- Email: [email protected]

S1063-4584(17)31130-5

DOI:

10.1016/j.joca.2017.07.022

Reference:

YJOCA 4061

To appear in:

Osteoarthritis and Cartilage

Received Date: 15 November 2016 Revised Date:

30 June 2017

Accepted Date: 31 July 2017

Please cite this article as: Inacio MCS, Paxton EW, Graves SE, Namba RS, Nemes S, Projected Increase in Total Knee Arthroplasty in the United States - an Alternative Projection Model, Osteoarthritis and Cartilage (2017), doi: 10.1016/j.joca.2017.07.022. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT 1

Projected Increase in Total Knee Arthroplasty in the United States - an Alternative

2

Projection Model

3

Maria C. S. Inacio, MS, PhD1 Elizabeth W. Paxton, MA,2 Stephen E. Graves, MBBS, PhD,3

4

Robert S. Namba, MD, 4 Szilard Nemes, PhD5

6

1. Maria C.S. Inacio (corresponding author)

RI PT

5

Quality Use of Medicines and Pharmacy Research Centre

8

Medicine and Device Surveillance Centre of Research Excellence

9

Sansom Institute, School of Pharmacy and Medical Sciences

11

GPO Box 2471, Adelaide 5001

12

South Australia, Australia

13

[email protected]

14

Phone: +61 8302 133 33

16

2. Surgical Outcomes and Analysis Department, Kaiser Permanente

EP

15

San Diego, CA, USA

18

Email: [email protected]

20 21 22

AC C

17

19

M AN U

University of South Australia

TE D

10

SC

7

3. Australian Orthopaedic Association, National Total Joint Replacement Registry Adelaide, SA, Australia Email: [email protected]

23 24

4. Department of Orthopedic Surgery, Kaiser Permanente, Orange County

1

ACCEPTED MANUSCRIPT 25

Irvine, CA, USA

26

Email: [email protected]

27

5. Swedish Hip Arthroplasty Register

29

Gothenburg, Sweden

30

Email: [email protected]

AC C

EP

TE D

M AN U

SC

31

RI PT

28

2

ACCEPTED MANUSCRIPT Abstract

33

Objective: The purpose of our study was to estimate the future incidence rate (IR) and

34

volume of primary total knee arthroplasty (TKA) in the United States from 2015 to 2050

35

using a conservative projection model that assumes a maximum IR of procedures.

36

Furthermore, our study compared these projections to a model assuming exponential

37

growth, as done in previous studies, for illustrative purposes.

38

Methods: A population based epidemiological study was conducted using data from US

39

National Inpatient Sample (NIS) and Census Bureau. Primary TKA procedures performed

40

between 1993 and 2012 were identified. The IR, 95% confidence intervals (CI), or prediction

41

intervals (PI) of TKA per 100,000 US citizens over the age of 40 years were calculated. The

42

estimated IR was used as the outcome of a regression modelling with a logistic regression

43

(i.e. conservative model) and Poisson regression equation (i.e. exponential growth model).

44

Results: Logistic regression modelling suggests the IR of TKA is expected to increase 69% by

45

2050 compared to 2012, from 429 (95%CI 374-453) procedures/100,000 in 2012 to 725

46

(95%PI 121-1041) in 2050. This translates into a 143% projected increase in TKA volume.

47

Using the Poisson model, the IR in 2050 was projected to increase 565%, to 2854 (95%CI

48

2278-4004) procedures/100,000 IR, which is an 855% projected increase in volume

49

compared to 2012.

50

Conclusions: Even after using a conservative projection approach, the number of TKAs in the

51

US, which already has the highest incidence rate of knee arthroplasty in the world, is

52

expected to increase 143% by 2050.

