Posts
The Custody Death Problem
Why Custody Deaths Concentrate the Failures
Every category of death the coroner system handles involves some risk of error, political influence, or resource inadequacy. Custody deaths concentrate all of those risks simultaneously — and add a fourth: direct structural conflict of interest between the death investigation function and the law enforcement function it is supposed to evaluate independently.
The stakes compound the concentration. A custody death misclassified as accidental or natural does not merely produce an inaccurate certificate. It ends civil litigation before it begins. It prevents criminal accountability for officers whose conduct contributed to the death. It tells a family that their person died of their own physiology, not of what was done to them. It enters the national data record as a non-homicide and disappears from the custody death count. Each downstream consequence is irreversible once the certificate is signed and the investigation is closed.
Four Compounding Risk Factors — Unique to Custody Deaths
Structural conflict of interest: In sheriff-coroner counties, the same elected official controls both the detention facility and the death investigation. In coroner-dominant counties, the elected coroner and the elected sheriff are both county row officers who share a political ecosystem, a voter base, and often a party affiliation. The independence that death investigation requires from the agency whose conduct is under examination is structurally absent in both arrangements.
Reliance on law enforcement accounts: Custody death investigations depend heavily on officer statements, facility records, and video evidence controlled by the agency under examination. A forensic pathologist conducting an autopsy without independent scene access, without unfiltered witness interviews, and without adversarial review of official accounts is working from a curated record. The curation is performed by the party with the highest interest in the outcome.
Classification tools that absorb the conflict: "Excited delirium," "positional asphyxia," and related classifications allow cause-of-death determinations that attribute death to the decedent's physiology — drug intoxication, agitation, cardiac arrhythmia — rather than to the restraint techniques applied. These classifications are not fabricated from nothing. They reflect genuine physiological processes. But they are applied disproportionately in custody deaths, and their application terminates accountability inquiries that might otherwise proceed.
Irreversibility under the architecture: Unlike prosecutorial charging decisions, which can be revisited, or judicial verdicts, which can be appealed, a manner-of-death classification has no formal review mechanism in most jurisdictions. Once "accident" or "natural causes" is signed onto the certificate, it requires extraordinary intervention — an external audit, a court order, or a high-profile public case — to revisit. The architecture provides no routine path for correction.
Thirty-Six Reclassifications — What the Numbers Mean
Maryland's Office of the Chief Medical Examiner is an appointed, physician-led office — nominally among the stronger models in the American system. David Fowler served as its chief from 2002 to 2019. He was not an elected coroner with minimal qualifications. He was an appointed forensic physician who headed one of the country's more professional death investigation offices.
In May 2021, Fowler testified as a defense expert in the murder trial of former Minneapolis police officer Derek Chauvin, charged in the death of George Floyd. Fowler offered alternative explanations for Floyd's death, suggesting that carbon monoxide exposure from vehicle exhaust, heart disease, and drug intoxication — rather than neck compression — were contributing or primary causes. The Chauvin jury rejected the defense theory. Chauvin was convicted.
Fowler's testimony triggered a response that his own office's procedures had not. More than four hundred physicians signed a letter calling for a review of Fowler's Maryland cases. The Maryland Attorney General commissioned an independent audit.
Scope: 87 in-custody death cases from Fowler's tenure (2002–2019) were reviewed by an independent panel of forensic pathologists.
Reclassifications: At least 36 cases — approximately 41% of the reviewed cases — were found to have been incorrectly classified. The independent reviewers determined these deaths should have been ruled homicide rather than accident, natural causes, or undetermined.
Pattern finding: Reviewers identified patterns consistent with potential pro-law-enforcement bias in the original classifications. Cases involving Black decedents showed higher rates of reclassification, suggesting potential racial bias in the original determinations.
Anton Black: One of the most prominent reclassified cases involved Anton Black, a 19-year-old Black man who died in September 2018 in Greensboro, Maryland, following a police restraint. Fowler's office had originally ruled his death accidental, citing excited delirium and a pre-existing heart condition. The independent audit found the death should have been classified as homicide. No criminal charges had been filed against the officers involved. The reclassification reopened questions about accountability that the original classification had foreclosed.
Implications: The 36 reclassifications are not administrative corrections. Each represents a death that was officially recorded as accidental or natural — foreclosing civil litigation, preventing criminal accountability, and telling a family an official version of events that the independent review found to be incorrect. For cases within the statute of limitations, the reclassifications create new legal possibilities. For those outside it, the only consequence is the correction of the historical record.
The Maryland audit is the most extensive formal review of a single ME office's custody death record in American history. What it revealed in a relatively strong, appointed system — 36 incorrect classifications out of 87 reviewed cases, representing a 41% error rate in the most politically sensitive death category — should be read alongside the architecture's distribution described in Post II. Maryland had a professional, appointed office. Most American deaths in custody are investigated by something with weaker structural insulation. The Maryland number is a floor, not a ceiling.
