Investigation performed at Royal Adelaide Hospital, Adelaide, South Australia, Australia

Transcription

1 1745 COPYRIGHT 2002 BY THE JOURNAL OF BONE AND JOINT SURGERY, INCORPORATED Comparison of Patient and Doctor Responses to a Total Hip Arthroplasty Clinical Evaluation Questionnaire BY MARGARET A. MCGEE, BSC, MPH, DONALD W. HOWIE, PHD, MBBS, FRACS, PHILIP RYAN, MBBS, BSC, FAFPHM, JOHN R. MOSS, MSOCSCI, BEC, MBBS, FCHSE, AND OKSANA T. HOLUBOWYCZ, PHD, MPH, BA Investigation performed at Royal Adelaide Hospital, Adelaide, South Australia, Australia Background: Surgeons traditionally undertake a prospective evaluation of patients undergoing total hip arthroplasty in order to determine outcomes. The validity of doctor-derived data is questionable because of the potential for interobserver error, reporting bias, and differences between the perceptions of doctors and patients. Also, the use of doctor-derived data necessitates the use of costly outpatient services. Consequently, there are likely to be benefits associated with the use of patient-derived clinical evaluation data. However, few studies have focused on whether data obtained from the patient and doctor differ. Methods: The agreement between patient and doctor responses on a sixteen-item total hip arthroplasty clinical evaluation questionnaire completed at more than 2900 clinical assessments was determined. Data from repeated assessments performed preoperatively and postoperatively enabled stratified analyses that were used to examine reasons for disagreement and factors influencing agreement. Agreement was measured with use of the kappa coefficient. Results: For twelve of the sixteen items, the patient responses had acceptable agreement with the doctor responses. Some important differences between patient-derived and doctor-derived data were found. If the patient had other joint or health problems, had a revision total hip arthroplasty, or reported mild or moderate pain, there was a greater chance of reduced agreement on the pain items. Younger patients demonstrated better agreement with doctors than older patients did. Conclusions: Patients perceptions of symptoms and outcomes after total hip arthroplasty are relatively similar to those of their doctor. There is minimum risk of misinterpreting outcomes data by replacing doctor-completed questionnaires with patient-completed questionnaires in uncomplicated total hip arthroplasty cases. For patients with comorbid joint problems or other health problems, and for those reporting substantial pain, direct physician involvement in the evaluation of pain is recommended. The selective use of patient-completed questionnaires has the potential to substantially reduce the costs of outcomes evaluation programs by minimizing doctor input. Pending revision of some of the items, the use of this patient-completed questionnaire is advocated. The importance of measuring health outcomes after total hip arthroplasty has become apparent because of the considerable variation in the quality, processes, and technology used in clinical practice and a realization that reports in the orthopaedic literature are often based on the results from tertiary referral centers rather than from general orthopaedic practice. Health-care outcomes instruments that have been developed to assess the physical symptoms and related health states of patients managed with total hip arthroplasty traditionally have been completed by the doctor 1-3. There are concerns about possible differences between the perceptions of the doctor and those of the patient as well as about the validity of doctor-derived data. There is also the potential for systematic or random interobserver error between doctors, including possible reporting bias by doctors. Consequently, the potential importance of patient-derived data is highlighted, and recent outcomes instruments have emphasized patient-reported outcomes 4-6. However, few studies have compared patient and doctor-derived data on the outcome of total hip arthroplasty. Through a process of consensus involving committees of three international orthopaedic groups, namely, the Commission on Documentation and Evaluation of the Société Internationale de Chirurgie Orthopédique et de Traumatologie (SICOT), the Hip Society, and the Task Force on Outcome Studies of the American Academy of Orthopaedic Surgeons, key outcome parameters for total hip arthroplasty were identi-

2 1746 TABLE I Total Hip Arthroplasty Clinical Evaluation Questionnaire 7 Clinical Parameters Patient Satisfaction Parameters 1. Level of activity and work 12. Whether operation increased function 2. Capacity to do work and level of activity in the preceding 3 months 13. Whether operation decreased pain 3. Degree of hip pain 14. Whether operation decreased need for pain medication 4. Occurrence of hip pain 15. Whether patient was satisfied with results 5. Ability to put on shoes and socks 16. State of hip compared with last visit 6. Ability to ascend and descend stairs 7. Ability to go from sitting to standing 8. Usual support needed when walking 9. Time walked without support 10. Time walked with support 11. Degree of limp when walking without support fied. A standard system of terminology for these parameters was then developed to facilitate the reporting of the clinical and radiographic results of total hip arthroplasty. This total hip arthroplasty evaluation system was described by Johnston et al. in , and the accompanying editorial 8 strongly recommended widespread adoption of these standard outcome measures to facilitate uniform reporting and therefore communication between orthopaedic groups as well as the conduct of comparative studies. We therefore decided to use these recommended parameters and terminology to develop a questionnaire for this study. The objectives of the present study were to compare patient and doctor responses to the items on this total hip arthroplasty clinical evaluation questionnaire, to