Abstract
Background: Over the past decade, there has been an increase in the
number of revision total hip and knee arthroplasties performed in the United
States. The purpose of this study was to formulate projections for the number
of primary and revision total hip and knee arthroplasties that will be
performed in the United States through 2030.
Methods: The Nationwide Inpatient Sample (1990 to 2003) was used in
conjunction with United States Census Bureau data to quantify primary and
revision arthroplasty rates as a function of age, gender, race and/or
ethnicity, and census region. Projections were performed with use of Poisson
regression on historical procedure rates in combination with population
projections from 2005 to 2030.
Results: By 2030, the demand for primary total hip arthroplasties is
estimated to grow by 174% to 572,000. The demand for primary total knee
arthroplasties is projected to grow by 673% to 3.48 million procedures. The
demand for hip revision procedures is projected to double by the year 2026,
while the demand for knee revisions is expected to double by 2015. Although
hip revisions are currently more frequently performed than knee revisions, the
demand for knee revisions is expected to surpass the demand for hip revisions
after 2007. Overall, total hip and total knee revisions are projected to grow
by 137% and 601%, respectively, between 2005 and 2030.
Conclusions: These large projected increases in demand for total hip
and knee arthroplasties provide a quantitative basis for future policy
decisions related to the numbers of orthopaedic surgeons needed to perform
these procedures and the deployment of appropriate resources to serve this
need.
Over the past thirteen years, there has been an increase in the number of
revision total hip arthroplasties and total knee arthroplasties performed in
the United States1,
while the revision burden—defined as the ratio of revision
arthroplasties to the total number of arthroplasties—has remained
relatively constant. Previous analyses of nationwide data on the procedures
have indicated that the revision burden for total hip arthroplasty was
approximately 17.5% between 1990 and 2002, whereas the revision burden for
total knee arthroplasty was approximately
8.2%1. In economic
terms, revision total hip arthroplasty is estimated to have consumed 19% of
the Medicare hip replacement expenditures between 1997 to 2003, whereas total
knee arthroplasty revisions consumed only 8% of the total annual Medicare
expenditures for knee
replacement2.
In Scandinavia, England, Australia, and Canada, the numbers of these
procedures that have been performed are readily obtainable from national
arthroplasty
registries3-6.
However, in the United States, the American Academy of Orthopaedic Surgeons
(AAOS)7 and other
investigators have had to rely on representative surveys of hospital discharge
records to provide estimates of the prevalence of primary and revision
arthroplasties1,8.
Given the many years required for the training of surgeons and the equally
complex task of planning for hospital capacity, reliable projections regarding
the demand for arthroplasties are crucial for policy makers in government,
education, and industry. Importantly, reliable projections of revision
arthroplasties would be particularly useful, as they consume greater economic
resources than do primary
procedures9.
For this study, we hypothesized that the demand for total hip and total
knee arthroplasties in the United States will increase substantially over the
next twenty-five years. To test this hypothesis, we performed statistical
projections of the number of primary and revision total hip and total knee
arthroplasties between 2005 and 2030 on the basis of the available historical
Nationwide Inpatient Sample (NIS) data from 1990 to
200310, compared
with projections assuming a constant prevalence.
Data Sources
The National Hospital Discharge Survey (NHDS) and the NIS are national
(United States) sample surveys of hospital discharge records. Comparisons of
the two surveys conducted by the Agency for Healthcare Research and Quality
have found that, on a year-by-year basis, the numbers of surgical procedures
estimated to have been performed from the NIS and the NHDS are similar,
differing by approximately
10%11. However, the
NIS collects a substantially larger number of discharge records than does the
NHDS and is therefore better suited to accurately quantify the prevalence of
inpatient arthroplasty procedures in the United States. Consequently, we
relied on the NIS data from 1990 to 2003 for our projections for the current
study. The NIS is a federal-state cooperative database designed to compile
annually a representative sample of hospital discharge records in the United
States. In 2003, the NIS had a sample size of about eight million discharge
records from approximately 1000 hospitals, which represent approximately 20%
of all United States community hospitals. All of the discharge records from
1990 to 2003 were examined for this study.
We obtained demographic data on the patients (e.g., age, gender, and race
and/or ethnicity) from the NIS. Disease diagnoses and surgical procedures
performed (if any) were recorded for the NIS with use of the Ninth Revision of
the International Classification of Diseases (ICD-9-CM). Specifically, primary
hip and knee arthroplasty are identified by the ICD-9-CM codes 81.51 and
81.54, respectively. For revisions, the corresponding codes are 81.53 and
81.55. From 1990 to 2003, the ICD-9-CM codes for these procedures were
consistent, thereby allowing the determination of longitudinal trends in the
prevalence of both primary and revision joint arthroplasty. We also used
projected population statistics for the nation and for individual states by
age, gender, and race and/or ethnicity through 2025 that were published by the
Census Bureau in
199712.
