The Missing Revision Burden

Abstract

Background:

Rising volumes of primary hip and knee replacements, coupled with longer lifespans, have led to projections of revision surgeries becoming a significant cost and healthcare burden. However, techniques and technology have improved. The purpose of this study was to determine whether revision hip and knee arthroplasty incidence has risen as previously predicted in the United States.

Methods:

This retrospective serial cross-sectional study analyzed data on primary and revision hip and knee replacements from the National Inpatient Sample, a comprehensive US hospitalization database, from 1996 to 2020. Primary outcome was the incidence of primary and revision arthroplasty. Secondary outcomes were comparisons between observed revision volumes and predictions, as well as trends in indications for revision surgery.

Results:

From 1996 to 2019, the incidence of primary hip replacements increased by 156%, from 128 to 328 per 100,000 person-years. The incidence of revision hip replacements increased by 41%, from 24 to 34 per 100,000 person-years. The incidence of primary knee replacements increased 136% from 220 to 520 per 100,000 person-years, while the incidence of revision knee replacements rose by 147%, from 19 to 47 per 100,000 person-years. Modeling a 10% failure rate at 10 years, revision hip replacement volumes were 41% lower than expected. Modelling a 7% failure rate at 10 years, revision knee replacement volumes were 33% lower than expected. The proportion of revisions due to loosening and implant wear decreased significantly over the last decade: for hip replacements, from 34.3% to 20.7% (p < 0.001), and for knee replacements, from 49.2% to 28.5% (p < 0.001).

Conclusion:

The burden of revision surgeries, especially due to implant wear, is significantly lower than anticipated. These data suggest improving longevity, which may support surgeon and patient confidence in arthroplasty durability.

Total hip and knee replacements produce some the highest improvements in health-related quality of life of all procedures^1,2. However, all hip and knee replacements have a finite durability. The polyethylene bearing surface can wear, eventually necessitating revision joint replacement.

The increasing number of arthroplasties being performed, coupled with longer lifespans, and younger ages at primary replacement, has led to major concerns that a large number of patients will require revision joint replacement^3-6. Revision hip and knee replacements are complex, require longer hospital stays, and incur significantly higher costs compared with primary replacements^5,7. Based on rising rates of arthroplasty in the 2000s, several authors have predicted a large burden of revision hip and knee replacement in the United States^3,8,9. In the United Kingdom, decreasing age at primary arthroplasty has led to a similar worry that revision rates will rise¹⁰.

Because the United States does not have nationwide implant registry and Medicare primarily has data on patients 65 and older, there have been few studies investigating if revision rates are in fact rising. Because our clinical experience has been that the number of revision arthroplasties, especially for wear, did not seem to be growing. we hypothesized that the incidence of revision hip and knee replacement in the United States has not been rising. Therefore, we used the National Inpatient Sample (NIS) to evaluate whether the volume and incidence of revision arthroplasties have risen as previously predicted.

Methods

The NIS contains discharge data for over 7 million inpatient stays annually. It uses a weighting strategy that allows for comprehensive national estimates of procedure volumes covering all payers and age ranges. Trend weighting files prepared by Agency for Healthcare Research and Quality adjust for sampling frame changes to allow accurate comparison across years. Primary hip and knee replacements were inpatient-only until January 1, 2020, and January 1, 2018, respectively¹⁰. After these years, volumes were supplemented with data from the Nationwide Ambulatory Surgery Sample to account for outpatient procedures that were not captured by the NIS. (Revision arthroplasty was inpatient only through the study period). The size of the US population 40 years and older was obtained from the US census data. 99% of arthroplasties in NIS were age 40 years and older.

After Institutional Review Board approval, primary and revision arthroplasties were identified by International Classification of Diseases Ninth Revision (ICD-9) procedure code until October 1, 2015, and by International Classification of Diseases Tenth Revision (ICD-10) procedure codes thereafter. In the ICD-10 era, we required revision cases to either have either the revision code or both an ICD-10 removal and replacement code. The coding strategy followed that of Upfill-Brown et al. and is available in the Supplemental Methods.^11,12 We did not use Current Procedural Terminology (CPT) codes. We confirmed that the number of primary and revision cases were very similar to the numbers identified by other researchers^3,8,9,11,12. Consistent with previous studies, conversion of previous hip surgery to total hip replacement would be considered a primary hip replacement in this schema.

