Administrative Databases Can Yield False Conclusions—An Example of Obesity in Total Joint Arthroplasty

Accepted Manuscript Administrative Databases Can Yield False Conclusions – An Example of Obesity in Total Joint Arthroplasty Jaiben George, MBBS, Jared M. Newman, MD, Deepak Ramanathan, MBBS, Alison K. Klika, MS, Carlos A. Higuera, MD, Wael K. Barsoum, MD PII:

S0883-5403(17)30084-0

DOI:

10.1016/j.arth.2017.01.052

Reference:

YARTH 55638

To appear in:

The Journal of Arthroplasty

Received Date: 14 November 2016 Revised Date:

9 January 2017

Accepted Date: 29 January 2017

Please cite this article as: George J, Newman JM, Ramanathan D, Klika AK, Higuera CA, Barsoum WK, Administrative Databases Can Yield False Conclusions – An Example of Obesity in Total Joint Arthroplasty, The Journal of Arthroplasty (2017), doi: 10.1016/j.arth.2017.01.052. 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

Administrative Databases Can Yield False Conclusions – An Example of Obesity in Total Joint Arthroplasty

RI PT

Jaiben George1, MBBS; Jared M. Newman1, MD; Deepak Ramanathan1, MBBS; Alison K. Klika1, MS; Carlos A. Higuera1, MD, Wael K. Barsoum1, MD

1

TE D

M AN U

SC

Department of Orthopaedic Surgery, Cleveland Clinic, 9500 Euclid Ave., Cleveland, OH 44195. [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]

AC C

EP

Corresponding Author: Alison K. Klika, MS Department of Orthopaedic Surgery Cleveland Clinic – A41 9500 Euclid Ave. Cleveland, OH 44195 216-444-4954 [email protected]

ACCEPTED MANUSCRIPT

1

Administrative Databases Can Yield False Conclusions – An Example of Obesity in Total

2

Joint Arthroplasty ABSTRACT

4

Background

5

Research using large administrative databases has substantially increased in recent years.

6

Accuracy with which comorbidities are represented in these databases has been questioned. The

7

purpose of this study was to evaluate the extent of errors in obesity coding and its impact on

8

arthroplasty research.

9

Methods

M AN U

SC

RI PT

3

18,030 primary total knee arthroplasties (TKA) and 10,475 total hip arthroplasties (THA)

11

performed at a single healthcare-system from 2004-2014 were included. Patients were classified

12

as obese/non-obese by two methods: 1) BMI ≥30 and 2) ICD-9 codes. Length of stay, operative

13

time, and 90-day complications were collected. Effect of obesity on various outcomes was

14

analyzed separately for both BMI- and coding-based obesity.

15

Results

16

From 2004-2014, the prevalence of BMI-based obesity increased from 54% to 63% and 40% to

17

45% in TKA and THA, respectively. The prevalence of coding-based obesity increased from

18

15% to 28% and 8% to 17% in TKA and THA, respectively. Coding overestimated the growth of

19

obesity in TKA and THA by 5.6 and 8.4 times, respectively. When obesity was defined by

20

coding, obesity was falsely shown to be a significant risk factor for deep vein thrombosis (TKA),

21

pulmonary embolism (THA) and longer hospital stay (TKA and THA).

22

Conclusion

23

The growth in obesity observed in administrative databases may be an artifact due to

24

improvements in coding over the years. Obesity defined by coding can overestimate the actual

25

effect of obesity on complications after arthroplasty. Therefore, studies using large databases

26

should be interpreted with caution, especially when variables prone to coding errors are

27

involved.

AC C

EP

TE D

10

1

ACCEPTED MANUSCRIPT

28 29

KEYWORDS: Obesity. Total joint arthroplasty, Coding, Administrative database

30 INTRODUCTION

RI PT

31

There is an increasing number of studies being performed that are using administrative

33

databases such as the Nationwide Inpatient Sample, national Medicare claims data, and State

34

Inpatient Database [1]. However, administrative databases depend on international classification

35

of disease, 9th edition (ICD-9) codes to identify procedures, comorbidities and complications.

