Patient-Reported Outcome

Patient-Reported Outcome in Trials, HTA and NHS Performance Measurement

A patient-reported outcome (PRO) is health information taken from the patient's own answer and recorded as given, whether the topic is pain on a numeric scale or the five EQ-5D dimensions. What makes an outcome patient-reported is the person giving the report, so the same symptom is a PRO when the patient rates it and a clinician-reported outcome when a doctor rates it. Health economics relies on PROs because the EQ-5D, NICE's preferred adult measure for QALYs, is a patient questionnaire, and NICE asks for health-related quality of life to be measured directly by patients. The economist's questions are whether an instrument fits the decision and what happens to the evidence when patients stop answering.

The source of the report defines a PRO

The FDA's 2009 guidance on PRO measures defines a PRO as a measurement based on a report that comes directly from the patient about the status of their health condition, without amendment or interpretation of the response by a clinician or anyone else. An interviewer may collect it, provided the interviewer records only the patient's response. The guidance advises a PRO instrument when the concept is best known by the patient or best measured from the patient perspective.

A proxy-reported outcome, from someone answering as if they were the patient, is not a PRO, and an observer such as a clinician or caregiver may add interpretation to what was seen. PROs form one branch of the clinical outcome assessment family, and the clinician-reported outcome page covers what changes when professional judgement enters the score.

The PRO is the concept, such as fatigue severity, and a patient-reported outcome measure (PROM) is the questionnaire that captures it, so one concept can have several competing measures. A patient-reported experience measure describes how care was delivered, not the patient's health.

How answers become scores

The FDA guidance asks for an instrument's conceptual framework, which defines the concepts measured and how items, domains and scores relate. It warns that answers relying on memory over a long period are likely to be influenced by the patient's state at the time of recall, so items on the current or recent state are usually preferable.

The EQ-5D is generic, describing health on five dimensions (mobility, self-care, usual activities, pain and discomfort, anxiety and depression) so that results can be compared across diseases and treatments. Condition-specific PROMs ask about one condition and are often more sensitive within it. Other generic instruments include the PROMIS-29 profile and the preference-based SF-6Dv2.

A second split matters more to economists. The EQ-5D questionnaire produces a profile of ticked levels, which is the patient's report. NICE's PMG51 statement describes a value set as a mathematical model that converts EQ-5D responses to utility values. The patient supplies the profile and a sample of the public supplies the weights, so a utility is a PRO combined with other people's preferences.

Showing that an instrument measures what matters to patients

The FDA guidance defines content validity as the extent to which the instrument measures the concept of interest, supported by qualitative evidence that items and domains are appropriate and complete for the intended concept, population and use. It states that evidence of other types of validity or reliability will not overcome problems with content validity, which should be established before other measurement properties are evaluated. An existing instrument cannot be assumed to have content validity if patients were not involved in developing it.

If many items ask about activities that most trial patients would not perform, the guidance warns, the measure is biased towards the null and can show no effect even when the treatment works. Further evidence comes from construct validity, test-retest reliability, responsiveness and measurement error. A ceiling effect, where many respondents report no problems at all, limits how much improvement an instrument can register.

PROs in trial protocols and reports

PRO evidence is often weakened by trial design and reporting rather than by the instrument. The SPIRIT-PRO extension of 2018 adds 11 extensions and 5 elaborations to the SPIRIT checklist for protocols in which PROs are a primary or key secondary outcome, including items on assessment time points, proxy completion and strategies to minimise missing data.

The CONSORT PRO extension of 2013 recommends five items for reporting such trials. The PRO should be identified as a primary or secondary outcome in the abstract, the hypothesis and relevant domains described, evidence of validity and reliability given or cited, the statistical approach to missing data stated, and PRO-specific limitations and generalisability discussed.

