**2. The American experience**

Literature emerging from the United States has provided both the largest volume of data and greatest variation in results. The landmark study by Kreder et al1 examining outcomes of total hip arthroplasty (THA) in Washington state between 1987-1991 demonstrated that low-volume surgeons experienced higher rates of mortality, infection, and revision arthroplasty when compared to high-volume surgeons. Low volume surgeons were defined as those performing less than 2 THA/year and high-volume surgeons as those performing greater than 10 THA/year. Complication rates were adjusted for patient age, comorbidity, gender and diagnosis. Patients of low-volume surgeons also had a longer acute duration of hospitalization, which was 0.8 days longer than patients of high-volume surgeons. However, this effect was reversed with respect to institutional volume with low-volume hospitals demonstrating a shorter duration of hospital stay. This could be due to a variety of factors, such as low volume centers (which do not have inpatient rehabilitation services) having to discharge patients to outpatient rehabilitation facilities. It is also possible that high volume centers have more complex cases or patients with a greater number of comorbidities – factors which may not have been fully accounted for in data analysis. Interestingly, average costs for hospital services were found to be greater in low-volume centers, due primarily to relatively higher charges for implants and the operations as a whole. Overall, the majority of THAs in this study were performed by low-volume surgeons who comprised 170 of 494 (34.4%) of the study population. The cut-off for center volumes was less than 16 for low-volume, and greater than 65 for high volumes. Volume categories for both surgeons and institutions were derived from the quartiles of the volume distributions. The second and third quartiles were collapsed into one group for comparison. Perhaps one of the most surprising results from this study is that over one third of all practicing surgeons performed an average of 2 or fewer THA/year. With such low average volumes, it is therefore not surprising that the complication rates among this group were significantly higher. Given the steady rise in rates of THA in the past decade, it is likely that the average numbers of low volume surgeons and centers have experienced a similar and parallel upward trend, possibly narrowing the difference in complication rates between the lowest and highest quartile surgeons.

Katz et al, who have extensively published on the volume-outcome relationship, studied the association between procedure volume and primary and revision THA outcomes in Medicare patients operated on in 1995-19962. Rather than use the data distributions to define procedure volume categories, hospital and surgeon volumes were stratified into what the authors described as clinically sensible categories. In their analysis, it was revealed that 52% of primary THAs and 77% of revisions were performed by surgeons doing 10 or fewer procedures a year. A strong association was found between low surgical volumes and increased complications. Surgeons in the low volume group (<12/year) compared to those in a high volume group (>50/year) experienced higher rates of all complications examined including death, pneumonia, pulmonary embolism, myocardial infarction and deep infection with significant lower rates of dislocation and deep hip infection. Perhaps one of the most striking observations was the discrepant mortality rate for patients who underwent primary or revision THA by a surgeon performing greater than 10, as opposed to fewer than 3, procedures per year. After adjusting for patient comorbidities, mortality rates for primary THA by a high volume surgeon were 0.7% versus 1.3% for low-volume surgeons. Mortality for revision THA was 1.5% for high volume surgeons as compared to 3.1% for low-volume

Literature emerging from the United States has provided both the largest volume of data and greatest variation in results. The landmark study by Kreder et al1 examining outcomes of total hip arthroplasty (THA) in Washington state between 1987-1991 demonstrated that low-volume surgeons experienced higher rates of mortality, infection, and revision arthroplasty when compared to high-volume surgeons. Low volume surgeons were defined as those performing less than 2 THA/year and high-volume surgeons as those performing greater than 10 THA/year. Complication rates were adjusted for patient age, comorbidity, gender and diagnosis. Patients of low-volume surgeons also had a longer acute duration of hospitalization, which was 0.8 days longer than patients of high-volume surgeons. However, this effect was reversed with respect to institutional volume with low-volume hospitals demonstrating a shorter duration of hospital stay. This could be due to a variety of factors, such as low volume centers (which do not have inpatient rehabilitation services) having to discharge patients to outpatient rehabilitation facilities. It is also possible that high volume centers have more complex cases or patients with a greater number of comorbidities – factors which may not have been fully accounted for in data analysis. Interestingly, average costs for hospital services were found to be greater in low-volume centers, due primarily to relatively higher charges for implants and the operations as a whole. Overall, the majority of THAs in this study were performed by low-volume surgeons who comprised 170 of 494 (34.4%) of the study population. The cut-off for center volumes was less than 16 for low-volume, and greater than 65 for high volumes. Volume categories for both surgeons and institutions were derived from the quartiles of the volume distributions. The second and third quartiles were collapsed into one group for comparison. Perhaps one of the most surprising results from this study is that over one third of all practicing surgeons performed an average of 2 or fewer THA/year. With such low average volumes, it is therefore not surprising that the complication rates among this group were significantly higher. Given the steady rise in rates of THA in the past decade, it is likely that the average numbers of low volume surgeons and centers have experienced a similar and parallel upward trend, possibly narrowing the difference in complication rates between the lowest and highest