Thirty-six deaths. Officially accidental. Officially natural. Now officially homicides — after the person who signed the original certificates retired, after a high-profile trial triggered external review, after an attorney general commissioned an audit that the system itself would never have produced. The architecture has no internal mechanism that generates what the audit found.
The Classification That Isn't a Diagnosis
"Excited delirium" is not recognized as a diagnosis by the American Medical Association, the American Psychiatric Association, or the World Health Organization. It does not appear in the Diagnostic and Statistical Manual of Mental Disorders. It is not a clinical entity that physicians diagnose in living patients.
It is, however, listed as a cause of death on thousands of American death certificates — almost exclusively in cases where the deceased was in law enforcement custody and physical restraint was applied in the period before death.
Origin: The term entered forensic medicine through a 1985 paper and was promoted by a small number of forensic pathologists with documented ties to TASER International (now Axon), the manufacturer of conducted energy weapons. TASER funded research, sponsored conferences, and cultivated relationships with the medical examiners and coroners whose testimony would determine whether TASER devices contributed to custody deaths. The classification served a specific institutional interest before it served a scientific one.
Application: When a person in police restraint dies, excited delirium as a cause of death attributes the death to an acute physiological crisis in the decedent — typically combining agitation, elevated temperature, apparent superhuman strength, and sudden cardiac arrest — rather than to the physical restraint applied by officers. It makes the death a consequence of the decedent's condition, not of what was done to them.
Scientific status: Multiple medical organizations have formally rejected excited delirium as a valid clinical or forensic entity. The American Medical Association adopted a policy in 2021 stating that "excited delirium" should not be used as a cause of death on death certificates. The National Association of Medical Examiners updated guidance to discourage its use. Some jurisdictions have formally prohibited its use in death classification.
Persistence: Despite formal rejection by major medical organizations, the classification continues to appear on custody death certificates in jurisdictions — particularly those with elected coroners or less-resourced ME offices — where the pressure to use it is high and the institutional capacity to resist that pressure is limited. Its persistence is a function of the architecture, not of its scientific merit.
Structural function: In FSA terms, excited delirium functions as a conversion mechanism — a classification tool that transforms a potentially unlawful death into a medical event, terminating accountability processes that a homicide ruling would initiate. It is insulation built into the vocabulary of forensic medicine.
How Many People Die in Custody — Nobody Knows
The United States does not have a reliable count of how many people die in law enforcement custody each year. This is not a statement about data quality at the margins. It is a statement about the fundamental architecture of custody death reporting — which is voluntary, fragmented, and designed in ways that structurally undercount the category it is supposed to measure.
The Bureau of Justice Statistics operates the Deaths in Custody Reporting Program — the primary federal mechanism for tracking custody deaths. Participation by state and local agencies is required under the Death in Custody Reporting Act of 2013. In practice, reporting is incomplete, delayed, and inconsistent. The BJS has documented that a substantial percentage of reportable deaths go unreported in any given year. The agency has been unable to certify the accuracy of its own counts.
Classification dependency: The Deaths in Custody Reporting Program counts deaths that are reported as custody deaths. If the coroner or ME classifies a custody death as natural causes or accident — removing the investigative flag that would prompt reporting — the death may not enter the custody death count at all. The undercount in the death certificate feeds the undercount in the national database.
Voluntary compliance failures: A 2016 BJS report found that approximately 1,000 deaths that should have been reported under the Death in Custody Reporting Act were not reported by participating agencies. The Department of Justice has limited enforcement mechanisms. The reporting gap is structural, not incidental.
Definitional inconsistency: "Custody" is not defined uniformly across reporting jurisdictions. Deaths that occur during a police encounter before formal arrest, deaths in transport, deaths shortly after release — their inclusion or exclusion varies by agency, by state, and by the judgment of the individual officer completing the report. The denominator of custody deaths is itself contested.
The Guardian and Washington Post counts: Beginning in 2015, independent journalism projects — The Guardian's "The Counted" and the Washington Post's ongoing police shooting database — began tracking law enforcement deaths through crowdsourcing, local news monitoring, and public records. Both consistently found totals substantially higher than official government figures. The gap between journalistic and official counts is itself a measure of the data architecture's failure.
Post-reclassification consequences: The Maryland audit reclassified 36 deaths from 2002–2019. Those deaths, originally recorded as accidental or natural, were presumably not counted in custody death totals during those years. The reclassification corrects the individual certificates. It does not retroactively correct the national count. The historical record remains distorted.
When the Jailer Signs the Death Certificate
The sheriff-coroner structure documented in Post II reaches its most acute expression in custody death cases. In the 49 California counties where the elected sheriff serves simultaneously as coroner, the official who oversees county jails and supervises the deputies whose conduct may have contributed to a death in custody also certifies the official cause and manner of that death.
The San Joaquin County audit of 2016 documented that forensic pathologists working under Sheriff-Coroner Steve Moore reported direct pressure to modify autopsy findings in law enforcement-involved deaths. Pathologists described being asked to change manner-of-death classifications — specifically, to move deaths from homicide to accident or undetermined — in cases where law enforcement conduct was at issue. The audit confirmed multiple instances.