36

ICD-9 codes were initially designed for obtaining information on important measures like

37

mortality and morbidity[2]. But, the application of ICD-9 codes gradually expanded over the

38

years to healthcare policy and health economy, and is currently used predominantly to determine

39

hospital reimbursements by payers including the Centers for Medicare & Medicaid Services

40

(CMS) [1,2]. With the increasing accessibility to billing claims and the presence of larger sample

41

sizes, administrative databases have emerged as a popular tool among researchers.

42

Unfortunately, the reliability of these databases in research is disputed as previous studies have

43

reported inaccuracies with ICD-9 codes[3–5].

TE D

M AN U

SC

32

Obesity is a widely prevalent (over 60%) comorbidity among patients undergoing total

45

joint arthroplasty (TJA) [6,7]. In addition to being an established risk factor for TJA, it is also a

46

known risk factor for a number of complications following TJA[8–10]. Therefore, most studies

47

include obesity and/or body mass index (BMI) as an important demographic variable and adjust

48

for its presence while performing statistical analyses[11,12]. Since obesity is an important

49

comorbidity in patients undergoing TJA, accurate representation of its presence is critical.

50

Unfortunately, it is one of the most common conditions that is under coded in administrative

51

databases[3]. While it has been shown that the sensitivity of obesity coding is low, the effects of

52

these coding discrepancies on the results of studies based on administrative databases are yet to

53

be reported[4].

AC C

EP

44

54

The objectives of this study were to: 1) evaluate the prevalence of obesity among primary

55

total knee arthroplasty (TKA) and total hip arthroplasty (THA) using BMI-based and ICD-9

2

ACCEPTED MANUSCRIPT

56

coding-based classifications; and 2) compare the effects obesity has on complications following

57

TKA and THA when it is defined by BMI versus by ICD-9 codes.

59

METHODS

60

Study design

RI PT

58

This retrospective study was reviewed and approved by the institutional review board.

62

All adult patients who underwent primary TKA and THA at a single health system from 2004-

63

2014 were identified using a query of the electronic medical records. Patient demographics,

64

comorbidities, surgical information, including operative time, length of hospital stay (LOS), and

65

90-day complications were obtained from the electronic medical charts. All patients were

66

classified as obese or non-obese using two methods: 1) BMI and 2) ICD-9 diagnosis codes. BMI-

67

based obesity was defined as having a BMI ≥30 kg/m2 at the time of the surgery in accordance

68

with the WHO classification of obesity[13]. Coding-based obesity was defined as the presence

69

of one or more of the following ICD-9 codes during the admission for TKA/THA: 278.0, 278.00,

70

278.01, 278.03, 649.10-14, 793.91, V85.30-39, V85.41-45, V85.54. These were the same ICD-9

71

codes used by Nationwide Inpatient Sample to identify obesity in the comorbidity classification

72

software[14]. The ICD-9 codes associated with a particular admission of TKA/THA were

73

obtained from the electronic query of discharge records of the respective admission. The

74

complications assessed in this study were obtained using ICD-9 diagnosis codes and include

75

deep infection/prosthetic joint infection (Codes: 996.66-7, 996.69) superficial infection (Codes:

76

682.6, 686.9, 890, 998.5, 998.51, 998.6, 998.83, 998.59), deep vein thrombosis (DVT) (Codes:

77

451.1, 451.11, 451.19, 451.2, 451.8, 451.9, 453.2, 453.4, 453.89, 453.9), pulmonary embolism

78

(PE) (Codes: 415.11, 415.12, 415.13, 415.19), wound dehiscence/drainage (Codes: 998.3,

79

998.31-2, 890, 890.1-2, 891, 891.1-2, 894, 894.1-2 ) and wound hematoma/bleeding into the

80

joint (Codes: 719.15, 719.16, 998.11-3).

