Near misses are data. Reporting systems capture only part of it

Incident reporting systems collect information about safety events, near misses and unsafe conditions. They support investigation, learning and accountability. In many organisations they are the established record for an event that requires review.

AHRQ describes voluntary event reporting as passive surveillance. It depends on somebody recognising an event, deciding that it meets a threshold, finding the route, completing the report and accepting whatever personal consequences they believe may follow. ¹

That chain shapes the resulting data. Reports can identify hazards and cases worth investigating. They do not provide a complete count of events, and they cannot estimate the epidemiology of safety problems without a reliable denominator.

A clear near miss is easy to describe after review: an unsafe event did not harm a patient because it was caught, did not reach them or caused no harm. The frontline moment can feel less certain. Someone sees an unusual workaround, a rushed handover or a medication process that almost failed. They may not know whether the observation qualifies.

Formal systems need categories and thresholds. Staff still make a personal calculation before opening the form. Was this serious enough? Am I certain? Will I be blamed? Has someone else reported it? Will anything happen?

AHRQ lists fear of repercussions, blame culture, poor integration and lack of feedback among the barriers to reporting. It recommends confidentiality, broad participation, timely review and communication of action plans. ^{1, 2}

Speaking up is often described as an individual act of courage. The literature points back to the conditions around the person. Kane and colleagues reviewed 294 articles and found inconsistent definitions of speaking up, then proposed a definition centred on raising a safety concern to someone with the power to address it. ³

O’Donovan and McAuliffe reviewed interventions intended to improve psychological safety, speaking up and voice behaviour in healthcare. They found only 14 interventions and substantial variation in design and evidence. Education appeared often, but education alone cannot guarantee that a concern receives a fair response. ⁴

A low-friction tool can reduce the practical cost of sharing. It cannot repair a punitive culture, create trust in leadership or protect somebody from consequences outside the system’s governance. Those responsibilities remain with the organisation.

High-reliability practice asks organisations to remain sensitive to operations and preoccupied with failure, including weak signs that normal work is becoming less safe. Serou and colleagues reviewed learning tools used in high-reliability organisations that could be adapted for healthcare. The review supports disciplined learning from incidents while showing that transfer between industries requires care. ⁵

Weak signals should not become a stream of unfiltered alerts. Leaders already work inside crowded information environments. The useful work is aggregation: finding repeated observations about the same handover, piece of equipment, process or time of day.

The resulting theme remains a prompt. Reviewers can inspect the underlying observations, compare them with formal events and operational measures, and ask the people doing the work what conditions they see.

A system that accepts concerns and produces silence trains people to stop using it. AHRQ identifies lack of feedback as a common barrier to reporting. Its guidance on event investigation calls for action plans to be communicated to reporters and other stakeholders. ^{1, 2}

Closing the loop does not require disclosing confidential investigations. It can be specific without naming people: the equipment check changed, the handover prompt was revised, the staffing concern went to the responsible executive, or the team reviewed the pattern and found no action was warranted.

The response also needs a clock. A same-shift safety issue and a gradually rising operational theme belong to different workflows. Governance should define who reviews each kind of signal and how quickly.

Pattern systems can create their own blind spot. A rare but severe observation may never form a cluster. A large theme may consist of low-consequence friction. Review design needs separate paths for urgent content and recurring operational signal.

Automated grouping should remain inspectable. Reviewers need access to the observations behind a label, the ability to correct grouping and a record of when the theme changed. A model can reduce reading effort. It should not become the final authority on safety relevance.

Alert thresholds also deserve restraint. If every new observation generates an executive notification, leaders will recreate the noise that aggregation was meant to solve. Review cadence should reflect risk, recurrence and the organisation’s ability to respond.

Patients and families notice identification checks, unexplained changes, communication gaps and conditions that feel unsafe. They can also recognise good catches and recovery. WHO’s patient-safety strategy places meaningful patient and family engagement among the work required for safer systems. ⁶

Their accounts should not be treated as clinical adjudication. They provide a perspective on what happened and how the system appeared from the receiving side. Reviewers can compare that account with records, formal reports and staff context.

The practical value grows when the organisation accepts different levels of certainty. A person should be able to say “something did not feel right” without selecting an event category. Human review can decide what the observation warrants.

A useful review begins with severity, not popularity. Did any observation suggest immediate harm, safeguarding risk or a reportable event? Those items leave the pattern workflow through the agreed restricted route. The group then reviews recurrence: which themes are new, rising, concentrated in one setting or appearing across several services?

The group should read a small sample of the underlying observations before discussing a label. It can ask what else changed during the same period, whether capture was promoted differently and which operational or formal data might challenge the pattern. Someone close to the work should describe local conditions without being asked to identify a contributor.

The meeting ends with decisions, not alerts. Each theme receives an owner, a next question, an action or a dated decision to watch. The record should state why the group chose that response. This cadence turns weak signal into organised inquiry while preserving the faster routes that serious events require.

PulseMD provides a short voice or text route for observations that may feel too small or uncertain for a formal report. It groups repeated signals into themes and trends so leaders can review patterns rather than respond to every fragment as an alert.

The product sits upstream of incident reporting. Emergencies, incidents, safeguarding matters and mandatory reports stay in their established systems. PulseMD does not classify an observation as a near miss, estimate incident rates or predict a safety event.

A pilot should document those boundaries before launch. Capture language, escalation instructions, review roles, retention and communication back to staff all need agreement. Lowering the effort to speak creates value only when the organisation is ready to listen.