Projection Methodology and Statistical Analyses
The annual prevalence of arthroplasty surgery was modeled with use of a
Poisson regression with age, gender, race and/or ethnicity, census region, and
calendar year as covariates to account for differences in prevalence among
population subgroups as well as changes over time. Age was categorized into
eight subgroups (less than forty-five, forty-five to fifty-four, fifty-five to
sixty-four, sixty-five to sixty-nine, seventy to seventy-four, seventy-five to
seventy-nine, eighty to eighty-four, and eighty-five or more years old), while
race and/or ethnicity was grouped into five categories (white, black, Asian,
Hispanic, and Native American). "Hispanic" included all patients
of Hispanic origin, regardless of race. Four census regions (Northeast, South,
Midwest, and West) and the two genders were also categorical covariates in the
analysis. Two-way interactions between age, gender, race, census region, and
calendar year were included in the regression model. Surgery prevalence was
calculated by dividing the number of procedures estimated from the NIS for
each population subgroup by the corresponding population from the Census
Bureau. The projected number of procedures was estimated by applying the
surgery prevalence estimated from the regression model to the projected
population data for each subgroup. The projected national total is the sum of
the projected number of procedures from each subpopulation. Independent models
were used for each type of primary and revision hip and knee arthroplasty.
To evaluate the methodological sensitivity of our results, we compared our
primary projections obtained from the NIS (in which the prevalence of surgery
is allowed to vary over time on the basis of the actual data) with projections
in which the prevalence of each population subgroup was held constant on the
basis of the 1990 to 2003 averages. Deviance and Pearson chi-square values
were determined to describe the goodness of fit for the Poisson regression
model for the various arthroplasty data. Additional detailed descriptions of
the statistical analyses are presented in the Appendix.
In 2003, the most recent year for which national inpatient procedure data
are currently available from NIS, a total of 202,500 primary total hip
arthroplasties and 402,100 primary total knee arthroplasties were performed
nationally in the United States. During the same year, a total of 36,000
revision total hip arthroplasties and 32,700 revision total knee
arthroplasties were performed.
Sensitivity of Projection Methodology
Between 1990 and 2003, the prevalence of primary and revision total hip and
knee arthroplasties all increased substantially over time. The overall
goodness of fit of the regression models, represented by the value of the
scaled Pearson chi square (a measure of the lack of fit between model and
data), averaged 1.11 (range, 1.03 for primary total knee replacement to 1.26
for revision total knee replacement procedures) (see Appendix). When the year
of surgery was excluded from the Poisson regression model to simulate a
constant prevalence over time, the models fitted with the remaining covariates
all showed a substantial increase in the deviance value (i.e., poorer fit),
especially for knee arthroplasty.
The projections of primary and revision total joint replacement were found
to be highly sensitive to assumptions regarding trends in the prevalence of
surgery. If the trends (i.e., increases in the prevalence of surgery) observed
from 1990 to 2003 were to continue, by 2030 the projections with use of the
NIS data could range from two to five times greater than the projections
assuming a constant surgery prevalence over time
(Table I). The projections for
primary and revision total knee surgery were more sensitive to modeling
assumptions than those for primary or revision total hip arthroplasty because
of the steep increase in the number of knee procedures from 1990 to 2003.
Projected Primary and Revision Arthroplasty Procedures with Use of
the Nationwide Inpatient Sample Baseline Model
Our projection model predicted substantial increases in the numbers of hip
and knee replacement procedures (Figs.
1 and
2). On the basis of the NIS
model, the demand for primary total hip arthroplasty was estimated to grow by
174%, from 208,600 (95% prediction interval, 193,300 to 224,600) in 2005 to
572,000 (95% prediction interval, 481,000 to 681,000) by 2030
(Fig. 1). If the number of
total knee arthroplasties performed continues at the current rate, the demand
for primary total knee arthroplasty is projected to grow by 673%, from 450,000
(95% prediction interval, 425,000 to 477,000) in 2005 to 3.48 million
procedures (95% prediction interval, 2.95 to 4.14 million) by 2030.