53

Keywords: total knee arthroplasty, projections, epidemiology, incidence rates

AC C

EP

TE D

M AN U

SC

RI PT

32

54

3

ACCEPTED MANUSCRIPT 55 56

Introduction The United States has the highest incidence rate of knee arthroplasty worldwide, with 235 procedures/100,000 habitants.1 In 2012 alone, the number of knee arthroplasties

58

performed in the United States was over 700,000, which was 9% of the 8 million non-

59

maternal and non-neonatal inpatient hospital stays where a procedures was performed.2

60

Because of the high incidence of knee arthroplasty and its historical increase,3-7

61

understanding the future healthcare demands of these procedures is important. Reliable

62

projections of future surgery demand can assist healthcare providers, leaders, and services

63

in preparing, through capacity building, staff training, financial commitment, and other

64

important resource building strategies.

65 66

M AN U

SC

RI PT

57

Three separate projections studies of future demand for knee arthroplasty in the US have been published.3-5 In 2007, Kurtz et al., estimated that from 2005 to 2030 a 673%

68

increase in the number of procedures would be observed, with an expected volume of 3.5

69

million procedures being performed yearly by 2030.3 In 2014, Kurtz et al. revaluated their

70

original models and new estimates were proposed, which were slightly lower for 2020,

71

however, no figures could be extrapolated from their published graphs.4 A third study, by

72

Bashnskaya et al. in 2012, proposed that a similar increase in volume as proposed by Kurtz et

73

al. 3 was expected for 2030, and projected that 3 million procedures would occur annually

74

by then.5 While these three studies have arrived at similarly high estimates of future

75

demand they have used different modelling approaches. The first study used a Poisson

76

regression model, which assumed an exponential growth in the incidence of knee

77

arthroplasty. The second and third studies used linear regression models, which also

78

assumed a continuous increase in demand. 4,5 Both models make projections that are not

AC C

EP

TE D

67

4

ACCEPTED MANUSCRIPT biologically plausible, or consistent with what we know about what drives the incidence of

80

these procedures. To illustrate this point, the estimated prevalence of knee osteoarthritis

81

according to the Global Burden of Disease 2010 study puts the global age-standardised

82

prevalence of symptomatic radiographically confirmed knee osteoarthritis around 3.8%,8

83

while the prevalence of knee osteoarthritis is estimated to be 10-13% among Americans.8,9

84

This relatively low number of people with osteoarthritis, the already high incidence of knee

85

arthroplasty in the US, and the fact that it is estimated that only about 50% of those with

86

symptomatic osteoarthrtitis have surgery in the United States10 suggest that there will be a

87

finite number of future patients undergoing surgery- and not a continuous projected growth

88

as proposed by other authors.

89

M AN U

SC

RI PT

79

There is no way to test which methodology is best suited for joint arthroplasty

90

projections, but biological plausibility and constraints to the healthcare system have

92

prompted researchers to explore alternative projection models for this patient population.11-

93

13

94

when overestimated and lack of preparedness when underestimated. Evaluating alternative

95

projection models will increase our understanding of the modelling limitations and

96

eventually reproducibility of findings will increase our confidence in projected estimates. In

97

this study we used the more conservative projection model proposed by Nemes et al. in

98

Sweden to estimate the future demand of knee arthroplasty in the US11,12 because there is

99

an upper limit to total knee arthroplasty (TKA) incidence that prior projections have not

100

considered. Failure of considering this upper limit, or asymptote, will ultimately lead to

101

biased projected rates.14 The model is a logistic model, which has more parameters than the

102

models used by the other US projection studies. The parameter of most interest in this

TE D

91

AC C

EP

Furthermore, inaccurate growth projection can lead to wasted or misplaced resources

5

ACCEPTED MANUSCRIPT model is the asymptote, which in this case is the estimated maximum incidence of knee

104

arthroplasty. This parameter has units that correspond to the outcome. The existence of

105

such a maximum incidence is a strong assumption and an untestable one; however,

106

employing such assumption is unlikely to attain numbers high enough to sustain the yearly

107

growth in knee arthroplasty proposed by the models in the current published projection

108

studies.

RI PT

103

109

The purpose of our study was to estimate the future incidence rate and volume of

111

primary TKA in the United States from 2015 to 2050 using an alternative projection model

112

that assumes a maximum incidence rate. Furthermore, our study compared these

113

projections to a model assuming exponential growth for illustrative purposes.

