Patient-Reported Outcome

How patient-reported outcomes capture the patient's perspective

A patient-reported outcome records information about health, functioning or wellbeing directly from the patient, without a clinician or another person interpreting the response. This page explains what patient-reported outcomes measure, how instruments are selected and administered, and how the resulting evidence can be used in clinical research, healthcare evaluation and health economics.

The defining feature is the source of the report rather than the topic being measured. A patient-reported outcome may describe a single symptom, several dimensions of functioning or a broad assessment of health-related quality of life.

What patient-reported outcomes can measure

Patient-reported outcomes are used when the patient's experience cannot be fully observed through laboratory tests, imaging, clinical examination or administrative records. They can capture effects that matter directly to patients and may reveal benefits or harms that are not visible in conventional clinical measures.

Patient-reported outcomes may measure:

• Symptoms such as pain, fatigue, nausea, breathlessness or anxiety.
• Physical, cognitive, emotional or social functioning.
• Ability to perform usual activities or fulfil important roles.
• Health-related quality of life.
• Treatment burden and adverse effects.
• Satisfaction with health status or treatment outcomes.
• Overall perceptions of health or change in health.

Patient experience measures and patient-reported outcomes are related but distinct. A patient-reported experience measure describes how care was delivered, while a patient-reported outcome describes the patient's health, functioning or wellbeing.

The difference between an outcome and an instrument

A patient-reported outcome is the concept being measured, such as pain severity or physical functioning. A patient-reported outcome measure is the instrument, questionnaire or scale used to collect and score the patient's report.

This distinction matters because one outcome may be measured by several instruments, and those instruments may not be interchangeable. They can differ in their questions, response options, recall periods, scoring rules, target populations and sensitivity to change.

An instrument may contain:

• Individual items or questions.
• Response options or rating scales.
• A specified recall period.
• Instructions for administration and completion.
• Rules for handling missing responses.
• A scoring algorithm.
• One or more domain or summary scores.
• Guidance for interpreting scores and changes.

Changing an instrument's wording, response scale, language, recall period or mode of administration can change what it measures. Modifications should therefore be justified and evaluated rather than treated as purely cosmetic.

Selecting an appropriate measure

Instrument selection should begin with the concept of interest and the intended use of the result. A widely used measure is not necessarily appropriate for every population, treatment or decision.

Selection should consider:

• Whether the instrument measures the outcome that matters for the research or decision question.
• Whether its content is relevant and understandable to the target population.
• Whether it has been evaluated in the language, culture and setting in which it will be used.
• Whether it can detect changes expected from the intervention.
• Whether the recall period matches the condition and timing of assessment.
• Whether the response burden is acceptable.
• Whether scoring and interpretation guidance are available.
• Whether licensing or use restrictions apply.
• Whether the instrument supports the intended clinical, regulatory or economic analysis.

The choice should be made before results are examined. Selecting an instrument or outcome after seeing which result is favourable increases the risk of selective reporting.

Generic and condition-specific measures

Generic patient-reported outcome measures can be used across diseases and populations. They support comparisons between conditions but may omit symptoms or functional changes that are particularly important within a specific disease.

Condition-specific measures focus on the effects of a particular disease, symptom or treatment. They may be more responsive to clinically important change but can be difficult to compare across conditions.

A study may use both when each serves a defined purpose. The measures should not be treated as duplicates merely because they both use patient reports.

Establishing content validity

Content validity concerns whether an instrument adequately represents the outcome it is intended to measure. It requires evidence that the items are relevant, comprehensive and understandable for the target population and context of use.

Patient involvement is central because researchers and clinicians may overlook effects that patients consider important. Qualitative interviews, focus groups and cognitive testing can help establish whether patients interpret the questions and response options as intended.

Content validity should be examined before relying on statistical performance. A highly reliable score is not useful if the instrument consistently measures the wrong or an incomplete concept.

Evaluating measurement properties

An instrument should have evidence supporting its use in the intended population and purpose. Measurement properties describe how well the instrument converts patient responses into scores that can support interpretation.

Important properties include:

• Reliability: Scores are sufficiently consistent when the patient's underlying status has not changed.
• Construct validity: Relationships with other measures and patient characteristics are consistent with prespecified expectations.
• Criterion validity: Scores agree with an appropriate reference standard when one exists.
• Responsiveness: The instrument can detect meaningful change over time.
• Measurement error: Random or systematic error is small enough for the intended interpretation.
• Cross-cultural validity: Translated or adapted versions measure the concept comparably across languages or cultures.
• Interpretability: The meaning of a score or change can be explained in clinically or personally relevant terms.

Floor and ceiling effects occur when many patients score at the lowest or highest possible level. These effects can prevent the instrument from distinguishing patients or detecting further deterioration or improvement.

Collecting patient-reported outcomes

Collection procedures can affect who responds and what is reported. Timing, setting, assistance, privacy and administration mode should therefore be planned and documented.

A collection process commonly includes:

1. Define the outcome and purpose before selecting the instrument.
2. Choose the validated version appropriate for the population, language and setting.
3. Schedule assessments at time points that can capture expected benefits, harms and recovery.
4. Use consistent administration procedures across participants and comparison groups.
5. Record who completed the measure and whether assistance was provided.
6. Monitor missing items and missed assessments while avoiding pressure that compromises voluntary reporting.
7. Apply the prespecified scoring rules for the correct instrument version.
8. Protect privacy and explain data use when responses are collected electronically or linked to other records.

