This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Letters to the Editor: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when Letters to the Editor are posted
Right arrow Alert me if a correction is posted
Services
Right arrow E-mail this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My File Cabinet
Right arrow Download to citation manager
Right arrowReprints and Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by COYTE, P. C.
Right arrow Articles by WRIGHT, J. G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by COYTE, P. C.
Right arrow Articles by WRIGHT, J. G.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us   Add to Technorati  
What's this?
The Journal of Bone and Joint Surgery 81:773-82 (1999)
© 1999 The Journal of Bone and Joint Surgery, Inc.

Rates of Revision Knee Replacement in Ontario, Canada*

PETER C. COYTE, M.A., PH.D.{dagger}, GILLIAN HAWKER, M.D., M.SC., F.R.C.P.(C){ddagger}, RUTH CROXFORD, M.SC.§ and JAMES G. WRIGHT, M.D., M.P.H., F.R.C.S.(C)# TORONTO, ONTARIO, CANADA

Investigation performed at the University of Toronto, Toronto


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Background: The present study was designed to measure the longevity of knee replacements and to assess the determinants of revision knee replacements in order to enhance the potential for informed decision-making. Methods: Data on all hospitalizations for knee replacement that occurred in Ontario, Canada, between April 1, 1984, and March 31, 1991, were acquired. To calculate the rates of revision knee replacement, two algorithms were developed: one distinguished primary knee replacements from revision knee replacements, and the second linked revision knee replacements to primary knee replacements. The Kaplan-Meier method was used to assess survivorship (absence of a revision) for primary knee replacement. A proportional-hazards regression model was estimated to assess the role of independent variables on the survival of primary knee replacements. Results: During the period of the study, 7.0 percent (1301) of 18,530 knee replacements were classified as revisions. Significant differences were identified between hospitalizations for primary and revision knee replacements in terms of the patient and hospital characteristics. Patients who were more than fifty-five years old, lived in a rural area, or had a diagnosis of rheumatoid arthritis had a significantly (p < 0.05) longer duration before revision than did other patients. Primary knee replacements performed in a teaching or specialty hospital had a significantly (p < 0.05) shorter duration before revision than did those performed in a nonteaching hospital. The long-term rates of revision were uniformly low. Estimates of the proportion of knee replacements that would need to be revised within seven years ranged from a low of 4.3 percent, with use of the algorithm for the longest time to revision, to a high of 8.0 percent, with use of the algorithm for the shortest time to revision. Conclusions: Revision of a primary knee replacement was a rare event that depended on a patient's age, gender, and place of residence as well as on the hospital where the primary knee replacement was performed. Estimates of the rates of revision knee replacement after almost seven years ranged from a low of 4.3 percent to a high of 8.0 percent.


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Musculoskeletal disorders, including arthritis and rheumatism, affect a substantial segment of the population31,52 increase exponentially with age1,2,14,16,17,19,44, are a leading cause of permanent incapacity3,35,55, and represent an important economic burden to Canadians13. Arthritis causes pain, loss of function, a reduced quality of life, and extensive use of health-care resources3,21,30,33,35,39,40,55. When medical therapy fails to control the disease adequately, a knee replacement has been shown to be a cost-effective treatment resulting in relief of pain and enhanced function at an affordable cost to society23,26,34,50,55.

Although the rate of knee replacement has increased exponentially in recent years11,12,28, one crucial area of clinical uncertainty concerns the likelihood of a subsequent major operation on the knee with the primary replacement. Indeed, studies on the outcome of knee replacements generally have been uncontrolled (most are case series) and have involved small samples. In addition, the type of prosthesis, the perioperative management, and the duration of follow-up have varied, and the studies have lacked specified measures of comorbidity4,5,18,22,25,36,41,45. More importantly, these series have reflected the outcomes of knee-replacement procedures performed in specialized tertiary-care centers. In a recent study, only 16.4 percent (196) of 1193 knee replacements in randomly selected patients in the United States and 44.9 percent (193) of 430 knee replacements in randomly selected patients in Ontario, Canada, were performed in a tertiary-care center9. Consequently, the results in the literature cannot be generalized with confidence to other patient populations or to community settings.

Despite a recent meta-analysis4 of 130 studies, involving 9879 patients, that demonstrated a 3.8 percent rate of revision four years after primary knee replacement, both referring physicians and orthopaedic surgeons have widely divergent opinions of the rate of revision10,54,57. In two studies by three of us and colleagues, 205 orthopaedic surgeons, sixty-six rheumatologists, and ninety-nine family practitioners who had managed patients with severe osteoarthritis of the knee during the previous year were surveyed to determine their estimates of the percentage of patients who would need a revision knee replacement10,57. The orthopaedic surgeons estimated median rates of revision of 1, 5, and 12 percent at one, five, and ten years after primary knee replacement, respectively (Table I). All clinicians agreed that the rates of revision knee replacement increased as the time since the initial operation increased10,57. Interestingly, the referring physicians (rheumatologists and family physicians) consistently estimated higher rates of revision knee replacement than did the orthopaedic surgeons, and the variation in opinion was lower among the orthopaedic surgeons than among the referring physicians10,57 (Table I).


View this table:
[in this window]
[in a new window]
  		TABLE I ESTIMATED RATES OF THE PERCENTAGES OF PATIENTS WHO WOULD NEED A REVISION AT ONE, FIVE, AND TEN YEARS AFTER A PRIMARY KNEE ARTHROPLASTY

 
In most studies that have assessed the rate of revision knee replacement, the results have been based on small samples of patients in tertiary-care settings and sometimes have had well controlled and costly prospective cohort designs, leading to results that have limited generalizability to community settings4. In contrast, population-based health-services research is concerned with the identification and assessment of patterns of practice for specific conditions in order to improve the effectiveness and appropriateness of care practices in the community20,43,49. In the present study, we adopted this population-based approach and reviewed the data on more than 18,000 hospitalizations for knee replacement operations. Specifically, the purpose of our study was to measure and assess the determinants of the rate of revision knee replacements in Ontario, Canada, in order to enhance the potential for informed clinical decision-making with respect to knee-replacement operations.

