Adverse Event

How an adverse event is identified and evaluated

An adverse event describes harm or an unfavourable medical occurrence associated with healthcare, treatment or participation in a study. This page explains how adverse events are defined in different contexts, how severity and causality are assessed, and how adverse-event evidence is used to improve safety and evaluate the consequences of healthcare.

The term does not have one universal boundary. Patient-safety investigations often use adverse event to mean harm caused by medical management rather than the underlying disease, while clinical-trial and regulatory systems may record any unfavourable medical occurrence after treatment without requiring a causal relationship.

The meaning depends on the context

The applicable definition should be stated before adverse events are counted or compared. Mixing definitions can create misleading rates and conclusions.

In a patient-safety context, an adverse event generally involves injury or harm resulting from healthcare management. This may include acts of commission, failures to act, complications of treatment or failures within the care system.

In a clinical-trial or pharmacovigilance context, an adverse event may include any unfavourable medical occurrence after a medicine, device or intervention is used. The event does not need to have been caused by the intervention to be recorded as an adverse event.

The difference matters because a broad surveillance definition supports detection of possible safety signals, while a patient-safety definition focuses more directly on harm associated with the delivery of care.

The pathway from occurrence to classification

An unfavourable outcome must be described and assessed before it can support a safety conclusion. A complete assessment separates what happened from judgements about seriousness, severity, causality, expectedness and preventability.

A typical pathway is:

1. Detect the occurrence through patient report, clinical observation, record review, monitoring data or another reporting source.
2. Describe the event using the signs, symptoms, diagnosis, timing and outcome.
3. Assess severity according to the intensity of the event or its effect on functioning.
4. Assess seriousness using the applicable regulatory or governance criteria.
5. Evaluate causality to determine whether the healthcare intervention may have contributed.
6. Assess expectedness by comparing the event with existing safety information.
7. Assess preventability when the purpose includes patient-safety improvement.
8. Record actions and outcomes including treatment, withdrawal, recovery and follow-up.
9. Report or escalate the event according to the applicable timeframe and governance process.

These steps may occur iteratively as new clinical information becomes available. Initial classification should be updated when the diagnosis, outcome or causal evidence changes.

Severity and seriousness are different

Severity describes the intensity of an event, such as mild, moderate or severe symptoms. Seriousness is a regulatory or governance classification based on the event's consequences or required intervention.

Depending on the applicable framework, a serious adverse event may involve:

• Death.
• A life-threatening occurrence.
• Inpatient hospitalisation or prolonged hospitalisation.
• Persistent or significant disability or incapacity.
• A congenital anomaly or birth defect.
• Another medically important event that requires intervention to prevent a serious outcome.

A severe headache may be intense without meeting a criterion for seriousness. Conversely, a clinically moderate event may be serious if it results in hospitalisation. The two terms should not be used interchangeably.

Assessing causality

Causality assessment asks whether the intervention, device, medicine or care process contributed to the event. Temporal association alone does not establish causation because illness, comorbidities, other treatments and chance can produce the same outcome.

Assessment may consider:

• Whether the event occurred after the relevant exposure.
• Whether the timing is biologically and clinically plausible.
• Whether the event is already associated with the intervention.
• Whether another disease or treatment provides a stronger explanation.
• Whether the event improves after the intervention is stopped.
• Whether it reappears after re-exposure, when such evidence exists and re-exposure is ethically acceptable.
• Whether the relationship strengthens with greater exposure.
• Whether objective tests support the proposed mechanism.

Causality categories depend on the reporting system and may include unrelated, unlikely, possible, probable or definite relationships. These categories contain judgement and should not be treated as direct measurements of probability unless the method explicitly supports that interpretation.

Expected and unexpected events

Expectedness concerns whether the nature or severity of an event is consistent with existing information about the intervention. Relevant reference information may include an approved product label, investigator brochure, risk-management plan or study protocol.

An event can be expected but serious, or unexpected but unrelated. Expectedness should therefore be assessed separately from seriousness and causality.

Unexpected events may require rapid review because they can indicate a previously unrecognised risk. A single report may generate a signal for investigation without establishing that the intervention caused the event.

Preventability

Preventability asks whether reasonable action could have avoided the event or reduced its severity. It is particularly important in patient-safety work because it directs attention toward changes in systems, processes and practice.

Potentially preventable events may involve:

• Incorrect treatment selection or dosing.
• Failure to recognise a contraindication.
• Inadequate monitoring.
• Delayed diagnosis or treatment.
• Communication or handover failure.
• Equipment or system design problems.
• Failure to follow an appropriate procedure.
• Lack of timely access to required care.

Not every adverse event is preventable. Some harm can occur despite appropriate care, and preventability judgements may change as evidence and standards of practice develop.

Detecting adverse events

No single detection method identifies every adverse event. Different methods capture different event types and can produce different estimates of frequency.

Detection methods include:

• Voluntary incident reports.
• Patient and caregiver reports.
• Clinician reports.
• Prospective trial monitoring.
• Structured medical-record review.
• Trigger tools.
• Administrative and claims data.
• Registries and surveillance systems.
• Laboratory or device-monitoring data.
• Automated screening of electronic health records.

Voluntary reporting can identify important events and system failures, but it usually undercounts events and should not be used alone to estimate incidence. Record review may identify more events but depends on documentation quality and reviewer judgement.

Measuring adverse-event frequency

Adverse-event counts require a meaningful denominator and observation period. A count of events cannot be compared across populations without considering how many people were exposed, how long they were observed and how actively events were sought.