Electronic, paper, telephone and interview administration may produce comparable results for some instruments, but equivalence should not be assumed. Accessibility needs should be addressed without silently altering the meaning of the measure.

Recall periods and timing

The recall period tells patients which period to consider when answering, such as today, the previous seven days or the previous month. It should match the variability of the outcome and the purpose of measurement.

Long recall periods may be affected by memory and by the most recent or severe experiences. Very short periods may miss intermittent symptoms or create excessive assessment burden. The schedule should also avoid relying on a single measurement when the outcome changes substantially over time.

Baseline assessment is important when the analysis concerns change. If baseline data are collected after treatment begins, responses may already reflect treatment effects or expectations.

Scoring and interpreting results

Patient-reported outcome scores may represent individual items, domains or a total scale. The direction and range of the score must be stated because a higher score can mean better health on one instrument and worse symptoms on another.

Interpretation should distinguish:

• A difference between groups.
• A change within a patient or group over time.
• Statistical significance.
• Measurement error.
• A difference that patients consider important.
• A threshold used for clinical classification or response.

A minimally important difference estimates the smallest difference or change considered important in a defined context. It may vary by population, baseline severity, direction of change and method of estimation, so it should not be treated as a universal property of the instrument.

Responder analysis classifies patients according to whether their change reaches a defined threshold. The threshold and handling of missing data should be prespecified because different rules can materially change the proportion classified as responding.

Missing patient-reported outcome data

Missing data can bias results when the likelihood of response is related to health, treatment burden, adverse events or dissatisfaction. Patients who are severely ill or who discontinue treatment may be less likely to complete follow-up assessments.

The analysis should report:

• The number expected to provide data at each time point.
• The number who provided complete and partial responses.
• Reasons for missing assessments when known.
• Differences between responders and non-responders.
• The assumptions used to handle missing data.
• Sensitivity analyses testing alternative assumptions.

Scoring rules that allow a total score from partially completed items address item-level missingness. They do not correct bias caused by patients missing an entire assessment or leaving the study.

Proxy and observer reports

A proxy-reported outcome is provided by another person on behalf of the patient, while an observer-reported outcome records observable signs or behaviours. Neither is a patient-reported outcome because the information does not come directly from the patient.

Proxy reporting may be necessary for young children, people with severe cognitive impairment or patients who are too unwell to respond. The record should identify the respondent and whether the proxy was asked to report the patient's perceived experience or the proxy's own judgement.

Proxy and patient scores may differ systematically, particularly for internal experiences such as pain, mood or fatigue. They should not be combined without considering the source of the report.

Using patient-reported outcomes in health economics

Patient-reported outcomes can provide evidence about symptoms, functioning, treatment burden and health-related quality of life for economic evaluations. Some instruments generate preference-based utility values directly, while others provide clinical or quality-of-life scores that cannot be used as utilities without additional methods.

Economic analyses may use patient-reported outcomes to:

• Estimate health-state utility values.
• Measure temporary effects of adverse events.
• Describe treatment benefits not captured by survival or clinical events.
• Inform model health states and transitions.
• Identify effects on daily activities and functioning.
• Support subgroup or heterogeneity analysis.

A patient-reported score should not be entered into a quality-adjusted life-year calculation merely because it is numerical. The measure must have an appropriate preference-based interpretation, or a validated mapping model must be used and its additional uncertainty represented.

A simplified example

Suppose a trial uses a validated fatigue scale ranging from 0 to 40, with higher scores representing worse fatigue. The mean score falls from 28 to 19 in the intervention group and from 27 to 24 in the comparator group.

The average change is:

$$ \Delta PRO_{intervention} = 19 - 28 = -9 $$

$$ \Delta PRO_{comparator} = 24 - 27 = -3 $$

The difference in mean change is:

$$ \Delta\Delta PRO = -9 - (-3) = -6 $$

Because lower scores represent less fatigue, the result favours the intervention by six points. Interpretation still requires evidence about measurement error, missing data and whether a six-point difference is important to patients in this population.

Common misunderstandings

Patient-reported outcomes provide direct evidence about the patient's perspective, but they are not automatically subjective, unreliable or free from measurement problems. Their credibility depends on the instrument, collection process and analysis.

Common misunderstandings include:

• A clinician's description of a patient's symptoms is not a patient-reported outcome.
• A patient-reported experience measure is not necessarily a patient-reported outcome measure.
• Any questionnaire completed by a patient is not automatically a validated measure.
• A total score does not explain which dimensions changed.
• Statistical significance does not establish that patients experienced an important benefit.
• A condition-specific score is not automatically a utility value.
• Proxy and patient reports are not interchangeable.
• Missing assessments should not be assumed to represent no change.
• Electronic collection does not remove recall, selection or response bias.

Interpreting patient-reported outcome evidence

A patient-reported outcome should be interpreted in relation to the precise concept measured, instrument version, population, timing, response rate and scoring rules. The result is strongest when the outcome was prespecified, the instrument is fit for purpose and missing data are addressed transparently.

Patient-reported outcomes add information that clinical and administrative measures cannot provide on their own. They are most useful when they are integrated with clinical outcomes, safety evidence and patient priorities without being treated as a substitute for every other form of evidence.