81

Study population

AC C

EP

TE D

M AN U

SC

61

82

From 2004-2014, a total of n=30,110 primary TKA/THA were identified. After excluding

83

n=1,605 surgeries missing information on BMI, 28,505 surgeries (TKA=18,030, THA=10,475)

84

were included. The mean age of the cohort was 64 (±12) years. There were 11,936 (42%) males 3

ACCEPTED MANUSCRIPT

and 16,569 (58%) females. Based on BMI, n=11,065 (61%) of TKAs and 4,608 (44%) of THAs

86

were performed in obese individuals, whereas based on codes, 4,102 (23%) of TKAs and 1,459

87

(14%) of THAs were performed in obese individuals.

88

Statistical analysis

RI PT

85

The annual trends in the prevalence of BMI-based and coding-based obesity among

90

primary TKA/THA patients were studied. The sensitivity of coding was defined as the

91

percentage of obese patients (based on BMI) correctly classified as obese by ICD-9 coding.

92

Similarly, specificity of coding was defined as the percentage of non-obese patients (based on

93

BMI) who did not receive an ICD-9 code for obesity. The effect of obesity on operative time,

94

LOS, and 90-day complications were analyzed separately for BMI-based and coding-based

95

obesity, i.e. whether the risk for any complication in obese patients was different depending on

96

the way obesity was defined.

M AN U

SC

89

Simple linear regression analysis was used to evaluate the yearly trends in the prevalence

98

of obesity based on BMI and ICD-9 coding. Univariate logistic regression analysis was used to

99

estimate the effect of obesity on categorical outcomes and univariate linear regression was used

100

to estimate the effect of obesity on continuous variables. Statistical analyses were performed

101

with R software (version 3.1.3; R Foundation for Statistical Computing, Vienna, Austria). A p-

102

value of less than 0.05 was used to determine statistical significance.

103 104

RESULTS

EP

TE D

97

The prevalence of BMI-based obesity increased significantly in TKA (p<0.001) and THA

106

(p<0.001), as did the prevalence of coding-based obesity in TKA (p<0.001) and THA (p<0.001).

107

From 2004 to 2014, the prevalence of BMI-based obesity in TKA increased from 54% to 63%,

108

which was a growth of 15% (Figure 1). During the same period, the prevalence of coding-based

109

obesity increased from 15% to 28%, a growth of 85%. ICD-9 coding overestimated the

110

percentage growth of obesity in TKA by 5.6 times. In THA, the prevalence of BMI-based

111

obesity increased from 40% to 45%, a growth of 13%, while coding-based obesity increased

112

from 8% to 17%, a growth of 110% (Figure 1). In THA, ICD-9 coding overestimated the

113

percentage growth of obesity by 8.4 times.

AC C

105

4

ACCEPTED MANUSCRIPT

Overall, the sensitivity of coding was 35% and 28% in TKA and THA, respectively. The

115

sensitivity was significantly higher in TKA compared to THA (35% versus 28%, p<0.001).

116

Linear regression analysis showed a significant annual increase in the sensitivity of obesity

117

coding in both TKA (p<0.001) and THA (p<0.001). The sensitivity of obesity coding increased

118

from 26% to 42% and 19% to 34% in TKA and THA, respectively (Figure 2). Overall, the

119

specificity of coding was 96% and 97% in TKA and THA, respectively. The specificity was

120

significantly higher in THA compared to TKA (97% versus 96%, p<0.001).