Overall, the total number of revision arthroplasty procedures performed in
2005 is expected to double by the year 2026 for revision total hip
arthroplasty and by 2015 for revision total knee arthroplasty. Although more
revision total hip arthroplasties than revision total knee arthroplasties are
currently performed, the number of revision total knee arthroplasties
performed were predicted to outnumber total hip arthroplasty revisions after
2007 (Fig. 2). Total hip
arthroplasty revisions were projected to grow from 40,800 (95% prediction
interval, 34,900 to 47,000) in 2005 to 96,700 (95% prediction interval, 72,100
to 130,000) in 2030 (an increase of 137%). If the trend observed from 1990 to
2003 were to continue, total knee arthroplasty revisions were projected to
grow from 38,300 (95% prediction interval, 32,600 to 44,300) in 2005 to
268,200 (95% prediction interval, 192,700 to 381,400) in 2030 (an increase of
601%).
On the basis of these estimates, the revision burden for total hip
replacements was projected to be 16.3% in 2005 and 14.5% in 2030. The
corresponding revision burden for total knee replacements was projected to be
7.8% in 2005 and 7.2% in 2030.
In this study, arthroplasty projections were derived by considering
temporal changes in arthroplasty rates, as well as in population subgroups. As
the official demographer of the United States, the Census Bureau has devoted
considerable effort to developing reliable projections of the future United
States population. In contrast, little information is available to quantify
the expected number of hip and knee revision arthroplasties in the future. For
example, the projection recently developed by the
AAOS7 was limited to
primary hip and knee replacements. The AAOS projections were found to have
underpredicted the expected utilization of primary joint replacement
surgery1 because
they were based on the NHDS survey, which has a much lower sample size than
the NIS. Additionally, the AAOS estimates assumed a constant prevalence of
surgery over time. In contrast, our results underscore the importance of
accounting for changes in the rate of surgery for future projections because
the prevalence of surgery is changing rapidly over time.
The present study provides, for the first time to our knowledge,
quantification of the demand for primary and revision hip and knee
arthroplasties in the United States through 2030. We project a massive
increase in demand for primary and revision total joint procedures over the
next two decades—a demand that, to be met, will need to be addressed
with a combination of increased economic resources, operative efficiency,
technical capacity (i.e., additional surgeons), and implant longevity.
The projections in this study are limited on the basis of an extrapolation
of historical procedural data. As demonstrated in this study, the
uncertainties inherent in such an extrapolation can be minimized by choosing a
suitably large set of historical data (e.g., NIS instead of NHDS), and by
incorporating as many covariates as possible into the model. Nevertheless,
these projections are limited by the quantity and quality of available data.
The trends established by historical data, even if accurate, may not persist
in the future because of improvements in implant technology, such as advanced
bearing materials or designs. Furthermore, it is impossible to anticipate, at
present, whether future orthopaedic treatment technologies or newer
pharmaceutical nonoperative interventions will lead to a reduced demand for
primary total joint replacements by 2030. Our model also does not incorporate
unforeseen changes in economic factors associated with these arthroplasties.
It is uncertain, for example, to what extent the United States health-care
system will be able to finance the future demand for arthroplasties
anticipated by the present study.
We selected a twenty-five-year time frame for the study, extending to 2030,
purely to facilitate comparisons with previous AAOS projections, which
employed an identical time frame. Intuitively, we appreciate that long-term
projections will be more prone to unexpected disruptions than those spanning a
near-term horizon. Nevertheless, such uncertainties in no way diminish the
value and necessity of conducting projections for the purpose of long-range
planning and policy-making.
Consequently, it is inevitable that the projections performed in the
present study will be superseded in the future as new years of procedure data
become available. Other methodological approaches to the prediction problem,
such as age-period-cohort models or generalized additive models, should also
be attempted to further validate the reliability of the projections
established by the present approach. It is relatively straightforward to
update the projections reported in this study with use of the present
methodology, but continued monitoring and updating will need to occur.
In addition, it is clear from the different trends observed that the
sensitivity of the projections appears to be procedure-dependent. For example,
because of the substantially higher rate of increase in knee arthroplasty
between 1990 and 2003, the models simulating a constant prevalence over time
produced a considerably poorer fit. Consequently, the specific findings for
total hip and knee replacement should not be generalized to other orthopaedic
procedures, which may exhibit entirely different historical growth histories.
However, this study establishes a methodology whereby an investigator can
systematically evaluate orthopaedic surgery projections in a generalized
statistically based framework. Although the data and projected number of
procedures can be updated regularly, the methodology we have developed is
expected to remain relevant for years to come.
The projections for revision procedures in this study were limited by the
generality of the ICD-9-CM codes in the existing data, which currently do not
yet discriminate between partial or total revision of an artificial joint. As
of October 2005, new ICD-9-CM codes had been introduced by the Center for
Medicare and Medicaid Services (CMS) for revision hip and knee arthroplasties.