Ontario, which is the most populous Canadian province and has more than eleven million residents, provides a unique setting for this study for three reasons. First, all hospitals in Ontario are required to report all operative procedures to a central registry. Second, Ontario residents have universal and comprehensive public health insurance for all medically necessary services, without financial impediments to utilization. Third, because supplementary private health insurance for publicly insured services is prohibited under the Canada Health Act, knee-replacement operations are available to all residents of Ontario on equal financial terms and conditions. Thus, unlike studies from the United States, where data on patients who are ineligible for Medicare are difficult to comprehensively retrieve or where financial impediments may limit utilization, the data reported in our study capture the entire set of knee-replacement operations performed on Ontario residents.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Data on all hospital discharges between fiscal years 1984 and 1990 for patients who had a knee replacement, including hemiarthroplasty, were acquired from the Canadian Institute for Health Information Abstract Master File, held by the Ontario Ministry of Health. Knee replacements were identified by Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures code51 93.41 (total knee replacement: primary or revision) in any procedure field on the hospital discharge record, thereby yielding 19,674 hospitalizations for a knee-replacement procedure.

Hospitalizations were excluded if the patient was not a resident of Ontario, if pertinent data (such as date of birth or place of residence) were missing, or if a knee-replacement procedure either was not performed or was miscoded, as evidenced by a procedure performed in a nonacute-care facility or by a discharge to home with self-care within three days after the procedure. (In Ontario, during the period covered by this study, patients were never discharged home with self-care within three days after knee replacement.) Application of these criteria resulted in an analysis file with data on 18,530 hospitalizations for knee replacement.

Classification of Primary and Revision Knee Replacements
The absence of procedure codes to distinguish a primary knee replacement from a revision knee replacement within the hospital discharge database required the development of a classification algorithm. Revisions were identified on the basis of the simultaneous occurrence of Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures code51 93.41 (total knee replacement: primary or revision) in any procedure field in the record and any one of the following ICD-9 (International Classification of Diseases—Ninth Revision56) diagnostic codes in any diagnostic field in the record: 996.4 (mechanical complication of internal orthopaedic device, implant and graft), 996.6 (infection and inflammatory reaction due to internal prosthetic device, implant and graft), or 996.7 (other complications of internal prosthetic device, implant and graft). All other knee replacements were classified as primary. Although this is the best possible algorithm, given the limitations of the data, it is possible that some conversion total knee replacements (for example, a failed internal fixation device that was subsequently converted to a total knee replacement) were erroneously labeled as revisions. Such an error was possible only for patients who also had an earlier primary knee replacement on the contralateral knee recorded in the database. (This is because an apparent revision operation—as identified by the procedure code for a total knee replacement in conjunction with a diagnostic code indicating a complication associated with a prosthetic device—could be used in the survival analysis only if the corresponding primary knee replacement was also found in the database. Without the date of a primary procedure, the time to revision of the original prosthesis could not be calculated.)

The Ontario Health Insurance physician fee service claims, which distinguish between primary and revision knee replacements, were used to assess the sensitivity and specificity of the algorithm for identification of revision knee replacement. Although insurance data were cross-sectionally linked to hospital discharge data, difficulties with the longitudinal linkage of the insurance data, and the absence of variables with which to assess survival of primary knee replacements, restricted the use of the insurance data in the estimation of the rates of revision knee replacements. Each hospitalization for knee replacement during the seven-year study period was cross-sectionally linked to the physician fee service claims captured in the detailed insurance file by use of the health number, birth date, and gender of the patient. Only services performed within fourteen days after the date of the inpatient procedure, as recorded in the hospital discharge record, were examined.

The so-called gold standard for the classification of revision knee replacements in the insurance claims data was the occurrence of particular fee service claims. Specifically, revision knee replacements were identified on the basis of the occurrence of fee service code R244A (revision total knee arthroplasty) or the simultaneous occurrence of E564A (revision of arthroplasty) and any one of the following fee service codes: R248A (total knee replacement with take-down fusion), R441A (total replacement—both compartments), R482A (hemiarthroplasty—single component), or R483A (hemiarthroplasty—double component).

The algorithm for identification of revision knee replacement achieved a sensitivity of 77.7 percent and a specificity of 97.6 percent, implying that the probability that the algorithm would correctly identify a revision knee replacement was 77.7 percent and the probability that it would correctly identify a primary knee replacement was 97.6 percent (Table II). A similar revision algorithm used on data from the United States achieved a sensitivity of 87.2 percent and a specificity of 99.0 percent24.


View this table:
[in this window]
[in a new window]
  		TABLE II SENSITIVITY AND SPECIFICITY OF THE ALGORITHM FOR REVISION KNEE REPLACEMENT*

 

Linkage of Revision Knee Replacement to Primary Knee Replacement
An additional difficulty in the derivation of the rates of revision knee replacement is the absence of coding for laterality (left or right side) in the hospital database. Without such codes, it is impossible to definitively link a revision knee replacement to a primary knee replacement on the same knee. To address this issue, we identified all potential primary and revision knee-replacement combinations during the seven-year study period and classified potential procedural linkages accordingly. For example, if we had classified the first two knee replacements in an individual who had had three knee replacements as primary replacements, referred to as P1 and P2, then the third was classified as a revision, referred to as R. For this individual, the shortest time to revision was the interval between the second primary procedure and the revision (P2, R), and the longest time to a revision was the interval between the first primary knee replacement and the revision (P1, R). The former was assigned to the shortest-time data set, whereas the latter was assigned to the longest-time data set.

