of another person, would this death have still occurred?
I think the problem is that NAME doesn’t specify what level of proof is required to get to a homicide. Is it beyond a reasonable doubt? A preponderance of the evidence? It doesn’t say. So that leaves all kinds of room for human judgment and potentially human biases, even for people who are trying to follow the guidelines carefully. Depending on which threshold you apply, you can reach a completely different manner of death.
So even if we assume everybody’s doing their best to be professional, unbiased, and ethical, there’s just too much room for differences of opinion built into the guidelines.
And of course all of this has been really complicated by the emergence of bullshit medical theories about about deaths in police custody, and particularly this theory of excited or agitated delirium.
Yes, I definitely want to ask you about that. But let’s get back to the letter that brought on this whole investigation.
Yes, back to that. So Roger Mitchell, who is the medical examiner in D.C., organized this letter with 500 signatures from medical professionals and sent it to the governor and the attorney general of Maryland. They looked it over and decided they needed to do some sort of investigation, but didn’t initially know how to go about it. This sort of thing — you might call it a post-mortem post-mortem — had never been done before. So they reached out to a number of different groups and organizations for ideas. Ultimately, they decided to appoint an advisory committee, which they called the Audit Design Team. Basically, the idea was to assemble a group of people with expertise on research methods, cognitive bias, and forensic pathology, get them all together and have them design a way to review these cases that would be robust and scientifically sound.
So you were part of that audit team?
Yes. There were initially four forensic pathologists and four people from the social sciences. Initially, it was a struggle getting the two sides to agree on a general design for the study. The social scientists had very different ideas about how to proceed than the medical examiners. It took about a year for us to produce a report. But finally in August 2022 we produced two. One was for the public, and one was a longer, more detailed report that had to remain confidential at the time because it contained background information on why the audit was necessary.
Interesting. So the initial confidentiality was so anyone who might be evaluating the old case wouldn’t be exposed to biasing information?
That was part of it. When we initially started looking at the data, we found all kinds of things that were suggestive of problems. But if we wrote about those problems in the public report, it might influence the evaluation process.
One of the first things we did as an audit team was to ask the current Office of the Chief Medical Examiner to identify and provide information on all death in custody cases over the period we were looking at. We fairly quickly got a large database with over 1,300 cases. Those cases included characterizations of cause and matter of death from medical reports. Then we spent some time looking at public databases and found a few additional cases that had been missed. We winnowed that down to a set of 87 cases in which someone died in police custody following physical restraint and the cause and manner of death appeared uncertain.
Our key goal was to find a scientifically rigorous way to review those cases in order to assess whether Maryland’s manner of death determinations were influenced by racial or pro-police bias. Eventually we came up with what I think was a pretty good approach, which involved the use of blinding and unmasking procedures.
I want to get to that, too. But can we talk a little bit about your background? How did you get interested in forensics, and why were you chosen for the audit team?
I’m a psychologist and a lawyer. I was trained in cognitive and social psychology, and I study human judgment and decision making. I’m really interested in the interface between forensic science statistics, computers, and human judgment. Although I have gotten involved in cases as a lawyer from time to time, I am primarily an academic. I have been at UC Irvine since 1983. I’m now retired with emeritus status, although I am still carrying on with some research.
I think I was chosen for the audit design team because I have published a number of articles about the potential for bias in forensic science and how to deal with it. The Maryland authorities were interested in assessing whether the medical examination system for the state was affected by racial or pro-police bias and, if so, how to fix that problem. They identified me as someone who might know how to do such an assessment in a rigorous, credible way.
You mentioned that few medical examiners signed the letter sent to the Maryland governor and attorney general. I’m sure you’re also aware of the furious backlash from the medical examiner community against the authors of a series of papers on cognitive bias in death investigations. Was it difficult to find people who were willing to second guess the work of Fowler and his staff on these old cases?
Yes, I understand that it was initially difficult for the attorney general’s office to find medical examiners willing to work on the Audit Design Team, although ultimately they assembled a truly distinguished group. The attorney general’s Office in Maryland reached out to NAME, which is the major professional organization for medical examiners in the U.S., but as I understand it, they received no cooperation. NAME thought such a review was unnecessary and wanted no part of it. In the end, three of the four forensic pathologists on the team were from outside the U.S. — Australia, Northern Ireland and Canada. The one American medical expert, Michael Freeman, was trained in forensic pathology in the U.S., and has been active as an expert witness, but is primarily an academic. He is particularly knowledgeable about the dangers of prone restraint, however, and also has training in epidemiology, so he has been a particularly valuable member of the team.