We then compared the volume of revision surgeries that had been previously predicted in the early 2010s by Schwartz et al. and Kurtz et al. to the actual volume that was performed in 2019^3,9.

We modeled the expected number of revision hip and knee replacement based on the previous years' volumes of primary replacements. We chose a simple arithmetical model. For hips, we assume the baseline rate of 26,000 revision hip replacements per year seen in 1996 to continue. In the 2000s, 10-year revision rate for hip replacements was reported to be 5.7% to 17.4%, with a mean of 9.1% (Supplemental Methods)¹³. Starting with the cohort of primaries done in 1996, the model assumes no revisions occur during the first 5 years. After this lag, 2% of that year’s primary hip replacements are revised annually for up to 15 years (i.e., 90% implant survival rate at 10 years). After 15 years, no further revisions are expected for that year’s cohort. Revisions are added cumulatively, meaning the revision volume in any given year includes revisions from multiple past primary cohorts (e.g., 1996 and 1997).

For knees, the same approach was applied. The 1996 baseline rate of 36,000 revision knees replacements per year was used. Because the 10-year revision rate was reported in the 2000s to be 4.0% to 12.6% (Supplemental Methods)¹⁴, we modelled a 1.4% annual revision rate after the 5-year lag (i.e., 93% implant survival rate at 10 years), continuing for up to 15 years.

We used the ICD9 or ICD10 diagnosis codes, as described by Upfill-Brown et al.,^11,12 to classify the indications for revision arthroplasty.

Finally, we assessed the severity of revision from 2010 to 2020, by assessing the extent of revision surgery. For a hip replacement, up to 3 components can be revised (femoral component, acetabular component, and polyethylene liner). For knees, up to 4 components can be revised (femoral component, tibial component, polyethylene liner, and patellar component). When specified by ICD-9 or ICD-10 procedure codes, we tallied the number of components that were revised up to maximum of 3 for hip and 4 for knees. The coding strategy can be found it the Supplementary Methods.

Results

Between 1996 and 2019 (excluding the COVID-19 pandemic), the volume of primary hip replacements per year increased from 140,000 (95% CI, 130,000-150,000) to 516,000 (95% CI, 495,000-538,000), while revision hip replacement increased from 26,600 (95% CI, 24,000-29,200) to 52,700 (95% CI, 49,900-55,500). Between 1996 and 2019, the volume of primary total knee replacement increased from 241,000 (95% CI, 224,000-257,000) to 818,000 (95% CI, 790,000-845,000), while revision knee volume increased from 20,600 (95% CI, 18,400-22,800) to 73,900 (95% CI, 70,200-77,600) (Fig. 1).

Fig. 1.

Volume of primary vs revision hip and knee replacement in the United States, 1996 to 2020. Analysis uses National Inpatient Sample data. Fig. 1-A Annual volume of primary joint arthroplasty vs revision arthroplasty. Fig. 1-B Annual volume of primary hip, primary knee, revision hip, and revision knee arthroplasty. All standard errors are less than ±4% of estimates.

The incidence (which accounts for the population growth) of primary hip replacement increased over the study period from 128 (95% CI, 119-137) per 100,000 persons to 328 (95% CI, 314-342) per 100,000 persons. The incidence of revision hip replacement increased slightly from 24 (95% CI, 22-27) per 100,000 persons to 34 (95% CI, 32-35) per 100,000 persons. The incidence of total knee replacement rose over the study period from 220 (95% CI, 205-236) per 100,000 persons to 520 (95% CI, 502-537) per 100,000 persons, while the incidence of revision total knee replacement rose in tandem from 19 (95% CI, 17-21) per 100,000 persons to 47 (95% CI, 45-49) per 100,000 persons (Fig. 2).

