When Someone Dies in Custody, the Evidence Clock Starts Immediately

Slug: death-in-custody-evidence-independent-investigation
Tags: Crime, Public Safety
Meta description: When someone dies in custody, the evidence clock starts immediately. Learn what an independent investigation must preserve, test and explain.

When a person dies in state custody, the first public question is often, “What caused the death?” A credible investigation must ask something wider: what happened before, during and after the fatal event; what the authorities knew; what they did; what evidence exists; and whether systems designed to protect life worked as intended.

Custody creates a special relationship of power and responsibility. The person cannot simply leave, choose another provider or independently control the conditions around them. That does not mean every death in custody proves wrongdoing. It means the state must support its account with evidence capable of independent testing.

This article is a general public-interest explainer, not an analysis of any named case. It does not presume criminal liability, misconduct or innocence. Its purpose is to show why evidence preservation, institutional independence and family participation matter from the earliest hours.

The international benchmark: investigate potentially unlawful death

The United Nations’ Minnesota Protocol on the Investigation of Potentially Unlawful Death is a leading international guide to the practical implementation of the duty to protect life and investigate potentially unlawful deaths. Its importance lies in method: an investigation should be prompt, effective, thorough, independent, impartial and transparent enough to command confidence.

Evidence: international and UK mechanisms recognise a duty to investigate deaths in custody independently. Interpretation: the investigation must examine more than the final medical mechanism. A finding such as cardiac arrest, suffocation, intoxication or injury describes part of what happened; it may not explain why the danger arose, whether it was recognised or whether a preventable failure contributed.

The evidence clock starts immediately

Some evidence degrades naturally. Memories change, digital systems overwrite recordings, rooms are cleaned, staffing patterns shift and informal messages disappear. A serious response therefore begins with preservation, not a premature theory.

Secure the physical environment

The relevant location may include a cell, vehicle, corridor, medical room, reception area or place of arrest. Investigators need a reliable record of layout, ventilation, temperature, doors, restraints, bedding, medication, food, water and other conditions that could matter. Items must be identified, photographed, collected where necessary and tracked through a documented chain of custody.

Preserve digital records before they are overwritten

CCTV, body-worn video, vehicle cameras, access-control logs, radio traffic, telephone recordings and computer audit trails can establish sequence and timing. Investigators should identify every relevant system, its clock settings, retention period, gaps and authorised users. A missing recording is not automatically proof of concealment, but its absence must be explained rather than ignored.

Separate witnesses before accounts converge

Staff, detainees, medical personnel and other witnesses may each hold a fragment of the timeline. Early independent accounts reduce the risk that later discussion—innocent or otherwise—reshapes memory into one shared narrative. Interviewers should distinguish what a person directly observed from what they heard afterwards.

Build one timeline from many clocks

A custody death is rarely understood from a single document. The investigation must reconcile:

arrest, arrival and booking times;
risk assessments and observation checks;
requests for help and responses;
meals, medication and medical examinations;
use of force, restraint or movement;
cell-door and access records;
camera footage and radio communications;
the first sign of distress, emergency call and clinical intervention;
transfer, hospital treatment and formal confirmation of death.

Different systems may use clocks that are seconds or minutes apart. That matters when an institution claims checks occurred at a particular interval or help arrived promptly. Investigators should state how timestamps were synchronised and where uncertainty remains.

Medical cause is not the same as institutional explanation

A post-mortem examination can identify injuries, disease, toxicology and the physiological cause of death. It should be informed by the scene and custody history, not treated as an isolated laboratory exercise. The questions may include whether restraint affected breathing, whether symptoms were visible, whether prescribed medication was available, whether intoxication or withdrawal required monitoring, and whether delays changed the outcome.

Investigators should also test the counterfactual carefully: if the relevant policy had been followed, is there credible evidence that the risk could have been reduced? This is not the same as claiming certainty about survival. It separates unavoidable tragedy from preventable organisational failure.

Independence must be structural, not ceremonial

The same organisation whose actions are under scrutiny should not control the decisive evidence, witnesses and conclusions without external oversight. In England and Wales, different mechanisms apply depending on the custodial setting. The Independent Office for Police Conduct publishes annual statistics and conducts independent investigations into deaths during or following police contact. The Prisons and Probation Ombudsman investigates deaths in prisons and several other forms of detention and supervision.

The PPO’s terms of reference connect fatal-incident investigations to the Article 2 requirement for independent investigation of deaths in custody. Its public explanation says investigators examine relevant records and policies and may interview staff, prisoners or residents. These investigations are distinct from a coroner’s work and from any criminal investigation, although the processes can interact.

An inquest answers defined questions; it is not a criminal trial

In England and Wales, a coroner investigates deaths that may be violent or unnatural, have an unknown cause or occur in prison or state detention. Official guidance explains that an inquest seeks to establish who died and how, when and where the person came by their death. It does not determine criminal or civil liability and does not function to convict a person.

This distinction protects accuracy. A finding that systems failed is not automatically a finding that a named individual committed a crime. Conversely, the absence of a criminal charge does not prove that care, policy, supervision or institutional design was adequate. Accountability has several layers: criminal, disciplinary, civil, regulatory, managerial and preventative.

The family is not an audience member

Bereaved relatives may hold important information about health, medication, communication and the person’s condition before detention. They also need timely information about investigative steps, access to representation and a meaningful opportunity to raise questions.

Family participation does not mean investigators must adopt the family’s theory. It means the process should not treat those closest to the deceased as passive recipients of a finished institutional account. The PPO describes family liaison as part of its death-investigation process, and official coroner guidance is designed to help bereaved people understand what to expect.

Seven questions that expose an incomplete investigation

Preservation: Which physical and digital evidence was secured, when and by whom?
Timeline: Have all clocks, logs and recordings been reconciled?
Health: What risks were known, assessed, communicated and monitored?
Force and environment: Did restraint, position, temperature, ventilation or cell conditions contribute?
Independence: Who controls the evidence and who has authority to challenge the institution?
Family participation: Were relatives informed, heard and able to raise relevant concerns?
Learning: Which concrete action follows, who owns it and how will implementation be checked?

Statistics require context, not dismissal

The IOPC’s latest annual page covers deaths during or following police contact between 1 April 2025 and 31 March 2026. It reports that the IOPC began 103 independent investigations into 105 deaths across the categories in its report. Those figures include more than deaths inside police cells: the published categories also cover road incidents, shootings, apparent suicides following custody and other deaths following police contact.

Therefore, a headline total must not be presented as though every death occurred in the same circumstances. Categories, definitions and case-level evidence matter. At the same time, classification should not be used to make institutional responsibility disappear. Every individual case still requires the appropriate investigation.

The goal is truth, accountability and prevention

A trustworthy investigation is not built around the question, “How can this institution defend itself?” Nor should it begin with an unsupported accusation. It begins by protecting evidence, testing competing explanations and identifying both individual actions and system conditions.

The public interest is larger than blame. If a death reveals a dangerous restraint method, failed observation routine, broken ventilation system, inaccessible medical service or culture of inaccurate record-keeping, the finding must lead to verifiable change. Otherwise, an investigation may explain the past while leaving the next person exposed to the same risk.

Featured image disclosure: The featured visual is an original conceptual AI illustration. It does not depict a real deceased person, investigator, institution or case.