AC C

EP

TE D

M AN U

SC

110

6

ACCEPTED MANUSCRIPT 114

Methods A population based epidemiological study was conducted using data from US

116

National (previously Nationwide) Inpatient Sample (NIS)15 and US Census Bureau.16 The NIS

117

was developed by the Healthcare Cost and Utilisation Project (HCUP), which is sponsored by

118

the Agency for Healthcare Research and Quality (AHQR). This ~20% stratified sample of

119

patients who were discharged from 1000 hospital in 44 states in the US has been shown to

120

be 95% representative of the US population.15

SC

RI PT

115

121

Primary TKA procedures performed between 1993 and 2012 (most recent year

123

available) were identified in the NIS using International Classification of Disease, 9th revision,

124

Clinical Modification (ICD-9-CM) code 81.54. The new trend weights released with the 2012

125

NIS data were used to estimate procedure yearly volumes.17

M AN U

122

TE D

126

The US Census historical population estimates for 1993-2012 were obtained from the

127

intercensal estimates of the resident population by age18 and projected population

129

estimates for 2015 to 2050 were obtained from the projections of the population for the

130

US.19

AC C

131

EP

128

Historical changes in the total number of TKA per year were modelled with

132

piecewise-linear regression splines. Piecewise-linear regression splines fit multiple

133

regression lines to the data connected by a change-point. Model fitting and change-point

134

estimation followed the algorithms outlined by Mueggo.20 The number of change points

135

were selected by minimizing the Akaike Information Criterion, with the restriction that

136

change-points cannot be placed at the extremes of the data. The incidence rate and 95%

137

confidence intervals (CI) or prediction intervals (PI) of TKA per 100,000 US citizens over the 7

ACCEPTED MANUSCRIPT age of 40 years were calculated. Historical data was summarized as incidence rates and

139

associated 95%CI, while projections were presented with 95%PI. The incidence in the

140

population 40 years or older was chosen because the number of patients with joint

141

arthroplasty younger than 40 is minimal (0.7% of TKAs). The estimated incidence rate was

142

used as the outcome of a regression modelling with logistic regression and Poisson

143

regression equation. Logistic regression is a conservative projection model that assumes a

144

maximum incidence level exists and changes in estimates grow exponentially from the

145

beginning to the half-way point of the maximum incidence and then decrease until

146

maximum incidence is reached. This model is based on the logistic growth model and

147

assumes a S-shape growth curve. The model parameters have clear meanings. According to

148

this model the incidence in a time point is directly dependent on the incidence a time-unit

149

before and can be modelled as

dy ' y = β y 1 − dt A

TE D

150

151

M AN U

SC

RI PT

138

where y observed incidence, A the asymptote that has units equal to incidence and β the growth rate with units equal to t (years in our case). As one can see if the observed

153

incidence is far from the asymptote the growth is accentuated. However, as the observed

154

incidence y approaches the asymptote A the growth slows down. The model parameters are

155

estimated using its integrated form and routine optimization methods. We used the

156

following re-parametrised form

AC C

EP

152

A 1 + e(γ −t )/ξ

157 158

Where the parameters of interest are A the asymptote that has units equal to

159

incidence and γ a numeric parameter representing the year value at the inflection point of

160

the curve. 8

ACCEPTED MANUSCRIPT 161 162

By comparison, Poisson regression assumes exponential growth through the whole time period. We used grid-search to estimate the parameters of the logistic regression

164

equation. Grid search is an exploratory Monte Carlo optimization method.21,22 Parameters

165

of the Poisson regression were estimated using maximum likelihood method. 95%PIs were

166

constructed with Monte Carlo simulation for the logistic regression and with bootstrapping

167

for Poisson regression. 95%PIs for incidence rates were calculated with the Wilson method.23

168

All analyses were conducted using R 3.3.1.

SC

RI PT

163

TE D

Review Board was not required.