It was somewhat easier to recruit U.S. medical examiners to review files in specific cases and give a second opinion than it was to recruit them for the audit, although everyone recognized that the study was politically sensitive, and some potential reviewers elected not to participate for that reason. The majority of our case reviewers were U.S.-based forensic pathologists. We promised them that their evaluations would be anonymous, and we have not reported their names.
The U.S. is the only country in the developed world in which medical examiners testify at trial as to manner of death. In other countries, they might make that determination for public health data, but not in a criminal trial. It tends to be made by someone with legal training, or left for judges or juries to decide. Did you have any complications in asking foreign medical examiners to review manner of death determinations made by U.S. pathologists?
Seven of our twelve reviewers were practicing forensic pathology in the United States; five were from other countries — three from the U.K., one from Canada, and one from South Africa. For the benefit of those without experience in the U.S., we started by giving them a training session on the NAME guidelines for manner of death determination. We brought in a prominent, well-regarded forensic pathologist who has been doing professional training on the standards for a long time. We also had them review some other materials. For example, there have been a lot of critiques of excited delirium. So we gave them a bibliography and some background articles on excited delirium — including articles defending the concept. We also gave them statements by various medical associations on the topic so they would be up to speed on the current thinking in the field.
Were there any statistical differences between U.S.-based reviewers and those from other countries? Was either group more likely to disagree with the Maryland Medical Examiner’s Office? Was either group more likely to classify these cases has homicides?
No. We looked at that, and we didn’t detect any differences based upon the national origin of the reviewers.
Just to play devil’s advocate, given press coverage of the audit, I’d imagine that most of your reviewers were aware of the origins of this project. Since we’re talking about preventing bias, do you think just knowing why they were being asked to review these cases in the first place might have nudged some reviewers toward ruling more cases as homicides?
I think they all understood why the attorney general was doing this. But they were told that we wanted an objective, independent, scientifically responsible opinion. In other words they were told, Do your best and give us the truth. Be unbiased. I don't think there is any evidence that any of them had a particular ax to grind. They were just giving us their opinion of what the correct determination should have been.
There were definitely differences among our reviewers in how often they found homicide, and I do have the sense that some of them applied a higher threshold than others. But there were three reviewers for each case, and when they disagreed we had consensus meetings to discuss their disagreements. There were only a few cases where they could not reach agreement.
People like Peter Neufeld argue that manner of death is a legal determination, not a medical one — and that the U.S. should join the rest of the world in leaving that question to fact-finders, not medical examiners. The fact that you can have so much disagreement between experts on such a profound and consequential question would seem to support that argument. How can we tell juries that manner of death is the product of training and “expertise” when it’s clearly so subjective?
I agree that it is problematic for medical examiners to testify about the manner of death in criminal trials. Manner of death determination is not just a medical finding; whether a case is a homicide or not also depends critically on the circumstances of the death. It may not be clear to the jury the extent to which the ME’s opinion depends on medical facts from the autopsy rather than circumstantial facts that the jurors may have already considered, or may view differently. So, an opinion on the manner of death may be difficult for the jury to evaluate properly. Given that, it might well be better to limit medical examiners to testimony about the medical findings of the autopsy when they testify in criminal trials, and not allow them to opine on whether the death is a homicide.
That said, I think manner of death determinations play a really important role in public health assessment. Most of these cases are not going to be taken to trial, and the major purpose of the of the medical examination isn’t to determine whether somebody should be prosecuted, it’s to identify public health problems that could be addressed, and few people are better positioned than medical examiners to tell us that this is a case where somebody died because of the actions of another human being.
I do think medical examiners need better contextual information than they often get, and I do think that they are relying too heavily on the police, particularly for in-custody death cases, where the police themselves are an interested party. But ultimately I believe that there’s value to the manner of death determination in public health, because we want to know when there are unnatural deaths that could have been prevented.
David Fowler’s testimony at Chauvin’s trial alarmed a lot of people, which of course is why you were asked to do this audit. But I’m curious what you make of the testimony by Andrew Baker, the Hennepin County medical examiner, and the doctor who did the autopsy on George Floyd. I’ve talked to a lot of people who found his testimony really frustrating.
