A retrospective cohort study was conducted with use of administrative health-care databases in Ontario, Canada (population, 12.4 million in 2004), and England (population, 50.1 million in 2004).
Data Sources
Ontario's hospitalization records were obtained from the Canadian Institute for Health Information (CIHI) Discharge Abstract Database, which contains detailed records of all hospital admissions, including day-surgery and inpatient admissions. England's hospitalization records were obtained from the Hospital Episode Statistics (HES) administrative databases, which include inpatient and day-surgery admissions, private patients treated in the National Health Service (NHS) hospitals, and independent-sector activity if commissioned by the NHS.
Index Knee Arthroscopy Cohorts
The hospitalization databases were used to construct four separate cohorts in Ontario and England. Each cohort consisted of individuals who had undergone their first knee arthroscopy procedure over a defined one-year time period. Patient accrual for the four cohorts began on April 1 of 1993, 1997, 2002, or 2004. Men and women twenty years of age or older who had received their first knee arthroscopy in the relevant HES or CIHI financial year were included. The patients from England were considered to have had a knee arthroscopy if they had an OPCS-4 (Classification of Surgical Operations and Procedures, 4th revision) code, in any of the operation-code fields, of W822 (resection of knee cartilage) or W852 (irrigation of knee joint). The side of the arthroscopy was also determined on the basis of the OPCS-4 code, with Z941 indicating a bilateral operation; Z942, a right-sided operation; and Z943, a left-sided operation. Patients were excluded if they had a diagnosis of cancer or trauma or had undergone a total knee replacement prior to the arthroscopy on the same side. To identify the Ontario patients, the CIHI records were searched for the presence of a procedure field containing a Canadian Classification of Diagnostic, Therapeutic and Surgical Procedures (CCP) code of 9285, indicating a knee arthroscopy. The ICD-10/CCI classification (The International Statistical Classification of Diseases and Related Health Problems, 10th Revision) was implemented in Ontario in 2002/03. It was not possible to determine the laterality of the operation or to distinguish a unicompartmental arthroplasty from a total knee replacement with use of the Ontario CIHI records.
Patient Diagnosis
The patient diagnosis—i.e., the diagnosis recorded as most responsible for the admission—was categorized as osteoarthritis/gonarthrosis (ICD-9 [715] and ICD-10 [M17]), internal derangement/dislocation of the knee (ICD-9 [7170 through 7175 and 8360 through 8362] and ICD-10 [M232, M233, and S832]), pain in joint (ICD-9 [7194] and ICD-10 [M255]), or "other."
Primary Total Knee Replacement Cohorts
We used the hospitalization databases to construct cohorts of individuals twenty years of age or older who underwent their first total knee replacement, in Ontario or England, in the HES or CIHI 2002 financial year (April 1, 2002, through March 31, 2003). Primary total knee replacements in England were identified by the OPCS-4 codes W40, W41, and W42 without a primary diagnosis indicating complications due to an internal prosthesis (ICD-9 [996.4, 996.6, or 996.7] and ICD-10 [T84 or T85]). The side of the primary total knee replacement was also determined by the OPCS-4 code, with Z941 indicating a bilateral operation; Z942, right-sided operation; and Z943, a left-sided operation. Primary total knee replacements in Ontario were identified by a procedure field containing the CCP code 93.41. Subsequent to 2002/2003, when the ICD-10/CCI classification was implemented, primary total knee replacements were identified with use of the rubric of 1.VG.53. In 2000, a new CCP code for revision total knee replacement was added (93.40); this was used to further identify and rule out revisions. The laterality of the operation could not be determined with use of the Ontario data. In both countries, total knee replacements performed for the treatment of cancer and fractures were excluded.
Statistical Methods
Descriptive statistics were used to examine the age and sex-standardized rates of knee arthroscopy per 100,000 population in 1993, 1997, 2002, and 2004 in Ontario and England, overall and by diagnosis, sex, age group (twenty to forty-four years, forty-five to sixty-four years, and sixty-five years and older), and socioeconomic status. In England, the Index of Multiple Deprivation (IMD) 2000 was used to estimate socioeconomic status. This index is based on a number of indicators covering a range of domains to create a single deprivation score for each ward. Only the income domain was used, to allow comparison with the Ontario data; income was split into quintiles, weighted according to ward population size. In Ontario, socioeconomic status was assessed on the basis of the median income of the enumeration area associated with the individual's residential postal code.
Cox regression analyses were conducted to examine the time to readmission for primary total knee replacement during a period of up to five years after the index knee arthroscopy. This was done for arthroscopic procedures performed in the 1993 HES or CIHI financial year (for which the five-year period of analysis was April 1, 1993, through March 31, 1999) and in the corresponding 1997 financial year (for which the five-year period of analysis was April 1, 1997, through March 31, 2003) to examine for differences over time. Patients treated with an arthroscopy were censored if they died or moved out of the province (Ontario cohort) during the follow-up period. Potential confounders that were controlled for in the analyses were age at the index knee arthroscopy (twenty to forty-four years, forty-five to sixty-four years, and sixty-five years and older), sex, and socioeconomic status (income quintiles). We identified the proportion of the patients who had undergone a primary total knee replacement in 2002 who had had an arthroscopy in the previous two years.