EP

171

This study uses publicly available de-identified data so review by an Institutional

AC C

170

M AN U

169

9

ACCEPTED MANUSCRIPT 172

Results Between 1993 and 2012 a total of 7.8 million primary TKAs were performed in US

173

and the number of surgeries increased by 224% (Figure 1, Appendix A). Historically, the

175

increase in yearly surgical volume was slower in the beginning of our study period then the

176

pace picked-up after year 2000 (95 % CI 1997-2002), with the average yearly increase up to

177

1999 of 10,520 (95% CI: 1490-19,540) cases/year. After 2000 the increase in yearly volume

178

accelerated to average of 32,730 (95% CI: 29,190 -36,270) procedures/year. The model had

179

good fit to the data, calendar year explaining 95.7 % of the observed variation in TKA

180

incidence.

M AN U

SC

RI PT

174

181

Logistic regression modelling suggests that the upper incidence of TKA/100,000

183

people over the age of 40 in the US is 746 (95%PI 117-1375). Using this model, the incidence

184

rate of TKA is expected to increase 69% by 2050 compared to 2012, from 429 (95%CI 374-

185

453) procedures/100,000 people over the age of 40 in 2012 to 725 (95%PI 121-1041) in 2050

186

(Figure 2). This change in incidence rate translates into a 143% projected increase in

187

procedure volume between 2012 and 2050 (Figure 3). The period between 1987- 2028

188

coincides with the fastest growth in yearly volume. This increase continuously accelerated

189

until 2007 after that growth rate while still increasing but the pace slows down (Appendix

190

B).

EP

AC C

191

TE D

182

192

Poisson regression modelling forecasted a greater increase of the incidence rate of TKAs

193

(Figure 2) than the logistic model. The incidence rate in 2050 using this model was 2854

194

(95%CI 2278-4004) procedures/100,000 people over the age of 40, which is a 565% increase

195

in incidence rate (Figure 2) and 855% projected increase in the volume (Figure 4) of 10

ACCEPTED MANUSCRIPT surgeries compared to 2012. According to the Poisson regression model every year there will

197

be 5% (95%CI: 4.49-5.44) more TKA surgeries than in the previous year (Appendix B).

AC C

EP

TE D

M AN U

SC

RI PT

196

11

ACCEPTED MANUSCRIPT 198

Discussion According to our projection model, a 143% increase in the number of TKAs

200

performed in the US between 2012 and 2050 is expected and we project that by 2050 the

201

number of TKAs performed annually will reach 1.5 million cases/year. The rate of growth of

202

these procedures started to slow down in 2007 and after 2028 it will decrease even more

203

substantially.

RI PT

199

204

For comparison with previous models, projections for 2020 from our logistic model

SC

205

suggests that 882,000 procedures will be performed annually (Appendix A). For the same

207

year Bashinskaya et al. forecasted 1.7 million yearly procedures, 5 Kurtz et al. 1.5 million in

208

their original projections,3 and around 1.4 million in their 2014 update.4 Additionally in our

209

Poisson regression, which is presented for illustrative purposes in this study, 1 million cases

210

were projected (Appendix A). The results of the previous projection analyses are much

211

higher compared to ours. Previous studies used linear or Poisson regression models. 3-5 The

212

different model used is the main reason for the differences in projections proposed in this

213

study and the other studies, but we highlight other notable differences among these studies

214

from ours. Bashinskaya et al. stratified modelling by age groups while Kurtz et al. used a

215

wider range of predictors, including age, gender, ethnicity, census region and calendar year

216

in one study, and added economic parameters in the other study. 3-5 Kurtz et al. in their

217

later publication4 followed the footsteps of Bashinskaya et al.5 and separated predictions for

218

each age, gender, ethnicity and census region and combined the projections. This resulted in

219

somewhat lower projected incidence, however, still resulted in exponential like growth

220

curves (Kurtz et al. figure 4).4 Generally, linear regression commands a constant growth thus

221

the exponential like growth curve of the aforementioned paper is likely to be the result of

AC C

EP

TE D

M AN U

206

12

ACCEPTED MANUSCRIPT the weighting attributed to the different stratum. Given the splines used by Bashinskaya et