RI PT

114

The mean BMI of coding-based obese patients was significantly higher than that of BMI-

122

based obese patients (38.7 ± 7.1 versus 36.8 ± 5.9, p<0.001). As BMI increased, the proportion

123

of patients receiving an ICD-9 code for obesity increased in TKA (p<0.001) and THA (p<0.001)

124

(Figure 3). For example, 63% of TKA patients and 62% of THA patents with BMI ≥ 45 kg/m2

125

received a code for obesity, while only 20% of TKAs and 17% of THAs in the BMI range 30-35

126

kg/m2 received an obesity code. In TKA, when obesity was defined by coding, obesity was

127

falsely shown to be a significant risk factor for DVT and increased LOS (Table 1). In THA,

128

when obesity was defined by coding, obesity was falsely shown to be a significant risk factor for

129

PE and increased LOS (Table 1). Although obesity increased the risk of other complications

130

irrespective of the way it was defined (BMI/ICD-9 coding), there were differences in the effect

131

size of obesity. For all of the outcomes assessed in this study, with the exception of DVT in

132

THA, the effect size was higher when coding-based obesity was used (Tables 1 and 2). For

133

example, the odds ratio for deep infection after TKA was 2.41 (95% confidence interval [CI],

134

1.85-3.13, p<0.001) with coding-based obesity, while it was only 1.53 (95% CI, 1.16-2.06,

135

p=0.003) when BMI-based obesity was used.

137 138

M AN U

TE D

EP

AC C

136

SC

121

DISCUSSION

Administrative databases are being increasingly used in orthopaedic surgery research and

139

these studies often provide valuable information to surgeons, hospital and payers in making

140

clinical and administrative decisions. However, as these databases rely on coding, the reliability

141

of the conclusions made from such studies are questionable. The present study was undertaken to

142

assess the extent of coding errors with respect to obesity in TJA and to evaluate the implications

143

of these errors on the conclusions of the studies. The results of the present study indicate that the 5

ACCEPTED MANUSCRIPT

accuracy of coding might have changed over the years and can influence the results of the studies

145

reporting trends. Additionally, the patients who are assigned an obesity code are likely to have a

146

higher grade of obesity. With the presence of such systematic errors in coding, the effects of

147

obesity on various outcome measures may be overestimated when a coding-based obesity is

148

used.

RI PT

144

The prevalence of obesity based on BMI was similar to other single institution studies

150

[6,15,16]. Kremers et al [16] reported that the prevalence of obesity among primary TKA was

151

above 50% in a single institutional study of more than 6,000 patients. In another study from the

152

same institution on THA patients, the reported prevalence of obesity was close to 40%[15].

153

Meanwhile, in the NIS database, the prevalence of obesity among primary TKA during the

154

period 2002-2009 was reported to be 15%[17]. This is similar to the estimates of the code-based

155

obesity estimated in the present study. The poor sensitivity of obesity coding reported in the

156

present study is in agreement with other studies[3,18]. As obesity is highly prevalent among the

157

TJA population, it is likely to be overlooked as a comorbidity by medical professionals. As

158

coders usually assign the obesity codes based on the diagnosis available in the charts, rather than

159

the BMI, obesity is highly prone to underreporting. Although the management of the patient is

160

unlikely to be affected by the lack of a diagnosis code for obesity, the consequences on research

161

could be prominent, as is evident from the present study.

TE D

M AN U

SC

149

The growth in the prevalence of obesity varied for BMI-based and coding-based obesity.

163

These changes in the prevalence of obesity over the years are consistent with the findings in

164

other studies[6,15–17]. In a NIS database study, the prevalence of obesity in primary TKA

165

increased by almost 100% from 2002 to 2009, while an increase of only 20% was reported in a

166

study from a single institution suggesting that the increase in prevalence observed in the NIS

167

database is likely due to coding errors[16,17]. With continued efforts to improve coding and the

168

recognition of obesity as an important comorbidity, more patients are correctly identified as

169

obese as denoted by the improvements in sensitivity over the years.