determine the level of agreement between the patient and doctor responses, and to examine which factors influenced agreement. Our hypothesis was that there are important differences between patientderived and doctor-derived data from questionnaires that are used to evaluate health outcomes after total hip arthroplasty. Materials and Methods clinical evaluation questionnaire was developed in two A formats, one to be completed by the patient and one to be completed by the doctor. The questionnaire had two components: a clinical evaluation component and a patient satisfaction component 7 (Table I). The clinical evaluation component comprised eleven items, including two on activity, two on pain, three on function, and four on gait. The patient satisfaction component was completed at postoperative assessments only and included five items related to patient satisfaction after total hip arthroplasty. The responses to the clinical evaluation items were ranked with use of ascending or descending ordinal categories. For most items, a higher ranking indicated a poorer response category. Most of the patient satisfaction items had dichotomous responses. The questionnaire was completed by both the patient and the doctor as part of the routine clinical assessment, both preoperatively and at postoperative follow-up outpatient visits. Additional information was collected on comorbid joint problems and other health problems, strength, mobility of the hip joint, and limb length. Radiographic review of the hip joint was also undertaken. Each patient was asked to complete the patient questionnaire in the waiting room prior to his or her consultation. The doctor completed the doctor questionnaire during the consultation with the patient, with use of conventional patient-interview techniques. To enable the doctor to respond to the question on limp, the patient may have been asked to walk a small distance. The patient was asked about, but was not actually required to perform, the other tasks referred to in the remaining function and gait items. More than 85% of the doctor questionnaires were completed by the patient s consultant surgeon or a trainee registrar. The remaining questionnaires were completed by trainee interns under the supervision of the consultant surgeon or trainee registrar. The data set included all clinical assessments performed over a period of approximately five years (from May 1991 to November 1996) during which both the doctor and the patient completed at least some aspect of their respective questionnaires. Approval for use of the patient-outcomes questionnaires was obtained from the institutional ethics committee. All data were prospectively entered into a joint replacement database and were exported into the SAS software package (SAS System for Windows, Release 6.12, ; SAS Institute, Cary, North Carolina) for statistical analyses. The kappa coefficient (κ) was used as the measure of agreement between the patient and doctor responses to each of the items in the questionnaires In most cases, the weighted κ was used, with the weights quantifying the relative differences between the response options. For items with di-

3 1747 TABLE II Covariates Used in the Stratified Analyses of Agreement Covariate Patient gender Male Female Strata Patient age* 45 yr (young) yr (middle-aged) >65 yr (elderly) Total hip arthroplasty procedure Primary Revision Other joint or health problems, including the back Absent Present Patient-reported hip pain None Mild Moderate Severe Type of assessor Consultant doctor Trainee doctor Assessment period Preoperative Each postoperative period *Categories were arbitrarily chosen. chotomous responses, unweighted κ values were calculated. To describe the relative strength of agreement, κ was graded with use of recommended categories 12. A κ value of >0.60 was considered to indicate acceptable agreement. To examine the influence of other factors on the relative agreement, stratified analyses were undertaken for covariates (Table II). Differences in the value of κ among the different strata were examined with use of a chi-square statistic 13. A value of 0.01 was chosen as the critical level of significance to minimize false-positive findings. On the basis of a sample-size calculation formula for the comparison of two coefficients of interobserver agreement 14, the number of assessments undertaken for the study gave sufficient numbers of assessments within each stratum to detect a significant difference, at this level of significance, with a power of 80%. The study data were obtained during 2934 clinical assessments, including 839 preoperative assessments and 2095 postoperative assessments. The postoperative assessments were performed three months to more than ten years after the procedure. Another 518 assessments were not included in the analyses because a questionnaire had not been at least partially completed by both the doctor and the patient. Questionnaire data were obtained at one or more clinical assessments of 1117 patients who underwent a total of 1273 total hip arthroplasty procedures (Table III). Consultant and trainee doctors performed similar total numbers of assessments. Two consultants, both of whom specialized in joint replacement surgery, performed 618 (57%) of the 1093 assess- TABLE III Data on the Patients Operative or Demographic Parameter No. of Patients No. of Procedures Type of hip procedure Unilateral procedure Bilateral procedure Multiple procedures on one hip* Bilateral procedure and two procedures on one hip Total Type of total hip arthroplasty Gender Age Primary 916 (72%) Revision 326 (26%) Other 31 (2%) Male 516 (41%) Female 757 (59%) 45 yr 72 (6%) yr 410 (32%) >65 yr 791 (62%) *Twenty-six patients had two operations, and one patient had three operations. Other = excision arthroplasty or other reoperation at the site of a total hip arthroplasty. The median age was sixty-nine years (range, sixteen to ninety-three years).