223

al. and the weighting strategy used by Kurtz et al. one cannot draw far reaching conclusions

224

of the appropriateness of the models. 4,5 However, because the previous studies used

225

models that assume ever increasing constant growth they may not reflect biological

226

plausibility, or consistency with what we know about the disease- that the need for the

227

procedures are limited by the prevalence of the disease and the subset with severe disease

228

requiring TKA. Other differences in the Poisson estimates between our study and previous

229

studies could be due to our inclusion of more years of historical annual incidence from the

230

NIS in our projections, the use of recently updated trend weights from NIS in our volume

231

calculations, and our restriction of the population to those 40 years and older. This age

232

restriction was chosen in an attempt to provide more realistic incidence rate estimates and

233

remove the impact of a greater population increase in the younger groups to affect our

234

population estimations.

236

SC

M AN U

TE D

235

RI PT

222

The logistic model we employed incorporates maximum incidence. So far only one prior study has compared these two regression methods. Nemes et al. projected the

238

incidence and number of total hip arthroplasty (THA) in Sweden from 2013 to 2030.11 As of

239

2016, there are publicly available data for 2013, 2014 and 2015 THA volume in Sweden,

240

which allowed us to compare the projections with actual surgeries. Using the Poisson

241

regression model in 2013 the authors projected 20,006 THAs, but only 16,348 were

242

performed, similarly, in 2014 20,891 THAs were projected but only 16,562 were observed,

243

and in 2015 21,816 THAs were projected and 16,617 were observed. Using an asymptotic

244

regression model, closer estimates to the observed values were projected: 16,021 for 2013,

AC C

EP

237

13

ACCEPTED MANUSCRIPT 245

16,318 for 2014 and 16, 595 for 2015. These numbers are slightly lower that the realized

246

numbers.

247 248

Our projections are inheritably limited by the modelling method used. The model only accounts for past and recent incidence of knee arthroplasty and projected population

250

growth, and therefore it does not account for future changes in patient, surgeon, healthcare

251

system, political, and other population factors. Some of these factors could directly affect

252

the estimations, for example, the prevalence of symptomatic osteoarthritis, which is

253

increasing nationally due to the aging and increasingly obese population9,24 Other factors

254

could affect the estimations indirectly, for example healthcare policy, which will certainly

255

change in the next 30 years, could change patients’ access to surgery. Also important are

256

general extraneous factors like economic recessions, which occurred in the US in the 1990s

257

and 2000s, which could have influenced the historical trends we observed. These factors

258

could not be accounted for in our analysis due to data availability and modelling constraints

259

and this likely means our projections could be underestimated. We support the conservative

260

projection approach carried out in this study, however, we understand that possible

261

underestimation of future volumes can contribute to a lack of preparedness and appropriate

262

resource allocation. We hope that the limitations of our strategy, as in other models, are

263

therefore clear. Our projections are also limited by the time period they cover, which means

264

that the longer term projections can be more unreliable.

AC C

EP

TE D

M AN U

SC

RI PT

249

265 266

Even after using a more conservative projection approach than previously proposed,

267

the number of TKA procedures in the US, which already has the highest incidence rate of

268

knee arthroplasty in the world, is expected to increase 143% by 2050. The projected increase 14

ACCEPTED MANUSCRIPT will translate into a projected yearly volume of 1.5 million cases by 2050. With the proposed

270

model and all assumptions made, while the rate of demand started to slow down in 2007,

271

and is expected to slow even more significantly after 2028, overall growth still continues.

272

And as we modelled incidence pre 100,000 individuals older than 40 the absolute numbers

273

of performed TKA will continue to increase in concordance in the aging of the population.

274

These projection models can be useful in future planning and resource allocation for the

275

delivery of necessary care to knee arthroplasty patients in the US.