For all observed patterns of hospitalization, the linkage algorithm produces two possible interpretations: one that interprets the available data on knee replacements in the best possible light with respect to revision and one that produces a worst-case scenario. Some patterns of hospitalization are also consistent with an interpretation somewhere between the two extremes. Thus, the estimated rates of revision knee replacement derived in the present study represent upper and lower-bound estimates of true rates of revision knee replacement.

Hospitalizations for primary knee replacement and those for revision knee replacement were compared with respect to the patient's age, gender, and place of residence; the type of arthritis; the number of diagnostic codes; the presence of comorbid conditions; the teaching status of the hospital; the annual volume of knee replacements at the hospital; the duration of hospitalization; the occurrence of bilateral knee replacement; and the discharge destination.

In an effort to control for differences in comorbidities among patients, the Patient Management Category software58, developed at the Pittsburgh Research Institute, was adapted for use with the ICD-9 diagnostic codes56 and the procedure codes of the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures51. Seven levels of severity were generated by the Patient Management Category software. Most patients had level-4 severity, indicating no (or limited) comorbidity. Level 1, 2, or 3 was assigned when a patient had either an unspecified type of arthritis or miscoded data; 1.8 percent (310) of 17,229 primary knee replacements and 0.7 percent (nine) of 1301 revision knee replacements were performed in patients who were assigned such a level. Because the levels of severity were based on diagnostic and procedure codes, the absence of a specific type of arthritis resulted in the assignment of a lower severity level. Thus, after consultation with the developers of the software, the patients who had level-1, 2, or 3 severity were grouped in an unclassified category. Patients with level-6 or 7 severity were also grouped together, as there were few observations of level 7. Level-4 severity was used as the reference category. Although the severity-level software program has not been used previously to predict revision knee replacements, it has been used to predict other outcomes37.

To further assess the comorbidity of the patients during the initial hospitalization for a knee-replacement operation, the Charlson index6, a medical record-based comorbidity index, was adapted for use with the ICD-9 diagnostic codes56 and the procedure codes of the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures39,47,51. Although this index originally was designed to predict the relative risk of one-year mortality on the basis of data in medical records, it has been used to predict other outcomes in claims data15,37,47,48. As very few patients who had a joint replacement had a comorbidity index of more than 1, only three comorbidity categories—0, 1, and 2 or more—were used.

Knee replacements are performed for an array of different diagnoses, but most records for primary knee replacements indicate a diagnosis of arthritis. A classification of rheumatoid arthritis was given if there was an ICD-9 diagnostic code56 in the range of 714.0 through 714.9 (rheumatoid arthritis and other inflammatory polyarthropathies) in any diagnostic field in the hospital record. Traumatic arthritis was designated if there was a diagnostic code of 716.1 (traumatic arthropathy), whereas osteoarthritis was indicated if the diagnostic code was in the range of 715.0 through 715.9 (osteoarthrosis and allied disorders). All other diagnoses were assigned to a miscellaneous category. (Because of the nature of the classification, if the hospital record contained a diagnostic code for both rheumatoid arthritis and osteoarthritis, the diagnosis was classified as rheumatoid arthritis.)

Hospitals were classified as teaching (a member of the Ontario Council of Teaching Hospitals) or nonteaching. The sole specialty hospital for the treatment of arthritis, the Orthopaedic and Arthritic Hospital, was classified as a teaching hospital.

Univariate analyses to test for significant differences between hospitalizations for primary and revision knee replacements were conducted before multivariate statistical modeling. A chi-square test was used for categorical descriptive variables, such as gender, type of arthritis, and type of hospital, and the Wilcoxon rank-sum test was used for continuous variables.

A Cox proportional-hazards regression model7,8,53 was specified to assess survivorship (the absence of a revision) for primary knee replacements after controlling for patient and hospital characteristics. Survival analysis was applied to all primary knee replacements, irrespective of when the primary knee replacement was performed and the duration of the follow-up27. A revision knee replacement was selected as the end point and was used in the calculation of the cumulative rate of survival. Observations were censored either by the patient's date of death (if an inpatient death was identified in the hospital discharge record) or by the day after the last insurance fee service claim, including those for follow-up visits, was submitted for the patient. (This was the latest date on which the patient could be presumed to be both alive and living in Ontario; if the patient had died or had moved from Ontario, subsequent revisions could not be observed and the survival time of the knee replacement was censored.) The data supported the assumption that the rates of revision were constant throughout the study period. Relative risk was reported for each independent variable along with its associated p value. Relative risk was interpreted as a multiplier yielding the constant proportional increase (or decrease) in the hazard of a revision knee replacement relative to the baseline. The Kaplan-Meier method27 was used to generate survival curves for primary knee replacement that depicted, with use of a step function, the probability of survival of primary knee replacements at various intervals during the seven-year study period. The probabilities of survival of primary knee replacements and the associated 95 percent confidence intervals for both the shortest time to revision and the longest time to revision were determined. The log-rank test was used to compare the survival curves associated with distinct patient populations38.