San Joaquin County is one county in one state. The structural conditions it represents — elected law enforcement official holding death investigation authority — exist in 49 California counties and in varying forms across multiple other states where sheriff-coroner arrangements persist. The audit happened because the San Joaquin case attracted scrutiny. Most sheriff-coroner counties have never been audited.
The architecture does not require that pressure be applied in every case. It only requires that the structural conditions for pressure exist — and that no mechanism for detecting its application operates routinely. Both conditions are met in every sheriff-coroner county in America.
The Downstream Consequences of Getting It Wrong
A manner-of-death classification is not only a medical finding. It is the predicate for a series of legal, financial, and institutional processes that flow from it — and that become effectively irreversible once the classification is accepted. Understanding what reclassification means in the custody context requires understanding what the original wrong classification foreclosed.
Criminal accountability: A homicide classification by the ME or coroner is not a criminal conviction — it is a forensic finding that the death resulted from the act of another person. It is, however, the necessary predicate for a prosecutor to evaluate whether criminal charges are warranted. An accident or natural causes classification terminates that evaluation before it begins. Officers whose conduct contributed to a death classified as accidental face no prosecutorial review. The classification is the gatekeeper.
Civil liability: Section 1983 civil rights litigation — the primary mechanism for family members to seek accountability for custody deaths — requires establishing that a constitutional violation caused the death. A natural causes or accidental classification complicates that showing significantly. Defense attorneys cite the official classification as evidence that the death was not caused by officer conduct. Plaintiffs must overcome the official record to proceed. Many cannot. Cases that would survive summary judgment with a homicide classification are dismissed with an accidental one.
Administrative accountability: Police departments conduct internal affairs investigations into officer conduct in custody deaths. The scope and seriousness of those investigations are shaped by the official manner-of-death determination. Accidental or natural causes classifications signal to internal investigators that the death does not require the level of scrutiny a homicide classification would demand. Officers whose conduct contributed to a death classified as accidental may receive no disciplinary review at all.
Family truth: For the families of people who die in custody, the death certificate is the official account of what happened to their person. An accidental or natural causes classification tells them that their family member died of their own physiology — that the officers present bore no responsibility. That account shapes grief, memory, and the family's understanding of what justice is possible. When the classification is wrong, the family has been told an official lie — and the architecture provides no routine path for correcting it.
Statute of limitations: Civil rights claims have statutes of limitations — typically two to three years from the date of death or discovery. When a wrong classification is discovered years later — as the Maryland reclassifications were — many cases are outside the limitations period. The reclassification corrects the record. It does not revive the foreclosed legal claims.
The Custody Death Layer — What the Evidence Establishes
Custody deaths are the category where the Coroner Architecture's structural vulnerabilities produce their most consequential and least correctable output. The conflict of interest is structural and documented. The classification tools that absorb it — excited delirium foremost among them — have been formally rejected by major medical organizations and continue to appear on death certificates. The national data record is demonstrably incomplete. The downstream consequences of wrong classification are irreversible in most cases.
The Maryland audit is the only large-scale independent review of custody death classifications that has been conducted in the United States. Its finding — a 41% error rate in a nominally stronger, appointed system — is the single most important data point in the series. It is a floor. The accuracy of custody death classifications in coroner-dominant jurisdictions, under sheriff-coroner arrangements, and in under-resourced offices without institutional quality review has never been subjected to comparable scrutiny.
The architecture does not produce this outcome in every county or every case. It produces it structurally — in predictable directions, under predictable conditions, at rates that a single audit of a single office found to be 41%. Post VI documents what the public health record looks like when custody deaths are embedded in a broader pattern of systematic misclassification across cause-of-death categories.
| Finding | Basis | Status |
|---|---|---|
| 49 of 58 California counties: elected sheriff simultaneously serves as coroner | California county charters and statutes | Documented |
| San Joaquin County 2016 audit: confirmed pressure on pathologists to alter law enforcement death findings | San Joaquin County audit report, 2016 | Documented |
| Maryland OCME audit 2025: at least 36 custody deaths reclassified to homicide from accident/natural/undetermined | Maryland Attorney General commissioned audit, 2025 | Documented |
| Anton Black death reclassified to homicide; originally ruled accidental under David Fowler | Maryland OCME audit findings; court and press record | Documented |
| "Excited delirium" not recognized by AMA, APA, or WHO as valid clinical diagnosis | AMA policy statement 2021; DSM-5; WHO ICD classifications | Documented |
| Deaths in Custody Reporting Program: documented underreporting; BJS unable to certify count accuracy | BJS reports; GAO review of DCRP; DOJ Inspector General findings | Documented |
| Wrong classification forecloses criminal accountability, civil litigation, and administrative review | Legal and procedural analysis; case record | Documented |
| Rate of custody death misclassification in coroner-dominant and sheriff-coroner jurisdictions | No comparable audit has been conducted — the gap itself is documented | Unaudited · Gap Documented |
No comments:
Post a Comment