4 1748 Fig. 1 Illustration depicting the agreement between patient and doctor responses to the sixteen items of the total hip arthroplasty clinical evaluation questionnaire. ments that were undertaken by consultants. The majority of assessors were male and were forty-five years old or less. The analyses of agreement were performed twice with use of different data sets. The complete data set, which included data from all assessments, was used for the first analysis and for subsequent stratified analyses. On the basis of the results of a previously reported study 15, it was anticipated that the severity of symptoms might influence agreement. Symptom severity is most often greatest at the preoperative assessment. As 29% (839) of the clinical assessments were performed preoperatively, it was important that agreement was assessed with use of questionnaire data from all preoperative and postoperative assessments. For the purposes of analysis, if patients completed questionnaires on more than one occasion, data from each questionnaire were treated as being independent. As patients were followed prospectively, they may have been exposed to the questionnaire more than once. Furthermore, some patients had bilateral total hip arthroplasty or multiple procedures on the same hip and completed the questionnaire on different occasions for these different procedures. To eliminate the potential for response bias due to previous exposure to the questionnaire, only data from the patients first encounter with the questionnaire were included for the second analysis of agreement. The patient s first encounter with the questionnaire was not necessarily during the preoperative assessment. Results he first analysis of agreement between patient and doctor Tresponses involving data from all preoperative and postoperative assessments revealed that there was almost perfect agreement for one of the sixteen items and substantial agreement for eleven (Fig. 1). The analyses of interrater agreement with use of the second data set (that is, the data from the patient s first exposure to the questionnaire) indicated that the agreement between doctor and patient responses was slightly higher than that determined in the first analysis for every item except time walked without support. The strength of agreement did not change for any item. Analyses of the first data set demonstrated that, for nine of the thirteen items with more than two response categories, the responses differed by no more than one category in at least 90% of the assessments. For the items capacity for work/activity and degree of pain, doctors gave more favorable responses than patients in more than 5% of the assessments whereas the reverse was true for ability to ascend and descend stairs (Fig. 2). Interrater disagreement for decreased need for pain medication resulted from the finding that patients were more likely to select the no or not applicable response whereas doctors selected the yes response. Patient age, but not gender, significantly influenced interrater agreement (p < 0.01). For fifteen of the sixteen items, the responses of younger patients (that is, those who were

5 1749 forty-five years old or less) were more often the same as the doctors responses when compared with responses of patients in the two older age-groups. Within this group of younger patients, there was 100% agreement about whether the patient was satisfied with the results of surgery and almost perfect agreement for another six items. The difference in agreement between the age-groups was significant (p < 0.01) for eight of the sixteen items. When the data were stratified according to whether the procedure was a primary or revision total hip arthroplasty, there was reduced agreement for the item degree of hip pain when the patient had had a revision total hip arthroplasty (p = 0.006). Significantly reduced agreement between patient and doctor responses was also evident for the items degree of hip pain (p = 0.001), occurrence of hip pain (p = 0.001), and ability to put on shoes and socks (p = 0.006) when the patient had other joint or health problems (Fig. 3). For many of the items regarding clinical parameters of outcome, agreement between patient and doctor responses decreased as patients self-reported rating of hip pain increased from none to moderate (that is, when the patient was active but had had to modify or give up some activities because of pain). However, when patients reported severe pain (that is, major pain and serious limitations), the level of agreement between the patient and doctor responses was more comparable to that when patients reported no pain. A comparison of κ between the four pain strata indicated significant differences for six of the remaining ten clinical parameters (Fig. 4). While there was no consistent pattern of agreement for the five patient-satisfaction items across the pain responses from no pain to moderate pain, agreement was strongest when patients reported severe pain. There were significant differences in agreement between the strata for the first three