Fig. 2.

Arthroplasty incidence per 100,000 person-years age 40 years and older. Fig. 2-A Hip replacement. Fig. 2-B Knee replacement.

Based on modeling a 10% revision rate at 10 years, the predicted volume of revision total hip replacement should be 88,900 cases per year, which is 69% higher than the 52,700 cases per year observed (Fig. 3). When we model a 7% revision rate at 10 years for knees, we calculate 110,100 surgeries in 2019, which is 49% higher than the observed number 73,900.

Fig. 3.

Actual and predicted annual volume of revision arthroplasty from 1996 to 2020. Actual revision volumes are derived from National Inpatient Sample data (shown in blue). We model predicted revision hip replacement volumes (shown in red) based on previous years' primary hip replacement volumes. The cone represents the maximum and minimum confidence range of volumes.

The proportion of revisions related to implant wear declined between 2010 and 2020. Revisions due to loosening and implant wear decreased significantly over the last decade: for hip replacements, from 34.3% to 20.7% (p < 0.001), and for knee replacements, from 49.2% to 28.5% (p < 0.001). Infection as a reason for revision rose for both revision hip (from 8.2% to 21.9%; p < 0.001) and revision knee (from 15.0% to 31.1%; p < 0.001). Periprosthetic fracture also rose as a cause for hip and knee revision (Fig. 4).

Fig. 4.

Indications for revision hip and knee replacement from 2010 to 2020. The number of cases where indications for revision arthroplasty could be determined are displayed below the year.

One measure of the severity of a revision arthroplasty is the number of artificial components that need to be replaced. In the ICD-9 era from 2006 to 2014, we observed a linear decrease in the number to total hip components revised (R = -0.69, p < 0.03). For revision knee replacement, a similar decrease in the number of components revised was observed (R = −0.85, p < 0.002). Such decline was not observed in the ICD-10 era (Supplemental Figures 1 and 2).

Discussion

Hip and knee replacements are 2 of the top 10 most common procedures in US hospitals with approximately $20 billion spent annually^15,16. A rise in revision procedures (which are complicated, expensive, and have high readmission rates⁵) would have an outsized effect on the healthcare system in the United States and globally. This study demonstrates that the volume of revision hip and knee replacements in the United States has not increased to the extent previously predicted, suggesting improved implant durability.

The number of revision knee replacements performed in 2019 (before the 2020 COVID-19 pandemic) was well below the number previously predicted by experts. Based on previous revision volume trends and an aging US population, Kurtz et al. predicted 120,000 revision knee replacement per year in 2019⁹. Similarly Schwartz et al. predicted 104,000 revision knee replacements in 2019³. We identified 73,900 (95% CI 70,200-77,600) revision knee replacements in 2019, well below estimates. It should be noted that the confidence intervals of the predictions were very wide ( ±50,000 or more). For revision hip replacements, the actual number was also below previously predicted volumes. We identified 52,700 (95% CI 49,900-55,500) revision hip replacements, while predictions were for 60,000 to 63,000^3,9. We believe that the literature’s predictions for revision hip replacements were more in line because the flattening trend in revision hip replacements was evident between 2010 and 2015.

To the best of our knowledge, this study is the first to attempt to model the revision arthroplasty volumes based on the known primary arthroplasty volumes. We selected a conservative simple arithmetical model using contemporaneously published US hip and knee revision rates. We found that the actual number of revisions is lower that what should be expected based on prevailing revision rates. This “missing revision burden” is most visible in hip arthroplasty, where the actual number revisions is almost half what should be expected (Fig. 3). While actual knee revision volumes were also 33% less than modelled, the divergence began after 2010. The 110,000 knee revisions predicted by our model is close to 104,000 to 120,000 predicted by Schwartz et al. and Kurtz et al., further supporting this modelling strategy.