I think Baker is an intelligent, well-meaning guy. But he’s also a political actor. He has played a big political role in NAME. He has long been on the front lines of opposing efforts to bring any regulation or oversight of medical examiners. He’s been particularly resistant to any suggestion that medical examiners could be biased, or that we should take steps to reduce their potential for bias.
For years I was a member of OSAC, the body charged with developing scientific standards for these fields. I chaired the “Human Factors” committee, which worked with various forensic fields to introduce bias reduction procedures into their standards. We got more resistance from medical examiners than any other field, and Andy Baker was a leader in those efforts. They seem to see any effort to minimize bias as anathema to medical practice and professional autonomy.
It’s notable that you were sure to implement those masking and anti-bias precautions in the review of the old cases in Maryland. You all were careful to follow your own advice.
It was a key feature of the study. We blinded our reviewers to the race of the decedents, for example, which I thought was essential. We didn’t want to bias the reviewers in the direction of saying black deaths in police custody should have been homicides. We first wanted a straight analysis, without any contextual information that could bias their conclusions.
We also blinded them to other things, such as what the medical examiner’s original determination had been. That some took some persuading, even with the medical examiners who agreed to do the review.
And you would then reveal more information as it became necessary, right?
Yes, that’s correct. We asked them to give us an initial opinion without knowing the race of the decedent or the findings of the Maryland Office of Chief Medical Examiner. We then gradually “unmasked” this information. For example, they gave us initial evaluations without seeing autopsy photos that revealed the decedent's skin tone. Then we allowed them to review the photos and asked them to evaluate the case again in light of this new information. Eventually the reviewers knew everything, other than the identity of those involved. By using this sequential process we were able to get initial opinions that were race-blind and then see whether — and why — those opinions changed.
As I understand the study, there were only a handful of cases in which exposure to non-medical information actually changed the reviewers’ conclusions. Is that correct?
Correct. And when it did change their conclusions, more information tended to make them more likely to think there had been police abuse. There was one case, for example, where there were injuries that weren’t apparent from the medical examiner’s written report. Once the reviewers saw the autopsy photos, they said okay, something else clearly happened here. This person was handled more roughly than the original report describes.
There were a couple cases in which there was also body camera footage, which was obviously an important factor in the George Floyd case. Did that have any effect on the reviewers’ conclusions?
I think so, but it’s hard to say for certain because we only had a couple cases with video. But body camera footage shows what actually happened, and can be really crucial to determine things like the exact duration of the restraint, and particularly the duration of pressured restraint in a prone position. Just exactly how long were the officers sitting on the guy before they let him up?
There was a real story — a real person behind all 87 of these cases. Part of my job was to help classify these cases according to who was primarily applying the restraint, whether it was primarily law enforcement or someone outside of law enforcement.
That couldn’t have been easy, delving into the details of 87 different people who died in police custody.
It is one of the more depressing things I've ever done. Just case after case where people end up dead. You have to ask, “Why did this happen?” and “Could it have been prevented?” Some of the cases are heartbreaking. The story, all too often, I think, is that somebody was behaving in a aggressive, out-of-control manner, often because of drugs or mental illness, or maybe they’re just really mad and they’re struggling. The police officers would come in and immediately they get the person down on the ground, get their hands behind them, and restrain their arms and legs. But the person continues to thrash around and, even though they’re restrained, they’re trying to kick at the officers. The officers won’t tolerate that, because they’re trained to force submission. So they keep sitting on the person, saying, “Stop struggling, stop resisting. You’re still resisting. Stop. Stop resisting. Stop resisting.”
And it’s during this period when the person reaches the point where prone restraint and the pressure police are applying overwhelm the respiratory process.
A lot of police training manuals warn that this can become a self-reinforcing feedback loop. The person starts thrashing while under restraint because they can’t breathe. Sometimes they’re just trying to create a small amount of space to allow themselves to inhale and exhale. The police interpret the thrashing as resistance, so they increase the pressure.
That’s right. Many times, the person at this point may also think the officers are trying to kill them. Or they’ve seen the stories about how other people have died under similar conditions. A lot of times there’s just a lot of confusion. Many of these people were mentally ill, and were having some sort of crisis. But I think the literature shows that once police put on the handcuffs and leg restraints, if they’d just put the person on their side, back off, and let them thrash around for half hour or so, many of the 87 people whose cases were reviewed would still be alive.
I think that’s an important point. When this audit was first announced, there was this initial reaction from law enforcement that you all were just trying to make it harder for cops to do their jobs. Or you’re just trying to make it easier to charge police with crimes. But this is really about preventing unnecessary deaths. It’s about how, with better training, maybe fewer people will die.