170

improvements in coding of obesity is desired, this could lead to overestimation of the growth in

171

prevalence of obesity when studies attempt to evaluate data across the years. Although the

172

present study evaluated obesity, any variable with changes in coding accuracy over time is likely

AC C

EP

162

6

Even though the

ACCEPTED MANUSCRIPT

173

to be affected. Furthermore, periodic revisions to the coding systems can also complicate the

174

analysis of longitudinal trends when older codes are replaced with more specific newer codes. The present study also suggests that the coding errors could be systematic and not

176

necessarily a random event. The proportion of patients receiving an obesity ICD-9 code

177

increased as the BMI increased, suggesting that more obvious cases of obesity are likely to be

178

correctly classified as obese. This means that the patient population classified as obese in the

179

administrative databases need not be a uniform representation of the actual obese population,

180

which in turn can affect the results of a study. As surgical outcomes are dependent on the

181

severity of obesity, with higher rates of complications reported in morbid and super obese

182

patients, it is not surprising that the effects of obesity are overestimated when a code based

183

obesity is used[19–21]. When using codes, obesity was falsely shown to be a significant risk

184

factor for certain outcomes like thromboembolic events and LOS. Moreover, the effect of obesity

185

on other outcomes like infection and wound related complications were more prominent when

186

code based obesity was employed. Numerous studies have reported mixed results about the risk

187

of various complications among obese patients, possibly due to variations in sample sizes,

188

criteria for definition of outcomes and the presence of undetected confounders[8–10].

189

Nevertheless, the present study suggests that the results of studies can be affected when coding-

190

based obesity is used.

TE D

M AN U

SC

RI PT

175

There are notable limitations to this study. The study was based on a single institution

192

healthcare data system and may not be generalizable to other centers. However, with the large

193

sample size of this study, the results are expected to be in agreement with the national estimates,

194

as detailed in the discussion. Moreover, the prevalence of coding-based obesity reported in this

195

study was similar to that reported in the actual NIS database, indicating that the accuracy of the

196

coding at this institution is comparable to other centers in the United States. This was a

197

retrospective study and the outcomes were obtained from the electronic query of medical records

198

using ICD-9 codes. Manual chart review of the entire patient cohort was beyond the scope of the

199

present study and hence, the accuracy of the outcomes is a concern. Additionally, the diagnostic

200

criteria and the severity of these outcomes are subject to significant variations across the years

201

and surgeons, and were not accounted for in this study. However, such heterogeneity is also

202

expected to be present in the outcomes identified in large administrative databases which involve

AC C

EP

191

7

ACCEPTED MANUSCRIPT

multiple hospitals and is unlikely to affect the conclusions of this study. Also, only a limited

204

number of outcomes were evaluated in this study and the effect of coding errors on other

205

outcomes remains unknown. The present study considered obesity as a single category and did

206

not evaluate the effects of varying severity of obesity. The codes used to define obesity

207

encompass a wide range of codes for various grades of obesity. The present study did not

208

evaluate the accuracy of each individual code as these codes are often grouped together in

209

research studies and administrative databases[14,17]. Golinvaux et al [4] showed that the

210

sensitivity of coding for morbid obesity was better than that of obesity in a single institutional

211

study on 2,000 patients undergoing spine surgery. Therefore, it is possible that the extent of

212

coding errors might have been different for codes specifically denoting morbid obesity or other

213

severe forms of obesity. Although the present study suggests that administrative databases can

214

yield false conclusions, administrative databases might still be useful tools when used

215

appropriately and interpreted carefully. They provide invaluable information about a large

216

number of patients and can answer many research questions, which would have otherwise been

217

impossible to answer. Also, the extent of errors observed in obesity coding may not be evident in

218

other diagnosis/procedure codes.

219

procedures/conditions,

220

reimbursements are likely to be underreported. Finally, with the introduction of ICD-10 codes,

221

the effect of coding inaccuracies in the future is uncertain[1].