AC C

EP

TE D

M AN U

SC

RI PT

269

15

ACCEPTED MANUSCRIPT Acknowledgements: This work was supported by an Australian Government National Health

277

and Medical Research Council (NHMRC) Centre of Research Excellence in Post-Marketing

278

Surveillance of Medicines and Medical Devices grant (GNT1040938). No financial support or

279

other benefits from commercial sources was received by any of the authors for the work

280

reported on in the manuscript. The funders played no role in the analysis or interpretation of

281

the data or in the preparation of this manuscript. The authors had full access to all of the

282

data in the study.

SC

RI PT

276

283

Authors’ contributions:

285

Conception and design: MCSI, SEG, EWP, RSN, SN

286

Acquisition of data: MCSI

287

Analysis and interpretation of data: MCSI, SN

288

Drafting and revising manuscript: MCSI, SEG, EWP, RSN, SN

289

Final approval: MCSI, SEG, EWP, RSN, SN

TE D

290

M AN U

284

Role of the funding source: This work was supported by an Australian Government National

292

Health and Medical Research Council (NHMRC) Centre of Research Excellence in Post-

293

Marketing Surveillance of Medicines and Medical Devices grant (GNT1040938). No financial

294

support or other benefits from commercial sources was received by any of the authors for

295

the work reported on in the manuscript.

AC C

EP

291

296 297

Conflict of interest: None of the authors have any conflict of interest.

16

ACCEPTED MANUSCRIPT 298

References

299

1.

countries. Osteoarthritis Cartilage. 2015;23(10):1664-1673.

300 301

Pabinger C, Lothaller H, Geissler A. Utilization rates of knee-arthroplasty in OECD

2.

Healthcare Cost and Utilization Project (HCUP). Nationwide inpatient sample (NIS).Statistical Brief #186. 2014; http://www.hcup-

303

us.ahrq.gov/reports/statbriefs/sb186-Operating-Room-Procedures-United-States-

304

2012.pdf. Accessed September 16th, 2015. 3.

Kurtz S, Ong K, Lau E, Mowat F, Halpern M. Projections of primary and revision hip

SC

305

RI PT

302

and knee arthroplasty in the United States from 2005 to 2030. The Journal of bone

307

and joint surgery. American volume. 2007;89(4):780-785.

308

4.

M AN U

306

Kurtz SM, Ong KL, Lau E, Bozic KJ. Impact of the economic downturn on total joint

309

replacement demand in the United States: updated projections to 2021. The Journal

310

of bone and joint surgery. American volume. 2014;96(8):624-630. 5.

Bashinskaya B, Zimmerman RM, Walcott BP, Antoci V. Arthroplasty Utilization in the

TE D

311 312

United States is Predicted by Age-Specific Population Groups. ISRN orthopedics.

313

2012;2012. 6.

Cram P, Lu X, Kates SL, Singh JA, Li Y, Wolf BR. Total knee arthroplasty volume,

EP

314

utilization, and outcomes among Medicare beneficiaries, 1991-2010. JAMA : the

316

journal of the American Medical Association. 2012;308(12):1227-1236.

317

7.

Khatod M, Inacio M, Paxton EW, et al. Knee replacement: epidemiology, outcomes, and trends in Southern California: 17,080 replacements from 1995 through 2004.

318

Acta orthopaedica. 2008;79(6):812-819.

319 320

AC C

315

8.

Cross M, Smith E, Hoy D, et al. The global burden of hip and knee osteoarthritis:

321

estimates from the global burden of disease 2010 study. Annals of the rheumatic

322

diseases. 2014;73(7):1323-1330.

17

ACCEPTED MANUSCRIPT 323

9.

2010;26(3):355-369.

324 325

Zhang Y, Jordan JM. Epidemiology of osteoarthritis. Clinics in geriatric medicine.

10.

Weinstein AM, Rome BN, Reichmann WM, et al. Estimating the burden of total knee replacement in the United States. The Journal of bone and joint surgery. American

327

volume. 2013;95(5):385-392. 11.

Sweden from 2013 to 2030. Acta orthopaedica. 2014;85(3):238-243. 12.

arthroplasty in Sweden. Acta orthopaedica. 2015:1-6.