Missing data is the largest threat. The FDA guidance notes that it can introduce bias because only a subset of the randomised population contributes and the groups may no longer be comparable, and it asks that patients who stop treatment keep providing PRO data. Rules for missing items should be specific to each instrument. For whole missed assessments the guidance names no preferred method, since every method rests on assumptions about why data are missing that usually cannot be verified, and it recommends two or more sensitivity analyses. Because patients in worse health may be less likely to complete questionnaires, missingness can depend on the unobserved scores, so PRO data can be missing not at random.

Patient answers as the basis of QALYs in NICE appraisal

The NICE manual (PMG36, section 4.3) sets how PRO data become QALYs in a cost-utility analysis. Changes in health-related quality of life should be measured directly by patients, and when that is not possible by the person who acts as their carer rather than by healthcare professionals. Their valuation should come from public preferences in a representative UK sample using a choice-based method, and the EQ-5D is the preferred measure in adults. Submissions should report whether measures were completed by patients themselves or on their behalf, for example by parents, carers or clinicians.

PMG51, NICE's interim methods statement published on 27 August 2026, makes the EQ-5D-5L descriptive system and the new UK value set the preferred approaches for adults, with the UK 5L value set used to derive all adult utility values in reference-case analyses. The EQ-5D-5L value set page covers that change. When a trial collected only a condition-specific PROM, mapping to the EQ-5D remains a departure from the reference case (PMG36, section 4.3.9), and PMG51 (sections 1.10 and 1.11) now asks for the mapping function to target the UK 5L value set, fully described and with its fit shown.

Routine PROMs and the comparison of NHS providers

Devlin, Parkin and Browne report that the Department of Health introduced routine collection of PROMs to measure the performance of healthcare providers, with generic EQ-5D and condition-specific PROMs collected before and after four surgical procedures from April 2009. They argue that summarising EQ-5D profiles by their index scores can hide the nature of change in patient-reported health.

Their Paretian Classification of Health Change borrows from the Pareto improvement in welfare economics. A patient's health counts as better if it is better on at least one dimension and worse on none, and as worse in the mirror case. The two other outcomes are no change and mixed change. In the pilot hip replacement data 356 of 434 patients (82%) improved, 21 showed no change, 18 got worse and 39 had mixed change. For cataract surgery more patients worsened (26.4%) than improved (21.0%), which Devlin, Parkin and Janssen caution may reflect the EQ-5D not capturing the benefits of cataract operations. The method needs no preference weights, but it ignores the size of each change and the importance of the dimension affected.

Comparing hospitals needs case-mix adjustment. The NHS method summarised by Devlin, Parkin and Janssen uses regression to estimate each provider's expected outcome for its mix of patients and compares it with the actual outcome, and adjustment changed the ranking of hospitals by health gain.

Worked example with one provider's hip replacement PROMs

All numbers in this example are illustrative. A provider performs 400 hip replacements and 300 patients return both the pre-operative and the six-month EQ-5D-5L questionnaire, a response share of 0.75. Profiles list the levels for mobility, self-care, usual activities, pain and discomfort and anxiety and depression, where 1 means no problems.

1. Classify individual changes. Patient A reports 32342 before and 11222 after, better on four dimensions and worse on none, so improved. Patient B reports 21331 then 11223, better on three and worse on anxiety and depression, so mixed. Patient C reports 21221 on both occasions, so no change. Patient D reports 21231 then 22231, worse on self-care only, so worsened.

2. Summarise the provider. Of the 300 returning patients, 246 improved (82%), 12 showed no change (4%), 15 worsened (5%) and 27 had mixed change (9%), shares that add to 100%.

3. Convert profiles to utilities. Applying a value set to the profiles gives a mean utility of 0.45 before surgery and 0.83 after among the 300 responders, a gain of 0.83 minus 0.45 = 0.38.