Statistical analysis was performed on a Silicon Graphics 4D/220 minicomputer (Silicon Graphics, Mountain View, California) with use of SAS software (version 6.07; SAS Institute, Cary, North Carolina) and S-PLUS software (version 3.3; MathSoft, Seattle, Washington). All p values were two-tailed. A p value of 0.05 was considered significant, and 95 percent confidence intervals were provided for the survival analysis32. The study was approved by the Ethics Review Committee at the University of Toronto.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
A comparison of primary and revision knee replacements, after application of the algorithm for revision knee replacement to the 18,530 hospitalizations for knee replacement, demonstrated many differences with respect to the characteristics of the patients and the hospitals (Table III). During the seven-year study period, 7 percent (1301) of the hospitalizations for knee replacement were classified as hospitalizations for revision knee replacement. Significant differences were found between the primary and revision knee replacements with respect to the type of arthritis (p < 0.0001). The overwhelming majority of the primary knee replacements were performed because of osteoarthritis, whereas only 29.4 percent (382) of the 1301 revision knee replacements were performed because of osteoarthritis. (The two types of procedures were comparable with regard to the proportion of patients who had diagnoses of rheumatoid and traumatic arthritis; 11.5 percent [1981] of the 17,229 primary knee replacements and 10.3 percent [134] of the 1301 revision knee replacements were performed in patients who had a diagnosis of rheumatoid arthritis, and 0.4 percent [sixty-nine] of the primary replacements and 0.5 percent [seven] of the revision procedures were done in patients who had a diagnosis of traumatic arthritis. Although only 2.8 percent [482] of the hospitalization records of patients with primary knee replacement did not indicate any diagnosis of arthritis at all, 59.3 percent [772] of the hospitalization records on the revisions revealed no diagnosis of arthritis.) The number of diagnoses recorded on the discharge records of the patients who had primary knee replacement differed significantly from the number of those who had revision (p < 0.0001). Although 70.8 percent (12,198) of the primary knee replacements were done in patients who had one or two diagnoses, 71.1 percent (925) of the revision knee replacements were done in patients who had three diagnoses or more. (The greater number of diagnostic codes for the revision knee replacements reflects the finding of an ICD-9 diagnostic code56 996.4, 996.6, or 996.7 as well as additional diagnostic codes for those hospitalizations.) Revision knee replacements were significantly more likely than primary replacements to have been performed at a teaching or specialty hospital (p < 0.0001) and to have been performed at a hospital with a higher volume of knee replacements (p < 0.0001). They were also associated with a significantly longer duration of hospitalization (p < 0.0001). Moreover, revision knee replacements were significantly more likely than primary knee replacements to have been performed on patients from an urban community (p = 0.010) and on patients who were assigned a lower severity level58 (p = 0.007). No significant differences were detected between hospitalizations for primary and revision knee replacement with respect to the age (p = 0.222) or gender (p = 0.223) of the patient, the Charlson comorbidity index6 (p = 0.278), the discharge destination (p = 0.306), or the performance of bilateral knee replacement (p = 0.676).


View this table:
[in this window]
[in a new window]
  		TABLE III CHARACTERISTICS OF HOSPITALIZATIONS FOR PRIMARY AND REVISION KNEE REPLACEMENT IN FISCAL YEARS 1984 TO 1990

 
Cox proportional-hazards regression was applied to the survivorship analysis of primary knee replacements during the seven-year study period in Ontario. Regression models pertaining to both the determinants of the shortest time to revision and the determinants of the longest time to revision were obtained (Tables IV and V).


View this table:
[in this window]
[in a new window]
  		TABLE IV SURVIVORSHIP ANALYSIS FOR SHORTEST TIME TO REVISION KNEE REPLACEMENT*

 

View this table:
[in this window]
[in a new window]
  		TABLE V SURVIVORSHIP ANALYSIS FOR LONGEST TIME TO REVISION KNEE REPLACEMENT*

 
In the regression model for the shortest time to revision, patients who had a diagnosis of rheumatoid arthritis had a significantly (27.1 percent) lower rate of revision (p = 0.03) than did those who had other diagnoses, and the survival of a primary knee replacement (that is, the time until a revision knee replacement) was longer for patients who had rheumatoid arthritis (Table IV). In other words, at any given point in time, for every 100 patients without rheumatoid arthritis who have a revision, it is estimated that only seventy-three comparable patients with rheumatoid arthritis will have a revision. The age of the patient was analyzed as a continuous variable in the regression. There was a significant inverted-u-shaped relationship between the age at the time of the primary knee replacement and the survival of the prosthesis. (A simultaneous test of the significance of age and age-squared yielded a p value of less than 0.0001.) The rate of revision knee replacement was greatest and the survival of the replacement was shortest for patients who were fifty-five years old. Patients who were more than fifty-five years old at the time of the primary knee replacement had a lower rate of revision and a longer duration until revision than did younger patients. By the age of sixty-five, the risk of revision had decreased to 85 percent of the risk at the age of fifty-five, and by the age of seventy-five the risk had decreased to 70 percent of that at the age of sixty-five (that is, 60 percent of the rate at the age of fifty-five). In other words, for every 100 revisions performed on fifty-five-year-old patients, only eighty-five would be performed on comparable sixty-five-year-old patients and only sixty would be performed on comparable seventy-five-year-old patients. Primary knee replacements performed in teaching or specialty hospitals had a significantly (24 percent) higher rate of revision (p = 0.02) and a shorter survival time than did comparable knee replacements performed in nonteaching hospitals. Residents of urban communities had a significantly (31.5 percent) higher rate of revision (p = 0.004) than did patients from rural communities. A larger number of diagnoses at the time of the primary knee replacement was significantly associated with a higher rate of revision (p = 0.0004) and a shorter time to revision. Each additional diagnosis increased the risk of a revision by 9.9 percent. No other variables, including gender, severity58, and comorbidity (the Charlson index6), were found to be significant (p = 0.05), with the numbers available.

A similar analysis of survivorship applied to the longest time to revision revealed only three significant variables (Table V). Primary knee replacements performed in teaching or specialty hospitals had a significantly higher rate of revision (p = 0.02) and a shorter time until revision than did primary knee replacements performed in nonteaching hospitals. Residents of urban communities had a significantly higher rate of revision (p = 0.01) than did patients from rural communities. Bilateral knee replacement was associated with a significantly higher rate of revision (p = 0.0008) and a shorter time until revision than was unilateral knee replacement.

The rate of revision knee replacements was low. With use of the algorithm for the longest time to revision, the analysis of 17,229 procedures revealed that 0.02 percent were followed by a revision operation within 500 days. The algorithm for the shortest time to revision revealed a rate of revision within 500 days of 1.6 percent (Table VI). Even at 2500 days (almost seven years) after the primary knee replacement, revision rates ranged from a low of 4.3 percent for the longest time to revision to a high of 8.0 percent for the shortest time to revision (Table VI). There was a gradual increase in the rate of revision knee replacement associated with the time since the primary knee replacement.