I think a lot about one case involving a guy with Down Syndrome. He loved going to the movies. He had bought a ticket and watched a movie. But when the movie was over, he wanted to stay in his seat for the next showing. I think there was some dispute where he arrived late but paid full price, so he didn’t get to see the beginning. The movie theater had a policy that you have to pay for each showing. So they told him to leave, and he refused. So the security guards come in and grab him and try to forcibly remove him. He starts fighting back, so they call the police. Everybody sits on top of the guy, and he ends up dead. All over the price of one movie ticket.
Yeah, it seems like some basic empathy and situational awareness could have prevented that.
Right. You’re really going to arrest, cuff, and restrain this intellectually disabled guy over the price of a movie ticket? I mean, I think his death could have been prevented with better training. But maybe you also just let him watch the movie.
There’s another case of a guy who who went to an urgent care-type clinic at night. He starts to get agitated because other people were being attended to before him. At some point he had been there for more than eight hours. He gets increasingly angry because he’s been there eight hours and nobody has called him in to see the doctor. So he goes up to the desk attendant and says something like, “You’ve got to let me in to see the doctor. Now. I insist, I insist.” They say no and call in security guards to throw him out. There’s a big fight, they end up restraining him, and he ends up dead. You could see how this guy would be legitimately angry, particularly when instead of addressing his problem they just bring in private security guards to throw him out.
In another case, a guy stole a bicycle and fled the police, who then restrained him until he died. Just so unnecessary. I would hate if somebody stole my bicycle. I would want them to be punished. But I would not want them to get the death penalty.
It’s always been curious to me that you almost never see an excited delirium diagnosis outside the context of police custody. You don’t see it diagnosed when someone dies during a bar fight, or a street brawl, or from a caretaker restraining a child. Did the audit look at any cases in which someone died from restraint that wasn’t applied by law enforcement?
Yes. We were able to compare cases where the restraint was primarily applied by police with cases where it was primarily applied by someone else, such as a security guard, family member, medical personnel, or bystander. In all of these cases the decedent was in police custody at the time of death, but in these particular cases, the police typically arrived and took over after the decedent had initially been restrained for a period of time by others. The interesting finding was that, relative to our reviewers, the Maryland medical examiners were less likely to call the case a homicide when the primary person applying restraint was a police officer. Even for the subset of cases that our reviewers unanimously labeled homicides, the Maryland examiners were less likely to find homicide when it was police, rather than non-police, who were primarily involved.
When I was looking into this the Fall of Minneapolis film and writing a response to it, I looked at training manuals for dozens of police departments across the country. Nearly all of them teach officers how critical it is to roll people over onto their side. Nearly all of them warn about prone restraint and putting weight on the backs of restrained suspects. So this is taught. It is policy. DOJ, NYPD, and LAPD have been teaching this for years. But then you start looking at court cases and there seems to be this huge disconnect between what’s taught as generally accepted practice in the law enforcement community and what policies actually get enforced in court.
In courts, you have people like Gary Vilke and interest groups like Axon pushing the idea that positional asphyxiation just isn’t something that ever happens. And they seem to have an outsized influence in the courts that isn’t reflected in training manuals and police policies. Why is there such a disconnect between these widely accepted law enforcement protocols and what happens in court when those protocols are violated?
I can really only speak to the 87 cases that I looked at in connection with this study. But my impression is that, in practice, these officers are trying to achieve dominance over and capitulation from the person who is fighting them. I think the officers view it as a fight. I think they view themselves at risk. In some of these cases, you have witnesses saying that the officer was continuing to say “stop struggling” or “stop resisting” well after the person was completely restrained, or even unconscious. I think it’s hard for police to back off and disengage once the adrenaline is flowing.
So in those cases, it’s going to take a lot of training. But it is not just poor training. We also have cases where the police tased someone after they were already completely restrained. Why would you do that? Even if the person is still struggling or thrashing, once they’re restrained they’re no longer a threat. It seems like they often do it out of some misguided desire to punish the person for struggling. Or it’s just part of police culture to force compliance — just no tolerance for anyone who’s showing any defiance at all. I think that’s more of a police culture problem than a training problem.