patient-satisfaction items (p < 0.003). For most items, there was no consistent change in interrater agreement between the preoperative and postoperative assessment periods or during the postoperative assessment period. If there was decreased agreement, it tended to be within the first postoperative year. For the two items on activity, there was reduced agreement at the six-month assessment, which may have contributed to the poor overall agreement for these items. For fifteen of the sixteen items, the strength of agreement between the patient and doctor responses was reduced when the assessor was a consultant rather than a trainee. This reduced agreement was significant (p < 0.01) for the two items on activity, the two items on pain, and the items on walking and limp. As noted earlier, interrater agreement was improved when patients reported severe pain. However, when patients reported severe pain, a disproportionately larger number of assessments were undertaken by the trainee rather than the consultant. These data were removed to partly control for differences in pain severity between the assessor strata, and the remaining data were reanalyzed to determine the effect of assessor type on agreement. In contrast with the earlier analysis, this analysis showed that doctor-patient agreement was significantly reduced when the assessor was a consultant rather than a trainee for only one item, namely, occurrence of hip pain (p = 0.009). Fig. 2 Illustration depicting the direction of disagreement between patient and doctor responses to the sixteen items of the total hip arthroplasty clinical evaluation questionnaire.

6 1750 Fig. 3 Illustration depicting the effect of the presence of other joint or heath problems on agreement between patient and doctor responses to items of the total hip arthroplasty clinical evaluation questionnaire. Discussion e used questionnaires designed to evaluate patient health Woutcomes after total hip arthroplasty to examine the hypothesis that there are important differences between patientderived and doctor-derived data. The hypothesis was at least partially refuted because the interrater agreement was substantial or better for twelve of the sixteen questionnaire items. For ten of these items, there was no tendency for either the doctor or the patient consistently to report more favorable responses, leading to the conclusion that for these items there were no important differences between patient-derived and doctor-derived data. These results support those of previous studies that have shown that patients managed with total hip arthroplasty can reliably assess their health status and the outcomes of surgery 6,16. Only one item in the questionnaire reflected almost perfect agreement between patient and doctor responses. This finding is not unexpected. Both interview-derived and questionnaire-derived data may be susceptible to numerous factors that affect reliability, including lack of knowledge, ignorance of other conditions, unwillingness to answer, answering to please the interviewer, forgetfulness, respect for the interviewer, restricted range of responses, different perceptions of health and illness, and response bias 17. In the questionnaire used in the present study as in many other total hip arthroplasty clinical evaluation instruments hip pain, function, and gait are described by selecting one of a number of ordinal responses, such as none, slight, moderate, or severe, which may be supplemented with descriptive phrases. These phrases may, however, decrease the distinction between categories by generating ambiguities, contradictions, or exceptions 18. With no objective criteria, it is therefore unlikely that perfect agreement between raters will be achieved 15. Furthermore, patients completed the questionnaire independently before seeing the doctor. In contrast, the doctor completed the questionnaire during a clinical interview with the patient and therefore may have elicited information that the patient may not have considered when answering the questionnaire. In the present study, the age of the patient, the degree of patient-reported pain, the type of procedure, the type of assessor, and the presence of comorbid joint or health problems significantly affected the extent of interrater agreement, whereas the gender of the patient and the assessment period did not. Previous studies have similarly shown that patients in younger age-groups have better agreement with doctors than do patients in older age-groups 19,20. The greater disagreement between older patients responses and those of their doctor may arise for a number of reasons. First, the difficulty that some older patients experience in answering questions independently may be alleviated during consultation with the doctor, resulting in a different response to the same question. For example, these patients may not see the relevance of some of the items, particularly those on work and activity. Furthermore, older patients are more likely to be affected by other joint and health problems, which have been shown to reduce agreement.