View this table:
[in this window]
[in a new window]
  		TABLE VI ESTIMATED PERCENTAGE OF PRIMARY KNEE REPLACEMENTS REVISED POSTOPERATIVELY*

 


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
The present study was designed to assess the determinants of the rates of revision knee replacement in order to enhance the potential for informed clinical decision-making. A credible assessment of the likelihood that a revision operation will be needed and the duration of time that a prosthesis is likely to last and how these factors vary with respect to individual patient-related characteristics is an important component of a patient's ability to make an informed decision about a knee-replacement operation. Our principal finding was that the rate of revision knee replacement was low, an observation that is consistent with that reported in a recent meta-analysis4, which demonstrated a 3.8 percent rate of revision four years after primary knee replacements in 9879 patients. In our study, rates of revision at four years, estimated with use of the Kaplan-Meier method27, ranged from a low of 1.1 percent (95 percent confidence interval, 0.9 to 1.4 percent), when the algorithm to determine the longest time to revision was used, to a high of 4.7 percent (95 percent confidence interval, 4.2 to 5.2 percent), when the algorithm for the shortest time to revision was used (Table VI). Moreover, despite the wide variation in clinical perceptions regarding the effectiveness of knee replacement reported in the literature10,54,57, the median rates of revision knee replacement reported by orthopaedic surgeons were encompassed by our lower and upper-bound estimates. In contrast, the median rates of revision knee replacement reported by rheumatologists and family physicians were systematically higher than our estimates10,57 (Table I). Thus, orthopaedic surgeons have widely different estimates of the rate of revision, and referring physicians generally have pessimistically inaccurate opinions. Dissemination of our estimated rates of revision may reduce variations in the opinions of physicians and thereby decrease regional variations in the rates of knee-replacement procedures.

The present study demonstrated that an array of factors, including the characteristics of the patient, the clinician, and the hospital, influence the rates of revision knee replacement. Patients and doctors may find this information about revision rates useful when they are deciding whether to proceed with a knee replacement. Furthermore, determining the appropriate indications and contraindications for referral and operative intervention for knee replacement may be helpful in reducing subsequent rates of revision. Patients who are at high risk for a subsequent revision may choose to forego the operation, whereas those who are at low risk may be reassured.

Several important patient and hospital characteristics were found to be the main determinants of the survival of primary knee replacements. Specifically, primary knee replacements in older patients survived for a significantly longer period than did such replacements in younger patients (p < 0.0001). These findings were consistent with those reported elsewhere in the literature42,46, and they may reflect the additional demands placed on the knee prosthesis by younger patients but they may also be due to more severe disease, such as loss of bone density, at the time of the primary knee replacement. Similarly, the finding that patients who have rheumatoid arthritis at the time of the primary knee replacement were less likely to have a revision than were patients who had other types of arthritis, such as osteoarthritis and traumatic arthritis, was also consistent with results from other studies42.

Residents of urban communities were found to have a significantly higher rate of revision knee replacement and, hence, a shorter interval to revision, than did residents of rural communities (p = 0.004 in the proportional-hazards analysis for the shortest time to revision; p = 0.0096 in the analysis for the longest time to revision). This finding may reflect variations in the accessibility of specialty care in Ontario. Specifically, since almost 75 percent (969) of the 1301 revision knee replacements were performed at a teaching or specialty hospital located in one of five urban counties with a teaching health-science center, the proximity of urban residents to these specialty-care centers might account for the higher rate of revision.

The finding that primary knee replacements performed in a teaching or specialty hospital had a shorter time to revision than did equivalent procedures performed in other hospitals is in contrast to the findings reported elsewhere46. There are several possible explanations for this finding in Ontario. First, the shorter time to revision might be attributed to the role of resident physicians in the performance of primary knee replacements at teaching centers. Second, because most patients who receive a primary knee replacement in a teaching health-science center are also inhabitants of the community where the center is located, their proximity to specialty care might increase the likelihood that they will be referred for consideration for revision knee replacement. Third, patients referred to teaching and specialty hospitals for primary knee replacement may be systematically different with regard to an array of characteristics, including the severity of the disease, the complexity of the case, and the preference for operative rather than medical therapy, than patients referred to other hospitals. Such unmeasured or imperfectly measured characteristics may account for the differences in the rates of revision.

There were a number of limitations associated with this study. First, the analysis was based on administrative data that did not include procedure codes to distinguish a primary knee replacement from a revision knee replacement. As a consequence, an algorithm for revision knee replacement was developed for this purpose. Given the number of primary knee replacements performed, the problem of false-positive results (primary knee replacement operations mistakenly classified as revisions, the rate of which was approximately 2.4 percent) was less important than that of false-negative results (revision procedures mistakenly classified as primary procedures, the rate of which was approximately 22.3 percent). Even though this large rate of false-negative results biased our rates of revision knee replacement downward, the finding that revision rates were low suggests that a more precise classification algorithm may have increased the rates of revision knee replacement by only one or two percentage points.

Second, the absence of information on the laterality of the knee replacement made the task of linking revision knee replacements to primary procedures difficult. Our solution was to develop lower and upper-bound estimates for rates of revision knee replacement. Such imprecision was a consequence of the lack of detail in the administrative data. Although there was an eightfold to twofold difference in the rates of revision knee replacement for as many as 2500 days after the primary knee replacement, such differences were small in absolute terms and never exceeded four percentage points.

Third, many important predictors of failure of the prosthesis, such as detailed information on the extent of obesity, functional demands placed on the knee, and the health status of the patient, were absent from the administrative data. The omission of relevant explanatory variables from the survival analysis is an example of a specification error, which may result in potentially biased coefficient estimates27. Indeed, this limitation of our analysis is a general limitation associated with the use of administrative data.

