A new study has revealed alarming findings about the misclassification of homicides committed by police officers in Maryland, according to a report by Radley Balko for The Watch.
The audit was ordered after former Minneapolis police officer Derek Chauvin was tried for the murder of George Floyd—an audit which found that 36 out of 87 in-custody deaths involving restraint should have been classified as homicides but were instead labeled as accidental, natural or undetermined.
“The results are disturbing,” Balko writes, highlighting how such misclassifications often shut down further investigations into police culpability.
The study employed a rigorous methodology to minimize bias, including sequential unmasking to shield reviewers from irrelevant details like the race of the deceased or their criminal history.
As Balko notes, the auditors found no racial disparity in their own determinations, unlike the Maryland Office of Chief Medical Examiner (OCME), which was less likely to classify deaths as homicides when the victim was Black.
“The audit used a process called sequential unmasking to expose the reviewers to only the evidence they needed to draw their conclusions,” The Watch explains, underscoring the importance of unbiased forensic analysis.
William Thompson, a co-author of the study and professor emeritus at UC Irvine, told Balko that the audit was a direct response to Dr. David Fowler’s testimony in the Chauvin case, where Fowler argued George Floyd’s death should have been labeled “undetermined.”
“A homicide finding doesn’t necessarily mean someone is criminally culpable, but it flags the case for prosecutors to investigate,” Thompson said.
The study also revealed how medical examiners’ reliance on dubious concepts like “excited delirium” or “sickle-cell trait” often excused police accountability.
The audit’s findings suggest systemic issues in how deaths in police custody are investigated nationwide.
Thompson noted that if Maryland’s results were extrapolated, “you’re looking at thousands of preventable deaths.”
Balko adds, “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.”
The study’s design, including blinding reviewers to biasing information, could serve as a model for other states to address similar biases.
Balko’s article also critiques the resistance within the medical examiner community to reforms aimed at reducing bias.
For instance, Andrew Baker, the Hennepin County medical examiner who autopsied George Floyd, initially opposed masking procedures but later adopted them in high-profile cases.
“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,” Thompson said.
“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.”
Despite this progress, Balko notes that the variability in manner-of-death determinations raises questions about their reliability as scientific evidence.
The study’s implications extend beyond Maryland, demonstrating a national crisis in how police-involved deaths are investigated.
Balko concludes by emphasizing that if “experts can’t agree on basic conclusions, should we really be telling jurors that this is expert testimony?”