The coding might be largely accurate for billable

minor

comorbid

conditions

which

seldom

determine

TE D

while

M AN U

SC

RI PT

203

In summary, the prevalence of obesity was significantly lower when estimated using

223

ICD-9 coding. Administrative databases might overestimate the growth of obesity due to

224

improvements in coding over the years. The results of studies may be influenced by the way

225

obesity is defined. Obesity defined by coding can overestimate the actual effects that obesity has

226

on complications after TJA. Therefore, studies using large databases should be interpreted with

227

caution, especially if variables prone to coding errors, such as obesity, are involved.

AC C

228

EP

222

229 230

REFERENCES

231 8

ACCEPTED MANUSCRIPT

232

[1]

Pugely AJ, Martin CT, Harwood J, Ong KL, Bozic KJ, Callaghan JJ. Database and

233

Registry Research in Orthopaedic Surgery: Part I: Claims-Based Data. J Bone Joint Surg

234

Am 2015;97:1278–87. doi:10.2106/JBJS.N.01260. [2]

O’Malley KJ, Cook KF, Price MD, Wildes KR, Hurdle JF, Ashton CM. Measuring

RI PT

235 236

diagnoses: ICD code accuracy. Health Serv Res 2005;40:1620–39. doi:10.1111/j.1475-

237

6773.2005.00444.x.

238

[3]

Martin B-J, Chen G, Graham M, Quan H. Coding of obesity in administrative hospital discharge abstract data: accuracy and impact for future research studies. BMC Health Serv

240

Res 2014;14:70. doi:10.1186/1472-6963-14-70. [4]

Golinvaux NS, Bohl DD, Basques BA, Fu MC, Gardner EC, Grauer JN. Limitations of

M AN U

241

SC

239

242

administrative databases in spine research: a study in obesity. Spine J 2014;14:2923–8.

243

doi:10.1016/j.spinee.2014.04.025.

244

[5]

Losina E, Barrett J, Baron JA, Katz JN. Accuracy of Medicare claims data for rheumatologic diagnoses in total hip replacement recipients. J Clin Epidemiol

246

2003;56:515–9.

247

[6]

TE D

245

Fehring TK, Odum SM, Griffin WL, Mason JB, McCoy TH. The obesity epidemic: its

248

effect on total joint arthroplasty. J Arthroplasty 2007;22:71–6.

249

doi:10.1016/j.arth.2007.04.014. [7]

Losina E, Thornhill TS, Rome BN, Wright J, Katz JN. The dramatic increase in total knee

EP

250

replacement utilization rates in the United States cannot be fully explained by growth in

252

population size and the obesity epidemic. J Bone Joint Surg Am 2012;94:201–7.

253 254

doi:10.2106/JBJS.J.01958.

[8]

255

Haverkamp D, Klinkenbijl MN, Somford MP, Albers GHR, van der Vis HM. Obesity in

total hip arthroplasty--does it really matter? A meta-analysis. Acta Orthop 2011;82:417–

256 257

AC C

251

22. doi:10.3109/17453674.2011.588859.

[9]

Liu W, Wahafu T, Cheng M, Cheng T, Zhang Y, Zhang X. The influence of obesity on

258

primary total hip arthroplasty outcomes: A meta-analysis of prospective cohort studies.

259

Orthop Traumatol Surg Res 2015;101:289–96. doi:10.1016/j.otsr.2015.01.011. 9

ACCEPTED MANUSCRIPT

260

[10]

Kerkhoffs GMMJ, Servien E, Dunn W, Dahm D, Bramer JAM, Haverkamp D. The influence of obesity on the complication rate and outcome of total knee arthroplasty: a

262

meta-analysis and systematic literature review. J Bone Joint Surg Am 2012;94:1839–44.

263

doi:10.2106/JBJS.K.00820.

