Oxford Knee Score

Oxford Knee Score in Knee Replacement Evaluation and NHS PROMs

The Oxford Knee Score is the knee-specific questionnaire that patients complete before and after knee replacement in England's national patient-reported outcome programme. Dawson and colleagues at the University of Oxford published it in 1998, two years after the hip questionnaire from the same group, with its own items and its own validation study in knee replacement patients. For a health economist the score matters because NHS England uses it to compare providers of knee replacement, and because published mapping algorithms turn it into EQ-5D utilities for studies without a generic measure. Dakin and colleagues estimated one such algorithm on 134,269 questionnaires.

Development and validation in knee replacement patients

The questionnaire was designed for people having total knee replacement. Its items and its evidence of validity belong to that population.

Dawson and colleagues studied 117 patients prospectively before total knee replacement and again six months afterwards. Patients completed the new questionnaire and the SF-36, some also completed the Stanford Health Assessment Questionnaire, and each knee was assessed on the clinician-rated American Knee Society clinical score. The single score showed high internal consistency and acceptable test-retest reliability, its validity was assessed against the comparison measures, and its effect size compared favourably with those of the relevant parts of the SF-36. The authors concluded that the questionnaire gives a measure of outcome for total knee replacement that is short, practical, reliable, valid and sensitive to clinically important changes over time.

Scoring rules, recall period and incomplete forms

The way the Oxford Knee Score is scored changed after publication, and studies use both versions. The developers' guide sets out the current method and how to handle incomplete or ambiguous answers.

The questionnaire asks about the past four weeks, and each item has five ordered responses. The 1998 paper scored each item 1 to 5, giving totals of 12 to 60. The developers' guide notes that many surgeons found this unintuitive and adapted the scoring, which led to considerable confusion, and it now recommends scoring each item 0 to 4 with 4 as the best outcome, so that totals run 0 to 48 with 48 best. An old total converts to the new scale when subtracted from 60. Where a patient marks two answers for one item, the convention is to take the worse response.

Missing answers are handled with a limited imputation rule. If one or two items remain unanswered after attempts to complete them, the mean of the patient's other answers fills the gaps, and if more than two are missing no total is calculated.

$$S = \sum_{i \in A} x_i + m \times \bar{x}_A, \qquad m \le 2$$

where $S$ is the imputed Oxford Knee Score total, $A$ is the set of answered items, $x_i$ is the 0 to 4 score of answered item $i$, $m$ is the number of missing items and $\bar{x}_A$ is the mean score of the answered items. With $m$ equal to zero the formula reduces to the plain sum of the twelve items.

Knee replacement in the national PROMs programme

England's programme asks patients having NHS-funded elective knee replacement to complete the Oxford Knee Score with EQ-5D before surgery and again afterwards. NHS England uses the results to compare providers.

NHS England's 2017 programme guidance names the Oxford Knee Score as the condition-specific measure for knee procedures, collected alongside the three-level EQ-5D then in use, with a post-operative questionnaire after a minimum follow-up of six months. NHS England Digital's final 2024/25 publication, released on 2 April 2026 as management information and not as official statistics, counted about 116,000 knee replacement episodes, with about 75,000 pre-operative and 35,000 post-operative responses. In that release 72.8 per cent of knee patients said they felt much better after surgery and a further 16.8 per cent a little better, against 85.8 and 9.4 per cent for hip patients.

The 2023/24 release showed a similar gap, 73.5 against 84.8 per cent. NHS England Digital's data quality statement for 2023/24 says that it is fair to compare scores across organisations for the same procedure and measure, but not to compare scores for the same measure across procedures, so the gap does not rank knee replacement below hip replacement. An Oxford Knee Score gain and an Oxford Hip Score gain come from different instruments, which gives a comparison of their points even less basis.

Mapping Oxford Knee Score responses to EQ-5D utilities

The total is a non-preference-based summary of knee symptoms, so the Oxford Knee Score cannot be used as a utility weight for QALY calculation. Economic evaluations of knee interventions that collected only the Oxford Knee Score need a mapping function estimated in data that hold both measures, and the general method is the subject of the mapping study concept.