331 332

Nemes S, Rolfson O, A WD, et al. Historical view and future demand for knee

SC

329 330

Nemes S, Gordon M, Rogmark C, Rolfson O. Projections of total hip replacement in

13.

Culliford D, Maskell J, Judge A, et al. Future projections of total hip and knee

M AN U

328

RI PT

326

333

arthroplasty in the UK: results from the UK Clinical Practice Research Datalink.

334

Osteoarthritis Cartilage. 2015;23(4):594-600.

335

14.

Park EW, Lim SM. Empirical estimation of the asymptotes of disease progress curves and the use of the Richards generalized rate parameters for describing disease

337

progress. Phytopathology. 1985;75(1):786-791.

338

15.

TE D

336

Overview of the National (Nationwide) Inpatient Sample (NIS). HCUP Databases. Healthcare Cost and Utilization Project (HCUP). 2016; https://www.hcup-

340

us.ahrq.gov/nisoverview.jsp. Accessed August 20, 2016. 16.

2014 to 2060. Current Population Reports, P25-1143 2014;

342

http://www.census.gov/content/dam/Census/library/publications/2015/demo/p25-

343

1143.pdf. Accessed August 20, 2016.

344 345

Colby SL, Ortman JM. Projections of the Size and Composition of the U.S. Population:

AC C

341

EP

339

17.

Healthcare Cost and Utilization Project (HCUP). National inpatient sample (NIS).

346

Trend Weights for HCUP NIS Data. 2015; https://www.hcup-

347

us.ahrq.gov/db/nation/nis/trendwghts.jsp. Accessed July 1st, 2015.

18

ACCEPTED MANUSCRIPT 348

18.

http://www.census.gov/popest/. Accessed July 1, 2015.

349 350

US Census Bureau. Population and Housing Unit Estimates. 2016;

19.

US Census Bureau. 2014 National Population Projections: Summary Tables. 2014; http://www.census.gov/population/projections/data/national/2014/summarytables.

352

html. Accessed July 1st, 2015.

353

20.

Muggeo VMR. Estimating regression models with unknown break-points. Statistics in Medicine. 2003;22:3055-3071.

354

RI PT

351

21.

Christian R, Casella G. Monte Carlo Statistical Methods.

356

22.

Liu JS. Monte Carlo Strategies in Scientific Computing. Springer.

357

23.

Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial proportion (with

M AN U

discussion). Statistical Science. 2001;16:101-103.

358 359

SC

355

24.

Losina E, Thornhill TS, Rome BN, Wright J, Katz JN. The dramatic increase in total knee replacement utilization rates in the United States cannot be fully explained by

361

growth in population size and the obesity epidemic. The Journal of bone and joint

362

surgery. American volume. 2012;94(3):201-207.

EP

364

AC C

363

TE D

360

19

ACCEPTED MANUSCRIPT Figure Legend

366

Figure 1.

367

Title: US Population and Primary Total Knee Arthroplasty Volume in Citizens Aged 40 Years

368

and Older, 1993-2012

369

Footnote: TKA=Total knee arthroplasty.

RI PT

365

370

Figure 2.

372

Title: Historical (1993-2012) and Projected (2015-2050) Incidence Rate (per 100,000) of

373

Primary Total Knee Arthroplasty Procedures in Citizens 40 Years and Older from Logistic and

374

Poisson Models

375

Footnote: TKA=total knee arthroplasty.

376

M AN U

SC

371

Figure 3.

378

Title: Historical (1993-2012) and Projected (2015-2050) Volume and 95% Prediction Intervals

379

of Primary Total Knee Arthroplasty Procedures in Citizens 40 Years and Older from Logistic

380

Model

381

Footnote: N=Number of total knee arthroplasty procedures.

EP

AC C

382

TE D

377

383

Figure 4.

384

Title: Historical (1993-2012) and Projected (2015-2050) Volume and 95% Prediction Intervals

385

of Primary Total Knee Arthroplasty Procedures in Citizens 40 Years and Older from Poisson

386

Model

387

Footnote: N=Number of total knee arthroplasty procedures.