Age and Sex-Standardized Arthroscopy Rates
The age and sex-standardized rates of arthroscopy and the numbers of arthroscopic procedures performed in each diagnostic group in England and Ontario are shown in Figure 1. The arthroscopy rates were higher in Ontario than in England, and the pattern of change over time differed between the two countries. In England, the overall rate rose steadily over the twelve-year period, whereas the rates in Ontario peaked in 1993. Utilization of arthroscopy for treatment of internal derangement or dislocation of the knee increased steadily in both countries, whereas the rates of use for osteoarthritis decreased steadily in Ontario after 1993 and decreased slightly in England in 2004 in comparison with the peak rate of use in 2002.
Age, Sex, and Income
The total number of procedures overall and by age, sex, and income quintiles in each of the four years studied is shown in Table I. There was a marked change in the age profile of people undergoing arthroscopy, with fewer individuals under forty-five and more individuals over sixty-five having the procedure in each country. The proportion of arthroscopic procedures performed in women in Ontario remained fairly constant over time (at approximately 40%), whereas, in England, the percentage gradually increased from 30.2% in 1993 to 37.8% in 2004. Those in the lowest income quintile were least likely to undergo arthroscopy, whereas those in the two highest income quintiles were most likely to undergo this procedure. This gap widened between 1993 and 2004 in both countries.
Table II shows some of the relationships between arthroscopy rates and age, sex, and income in the two largest diagnostic groups (osteoarthritis and internal derangement or dislocation of the knee) in each country in 2004. As expected, arthroscopic procedures for treatment of internal derangement or dislocation of the knee were performed predominantly in younger men, whereas arthroscopic procedures for treatment of knee osteoarthritis were performed more frequently in older people and in equal numbers of men and women; these patterns were similar for Ontario and England. The differences based on socioeconomic status persisted within each of the diagnostic groups, with no changes over time.
Readmission for Primary Total Knee Replacement Following Knee Arthroscopy
The crude rates of readmission for primary total knee replacement in the five-year period following the arthroscopic procedures performed in 1993 and 1997 are shown in Table III. The rates were similar for the two countries, increasing from approximately thirteen readmissions per 1000 person-years in 1993 to approximately seventeen readmissions per 1000 person-years in 1997. The rate of readmission was highest among people with osteoarthritis and lowest for individuals with internal derangement or dislocation. Crude readmission rates were higher for women than men and higher for those with a higher socioeconomic status than for those with a lower status. Between 1993 and 1997, the crude rate of readmission (per 1000 person-years) for total knee replacement among people with osteoarthritis increased slightly (from 35.6 to 43.4 in England and from 45.5 to 49.3 in Ontario), as did the rate of readmission for individuals sixty-five years of age or older (from 54.4 to 61.9 in England and from 80.9 to 84.0 in Ontario).
Approximately one-fifth of individuals with osteoarthritis (16.0% in England and 19.6% in Ontario) received a total knee replacement within five years after an index knee arthroscopy performed in 1993. Marginally higher proportions of individuals with osteoarthritis and an index knee arthroscopy performed in 1997 received a total knee replacement within five years (19.1% in England and 20.8% in Ontario). It was also found that 5.2% of individuals in England and 7.0% in Ontario with osteoarthritis and an index knee arthroscopy done in 1993 had a primary total knee replacement within one year. These proportions were similar to that for the index arthroscopic procedures done in 1997 in England (4.8%) but lower than the proportion for the index arthroscopic procedures done in 1997 in Ontario (8.5%). Substantially fewer individuals (=5%) with other diagnoses received a total knee replacement within five years subsequent to the arthroscopy.
Predictors of Readmission for Primary Total Knee Replacement Following Arthroscopy
Unadjusted analyses showed the risk of readmission for primary total knee replacement within five years after an index knee arthroscopy to be higher among those with a diagnosis of osteoarthritis, "pain in joint," or "other" than it was for those with an internal derangement or dislocation. The risk of readmission was also higher for women than for men, and it increased with increasing age at the time of the index knee arthroscopy. Socioeconomic status was not seen to have an effect. Results were similar for England and Ontario, and for patients treated with the index arthroscopy in 1993 and those treated in 1997.
In the adjusted analyses, the risk of readmission for total knee replacement was found to be independently associated with a diagnosis of osteoarthritis, "pain in joint," or "other" compared with internal derangement or dislocation. It was also associated with female sex and increasing age (Table IV). The gradient of effect for age group was greater in England than in Ontario. A borderline significant effect was also seen for income level in England; those with the highest levels of income were more likely to be readmitted for total knee replacement than those with the lowest income, after adjustment for diagnosis, sex, and age.