I remember one case where a mentally disabled kid ran from police, ran back to his own house, got into his parents’ car, and locked himself inside. At that point he isn’t a threat to anyone. Why not just wait him out? But of course that isn’t what they did. They smashed the windows and hauled him out of the car. He struggled, so they put him on the ground, restrained him, put weight on him, and he died.
So what’s the rationale for that? I mean, I guess they might think that this guy is still dangerous, but isn’t he less dangerous inside the car? Even if you think he might have a gun, aren’t you going to get shot when you break the windows? It just doesn’t make a lot of sense to me. Sometimes it really does seem to just be about dominance and punishing defiance rather than rationally figuring out how to deal with a possible threat.
I’m struck by the fact that all we have is this one review in one state, and even that only happened because of this really unusual series of events — a death in police custody in another state, national protests and outrage, a trial in which a former state medical examiner gave disturbing testimony, all of which resulted in an audit because of an intervention by hundreds of medical professionals, a governor, and state attorney general who happened to be sympathetic. And still, this audit found that three dozen deaths in police custody that should have been homicides were wrongly classified in a way that cleared the police and precluded any further investigation. If you start to extrapolate those results and scale them nationally, you’re looking at thousands of preventable deaths. Do you think it’s appropriate to think about this study that way?
I don’t see any reason to think that other states would be all that different. There are some indications that the Maryland office was more likely to diagnose excited delirium. Fowler was coauthor of an article that we cite in our report in which he discusses an office policy stating that, in any case involving excited delirium, the manner of death would be classified as undetermined. So Fowler might have been more of a champion of excited delirium than other medical examiners.
But I certainly think that’s something we need to look at. The statistic from our study that I find most alarming is that our reviewers were four times as likely to classify these cases as homicides as the state medical examiner’s office. So of the 87 cases, they classified 12 as a homicide. Our reviewers said 48 were homicides. If that’s happening nationally, that’s a really big deal.

Does the Maryland medical examiner’s office still have that policy?
When we got there in 2022 and were trying to design this study, we interviewed the then-acting state medical examiner, the chief medical examiner, and an assistant, and asked them their policy on excited delirium. They said it’s up to the judgment of each individual medical examiner.
There was a new medical examiner appointed in 2023 and, as I understand it, she has announced they no longer recognize excited delirium, in part in reaction to our report. But we recommended that there be a five year follow-up to see if things have changed in Maryland.
But the study design we came up with — with the masking, having an independent panel of experts review selected cases in a redacted way — that could be done anywhere. Any state official who is seriously interested in testing to see if their own medical examiner’s office has the sort of racial and pro-police bias that we that we saw in this audit could do a similar audit in their own state.
You’d just have to get your medical examiner’s office to pull out their most recent death in custody cases, have those cases redacted the way we did, and have them reviewed by an independent panels of experts.
Did you get any feedback from the foreign reviewers about excited delirium? My understanding is that it isn’t a diagnosis used very often in other countries.
That isn’t something we really looked into. We did ask for feedback, and there are a lot of quotations at the end of the report, some of which address excited delirium. They mostly said that it isn’t a valid diagnosis, and that it’s often used to excuse police abuse or explain away deaths that would otherwise raise uncomfortable questions. But the comments were anonymous because we wanted people to speak freely. So I couldn’t tell you which were from foreign reviewers and which were from reviewers based in the U.S.
I interviewed Itiel Dror a couple years ago. One thing he told me is that medical examiners conduct a lot more tests when there’s a death in police custody. They’ll be more thorough in order to find some kind of underlying health condition to blame the death on — which deflects responsibility away from the police. But in these cases it seems like it was the other way around. I saw a couple comments from reviewers about how they expected to see more histological slides and other testing.
That’s true, but we don’t have comparison cases from the same medical examiner’s office involving deaths that weren’t in police custody. It could just be that the OCME doesn’t do as much testing in general.
I’d also point out that there were some things the office did well. They did pretty standard toxic screening, but our reviewers thought that the office was good at getting outside neurological consults — that they went further on that than many medical examination offices do. But they did find deficiencies in a lot of other areas. So I would say our findings with regard to how much testing they did was a mixed bag.
Itiel was part of this project initially, but was not able to continue after the first few months. I think he did suspect that if we looked, we would find that the MEs did more testing in police-involved cases than in other suspicious deaths. But it would have been hard to design the study in such a way that we could test that hypothesis in a rigorous way.