4. Allow for the missing quarter. Suppose the 100 patients who did not return the second questionnaire gained only 0.25 on average.

$$\bar{\Delta} = r\,\bar{\Delta}_{obs} + (1-r)\,\bar{\Delta}_{mis} = 0.75 \times 0.38 + 0.25 \times 0.25 = 0.3475$$

where $\bar{\Delta}$ is the mean utility gain across all 400 patients, $r$ is the response share, $\bar{\Delta}_{obs}$ is the mean gain among responders and $\bar{\Delta}_{mis}$ is the assumed mean gain among non-responders.

5. Turn the gain into cost per QALY. Assume the gain lasts one year, that health would not have changed without surgery, and that the extra cost of surgery is GBP 7,000 per patient.

$$\frac{C}{\bar{\Delta} \times t} = \frac{7{,}000}{0.3475 \times 1} \approx 20{,}144$$

where $C$ is the extra cost per patient in GBP and $t$ is the duration of the gain in years. Using responders alone gives 7,000 / 0.38 = GBP 18,421 per QALY.

6. Interpretation. The complete-case figure is about 9% more favourable (20,144 / 18,421 = 1.094) because it assumes the missing patients did as well as those who answered. The before and after gain is an incremental effect only if health would have stayed the same without surgery, and Devlin, Parkin and Janssen recommend specifying that counterfactual or measuring it with a control group. The 9% of mixed cases are patients who improved on some dimensions and worsened on another, which a mean utility gain does not reveal.

Boundaries with nearby concepts

A PRO is a report by the patient, while health-related quality of life is one concept a PRO can measure, so a single pain rating is a PRO without being a quality of life measure. A health state utility value combines a patient-reported profile with public preferences and is not itself a pure PRO. The smallest score change patients regard as important is the minimal clinically important difference, and an agreed minimum set of outcomes for a condition is a core outcome set.

Where PRO evidence misleads

Many errors arise when PRO results move into models and performance tables. Each can change a cost per QALY or a provider ranking.

1. Counting a proxy answer as the patient's. The FDA does not treat a proxy report as a PRO, and NICE asks submissions to say who completed each measure.
2. Analysing only complete cases. If patients who gain less stop answering, the observed gain overstates the true gain, as steps 4 to 6 show.
3. Reading a before and after change as a treatment effect. Without a control group the result depends on the assumed counterfactual.
4. Trusting a generic measure everywhere. The cataract figures suggest a generic instrument can miss the benefit a treatment is meant to deliver.
5. Reducing a profile to one number too early. An index score hides mixed changes.

Sources

• Calvert M, Blazeby J, Altman DG, Revicki DA, Moher D, Brundage MD; CONSORT PRO Group. Reporting of patient-reported outcomes in randomized trials: the CONSORT PRO extension. JAMA. 2013;309(8):814-822. https://doi.org/10.1001/jama.2013.879
• Calvert M, Kyte D, Mercieca-Bebber R, et al.; SPIRIT-PRO Group. Guidelines for inclusion of patient-reported outcomes in clinical trial protocols: the SPIRIT-PRO extension. JAMA. 2018;319(5):483-494. https://doi.org/10.1001/jama.2017.21903
• Devlin NJ, Parkin D, Browne J. Patient-reported outcome measures in the NHS: new methods for analysing and reporting EQ-5D data. Health Economics. 2010;19(8):886-905. https://doi.org/10.1002/hec.1608
• Devlin N, Parkin D, Janssen B. Methods for Analysing and Reporting EQ-5D Data. Cham: Springer; 2020. Chapters 1, 2 and 5. https://doi.org/10.1007/978-3-030-47622-9
• Food and Drug Administration. Guidance for Industry. Patient-Reported Outcome Measures: Use in Medical Product Development to Support Labeling Claims. Silver Spring, MD: FDA; December 2009. https://www.fda.gov/media/77832/download
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 31 January 2022, last updated 13 October 2026. Section 4.3. https://www.nice.org.uk/process/pmg36
• National Institute for Health and Care Excellence. Interim methods statement: implementing the EQ-5D-5L value set (PMG51). Published 27 August 2026. https://www.nice.org.uk/process/pmg51