Knee Arthroscopy Before Primary Total Knee Replacement
The study showed that 2.7% of those who received a primary total knee replacement in 2002 in England and 5.7% of those in Ontario had undergone knee arthroscopy within the prior year, and 7.2% and 9.7%, respectively, had undergone knee arthroscopy within the previous two years (Table V). Osteoarthritis was the reason for the total knee replacement in >90% of those treated with a total knee replacement within two years after arthroscopy. More than half of the individuals who underwent knee arthroscopy prior to total knee replacement were women. In general, the likelihood of prior arthroscopy increased with higher socioeconomic status and increasing age.
Our objectives were to examine patterns of use of knee arthroscopy for possible inequalities and overutilization in England and Ontario. In both countries, arthroscopy is performed mainly for the treatment of internal knee derangement or dislocation and osteoarthritis. However, patterns of utilization over the study period differed between the countries, with the rates rising in England but remaining relatively stable in Canada. A greater proportion of older people received an arthroscopy in 2004 than in the previous years studied. The rates of arthroscopy differed by sex and income in each country; these differences remained stable over time and appeared to be independent of diagnosis. Individuals who underwent arthroscopy for the treatment of osteoarthritis were more likely than those in whom the arthroscopy was performed for another diagnosis to have a total knee replacement within a few years; conversely, approximately 7% to 10% of those who underwent a total knee replacement for the treatment of osteoarthritis had undergone knee arthroscopy in the previous two years.
We believe that this is the first study to document differences in rates of arthroscopic surgery according to sex or socioeconomic status. Although osteoarthritis affects about twice as many women as men7, and many more women than men receive a total knee replacement, studies have suggested that there remains a bias in favor of men receiving the operation7,24. In this study, rates of arthroscopy for the treatment of osteoarthritis were similar for men and women in both countries. Whether this reflects appropriate use of this procedure, however, could not be determined. The differential provision of arthroscopy according to socioeconomic status is also striking and warrants further study. There are several potential explanations, including differences based on income in the rates of sports-related trauma causing internal knee derangement and/or in an individual's willingness to consult an orthopaedic surgeon or consider arthroscopy.
We also sought to explore whether arthroscopy is being overused. Our data cannot answer this question. Crude rates of readmission for primary total knee replacement within one or two years after the knee arthroscopy were low overall. However, as many as 8.5% of patients who underwent an arthroscopy for the treatment of osteoarthritis went on to have a total knee replacement within one year, and almost 10% of those who had a total knee replacement had undergone a knee arthroscopy in the previous two years. Whether the arthroscopic procedures delayed the need for total knee replacement or resulted in a period of substantial symptomatic relief is unclear.
In one previous study, the prevalence of primary total knee replacement following knee arthroscopy was examined in individuals with knee osteoarthritis who were fifty years of age or older16. Although the data source used to identify the patients with knee arthroscopy (the Ontario physician claims database) differed from ours, the results were similar to ours. Of patients who had undergone arthroscopic débridement between 1992 and 1996, 9.2% received a total knee replacement within a year and 18.4%, within three years. Consistent with our findings, the rates of total knee replacement following arthroscopy were higher for those who were older at the time of the arthroscopy; one-third of individuals who were seventy years of age or older underwent total knee replacement within three years. We conclude from these studies that the use of arthroscopic surgery in the management of knee osteoarthritis requires closer evaluation, particularly in older people.
The strengths of our study include the large sample sizes and the ability to examine trends over time in two large areas. However, there are also potential limitations. It was possible to determine the side of the operation from the English HES data, enabling us to ascertain when knee arthroscopy and total knee replacement had been performed on the same knee, but it was not possible with use of the Canadian data. Wai et al.16 examined the medical records of all patients who had undergone arthroscopic débridement for the treatment of knee osteoarthritis in a given time period and subsequently had a total knee replacement within three years and found that only one of twenty-one patients so identified had undergone the total knee replacement and arthroscopy in different knees. In addition, concerns have been raised about the accuracy of data collected for administrative rather than research purposes25,26. In Canada and the United Kingdom, coding for total knee arthroplasty has been shown to be highly (97% to 98%) accurate27,28. Furthermore, as the databases in both of these countries are essential for health policy-makers to evaluate care and allocate resources, extensive checks are made to ensure the completeness and accuracy of the data; false-positive and negative procedure rates have been found to be extremely low29,30. Individual-level information on socioeconomic class is not available in the databases used for this study; hence, we used an ecological measure of socioeconomic status. Ecological bias could therefore explain our findings on socioeconomic status; the association observed at an area level could be the same, stronger, or even in the opposite direction at the individual level. Finally, the diagnostic codes used in this study had been recorded by the surgeon following the procedure. Thus, these codes may not accurately reflect the true indication for the arthroscopy surgery.
In conclusion, our findings indicate that the utilization of knee arthroscopy varies by age, sex, income, and diagnosis. Whether these differences represent provision out of step with need is unclear, and this issue warrants additional study. Rates of total knee replacement following knee arthroscopy were low overall, but they were highest among older individuals with osteoarthritis. Trials are needed to ascertain whether there are subgroups of patients with osteoarthritis in whom the use of arthroscopy is beneficial, and guidelines for the procedure and its indications should be developed. 