Katz et al, who have extensively published on the volume-outcome relationship, studied the association between procedure volume and primary and revision THA outcomes in Medicare patients operated on in 1995-19962. Rather than use the data distributions to define procedure volume categories, hospital and surgeon volumes were stratified into what the authors described as clinically sensible categories. In their analysis, it was revealed that 52% of primary THAs and 77% of revisions were performed by surgeons doing 10 or fewer procedures a year. A strong association was found between low surgical volumes and increased complications. Surgeons in the low volume group (<12/year) compared to those in a high volume group (>50/year) experienced higher rates of all complications examined including death, pneumonia, pulmonary embolism, myocardial infarction and deep infection with significant lower rates of dislocation and deep hip infection. Perhaps one of the most striking observations was the discrepant mortality rate for patients who underwent primary or revision THA by a surgeon performing greater than 10, as opposed to fewer than 3, procedures per year. After adjusting for patient comorbidities, mortality rates for primary THA by a high volume surgeon were 0.7% versus 1.3% for low-volume surgeons. Mortality for revision THA was 1.5% for high volume surgeons as compared to 3.1% for low-volume

**2. The American experience** 

quartile surgeons.

surgeons. Similar increased complication rates were found for low volume surgical centers (<25 TKA/year) when compared to high volume centers (>200/year). While this study demonstrated a clear trend toward better outcomes with higher volumes, no specific recommendations were given on what might constitute an acceptable minimum number of procedures.

Shortly thereafter, Katz's group used Medicare data from 1995-1997 to study rates of revision within 3 years of primary THA6. Hospitals were stratified into 4 volume groups: low (<25/year), medium (26-50/year and 51-100/year) and high (>100/year). Low volume surgeons were defined as those performing fewer than 12 primary THA/year. At the time of the study, 75% of all total hip replacements in the US were performed by surgeons doing less than 25/year, with centers performing over 100 cases a year accounting for only 10% of all hip replacements done. Comparing high volume surgeons (>12/year) in high volume hospitals (>100/year) to low volume surgeons (<12/year) and low volume centers (<25/year), the early failure rates were 3.3% vs 4.9%, or approximately a 50% increase in revision surgery. The highest rates of revisions for low volume surgeons were found within the first 18 months.

SooHoo and Lieberman examined the effect of hospital volume on outcomes of total knee arthroplasty (TKA) in California from 1991-20018. They divided hospitals into three groups: low volume (bottom 40th percentile), intermediate volume (middle 40th percentile) and high volume (top 20th percentile). Interestingly, the lowest volume centers (which accounted for 40% of the hospitals examined) averaged only 13 (standard deviation (SD) 5) TKA/year. This is in stark contrast to the intermediate group (50, SD 15) and high volume group (145, SD 47). Outcome measures included complications within the first 90 days and 365 days postoperatively. Statistically significant higher complication rates were found for low-volume centers across the outcomes measured, including mortality, readmission for infection, pulmonary embolism and thromboembolism.. The largest difference between the low and high volume centers was for infection (1.13% versus 0.65%, respectively; p = 0.004). However, the overall postoperative complication rates remained relatively low, even among low-volume centers, leading authors to question the rationale behind regionalization for TKA.

Hervey et al examined provider volumes and patient outcomes in an analysis of primary and revision TKAs3. Using the Healthcare Cost and Utilization Project Nationwide Inpatient Sample (HCUP-NIS), they were the first to create a national data set from multiple states and hospital settings in order to generate more comprehensive and robust estimates. Hospitals were divided by volume cutoffs into 4 groups: those performing <85, 85-149, 150- 250, and >250 TKA/year. Surgeons were stratified from lowest to highest volumes into 4 groups: those performing <15, 15-30, 30-59, and >60 procedures/year. Interestingly, surgeons performing 30 or fewer primary TKAs/year were responsible for nearly half the surgical volume of knee replacements in the US. Using multivariate regression analysis, higher volume surgeons and hospitals had lower mortality rates. Rates of other complications were elevated in the low volume groups, but did not differ statistically from those in the highest-volume groups. The other major finding was that low volume surgeons and centers tended to discharge patients later than their higher volume counterparts, in contrast to the early findings of Kreder et al1.