7 1751 Nevertheless, some important differences between patient-derived and doctor-derived data were identified. Specifically, in 17% (505) of the assessments, the doctor underrated the patient s pain. Because the primary indications for total hip arthroplasty include joint pain and the functional restrictions caused by this pain 7,21,22, an accurate evaluation of hip pain as well as other sources of pain is important, not only during the initial assessment but also when reviewing the outcomes of total hip arthroplasty. While bias has been used to explain disagreement in the assessment of pain and other clinical variables 15,19,20,23,24, this possibility remains a matter of conjecture as there is no objective so-called gold standard measure of outcome with which both patient and doctor responses may be compared. While it has been shown that pain can be recorded independently of the surgeon 25, it has also been reported that comorbidity, either from other affected joints or from other health problems, may confound assessments of patient pain and function 7,21,26. The results of the present study indicate that patient-doctor agreement for the two pain items was significantly influenced by the presence of other conditions (p = 0.001) and that the item on the degree of pain was influenced by whether the patient had undergone a revision total hip arthroplasty (p = 0.006). When there was disagreement, the patient often reported a greater degree and frequency of hip pain compared with those reported by the doctor. It is not uncommon for patients with back or knee problems to experience referred pain in the hip or thigh. These results suggest that some patients may confuse referred pain with hip or thigh pain, whereas doctors may be better able to differentiate between the two types of pain. This possibility highlights the need for direct physician involvement in the evaluation of pain to establish its importance when patients report substantial pain. A possible limitation of the present study is that the categories that were used to grade the reliability coefficient of agreement may be too lenient. It has been suggested that this factor may lead to premature acceptance of the precision of outcome measures 27. That the data set included repeated observations of outcome parameters for the same individuals over a period of time may be considered a further limitation. However, it was considered appropriate to include all available data in order to attain the sample sizes that were required for stratified analyses of the factors influencing agreement. Multivariate longitudinal analysis would be more appropriate for this dataset, but to the best of our knowledge longitudinal data models for measures of agreement do not yet exist. The motivation for this study was the perceived need to reduce the costs associated with monitoring outcomes following total hip arthroplasty. If patients could be shown to reliably assess the severity of hip symptoms and outcomes following total hip arthroplasty, it follows that surgeon input into completing questionnaires, and hence cost, could be reduced. Furthermore, a patient-completed clinical evaluation questionnaire provides the opportunity to undertake some Fig. 4 Illustration depicting the effect of the degree of patient-reported hip pain on agreement between patient and doctor responses to the clinical items of the total hip arthroplasty clinical evaluation questionnaire.

8 1752 follow-up assessments in the home setting rather than in the outpatient clinic, particularly for patients who have not had problems following surgery. Elderly patients, patients with long-term follow-up, and patients who live some distance from the hospital have been shown to be more inclined to respond to mailed questionnaires than to attend an outpatient clinic 28. Electronically transmitted questionnaires also may gain favor in outcomes-assessment programs in the future as access to these tools increases. These alternative approaches could reduce the cost of maintaining an outcomes program but will require further validation. The present study demonstrated that patients were able to provide an assessment of the severity of their symptoms and the results of total hip arthroplasty that was reasonably consistent with what their doctors reported. However, agreement was reduced for patients who reported mild and moderate pain, and agreement about pain was further compromised when the patient had comorbid joint or health problems. For these reasons, and because pain may indicate problems such as infection or loosening, a doctor s assessment of hip pain should also be included when the patient reports pain or has comorbid problems. NOTE: The authors thank Ms. K. Willson for statistical programming advice, Mrs. K. Costi for assistance with data retrieval, and Mr. J. Roper and Ms. D. Liebert for providing hospital outpatient attendance data. Margaret A. McGee, BSc, MPH Donald W. Howie, PhD, MBBS, FRACS Oksana T. Holubowycz, PhD, MPH, BA Orthopedic