Dakin and colleagues estimated such algorithms in 2013 to support cost-utility analysis of studies that use the Oxford Knee Score without a generic health measure. They pooled the Knee Arthroplasty Trial and the UK PROMs data set, estimated models on 134,269 questionnaires with a quarter of the data held out for internal validation, and tested the results externally on 10,002 observations from a London registry. Among ordinary least squares, generalised linear, two-part, three-part and response mapping models, a multinomial response mapping model that used Oxford Knee Score responses to predict EQ-5D response levels gave the best prediction accuracy, with a mean squared error of 0.033 and a mean absolute error of 0.129 in external validation. Coefficients differed slightly but significantly between the trial and PROMs data sets.

Because the response mapping model predicts EQ-5D-3L dimension levels and not a single index, the predicted levels can be scored with any EQ-5D-3L value set, though not directly with a 5L value set. NICE's technology appraisal manual (PMG36, section 4.3.9) allows EQ-5D values to be estimated by mapping other health-related quality-of-life measures when EQ-5D data are not available. Under the interim methods statement PMG51, published on 27 August 2026, such measures should be mapped to utility values based on the UK EQ-5D-5L value set, which may mean mapping onto the 3L value set first and then onto the 5L value set. A knee score algorithm that predicts 3L levels therefore needs that second step in a NICE reference case, and the coefficient differences between data sets show that a mapping function carries its estimation sample into every model that uses it.

Thresholds for meaningful change in knee scores

Thresholds for the Oxford Knee Score are slightly lower than those for the hip score. Beard and colleagues estimated them in 2015 from anchor-based and distribution-based methods.

Which threshold applies depends on the minimal clinically important difference question being asked. A between-group difference of 4 points in a knee trial falls below the between-group threshold of about 5 points, while an individual change of 7 points meets the individual threshold.

Illustrative worked example of imputation and provider comparison

All numbers in this example are illustrative. One patient returns a six-month Oxford Knee Score form with two items blank, and the ten answered items sum to 30. The patient's pre-operative total was 18. The provider's 120 patients with both questionnaires returned show a mean gain of 15 points, against a national mean gain of 16 points for the same procedure.

1. Impute the missing items. The mean of the answered items is 30 / 10 = 3. With two items missing, the imputed total is 30 + 2 × 3 = 36, which the rule allows because no more than two items are blank.

2. Judge the individual change. The change is 36 − 18 = 18 points, more than twice the individual threshold of 7 points.

3. Compare the provider with the national figure. The provider's mean gain is 16 − 15 = 1 point below the national mean. The data quality statement supports this kind of comparison across organisations for the same procedure and measure, though a raw gap of 1 point does not by itself show a difference beyond chance.

4. Count the patients behind the average. Adding this patient's imputed gain of 18 points raises the provider's sample to 120 + 1 = 121 and its mean gain to (120 × 15 + 18) / 121 = 1,818 / 121, or about 15.02 points.

Had a third item been blank, the patient would have no valid total and would drop out of the provider's average gain. Many such forms would shrink the sample on which the provider comparison rests and would leave it describing only the patients with complete answers.

Where the Oxford Knee Score sits among knee and generic measures

The Oxford Knee Score belongs to a family of joint-specific questionnaires and is often collected with generic and clinician-rated measures. Each neighbour answers a different question.

The Oxford Hip Score uses the same response format and scoring but different items and its own validation in hip replacement patients. The WOMAC is a separate osteoarthritis index for hip and knee with different items and scoring, so pooling WOMAC results with the Oxford Knee Score needs a validated crosswalk. The American Knee Society score, used as a comparator in the 1998 validation, is clinician-rated, not patient-reported. EQ-5D is a generic, preference-based measure of health-related quality of life scored with a value set, which is why it is collected alongside the knee score and why it, not the knee score, supplies QALY weights.

Kang reported in 2021, using NHS PROMs data linked to hospital records for 191,379 primary knee and 181,424 primary hip patients, that the Oxford Knee Score showed great univariate and multivariate responsiveness while the EQ-5D-3L showed moderate effects. Pain and function can also be reported separately, as Harris and colleagues did when they analysed Oxford Knee Score subscales.