Fourth, the 17,229 primary knee-replacement procedures included in our study were performed on 14,746 patients. Of those patients, 12,263 (83.2 percent) had one primary knee replacement and 2483 had two. The lack of independence in some of the observations that were included in the survival analysis may have resulted in biased coefficient estimates. However, given the small proportion of individuals who had multiple knee-replacement procedures during the study period, these potential problems may be small.

In conclusion, although the survival of primary knee replacements was found to be shorter if the patient was younger, resided in an urban community, and had the replacement at a teaching or specialty hospital, revision knee replacement was shown to be a relatively rare event. These results support the findings in the literature, which have demonstrated the cost-effectiveness of knee replacement in individuals for whom medical therapy has failed23,26,34,50,55. Such results suggest that the beliefs reported by some rheumatologists and family physicians regarding unfavorable outcomes after knee replacement were unfounded. As such, a process of effective dissemination of information on the outcome of knee replacement might reduce the variance of clinical opinions, increase the propensity of rheumatologists and family physicians to refer patients for knee replacement, and thereby improve relief of pain and enhance function for many residents of Ontario.


   Footnotes
 
*Although none of the authors has received or will receive benefits for personal or professional use from a commercial party related directly or indirectly to the subject of this article, benefits have been or will be received but are directed solely to a research fund, foundation, educational institution, or other nonprofit organization with which one or more of the authors is associated. Funds were received in total or partial support of the research or clinical study presented in this article. The funding source was United States Agency for Health Care Policy and Research Grant 06432.

{dagger}Department of Health Administration and Institute for Policy Analysis, Second Floor, McMurrich Building, Faculty of Medicine, University of Toronto, Toronto, Ontario M5S 1A8, Canada. E-mail address for Dr. Coyte: coyte@chass.utoronto.ca. Please address requests for reprints to Dr. Coyte.

{ddagger}Division of Rheumatology and Clinical Epidemiology, Department of Medicine, Women's College Hospital, 76 Grenville Street, Toronto, Ontario M5S 1B2, Canada.

§Clinical Epidemiology Unit, Sunnybrook and Women's College Health Science Centre, G-Wing, 2075 Bayview Avenue, Toronto, Ontario M4N 3M5, Canada.