One of the frustrating things about the Floyd case was that his death received the sort of attention that all in-custody deaths should get, but don’t. They brought in outside consults. Baker made preliminary conclusions based only on medical information, then revised after viewing the body camera footage and cell phone videos — exactly the sort of masking of information reformers have advocated for (and that, as you point out, Baker and other medical examiners have resisted).
But the political right saw all of this as confirmation of a conspiracy to frame Chauvin. I think what your study shows is that if state officials gave the proper amount of attention and analysis to these cases, we’d find a lot more in-custody deaths were actually homicides, and we could pass policies aimed at preventing them. So is there any effort, in Maryland or anywhere else, to make these precautions the standard way of investigating deaths in police custody?
There has been a lot of talk in the OSACs and the American Academy of Forensic Sciences about just that. Barry Scheck and I gave a proposal at AAFS about 10 years ago that there should be this kind of masking at the very least in high-profile cases, that there should be more screening, and that someone should be filtering out potentially biasing information that isn’t medical in nature. The leading critic at the time was Andy Baker. He was absolutely adamant that it was the wrong thing to do because it would compromise medical professionalism. He even went so far as to claim it would be medical malpractice.
But then, when Baker had the Floyd case land in his lap, as the medical examiner of Minneapolis, he did try to reduce his own potential for bias. As I understand it, he purposely chose not to watch the body cam video until after he evaluated the medical findings of the autopsy.
But it isn’t that he didn’t consider the body camera footage, right? He just did the medical exam part of the autopsy first, based only on medical information. Then he watched the body camera footage and factored that into his manner of death determination.
That’s right. And I think that’s fine. He had to watch the video before making the ultimate call on manner of death because the surrounding circumstances were crucial to that assessment. He apparently thought that seeing the video might have a biasing effect on how he interpreted the medical findings, and delayed watching to avoid that possibility. Controlling the order in which information is considered by an expert is exactly what psychologists who study expert decision making have been recommending as a way to reduce bias in forensic science.
I will say that Baker is a charming guy in person, I've had meals with him. He’s enormously likable, and I think, really well meaning. He did not respond well, at least not initially, to the suggestion from myself and others that medical examiners might be biased and should do things to reduce their bias. That just did not go down well with him. So, it was encouraging to see him taking steps to minimize his own bias when he encountered this really divisive case.
So the sort of information masking you did in the audit, is that scaleable? Would it be possible to implement it at medical examiner offices around the country?
So it’s definitely time consuming. It took us a lot of time to redact each file. So I think it would be hard to do that routinely for every case. Perhaps you'd want to do it only for particularly challenging cases involving in-custody deaths or high profile cases where there are concerns about bias.
But if I were running a medical examiner’s office, I would just start by telling death investigators that there are certain things they could exclude from the file so the medical examiners don’t see them. Do we really need to know the prior criminal record of the decedent? Probably not.
There could be certain things that are routinely redacted from every case file. And then, in high-profile or controversial cases, there could be an additional review process, to remove potentially biasing information that is not medically relevant, and to control the sequence in which information is presented in ways designed to minimize bias.
I was struck by the fact that after Floyd’s death NAME actually put out a statement discouraging getting a second opinion on autopsy results. They later retracted it, but still.
I think one of the main takeaways from our study is that we can expect a very low level of reliability or consistency among medical examiners in manner of death determinations.
Doesn’t that suggest that manner of death is too subjective to be considered science? I mean, you don’t have that sort of variability when it comes to single source DNA sequencings, or blood type. If experts can’t agree on basic conclusions, should we really be telling jurors that this is expert testimony?
There’s a saying among statisticians that validity cannot exceed reliability. So if only half the time they agree on whether a death is a homicide, that means half the time, one of the two experts is wrong. That should be alarming.
A nice report on and analysis of not simply individual ME performance failures, but of the potential of "event reviews" that show a system keeping secrets from itself. A reminder that these determinations are significant "upstream" from the courtroom moments that draw our attention.
Dear Radley, excellent article. The Chico, CA, police shot, choked, K9-bitten, and tased my 34-year-old son who was in a confined bathroom. The Butte County coroner testified at the trial and lied. The officers lied. The judge steered the jury to make a quick ruling. The fix was in. Lexipol, Axon, Force Science Institute, cronyism, the public's blind support of rogue cops, and money all play into the unnecessary deaths of civilians. I had a panic attack reading the article. I was shouting out 'yes, yes, yes,' but who is listening? Feckless politicians and prosecutors are way too cozy. Killings in rural areas are more corrupt than in urban areas. What can be done?