The proportion of exposed patients experiencing at least one event can be calculated as:

$$ Risk = \frac{Number\ of\ patients\ with\ at\ least\ one\ event}{Number\ of\ patients\ exposed} $$

An incidence rate that allows for different follow-up times can be expressed as:

$$ Incidence\ rate = \frac{Number\ of\ new\ events}{Total\ person\text{-}time\ at\ risk} $$

These measures answer different questions. Risk describes the proportion of patients affected during a defined period, while an incidence rate describes the number of events relative to accumulated time at risk.

Analyses should also state whether repeated events in the same patient are counted, how ongoing events are handled and whether the denominator includes all eligible patients or only those with complete follow-up.

Comparing adverse events between treatments

Comparisons should use consistent event definitions, ascertainment methods, follow-up periods and denominators. One treatment may appear to have more events because patients are followed longer or monitored more intensively.

Interpretation should consider:

• Baseline differences in patient risk.
• Exposure duration.
• Treatment switching and discontinuation.
• Competing risk of death.
• Differences in reporting and coding.
• Missing follow-up.
• Multiple event types and repeated events.
• Whether the study was large enough to detect rare harm.

Randomisation can improve comparisons of common events within a trial, but trials may be too small or short to identify rare or delayed harms. Observational and pharmacovigilance evidence may therefore provide essential complementary information.

Learning from adverse events

The purpose of patient-safety investigation is to understand how harm occurred and reduce future risk. Analysis should examine the interaction between people, tasks, technology, environment and organisational systems rather than stopping with the individual closest to the event.

An investigation may examine:

• The sequence of events.
• The patient's condition and care needs.
• Decisions and information available at each stage.
• Communication and handovers.
• Staffing, workload and supervision.
• Equipment, software and interface design.
• Policies and procedures.
• Environmental and organisational conditions.
• Barriers that prevented or limited harm.

Corrective actions should address the contributing mechanism. Reminding staff to be careful is unlikely to resolve a recurring event caused by confusing design, inadequate capacity or unreliable processes.

Using adverse-event evidence in health economics

Adverse events can affect both health outcomes and costs. Economic evaluations should include events that differ meaningfully between alternatives and could influence the decision.

Relevant consequences include:

• Diagnostic tests and additional treatment.
• Emergency visits and hospitalisation.
• Longer hospital stays.
• Rehabilitation and long-term care.
• Treatment discontinuation or switching.
• Reduced health-related quality of life.
• Disability and mortality.
• Patient and caregiver time.
• Lost productivity.
• Monitoring and prevention costs.

For event type (i), expected adverse-event cost may be represented as:

$$ Expected\ AE\ cost = \sum_{i=1}^{n} p_i \times C_i $$

where:

• (p_i) is the probability of adverse event (i).
• (C_i) is the cost associated with that event.
• (n) is the number of event types included.

Health effects can be represented through event-specific utility decrements, changes in health states or direct effects on survival. The model should avoid double-counting when adverse-event consequences are already reflected in health-state costs or utilities.

A simplified example

Suppose 1,000 patients receive Treatment A and 800 receive Treatment B during one year. Sixty patients receiving Treatment A and 32 receiving Treatment B experience at least one defined serious adverse event.

The one-year risks are:

$$ Risk_A = \frac{60}{1{,}000} = 0.06 = 6% $$

$$ Risk_B = \frac{32}{800} = 0.04 = 4% $$

The risk difference is:

$$ Risk\ difference = 6% - 4% = 2% $$

The observed risk is two percentage points higher with Treatment A. This comparison does not establish causality unless the study design, patient characteristics, follow-up and event ascertainment support that interpretation.

Adverse events and related concepts

Several safety terms overlap but should not be treated as exact synonyms.

• An adverse drug reaction is a harmful and unintended response for which a causal relationship with a medicine is at least reasonably possible under the applicable definition.
• A side effect is an effect related to the pharmacological properties of a treatment and may be harmful, neutral or beneficial depending on context.
• A medical error is a failure in the process of care and may or may not result in harm.
• A near miss is an incident that could have caused harm but did not reach the patient or did not produce harm.
• A complication is an unfavourable development in a disease or care pathway and is not automatically caused by error.
• A safety signal is information suggesting a new or changed association that warrants further investigation.

Using the correct term helps separate occurrence, causality, process failure and harm.

Common misunderstandings

An adverse event does not automatically prove negligence, error or causation. The meaning depends on the context and the evidence supporting each classification.

Common misunderstandings include:

• Every event occurring after treatment is not necessarily caused by treatment.
• Severe and serious do not mean the same thing.
• A known event can still be serious and clinically important.
• Absence of voluntary reports does not prove that no events occurred.
• More reports do not always mean that care is less safe because reporting culture and surveillance intensity differ.
• Preventable harm is not limited to an individual's mistake.
• Trial evidence alone may not identify rare or delayed events.
• Administrative codes do not automatically provide validated adverse-event diagnoses.
• A safety signal does not by itself establish a causal relationship.

Interpreting adverse-event evidence

Adverse-event evidence should be interpreted using the applicable definition, method of detection, population, exposure period and completeness of follow-up. Counts and rates should not be compared when the underlying surveillance systems or event criteria differ materially.

A useful safety assessment separates the occurrence from judgements about seriousness, severity, causality, expectedness and preventability. This structure supports appropriate reporting, fair comparison and system learning without claiming more certainty than the evidence provides.