Care worker safety, lone working and incident management

Start with the risk – not the injury count

Fact: HSE defines work-related violence broadly enough to include abuse, threats and assault connected with work. Its health and social care guidance specifically identifies high under-reporting and says workers may accept violence or aggression as part of the job. HSE also warns that apparently minor events such as hair pulling, pinching or verbal abuse can form an escalating pattern.

Analysis: a provider that records only physical injury is measuring the end of the pathway. Threats, blocked exits, sexual or discriminatory conduct, repeated intimidation, aggressive calls, unsafe visitors, near misses and failed attempts to get help can all reveal a control weakness before somebody is hurt. Low reporting is therefore not a safety outcome on its own; it needs to be tested against staff confidence, reporting accessibility and evidence that managers respond.

Interpretation for care: behaviour may arise in distressing or clinically complex circumstances. A proportionate response should consider the person’s communication, health, pain, environment and care needs without assuming that a diagnosis or service-user group predicts violence. Understanding context and protecting workers are compatible duties, not competing values.

Lone working is a condition of work, not a job title

Homecare makes the exposure easy to see: workers travel alone, enter private homes and may not know who else will be present. Isolation also occurs in supported living, on a night shift, during a medication round, on a community escort, in maintenance work or when a manager responds out of hours. A worker can be inside a staffed building yet functionally alone if colleagues cannot hear, locate or reach them.

HSE says lone working does not automatically mean a higher likelihood of violence, but the absence of nearby support makes a worker more vulnerable. Employers must manage the risk, train, supervise and monitor lone workers, keep in touch and respond to incidents. A useful assessment therefore asks where isolation occurs, how circumstances change, how quickly help can arrive, and what happens when a phone is unreachable, the worker cannot speak or the network fails.

Build one chain: before, during and after

Before an incident, general risk assessment should connect to person- and task-specific planning. Relevant controls can include triggers and early signs, prior incidents, household or visitor risks, staffing competence, safe exit, visit timing, check-in rules, communication coverage and clear authority to pause, withdraw or request a second worker. The control has little value if an agency worker cannot see it or if a material change does not reach the next visit or shift.

During an incident, the product is only the entry point. The provider should know who receives an alert, what location and worker information they see, what response time is expected, who can call emergency services, how a missed check-in is escalated and what fallback works without data coverage. Staff also need permission to act early: fear of criticism for a false alarm can turn a fast control into a delayed one.

After an incident, immediate needs come first – medical attention, safety, contact with police where appropriate, a confidential conversation and psychological support. HSE then expects incidents to be recorded and reviewed for further controls. Work-related violence is RIDDOR-reportable where it causes death, a specified injury or a physical injury leaving a worker unable to perform normal duties for more than seven days; over-three-day injuries must still be recorded. Care planning, staffing, visit arrangements, safeguarding and technology performance may all require review even where RIDDOR does not apply.

A device is a component. The control is the tested chain from activation to effective help.

Make the incident system the organisation’s memory

A useful record captures conditions as well as narrative: whether the worker was alone, prior warning signs, planned versus actual staffing, worker experience, alarm availability and use, response time, immediate support, corrective action and completion. HSE recommends a quick, accessible route – including paper, online or mobile options – because complex reporting suppresses the evidence leaders need.

The management task is aggregation. One event may be hard to predict; five events linked to a late visit, one task, one address, a failed device or an unfinished action form a pattern. Boards should see meaningful indicators such as reporting confidence, repeat exposure, response-time exceptions, overdue actions and verified control tests – not just a total incident count. Closing the form is not the same as closing the risk.

A 30 October legal change belongs in the same operating system

Since 26 October 2024, the Equality Act framework has required employers to take reasonable steps to prevent sexual harassment of employees. The Employment Rights Act 2025 strengthens that wording to all reasonable steps and adds employer liability for harassment by third parties where the employer failed to take all reasonable steps to prevent it. The government’s timetable, updated on 25 August 2026, schedules those measures for 30 October 2026 and says future dates remain subject to parliamentary processes.

For care providers, a third party can include a person receiving care, relative, visitor or member of the public. This does not turn every difficult interaction or act of aggression into Equality Act harassment. It does mean that sexual and protected-characteristic harassment cannot sit in an HR silo detached from visit planning, reporting, repeat-pattern analysis and workforce support. Providers should take specific legal advice on scope and reasonable steps, while making sure the operational evidence is joined up.

Buy a response capability, not a feature list

Monitored apps, wearables, duress alarms, man-down detection, timed check-ins, location sharing and mobile incident reporting can strengthen protection. Selection should begin with operating conditions: indoor and rural coverage, discreet activation, location accuracy, battery and device ownership, 24/7 monitoring, temporary-worker access, accessibility, data protection, testing frequency, response service levels and a documented failure mode.

The neutral procurement test is outcome-based: can workers obtain effective help in the places and at the times they actually work, and can the provider prove the control was tested? Suppliers can support design, monitoring, reporting and implementation, but they cannot own the employer’s risk assessment, culture or learning. Contract governance should connect alert logs to incident review without creating a second, isolated dataset.