New ICD-9-CM codes also had been introduced to track the type of bearing
(ceramic, metal, or polyethylene) used for total hip replacements. However,
2006 will provide the first full year of data incorporating this new coding
scheme, and there is a two and a half-year lag between the end of the calendar
year and the production of the corresponding NIS dataset. Furthermore, at
least four years of data would be necessary to perform even the most
rudimentary projection. Consequently, it will be well into the second decade
of this century that sufficient years of information will be available for
mathematically sound projections with use of the recently introduced ICD-9-CM
procedure codes.
We modeled revision hip and knee replacement as independent orthopaedic
procedures for this study, although it is feasible to construct a predictive
model for future revisions on the basis of the number of primary procedures
performed and an assumed Kaplan-Meier-type survivorship model for the
different population subgroups. However, the necessary survival data for such
a model can only be derived from a longitudinal database, such as a national
implant registry, which does not yet exist in the United States. The only
rationale for developing such a complex model would be historical evidence
that the revision burden was changing over time, as has been documented in
Sweden. However, available data in the United States between 1990 and 2003 do
not support such a hypothesis; indeed, the revision burden has remained
essentially unchanged in this country for over a
decade1. Between
1990 and 2002, the national revision burden for total hip arthroplasty ranged
between 15.2% and 20.5% (average, 17.5%); for total knee arthroplasty, the
revision burden varied between 7.3% and 9.7% (average,
8.2%)1. Without
compelling evidence that either implant technology or surgical technique have
improved the survival of primary replacements at a national level, it is
difficult to say whether a more sophisticated projection method for revisions
than the one we employed in the present study would be of use. The revision
projections in the current study may be interpreted as a conservative upper
bound for what awaits the orthopaedic community if improvements in primary
implant survival cannot be achieved at a national level.
The recent ICD-9-CM coding changes for hip and knee revisions in the
United States were accompanied by 26.5% increases in reimbursement by the CMS
and the formation of separate diagnosis-related group codes for primary and
revision procedures (544 and 545,
respectively)13.
The changes in coding and reimbursement reflect heightened awareness and
acceptance by CMS of the greater burden that revisions place on patients,
surgeons, and hospitals. The revision projections in the current study provide
the necessary foundation for future cost-benefit analyses at a national level,
to quantify the increasing societal impact of revision arthroplasty in the
United States.
A table showing the summary of multivariate Poisson regression analysis
results and a detailed description of the projection methodology and
statistical analysis are available with the electronic versions of this
article, on our web site at
(go to
the article citation and click on "Supplementary Material") and on
our quarterly CD-ROM (call our subscription department, at 781-449-9780, to
order the CD-ROM). ?
Kurtz S, Mowat F, Ong K, Chan N, Lau E,
Halpern M. Prevalence of primary and revision total hip and knee arthroplasty
in the United States from 1990 through 2002. J Bone Joint Surg
Am. 2005;87:
1487-97.871487
2005
[PubMed][CrossRef]
Ong KL, Mowat FS, Chan N, Lau E, Halpern
MT, Kurtz SM. Economic burden of revision hip and knee arthroplasty in
Medicare enrollees. Clin Orthop Relat Res.
2006;446:
22-8.44622
2006
[PubMed][CrossRef]
National Joint Registry. National
Joint Registry for England and Wales. 2nd Annual Report.
2005.
2005
Malchau H, Herberts P, Eisler T,
Garellick G, Soderman P. The Swedish Total Hip Replacement Register. J
Bone Joint Surg Am. 2002;84Suppl 2: 2-20. Erratum in: J Bone
Joint Surg Am. 2004;86:363.842
2002
Canadian Institute for Health Information, Canadian
Joint Replacement Registry (CJRR). 2004 report: total hip
and total knee replacements in Canada. 2004.
2004
Australian Orthopaedic Association.
National Joint Replacement Registry annual report 2005.
2005.
2005
Mahomed NN, Barrett JA, Katz JN,
Phillips CB, Losina E, Lew RA, Guadagnoli E, Harris WH, Poss R, Baron JA.
Rates and outcomes of primary and revision total hip replacement in the United
States medicare population. J Bone Joint Surg Am.
2003;85:
27-32.8527
2003
[PubMed]
Bozic KJ, Katz P, Cisternas M, Ono L,
Ries MD, Showstack J. Hospital resource utilization for primary and revision
total hip arthroplasty. J Bone Joint Surg Am.
2005;87:
570-6.87570
2005
[PubMed][CrossRef]