A preponderance of evidence points to improvements in technology as the source for the lower-than-expected revision rates. Total hip and knee replacements tend to fail over time due to mechanical wearing of the polyethylene bearing surface (also known as the liner). Wear particles cause bone resorption (osteolysis) and thus implant loosening. Beginning in 2000s, researchers identified techniques to improve the polyethylene components of first hip and then knee replacements¹⁷. Ultra highly crosslinked and vitamin E doped polyethylene have been shown to have markedly lower wear rates in prospective clinical studies^18,19. We observed a significant reduction in the proportion of revisions that were attributable to osteolysis, loosening, and implant wear. We further observed that when revision arthroplasty was performed, fewer individual components needed to be revised, consistent with improved technology. It is likely that the trend toward protocol standardization, concentration in high volume replacement centers, and fellowship training of surgeons, along with other factors, all contribute to improved outcomes^20-22. Evidence from registry data in the United Kingdom also indicates that procedures performed in later years have a lower revision rate²³.

Although the incidence of revision arthroplasty may be flattening, the absolute number does continue to rise. Between 1996 and 2019, revision knee replacement volume grew 258% and revision hip replacement volume grew 98%. However, Jella et al. showed that the number surgeons who perform revision hip replacement actually decreased by 40% from 2013 to 2019²⁴. Meanwhile, the number of orthopaedic surgeons in the United States is flat (or shrinking)²⁵. This mismatch between rising procedure volumes and a stagnant workforce suggests that access to care and operative workload per surgeon will become increasingly burdensome. Future resource planning should consider both absolute numbers and the distribution of cases across the surgical workforce.

For osteoarthritis of the hip and knee, several evidence-based treatments are recommended, including physical therapy, NSAIDs, and exercise²⁶. In addition, there are popular but less well-supported interventions, such as platelet-rich plasma and hyaluronic acid injections^27,28. The improved durability of modern implants strengthens the case for considering joint replacement for patients with severe osteoarthritis who have failed conservative management. This is particularly relevant for patients who may be hesitant due to concerns about future revisions. Hip and knee replacement procedures remain underutilized among blacks despite higher levels of osteoarthritis^29,30.

Strengths

The NIS provides a powerful method to reliably assess procedure volumes and is comparable across the years. The NIS provides data on a very large number of primary and revision arthroplasty. The number of procedures from 2 years of the NIS (2.4 million) is as much as the complete Australian joint registry (2.1 million). Data are reliable and comparable from year to year. Ours is the first study to span several decades of data and incorporates outpatient procedures from NASS after 2018. We show evidence of a missing revision burden both by comparing to expert estimates and modelling.

Limitations

Database studies by definition contain less granular data on individual patients and rely on coding accuracy. Although we can document a revision arthroplasty, we have no data on how long that arthroplasty was in place. The NIS does not allow tracking individual patients across the years nor does it contain data on individual surgeon volume. It may be that there is a large population of patients who are in need revision arthroplasty but are unable to receive the procedure for access or medical reasons. However, the fact that revision procedure volumes continue to rise shows the system is able to accommodate revisions and makes this less likely. The ICD-10 coding structure resulted in increased coding complexity, and we may be undercounting revisions³¹. However, for hip specifically, the missing revision burden predates the 2015 switch to ICD-10. Finally, the evidence of a missing revision burden is strongest for total hip arthroplasty where the incidence of revision has been flattest for the longest period of time. The incidence of revision knee arthroplasty has been rising in line with the incidence of primary knee arthroplasty, although the volume is less than previously predicted.

Conclusion

Each year, there are approximately 75,000 fewer revision joint surgeries than expected. These data demonstrate that the volume of revision hip and knee replacements has not increased to the extent previously predicted or, to the extent, that there should be based on primary volume trends. These data suggest improved implant durability in large population data sets. However, revision rates must be interpreted with consideration of patient outcomes, access to revision services, and workforce limitations.

Appendix

Supporting material provided by the authors is posted with the online version of this article as a data supplement at jbjs.org (http://links.lww.com/JBJSOA/A937). This content was not copyedited or verified by JBJS.