Limitations and common misreadings

The Oxford Knee Score is most often misused when old and new scales are mixed or when knee results are read as utilities. Debate over its ceiling and the loss of post-operative responses add further limits.

• Scoring direction. Studies that kept the 1998 scale report lower totals for better knees, and pooling them unconverted with 0 to 48 scores reverses effects.
• Ceiling debate. Harris and colleagues analysed 72,154 knee replacement patients in the NHS PROMs data set and found that 2.7 per cent reached the maximum score after surgery, concluding that the score and its pain and function subscales show no ceiling effect or floor effect. In the subgroup with a pre-operative score above 41, however, 13.7 per cent reached the ceiling, so studies of patients with good pre-operative function have less room to show gain.
• Points are not utilities. A knee score gain cannot be multiplied by time to give QALYs without mapping.
• Cross-procedure comparison. Knee and hip score gains are not comparable as measures of benefit, as the national data quality statement warns.
• Non-response. Fewer than half as many post-operative as pre-operative responses arrived in 2024/25, so linked before and after analyses describe responders only.

Sources

• Beard DJ, Harris K, Dawson J, Doll H, Murray DW, Carr AJ, Price AJ. Meaningful changes for the Oxford hip and knee scores after joint replacement surgery. Journal of Clinical Epidemiology. 2015;68(1):73-79. doi:10.1016/j.jclinepi.2014.08.009
• Dakin H, Gray A, Murray D. Mapping analyses to estimate EQ-5D utilities and responses based on Oxford Knee Score. Quality of Life Research. 2013;22(3):683-694. doi:10.1007/s11136-012-0189-4
• Dawson J, Fitzpatrick R, Murray D, Carr A. Questionnaire on the perceptions of patients about total knee replacement. Journal of Bone and Joint Surgery (British volume). 1998;80(1):63-69. doi:10.1302/0301-620x.80b1.7859
• Harris K, Lim CR, Dawson J, Fitzpatrick R, Beard DJ, Price AJ. The Oxford knee score and its subscales do not exhibit a ceiling or a floor effect in knee arthroplasty patients: an analysis of the National Health Service PROMs data set. Knee Surgery, Sports Traumatology, Arthroscopy. 2017;25(9):2736-2742. doi:10.1007/s00167-015-3788-0
• Kang S. Assessing responsiveness of the EQ-5D-3L, the Oxford Hip Score, and the Oxford Knee Score in the NHS patient-reported outcome measures. Journal of Orthopaedic Surgery and Research. 2021;16:18. doi:10.1186/s13018-020-02126-2
• NHS England. National Patient Reported Outcome Measures (PROMs) Programme Guidance. 2017. https://www.england.nhs.uk/wp-content/uploads/2017/09/proms-programme-guidance.pdf
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). 2022, updated 2026. Section 4.3.9. https://www.nice.org.uk/process/pmg36/chapter/economic-evaluation-2
• National Institute for Health and Care Excellence. Interim methods statement: implementing the EQ-5D-5L value set (PMG51). Published 27 August 2026. Sections 1.10 and 1.11. https://www.nice.org.uk/process/pmg51/chapter/statement-on-using-the-eq-5d-5l-value-set
• NHS England Digital. Patient Reported Outcome Measures (PROMs) in England, Final 2024/25 data. Published 2 April 2026. https://digital.nhs.uk/data-and-information/publications/statistical/patient-reported-outcome-measures-proms/final-2024-25-data
• NHS England Digital. Patient Reported Outcome Measures (PROMs) in England, Final 2023/24 data, data quality statement. Published 13 February 2025. https://digital.nhs.uk/data-and-information/publications/statistical/patient-reported-outcome-measures-proms/final-2023-24-data/data-quality-statement
• Oxford University Innovation. The Oxford knee score: a guide to the new scoring system. 2016. https://innovation.ox.ac.uk/wp-content/uploads/2016/05/OKS-scoring-guide_2016-01-15.pdf