1010 THE JOURNAL OF BONE AND JOINT SURGERY ANNOTATION The use of the Oxford hip and knee scores D. W. Murray, R. Fitzpatrick, K. Rogers, H. Pandit, D. J. Beard, A. J. Carr, J. Dawson From the University of Oxford, Oxford, England D. W. Murray, MD, FRCS(Orth), Professor H. Pandit, FRCS(Orth), Research Fellow D. J. Beard, DPhil, University Research Lecturer A. J. Carr, FRCS, Nuffield Professor of Orthopaedic Surgery Nuffield Department of Orthopaedic Surgery University of Oxford, Headington, Oxford OX3 7LD, UK. R. Fitzpatrick, PhD, Professor K. Rogers, BA (Hons), Statistical Analyst J. Dawson, DPhil, Senior Research Scientist Department of Public Health University of Oxford, Old Road Campus, Oxford OX3 7LF, UK. Correspondence should be sent to Professor D. W. Murray e-mail: David.Murray@noc.anglox.nhs. uk ��2007 British Editorial Society of Bone and Joint Surgery doi:10.1302/0301-620X.89B8. 19424 $2.00 J Bone Joint Surg [Br] 2007 89-B:1010-14. The Oxford hip and knee scores have been extensively used since they were first described in 1996 and 1998. During this time, they have been modified and used for many different purposes. This paper describes how they should be used and seeks to clarify areas of confusion. It is now almost ten years since the Oxford hip and knee scores were introduced.1,2 The use of patient-reported outcomes in orthopaedics at that time was extremely limited. The scores were developed to assess the outcome of hip and knee replacements in randomised trials and were designed to be completed by the patient in order to minimise potential bias unwittingly introduced by surgeons when assessing the results themselves.3 An additional advantage in using patient-reported outcome measures was that they could be completed at a remote loca- tion by post, thereby avoiding inconvenience and cost. The Oxford hip and knee scores were devised as joint specific instruments aimed to minimise the influence of comorbidity. They underwent rigorous assessment of reliability, validity and responsiveness in prospective studies. Their use has steadily increased and they are now widely employed. They have been used in cohort studies and audits and in national joint replace- ment registries, including those in England, New Zealand and Sweden. They have also been applied to other disorders of the joints and in surgical management other than arthroplasty. A number of issues have been raised and various groups have made modifications to the scoring system which has resulted in some con- fusion. We therefore now review their use and recommend ways in which they should be best employed. If the scores are used in a standard fashion, data from different series can be more easily compared which is one of the main reasons for developing and using such systems. Wording of questions and response categories While the response rates for the Oxford hip and knee scores are generally higher than those for many other measures of health status,1,2,4 we and others5,6 initially found that some patients had difficulty in answering specific questions. For example, in question 4 in the knee score, where patients are asked how long they are able to walk before pain becomes severe, the extreme response ���not at all��� was sometimes wrongly inferred by the patient to mean the opposite of that intended. For clarifi- cation, we made small amendments to the response categories for two items. The recom- mended format for questions and their response categories can be found on the Patient-Reported Health Instruments website.7 A different problem is typified by question 7 which regards kneeling. This causes particular difficulty when patients have been told not to kneel. The item reads ���Could you kneel down and get up again afterwards?��� The word ���could��� is in bold type, suggesting that patients should answer this item hypothetically if they have been told not to kneel. This is our recom- mendation, which can be communicated to the patient. A similar approach should be taken for the question about stairs by patients living in bungalows and about housework by those who have housekeepers or live in residential care. If, after clarification, an item is still unanswered, it should be dealt with as missing data. It has been further suggested that the Oxford scores should be extended by adding two questions on the need for walking aids and, in the case of the hip, sexual problems.8 We think that this would be more detrimen- tal than helpful for three reasons. First, the questions should be relevant to most patients, and these issues were not found to be generally important in the initial inter- views from which the Oxford hip score Annotation