264

[11]

RI PT

261

Cram P, Lu X, Kaboli PJ, Vaughan-Sarrazin MS, Cai X, Wolf BR, et al. Clinical

265

characteristics and outcomes of Medicare patients undergoing total hip arthroplasty, 1991-

266

2008. JAMA 2011;305:1560–7. doi:10.1001/jama.2011.478. [12]

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

SC

267

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

269

2012;308:1227–36. doi:10.1001/2012.jama.11153. [13]

271 272

Report of a WHO Expert Consultation. Geneva: 1995. [14]

273 274

World Health Organization. Physical status: the use and interpretation of anthropometry.

Nationwide Inpatient Sample (NIS). Healthcare Cost and Utilization Project (HCUP). 2000-2010. Agency for Healthcare Research and Quality, Rockville, MD. n.d.

[15]

Maradit Kremers H, Visscher SL, Kremers WK, Naessens JM, Lewallen DG. Obesity

TE D

270

M AN U

268

275

Increases Length of Stay and Direct Medical Costs in Total Hip Arthroplasty. Clin Orthop

276

Relat Res 2013;472:1232–9. doi:10.1007/s11999-013-3316-9.

277

[16]

Kremers HM, Visscher SL, Kremers WK, Naessens JM, Lewallen DG. The effect of obesity on direct medical costs in total knee arthroplasty. J Bone Joint Surg Am

279

2014;96:718–24. doi:10.2106/JBJS.M.00819. [17]

Odum SM, Springer BD, Dennos AC, Fehring TK. National obesity trends in total knee

283

AC C

280

EP

278

284

doi:10.1016/j.spinee.2014.04.025.

281 282

285

arthroplasty. J Arthroplasty 2013;28:148–51. doi:10.1016/j.arth.2013.02.036.

[18]

Golinvaux NS, Bohl DD, Basques BA, Fu MC, Gardner EC, Grauer JN. Limitations of

administrative databases in spine research: a study in obesity. Spine J 2014;14:2923–8.

[19]

McCalden RW, Charron KD, MacDonald SJ, Bourne RB, Naudie DD. Does morbid

286

obesity affect the outcome of total hip replacement?: an analysis of 3290 THRs. J Bone

287

Joint Surg Br 2011;93:321–5. doi:10.1302/0301-620X.93B3.25876. 10

ACCEPTED MANUSCRIPT

288

[20]

Werner BC, Evans CL, Carothers JT, Browne JA. Primary Total Knee Arthroplasty in Super-obese Patients: Dramatically Higher Postoperative Complication Rates Even

290

Compared to Revision Surgery. J Arthroplasty 2015;30:849–53.

291

doi:10.1016/j.arth.2014.12.016.

292

[21]

RI PT

289

D’Apuzzo MR, Novicoff WM, Browne JA. The John Insall Award: Morbid obesity

293

independently impacts complications, mortality, and resource use after TKA. Clin Orthop

294

Relat Res 2015;473:57–63. doi:10.1007/s11999-014-3668-9.

AC C

EP

TE D

M AN U

SC

295

11

ACCEPTED MANUSCRIPT

Table 1. Effect of obesity on outcomes following primary total knee arthroplasty. Outcomes

Obesity defined by BMI

Length of stay,

Non-obese

Statistic (95%

Obese

(n=11,065) (n=6,965)

CI)*

(n=4,102) (n=13,928)

CI)*

3.5±6.9

3.4±2.4

0.1 (0.0-0.3)

3.5±6.4

0.2 (0.0-0.3)†

210±55

206±63

3.9 (0.3-7.6)†

273 (2.5)

169 (2.4)

90-day Complications Deep vein thrombosis, n (%) Pulmonary

175 (1.6)

78 (1.1)

embolism, n (%) 165 (1.5)

(%) Superficial

279 (2.5)

infection, n (%) 114 (1.0)

dehiscence, n (%)

n (%)

47 (0.4)

AC C

Wound hematoma,

103 (1.5)

33 (0.5)

EP

Wound

68 (1.0)

1.02 (0.84-1.24)

126 (3.1)