Outcomes Unit, Department of Orthopedics and Trauma, L4 Bice Building, Royal Adelaide Hospital and University of Adelaide, North Terrace, Adelaide, South Australia 5000, Australia. address for M.A. McGee: mmcgee@mail.rah.sa.gov.au Philip Ryan, MBBS, BSc, FAFPHM John R. Moss, MSocSci, BEc, MBBS, FCHSE Department of Public Health, University of Adelaide, Adelaide, South Australia, Australia In support of their research or preparation of this manuscript, one or more of the authors received grants or outside funding from Australian Medicare Initiative (Government), Royal Adelaide Hospital, Adelaide University, Adelaide Bone and Joint Research Foundation. None of the authors received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, educational institution, or other charitable or nonprofit organization with which the authors are affiliated or associated. References 1. Harris WE. Traumatic arthritis of the hip after dislocation and acetabular fractures: treatment by mold arthroplasty. An end-result study using a new method of result evaluation. J Bone Joint Surg Am. 1969;51: Charnley J. The long-term results of low-friction arthroplasty of the hip performed as a primary intervention. J Bone Joint Surg Br. 1972;54: Wilson PD Jr, Amstutz HC, Czerniecki A, Salvati EA, Mendes DG. Total hip replacement with fixation by acrylic cement. A preliminary study of 100 consecutive McKee-Farrar prosthetic replacements. J Bone Joint Surg Am. 1972; 54: Johanson NA, Charlson ME, Szatrowski TP, Ranawat CS. A self-administered hip-rating questionnaire for the assessment of outcome after total hip replacement. J Bone Joint Surg Am. 1992;74: Gogia PP, Christensen CM, Schmidt C. Total hip replacement in patients with osteoarthritis of the hip: improvement in pain and functional status. Orthopedics. 1994;17: Dawson J, Fitzpatrick R, Carr A, Murray D. Questionnaire on the perceptions of patients about total hip replacement. J Bone Joint Surg Br. 1996;78: Johnston RC, Fitzgerald RH Jr, Harris WH, Poss R, Muller ME, Sledge CB. Clinical and radiographical evaluation of total hip replacements. A standard system of terminology for reporting results. J Bone Joint Surg Am. 1990;72: Galante J. Evaluation of results of total hip replacement [editorial]. J Bone Joint Surg Am. 1990;72: Cohen JA. A coefficient of agreement for nominal scales. Educat Psychol Measure. 1972;20: Cohen J. Weighted kappa. Nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull. 1968;70: Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York: Wiley; Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33: SAS Institute. SAS/STAT Software: changes and enhancements through release Cary, NC: SAS Institute; p Donner A. Sample size requirements for the comparison of two or more coefficients of inter-observer agreement. Stat Med. 1998;17: Lieberman JR, Dorey F, Shekelle P, Schumacher L, Thomas BJ, Kilgus DJ, Finerman GA. Differences between patients and physicians evaluations of outcome after total hip arthroplasty. J Bone Joint Surg Am. 1996;78: Fitzpatrick R, Fletcher A, Gore S, Jones D, Spiegelhalter D, Cox D. Quality of life measures in health care. I: Applications and issues in assessment. BMJ. 1992;305: Bulstrode CJK. Outcome measures and their analysis. In: Pynsent PB, Fairbank JCT, Carr A, editors. Outcome measures in orthopaedics. Boston: Butterworth-Heinemann; p Wright JG, Feinstein AR. Improving the reliability of orthopaedic measurements. J Bone Joint Surg Br. 1992;74: Dolan P, Kind P. Inconsistency and health state variations. Soc Sci Med. 1996;42: Wyller TB, Sveen U, Bautz-Holter E. The Barthel ADL index one year after stroke: comparison between relatives and occupational therapist s scores. Age Ageing. 1995;24: Bryant MJ, Kernohan WG, Nixon JR, Mollan RA. A statistical analysis of hip scores. J Bone Joint Surg Br. 1993;75: Keller RB, Rudicel SA, Liang MH. An Instructional Course Lecture, American Academy of Orthopaedic Surgeons. Outcomes research in orthopaedics. J Bone Joint Surg Am. 1993;75: Hoher J, Bach T, Munster A, Bouillon B, Tiling T. Does the mode of data collection change results in a subjective knee score? Self-administration versus interview. Am J Sports Med. 1997;25: Stephens RJ, Hopwood P, Girling DJ, Machin D. Randomized trials with quality of life endpoints: are doctors ratings of patients physical symptoms interchangeable with patients self ratings? Qual Life Res. 1997;6: Britton AR, Murray DW, Bulstrode CJ, McPherson K, Denham RA. Pain levels after total hip replacement: their use as endpoints for survival analysis. JBone Joint Surg Br. 1997;79: Dawson J, Fitzpatrick R, Murray D, Carr A. The problem of noise in monitoring patient-based outcomes: generic, disease-specific and site-specific instruments for total hip replacement. J Health Serv Res Policy. 1996;1: Shrout PE. Measurement reliability and agreement in psychiatry. Stat Methods Med Res. 1998;7: Berry DJ, Kessler M, Morrey BF. Maintaining a hip registry for 25 years. Mayo Clinic experience. Clin Orthop. 1997;344:61-8.