THE USE OF THE OXFORD HIP AND KNEE SCORES 1011 VOL. 89-B, No. 8, AUGUST 2007 resulted.2 Secondly, a new questionnaire would no longer be comparable with the old and thirdly, it would mean that further work on validation would have to be under- taken. It is wrong to imagine that there will ever be a per- fect questionnaire which suits all people at all times. Scoring When the Oxford hip and knee score systems were orig- inally devised they were made as simple as possible in order to encourage their use. Thus, each question was scored from 1 to 5, with 1 representing the best outcome/ least symptoms. The scores from each question were added so that the overall figure lies between 12 and 60, with 12 being the best outcome. Subsequently, many sur- geons have found this scoring to be unintuitive and have modified it, particularly when using the knee score. These changes have led to considerable confusion. For example, the use of a system9 whereby each question has been scored between 0 and 4, with 4 being the best out- come, produces overall scores running from 0 to 48, with 48 being the best outcome. The 60 to 12 system may be converted to the 0 to 48 score and vice versa by subtract- ing the score from 60.9 It has also been suggested that the score could be modified to range between 0 and 100 with either 100 or 0 being the best.10 Although it is possible to convert between one system and another it would sim- plify the situation considerably if a standard form was used. Our view is that both the hip and knee scores should be used from 0 to 48, with 48 being the best. The method used should always be clearly stated. Comparison with other scoring systems The Oxford scores were designed to be joint-specific in order to increase their sensitivity to the outcome of the joint replacement as far as possible and to be influenced as little as possible by other comorbidities. Although the scores are influenced by pathology, such as strokes and back problems11 elsewhere, they seem to be influenced to a less extent than is the case for other patient-reported outcome measures, used in this context.12,13 This feature of a score ��� its specificity ��� influences its responsiveness or sensitivity to change, which is the most important aspect in relation to prospective outcome studies.14,15 The Oxford hip and knee scores have been shown to have particularly high responsiveness.1,2 Since the Oxford hip and knee scores have been eval- uated independently and found to be the best and most reliable systems for the assessment of hip and knee replacement, respectively,4,16-19 there is some justifica- tion for using these scores in isolation. However, if it is important to compare the improvement resulting from hip or knee procedures with those occurring at other sites, it is sensible to use a general health measure, such as the SF-1220 as well as the Oxford score. If information on health economics is needed the EuroQol21 is valuable. If specific clinical and surgical data, such as range of movement, are required then a formal clinical assessment would be necessary. Use of the scores While the scores were designed to be primary measures of outcome in randomised, controlled trials, they have been much more widely used in cohort studies and audits. It has become apparent that one of the biggest determinants of outcome after a joint replacement is the pre-operative score.11,18,22 If the treatment of different cohorts of patients is being studied in a non-randomised setting, it is essential that both the pre-operative and post-operative scores are obtained. The change in the score should be analysed in addition to the post-operative score. Likewise, if a multi- variate analysis of outcomes is undertaken, this should take into account the pre-operative score.23 After joint replace- ment, most improvements in function and in the Oxford scores occur within the first year.8 It is therefore not unreasonable to assess the outcome at one year. Given that the score at one year is related to the pre-oper- ative score when using the scores for audit purposes, it is useful to know approximately what outcome would be expected after a total joint replacement given a particular pre-operative score. A publication based on data from a national audit22 has reported the mean Oxford hip score before and after total hip replacement (THR), presented by 10% bands (deciles) of the pre-operative score. Table I sum- marises these data converted to the 0 to 48 scoring system. A similar analysis has been undertaken on data from a large multicentre randomised, controlled trial of total knee replacement (TKR) and is presented in Table II.24,25 It should, however, be noted that, like all outcome scores, the absolute score tends to decrease with age.26 Therefore, in elderly patients, a ���normal��� score may be somewhat less than 48. Although we have achieved very high rates of response when the Oxford hip and knee scores have been sent to patients, other authors have not been as successful.27 Practical approaches to maximising rates of response have been recommended elsewhere.28 These include using care- fully-worded covering letters, sending reminders with second copies of the questionnaire using pre-paid reply envelopes and contacting patients by telephone. When patients have bilateral joint problems we favour giving two questionnaires, one for each side, rather than modify- ing the questionnaires to include both sides.10 Surgeons are attracted to systems of categorisation, grouping the results according to whether they are consid- ered to be excellent, good, fair or poor, rather than simply using a score. However, the cut-off points in categories are always very approximate and are likely to vary from one population to another. Work is currently in progress to pro- duce categories based on large international data sets. Until this is available we believe that surgeons should avoid cat- egorisation. If, in spite of this, surgeons are still keen to cat- egorise, then they can use published cut-off points. For