#Division of Orthopaedics, Hospital for Sick Children, 555 University Avenue, Toronto, Ontario M5G 1X8, Canada.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Badley, E. M.; Arnold, J.; and Webster, G.: Arthritis in Ontario: A Study of Intra-Provincial Variation. Toronto, Arthritis Community Research and Evaluation Unit, Working Paper 94-1, Feb. 1994.
  2. Badley, E. M.; Webster, G. K.; and Rasooly, I.: The impact of musculoskeletal disorders in the population: are they just aches and pains? Findings from the 1990 Ontario Health Survey. J. Rheumatol., 22: 733-739, 1995.[Medline]
  3. Bridges-Webb, C.; Britt, H.; Miles, D. A.; Neary, S.; Charles, J.; and Traynor, V.: Morbidity and treatment in general practice in Australia 1990–1991. Med. J. Australia, 157 (Supplement): 1-S56, 1992.
  4. Callahan, C. M.; Drake, B. G.; Heck, D. A.; and Dittus, R. S.: Patient outcomes following tricompartmental total knee replacement. A meta-analysis. J. Am. Med. Assn., 271: 1349-1357, 1994.[Abstract/Free Full Text]
  5. Callahan, C. M.; Drake, B. G.; Heck, D. A.; and Dittus, R. S.: Patient outcomes following unicompartmental or bicompartmental knee arthroplasty. A meta-analysis. J. Arthroplasty, 10: 141-150, 1995.[Medline]
  6. Charlson, M. E.; Pompei, P.; Ales, K. L.; and MacKenzie, C. R.: A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J. Chronic Dis., 40: 373-383, 1987.[Medline]
  7. Collett, D.: Modelling Survival Data in Medical Research, pp. 1-106, 149-197. New York, Chapman and Hall, 1994.
  8. Cox, D. R.: Regression models and life-tables. J. Roy. Statist. Soc., Series B., 34: 187-202, 1972.
  9. Coyte, P. C.; Wright, J. G.; Hawker, G. A.; Bombardier, C.; Dittus, R. S.; Paul, J. E.; Freund, D. A.; and Ho, E.: Waiting times for knee-replacement surgery in the United States and Ontario. New England J. Med., 331: 1068-1071, 1994.[Abstract/Free Full Text]
  10. Coyte, P. C.; Hawker, G.; Croxford, R.; Allard, C.; and Wright, J. G.: Variations in rheumatologists' and family physicians' perceptions of the indications for and outcomes of knee replacement surgery. J. Rheumatol., 23: 730-738, 1996.[Medline]
  11. Coyte, P. C.; Hawker, G.; and Wright, J. G.: Variations in knee replacement utilization rates and the supply of health professionals in Ontario, Canada. J. Rheumatol., 23: 1214-1220, 1996.[Medline]
  12. Coyte, P.; Wang, P. P.; Hawker, G.; and Wright, J. G.: The relationship between variations in knee replacement utilization rates and the reported prevalence of arthritis in Ontario, Canada. J. Rheumatol., 24: 2403-2412, 1997.[Medline]
  13. Coyte, P. C.; Asche, C. V.; Croxford, R.; and Chan, B.: The economic cost of musculoskeletal disorders in Canada. Arthrit. Care and Res., 11: 315-325, 1998.
  14. Cunningham, L. S., and Kelsey, J. L.: Epidemiology of musculoskeletal impairments and associated disability. Am. J. Pub. Health, 74: 574-579, 1984.[Abstract/Free Full Text]
  15. Deyo, R. A.; Cherkin, D. C.; and Ciol, M. A.: Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J. Clin. Epidemiol., 45: 613-619, 1992.[Medline]
  16. Dodge, H. J.; Mikkelsen, W. M.; and Duff, I. F.: Age-sex specific prevalence of radiographic abnormalities of the joints of the hands, wrists and cervical spine of adult residents of the Tecumseh, Michigan, Community Health Study Area, 1962–1965. J. Chronic Dis., 23: 151-159, 1970.[Medline]
  17. Felson, D. T.; Naimark, A.; Anderson, J.; Kazis, L.; Castelli, W.; and Meenan, R. F.: The prevalence of knee osteoarthritis in the elderly. The Framingham Osteoarthritis Study. Arthrit. and Rheumat., 30: 914-918, 1987.
  18. Goldberg, V. M.; Figgie, M. P.; Figgie, H. E., III; and Sobel, M.: The results of revision total knee arthroplasty. Clin. Orthop., 226: 86-92, 1988.
  19. Gordon, T.: Osteoarthritis in US adults. In Population Studies of the Rheumatic Diseases. Proceedings of the Third International Symposium, New York, N.Y., June 5–10, 1966, edited by P. H. Bennett and P. H. N. Wood. International Congress Series No. 148, pp. 391-397. Amsterdam, Excerpta Medica Foundation, 1968.
  20. Greenfield, S.: The state of outcome research: are we on target [editorial]?. New England J. Med., 320: 1142-1143, 1989.[Medline]
  21. Hadler, N. M.: Osteoarthritis as a public health problem. Clin. Rheumat. Dis., 11: 175-185, 1985.
  22. Hamilton, L. R.: UCI total knee replacement. A follow-up study. J. Bone and Joint Surg., 64-A: 740-744, June 1982.[Abstract/Free Full Text]
  23. Hawker, G.; Wright, J.; Coyte, P.; Paul, J.; Dittus, R.; Croxford, R.; Katz, B.; Bombardier, C.; Heck, D.; and Freund, D.: Health-related quality of life after knee replacement. Results of the Knee Replacement Patient Outcomes Research Team Study. J. Bone and Joint Surg., 80-A: 163-173, Feb. 1998.[Abstract/Free Full Text]
  24. Heck, D. A.; Melfi, C. A.; Mamlin, L. A.; Katz, B. P.; Arthur, D. S.; Dittus, R. S.; and Freund, D. A.: Revision rates after knee replacement in the United States. Med. Care, 36: 661-669, 1998.[Medline]
  25. Insall, J. N.; Hood, R. W.; Flawn, L. B.; and Sullivan, D. J.: The total condylar knee prosthesis in gonarthrosis. A five- to nine-year follow-up of the first one hundred consecutive replacements. J. Bone and Joint Surg., 65-A: 619-628, June 1983.[Abstract/Free Full Text]
  26. Jonsson, B., and Larsson, S.-E.: Functional improvement and costs of hip and knee arthroplasty in destructive rheumatoid arthritis. Scandinavian J. Rheumatol., 20: 351-357, 1991.[Medline]
  27. Kaplan, E. L., and Meier, P.: Nonparametric estimation from incomplete observations. J. Am. Statist. Assn., 53: 457-481, 1958.
  28. Katz, B. P.; Freund, D. A.; Heck, D. A.; Dittus, R. S.; Paul, J. E.; Wright, J.; Coyte, P.; Holleman, E.; and Hawker, G.: Demographic variation in the rate of knee replacement: a multi-year analysis. Health Serv. Res., 31: 125-140, 1996.[Medline]
  29. Kmenta, J.: Elements of Econometrics. Ed. 2, pp. 443-446. New York, Macmillan, 1986.
  30. Kramer, J. S.; Yelin, E. H.; and Epstein, W. V.: Social and economic impacts of four musculoskeletal conditions. A study using national community-based data. Arthrit. and Rheumat., 26: 901-907, 1983.
  31. Lee, P.: The economic impact of musculoskeletal disorders. Qual. Life Res., 3 (Supplement 1): 85-S91, 1994.
  32. Lettin, A. W. F.; Ware, H. S.; and Morris, R. W.: Survivorship analysis and confidence intervals. An assessment with reference to the Stanmore total knee replacement. J. Bone and Joint Surg., 73-B(5): 729-731, 1991.
  33. Liang, M. H.; Larson, M.; Thompson, M.; Eaton, H.; McNamara, E.; Katz, R.; and Taylor, J.: Costs and outcomes in rheumatoid arthritis and osteoarthritis. Arthrit. and Rheumat., 27: 522-529, 1984.
  34. Liang, M. H.; Cullen, K. E.; Larson, M. G.; Thompson, M. S.; Schwartz, J. A.; Fossel, A. H.; Roberts, W. N.; and Sledge, C. B.: Cost-effectiveness of total joint arthroplasty in osteoarthritis. Arthrit. and Rheumat., 29: 937-943, 1986.
  35. McAlindon, T. E.; Cooper, C.; Kirwan, J. R.; and Dieppe, P. A.: Knee pain and disability in the community. British J. Rheumatol., 31: 189-192, 1992.[Abstract/Free Full Text]
  36. Marmor, L.: Unicompartmental arthroplasty of the knee with a minimum ten-year follow-up period. Clin. Orthop., 228: 171-177, 1988.
  37. Melfi, C.; Helleman, E.; Arthur, D.; and Katz, B.: Selecting a patient characteristics index for the prediction of medical outcomes using administrative claim data. J. Clin. Epidemiol., 48: 917-926, 1995.[Medline]
  38. Peto, R., and Peto, J.: Asymptotically efficient rank invariant test procedures. J. Roy. Statist. Soc. Series A, 135: 185-206, 1972.
  39. Peyron, J. G.; and Altman, R. D.: The epidemiology of osteoarthritis. In Osteoarthritis: Diagnosis and Medical/Surgical Management. edited by R.W. Moskowitz, D. S. Howell, V. M. Goldberg, and H. J. Mankin. Ed. 2, pp. 15-37. Philadelphia, W. B. Saunders, 1992.
  40. Pincus, T.; Mitchell, J. M.; and Burkhauser, R. V.: Substantial work disability and earnings losses in individuals less than age 65 with osteoarthritis: comparisons with rheumatoid arthritis. J. Clin. Epidemiol., 42: 449-457, 1989.[Medline]
  41. Ranawat, C. S., and Boachie-Adjei, O.: Survivorship analysis and results of total condylar knee arthroplasty. Eight- to 11-year follow-up period. Clin. Orthop., 226: 6-13, 1988.
  42. Rand, J. A., and Ilstrup, D. M.: Survivorship analysis of total knee arthroplasty. Cumulative rates of survival of 9200 total knee arthroplasties. J. Bone and Joint Surg., 73-A: 397-409, March 1991.[Abstract/Free Full Text]
  43. Relman, A. S.: Assessment and accountability: the third revolution in medical care. New England J. Med., 319: 1220-1222, 1988.[Medline]
  44. Reynolds, D. L.; Chambers, L. W.; Badley, E. M.; Bennett, K. J.; Goldsmith, C. H.; Jamieson, E.; Torrance, G. W.; and Tugwell, P.: Physical disability among Canadians reporting musculoskeletal diseases. J. Rheumatol., 19: 1020-1030, 1992.[Medline]
  45. Riley, D., and Woodyard, J. E.: Long-term results of Geomedic total knee replacement. J. Bone and Joint Surg., 67-B(4): 548-550, 1985.
  46. Robertsson, O.; Knutson, K.; Lewold, S.; Goodman, S.; and Lidgren, L.: Knee arthroplasty in rheumatoid arthritis. A report from the Swedish Knee Arthroplasty Register on 4,381 primary operations 1985–1995. Acta Orthop. Scandinavica, 68: 545-553, 1997.[Medline]
  47. Romano, P. S.; Roos, L. L.; and Jollis, J. G.: Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J. Clin. Epidemiol., 46: 1075-1079, 1993.[Medline]
  48. Romano, P. S.; Roos, L. L.; and Jollis, J. G.: Response. Further evidence concerning the use of a clinical comorbidity index with ICD-9-CM administrative data. J. Clin. Epidemiol., 46: 1085-1090, 1993.
  49. Salive, M. E.; Mayfield, J. A.; and Weissman, N. W.: Patient outcomes research teams and the agency for health care policy and research. Health Serv. Res., 25: 697-708, 1990.[Medline]
  50. Scott, W. N.; Rubinstein, M.; and Scuderi, G.: Results after knee replacement with a posterior cruciate-substituting prosthesis. J. Bone and Joint Surg., 70-A: 1163-1173, Sept. 1988.[Abstract/Free Full Text]
  51. Statistics Canada: Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures. Ottawa, Statistics Canada, 1986.
  52. Statistics Canada: Health Status of Canadians: Report of the 1991 General Social Survey. Catalogue No. 11-612E, No. 8. Ottawa, Statistics Canada, 1994.
  53. Tibshirani, R.: A plain man's guide to the proportional hazards model. Clin. and Invest. Med., 5: 63-68, 1992.
  54. Tierney, W. M.; Fitzgerald, J. F.; Heck, D. A.; Kennedy, J. M.; Katz, B. P.; Melfi, C. A.; Dittus, R. S.; Allen, D. I.; Freund, D. A.; and the Knee Replacement Patient Outcomes Research Team: Tricompartmental knee replacement. A comparison of orthopaedic surgeons' self reported performance rates with surgical indications, contraindications, and expected outcomes. Clin. Orthop., 305: 209-217, 1994.
  55. Verbrugge, L. M.: Physical and social disability in adults. In Functional and Health Status Measures for Primary Care, pp. 31-57. Edited by H. H. Hibbard, P. A. Nutting, and M. L. Grady. Washington, D. C., United States Department of Health and Human Services, 1991.
  56. World Health Organization: Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death. Geneva, World Health Organization, 1977.
  57. Wright, J. G.; Coyte, P.; Hawker, G.; Bombardier, C.; Cooke, D.; Heck, D.; Dittus, R.; and Freund, D.: Variation in orthopaedic surgeons' perceptions of the indications for and outcomes of knee replacement. Canadian Med. Assn. J., 152: 687-697, 1995.[Abstract]
  58. Young, W. W.: Incorporating severity of illness and comorbidity in case-mix measurement. Health Care Financ. Rev., (Supplement): 23-31, 1984.