1.42 (1.09-1.87)† 80 (2.0)

1.53 (1.16-2.06)† 96 (2.3)

TE D

Deep Infection, n

216±64

M AN U

mean ± SD (min)

26 (0.4)

3.4±2.3

SC

mean ± SD (days) Operative time,

Non-Obese Statistic (95%

RI PT

Obese

Obesity defined by Codes

205±46

10.1 (6.7-13.5)†

316 (2.3)

1.37 (1.101.68)†

173 (1.2)

1.58 (1.212.06)†

137 (1)

2.41 (1.853.13)†

1.72 (1.38-2.17)† 136 (3.3)

246 (1.8)

1.91 (1.542.35)†

2.19 (1.50-3.27)† 63 (1.5)

84 (0.6)

2.57 (1.843.56)†

1.14 (0.71-1.84)

22 (0.5)

51 (0.4)

1.47 (0.87-2.39)

*Odds ratios are reported for binary outcomes. For continuous variables, differences in means are reported. †Denotes statistical significance (p<0.05). Shaded rows indicate the results are different for BMI and coding. CI- Confidence Interval, SD- Standard Deviation.

ACCEPTED MANUSCRIPT

Table 2. Effect of obesity on outcomes following primary total hip arthroplasty. Obesity defined by BMI

Length of stay,

Obese

Non-Obese

Statistic(95%

Obese

(n=4608)

(n=5867)

CI)*

(n=1459) (n=9016)

CI)*

3.7±2.2

3.7±2.6

0.0(-0.1 - 0.1)

3.9±2.7

3.6±2.4

0.2 (0.1 - 0.4)†

221±36

223±39

-2(-4 - 2)

222±37

222±37

0 (-4 - 3)

91(2%)

91(1.6%)

1.28(0.95 - 1.72)

60(1.3%)

58(1.0%)

mean ± SD (days) Operative time,

SC

mean ± SD (min)

thrombosis, n (%) Pulmonary embolism, n (%) Deep Infection, n

48(1.0%)

22(0.4%)

80(1.7%)

57(0.97%)

Wound dehiscence,

28(0.6%)

n (%)

n (%)

22(0.5%)

86 (1.0)

2.32 (1.52 3.47)†

39 (0.4)

4.99 (3.09-8.02)†

1.80(1.28-2.55)†

34 (2.3)

103 (1.1)

2.06 (1.38-3.02)†

12(0.2%)

2.98(1.55-6.10)†

18 (1.2)

22 (0.2)

5.11 (2.70-9.53)†

19(0.32%)

1.48(0.80-2.76)

15 (1.0)

26 (0.3)

3.59 (1.85-6.71)†

EP

Wound hematoma,

32(2.2)

1.08 (0.70 - 1.60)

31(2.1)

TE D

infection, n (%)

1.32(0.92 - 1.90)

155 (1.7)

2.79(1.71-4.73)†

(%) Superficial

27(1.9)

M AN U

90-day Complications Deep vein

Non-Obese Statistic(95%

RI PT

Outcomes

Obesity defined by Codes

AC C

*Odds ratios are reported for binary outcomes. For continuous variables, differences in means are reported. †Denotes statistical significance (p<0.05). Shaded rows indicate the results are different for BMI and coding. CI- Confidence Interval, SD- Standard Deviation.

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

ACCEPTED MANUSCRIPT

FIGURE LEGENDS Figure 1. The prevalence of obesity as defined by ICD-9 codes was lower than that defined by BMI throughout the study period in total knee arthroplasty (TKA) (A) and total hip arthroplasty

RI PT

(THA) (B) Figure 2. The sensitivity of coding increased from 2004 to 2014 in both TKA (p<0.001) and THA (p<0.001).

AC C

EP

TE D

M AN U

both TKA (p<0.001) and THA (p<0.001).

SC

Figure 3. The proportion of patients receiving an obesity code increased as BMI increased in