Finally, Katz et al analyzed Medicare data on TKAs performed in Tennessee, Ohio, Illinois and North Carolina in 2000 to examine the effect of provider volume on patient function post-operatively7. Using a random sample of TKAs from various hospitals, patients were

Provider Volumes and Surgical Outcomes in Total Hip and Knee Replacement 39

Kreder et al performed a study of THA using Ontario data from 199212. Surgeons and institutions were divided into 3 groups defined by the procedure volume distributions: <40th, 40-80th, and >80th percentile. The low surgeon volume group performed on average 9 THAs compared to the high volume group who performed on average over 27 THAs /year. Low volume centers performed <40 annual THAs compared to >80/year in the high volume groups. Though Kreder attempted to use the procedure volume cut-offs from his earlier Washington study1, Canadian provider volumes were such that the vast majority of Canadian surgeons would have been labeled high-volume (>10 procedures/year). The major finding of this study was that higher volume surgeons and centers discharged their patients much earlier (2.4 days earlier). No volume-outcome associations were found. Kreder et al later looked at provider volume-outcome relations in Ontario TKAs from 1993- 1996 using similar methods to the above study10. Again, patients of lower-volume surgeons

had longer lengths of stay in hospital. They also had double the risk of reoperation.

hospital, although no similar relation was observed with hospital volumes.

have relatively shorter lengths of hospital stay.

individual surgeon.

**5. Discussion** 

high volume centers1,10,12,14.

Finally, Paterson et al performed a more recent, larger study of Ontario patients including over 20,000 THA patients and 27,000 TKA patients14. In the 10 years since Kreder's previous analysis,12? hip replacement volumes had grown by 50%, and total knee replacement volumes by 100%. The number of arthroplasty surgeons during that time period also increased, but at a much slower rate, resulting in an increased volume of cases per

In this study, low-volume hip surgeons performed an average of 2-25 cases per annum, compared to 2-35 cases per annum for low-volume knee surgeons. This contrasted to the highvolume groups which were greater than 60 cases per annum for hip and greater than 71 cases per annum for knee. Of note, extremely low volume surgeons (defined as those performing less than 2 procedures/year) and low volume centers (defined as those performing less than 10 procedures/year) were excluded from analysis. The results of this study corroborated those of Kreder's earlier work in that no consistent relationships were found between complication rates and hospital volumes for either TKA or THA. Surgeons in the lowest volume quartile demonstrated a 30% increased rate of complications for THA but no similar relationship was seen for TKAs. Surgeon volume was also related to increased risk of revision for THA but not for TKA. Again, patients of lower-volume surgeons had relatively longer lengths of stay in

Overall, the Canadian experience with respect to TJR surgical volume-outcome relationships is mixed. The differences in adverse outcome as related to provider volumes is much less pronounced than in the US, with the most recent Canadian study showing essentially no relations for TKA and inconsistent findings for THA. The one consistent finding among the Canadian studies is that, adjusting for hospital volume, patients of higher-volume surgeons

In summary, while the literature suggests that lower surgeon and hospital procedure volumes are associated with poorer short-term patient outcomes after TJR,1-3,6,7*,*8,10,12,14 this relationship appears to exist only when provider volumes are extremely low. The data among moderate and high volume providers is otherwise inconsistent with no clear relationship between surgeon or center experience and patient outcomes. Shorter hospital stays have also been observed among high volume surgeons and to a lesser extent among

sent questionnaires capturing both self-reported WOMAC scores and surgery satisfaction scores. Poor WOMAC functional status scores (< 60) were noted to be more frequent among patients of surgeons who performed fewer than 12 TKA/year (20%), and centers performing fewer than 10 TKA/year (19%), as compared to their higher-volume counterparts (12% and 10% amongst high-volume surgeons and hospitals, respectively). These findings were consistent across the country, even after adjusting for medical comorbidities and age. Overall, the American experience is quite unique in that across the country there is a virtual myriad of low volume centers that coexist with some of the largest and highest volume centers in the world. The current body of data indicates that despite increasing volume trends for THA and TKA, the majority of these procedures in the United States are still being performed by surgeons who perform fewer than 30/year and centers performing fewer than 100/year. Several studies have demonstrated increased complications rates, including increased risk of infection, revision, and mortality among low-volume surgeons and centers. In particular, it is evident that very low provider volumes (fewer than 10/year) are more frequently associated with increased morbidity and mortality. However, consensus on a minimum individual or institutional volume threshold for optimal patient outcomes has not been reached. The data are less conclusive with respect to the relationship between provider volumes and length of hospital stay, with early studies showing shorter stays in low-volume centers and subsequent studies the reverse. Little has been published on the financial costs of care in high versus lowvolume centers. Further such studies are needed.