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us   Add to Technorati Technorati    What's this?


This article has been cited by other articles:


Home page
 JBJSHome page
R. L. Kane, K. J. Saleh, T. J. Wilt, and B. Bershadsky
The Functional Outcomes of Total Knee Arthroplasty
J. Bone Joint Surg. Am., August 1, 2005; 87(8): 1719 - 1724.
[Abstract] [Full Text] [PDF]

Home page
 BMJHome page
G. Holt and A G Gregori
Computer assisted knee arthroplasty is here today
BMJ, January 22, 2005; 330(7484): 197 - 197.
[Full Text]

Home page
 Ann Rheum DisHome page
M Weigl, F Angst, G Stucki, S Lehmann, and A Aeschlimann
Inpatient rehabilitation for hip or knee osteoarthritis: 2 year follow up study
Ann Rheum Dis, April 1, 2004; 63(4): 360 - 368.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Letters to the Editor: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when Letters to the Editor are posted
Right arrow Alert me if a correction is posted
Services
Right arrow E-mail this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My File Cabinet
Right arrow Download to citation manager
Right arrowReprints and Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by COYTE, P. C.
Right arrow Articles by WRIGHT, J. G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by COYTE, P. C.
Right arrow Articles by WRIGHT, J. G.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us   Add to Technorati  
What's this?