388 20

ACCEPTED MANUSCRIPT Appendix A. US Population, Primary Total Knee Arthroplasty Volume, and Annual Crude Incidence Rate per 100,000 Citizens Aged 40 Years or Older, 1993-2012 Years

US

US Population

Volume of

Population

>40 years old

TKA, N

IR of TKA in US Population >40 years old (95%CI)

259,918,588

101,979,186

194,914

1994

263,125,821

104,326,960

208,440

1995

266,278,393

106,788,713

217,725

1996

269,394,284

109,298,512

241,687

1997

272,646,925

112,003,174

252,979

226 (197-256)

1998

275,854,104

114,684,128

245,763

214 (187-244)

1999

279,040,168

117,347,814

256,428

219 (190-248)

2000

282,162,411

120,015,599

274,463

229 (192-251)

2001

284,968,955

122,597,688

305,572

249 (211-272)

2002

287,625,193

125,107,164

339,681

272 (232-295)

2003

290,107,933

127,582,475

369,985

290 (249-314)

2004

292,805,298

130,064,179

419,774

323 (278-347)

2005

295,516,599

132,494,084

483,067

365 (317-390)

298,379,912

134,682,947

482,689

358 (310-383)

301,231,207

136,726,423

533,602

390 (339-415)

200 (173-228) 204 (176-232) 221 (193-252)

SC

M AN U

TE D

AC C

2007

EP

2006

191 (165-220)

RI PT

1993

2008

304,093,966

138,770,154

592,323

427 (373-452)

2009

306,771,529

140,875,878

597,541

424 (370-449)

2010

309,349,689

143,116,396

632,862

442 (386-467)

2011

311,582,564

145,220,219

618,604

426 (371-450)

2012

313,873,685

147,025,546

631,214

429 (374-453)

TKA=Total knee arthroplasty. IR=Incidence rate. CI=Confidence interval.

ACCEPTED MANUSCRIPT Appendix B. Projected Annual Incidence Rate per 100,000 Citizens Aged 40 Years or Older and Volumes, 2015-2050 Projected TKA IR (95%

Projected US

Approach

PI)

Population >40 Years Old

in TKA Volume

152,041,000

733,021 (178,854-760,519)

2020

548 (117-581)

160,871,000

882,034 (187,892-934,370)

20.3

2025

603 (115-661)

170,446,000

1,027,494 (196,487-1,127,105)

16.5

2030

645 (117-741)

180,381,000

1,163,697 (211,391-1,336,413)

13.3

2035

676 (120-819)

2040

699 (118-896)

2045

714 (118-970)

2050

725 (121-1041)

2015

522 (471-588)

152,041,000

794,211 (716,075-894,001)

2020

666 (591-766)

160,871,000

1,071,061 (950,707-1,232,312)

34.9

2025

849 (747-1011)

170,446,000

1,446,387 (1,273,189-1,723,209)

35.0

1082 (940-1331)

180,381,000

1,950,967 (1,695,536-2,400,916)

34.9

2030

M AN U

SC

482 (118-500)

190,195,000

1,286,531 (228,050-1,557,975)

10.6

198,034,000

1,383,809 (234,512-1,774,031)

7.6

204,851,000

1,463,313 (242,484-1,987,381)

5.7

211,259,000

1,531,566 (254,622-2,198,785)

4.7

TE D

EP

Poisson

% Change

2015

AC C

Logistic

Projected TKA Volume, N (95%PI)

RI PT

Modelling Year

2035

1379 (1175-1748)

190,195,000

2,621,920 (2,234,791-3,324,704)

34.4

2040

1757 (1467-2307)

198,034,000

3,479,536 (2,905,109-4,568,941)

32.7

2045

2239 (1825-3034)

204,851,000

4,5875,52 (3,738,428-6,215,640)

31.8

2050

2854 (2278-4004)

211,259,000

6,030,029 (4,812,322-8,459,655)

31.4

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Copyright © 2021 COEK.INFO. All rights reserved.