In-Hospital Complications Following Ankle Arthrodesis Versus Ankle Arthroplasty a Matched Cohort Study

1469

In-Hospital Complications Following Ankle Arthrodesis Versus Ankle Arthroplasty A Matched Cohort Study

Susan M. Odum, PhD, Bryce A. Van Doren, MPA, MPH, Robert B. Anderson, MD, and W. Hodges Davis, MD

Investigation performed at OrthoCarolina Research Institute, Charlotte, North Carolina

Background: Ankle arthrodesis has been the traditional surgical treatment for end-stage hindfoot arthritis. However, utilization of total ankle arthroplasty (TAA) is increasing as surgical techniques and implants have substantially improved. The purpose of this study was to compare the U.S. national rates of perioperative (in-hospital) complications between a statistically matched cohort of patients who underwent either an ankle arthrodesis or a TAA. Methods: Data from the 2002 to 2013 Nationwide Inpatient Sample releases were analyzed. The International Clas- sification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) procedure codes were used to identify 4,192 patients treated with TAA (ICD 81.56) and 16,278 treated with ankle arthrodesis (ICD 81.11). ICD-9-CM diagnosis codes were utilized to identify major and minor in-hospital complications, and mortality was determined using the Uniform Bill patient disposition. The arthrodesis and TAA groups were matched with regard to age, sex, race, surgery year, hospital type, comorbidities, adjunctive procedures, and surgical indication. Unadjusted and adjusted in- hospital complication risks were compared between groups using the Fisher exact test and multiple logistic regression analysis. Results: We were able to statistically match 1,574 patients who underwent a TAA (37.5%) with a patient who underwent arthrodesis. A major in-hospital complication occurred in 8.5% (134) of the 1,574 patients in the ankle arthrodesis group compared with 5.3% (84) of the 1,574 in the TAA group (p < 0.001) whereas a minor complication was found in 4.7% (74) in the ankle arthrodesis group compared with 5.9% (93) in the TAA group (p = 0.14). There were no deaths in either group. After adjusting for case mix, we found that ankle arthrodesis was 1.8 times more likely to be followed by a major complication (odds ratio [OR] = 1.78, 95% confidence interval [CI] = 1.32 to 2.39) whereas the minor complication rate was 29% lower in that group (OR = 0.71, 95% CI = 0.45 to 1.13). Conclusions: In a matched cohort of 3,148 patients treated with either TAA or ankle arthrodesis, ankle arthrodesis was associated with a 1.8 times higher risk of a major complication but a 29% lower risk of a minor complication. Our findings are consistent with other studies that have shown TAA to be a safe procedure in the inpatient environment. Level of Evidence: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

nkle arthrodesis has historically been the preferred TAA procedures doubled between 2007 and 2011. Raikin A surgical treatment for end-stage arthritis in the hind- et al.4 reported that, while ankle arthrodeses were performed foot. However, in the U.S., utilization of total ankle 6 times more often than TAA, the frequency of the latter arthroplasty (TAA) has increased substantially since 1992, increased by 234% in a decade, from 763 in 2000 to 2,547 in as surgical technique and implants have substantially im- 2010. proved over the last 2 decades1-4.Whileanklearthrodesis Studies have demonstrated that ankle arthrodesis success- remains more common, utilization of that procedure seems fully reduces pain and improves function5-10.However,theloss to have plateaued4.Zhouetal.1 reported that the number of of motion with fusion increases stress on other joints, resulting in

Disclosure: The study was internally funded. No outside funding was received for this study. On the Disclosure of Potential Conﬂicts of Interest forms, which are provided with the online version of the article, one or more of the authors checked “yes” to indicate that the author had a relevant ﬁnancial relationship in the biomedical arena outside the submitted work (http://links.lww.com/JBJS/E207).

J Bone Joint Surg Am. 2017;99:1469-75 d http://dx.doi.org/10.2106/JBJS.16.00944 1470

T HE J OURNAL OF B ONE &JOINT SURGERY d JBJS. ORG IN -HOSPITAL C OMPLICATIONS F OLLOWING ANKLE ARTHRODESIS VOLUME 99-A d N UMBER 17 d S EPTEMBER 6, 2017 VERSUS ANKLE ARTHROPLASTY increased degeneration at the midfoot and forefoot areas11.Such Which procedure has superior outcomes has been the problems at adjacent joints can lead to increased pain, loss of subject of debate5-10,14. Using a non-inferiority design, Saltzman function, and the need for additional surgical procedures11. et al.6 showed that patients who underwent TAA had greater TAA preserves motion at the ankle joint, which reduces stress functional improvement and equivalent pain relief, but higher in other areas; however, there is a risk of revision surgery due to rates of complications, compared with patients who had an ar- implant failure5,10,12. The ﬁrst-generation TAA designs, dating throdesis. In a nonrandomized cohort study, Daniels et al.9 found back to the 1970s, were prone to failure and had higher com- that adjusted pain and function scores were comparable between plication rates compared with ankle arthrodesis5,12. However, procedures but functional improvement was greater in their modern implant designs have solved many of these earlier TAA group. In a systematic review, Haddad et al.5 reported problems13. that American Orthopaedic Foot & Ankle Society (AOFAS)

TABLE I Study Sample Characteristics

Variable TAA Arthrodesis P Value

Time of surgery 0.999 2002-2005 318 (20.2%) 318 (20.2%) 2006-2009 409 (26.0%) 409 (26.0%) 2010-2013 847 (53.8%) 847 (53.8%) Age quartile 0.999 <48 yr 177 (11.2%) 177 (11.2%) 48-58 yr 441 (28.0%) 441 (28.0%) 59-67 yr 492 (31.3%) 492 (31.3%) >67 yr 464 (29.5%) 464 (29.5%) Race 0.999 White 1,114 (70.8%) 1,114 (70.8%) Non-white 460 (29.2%) 460 (29.2%) Sex 0.999 Male 890 (56.5%) 890 (56.5%) Female 684 (43.5%) 684 (43.5%) Diabetes 0.999 No 1,503 (95.5%) 1,503 (95.5%) Yes 71 (4.5%) 71 (4.5%) No. of diagnoses 0.511 1-3 572 (36.3%) 613 (38.9%) 4-6 636 (40.4%) 608 (38.6%) 7-9 276 (17.5%) 268 (17.0%) >9 90 (5.7%) 85 (5.4%) Hospital location/teaching status 0.999 Rural 39 (2.5%) 39 (2.5%) Urban non-teaching 574 (36.5%) 574 (36.5%) Urban teaching 961 (61.0%) 961 (61.0%) Hospital region <0.001 Northeast 192 (12.2%) 306 (19.4%) Midwest 426 (27.1%) 390 (24.8%) South 530 (33.7%) 469 (29.8%) West 426 (27.1%) 409 (26.0%) Indication 0.999 Osteoarthritis 1,079 (68.5%) 1,079 (68.5%) Trauma 301 (19.1%) 301 (19.1%) Rheumatoid arthritis 31 (2.0%) 31 (2.0%) Other 163 (10.4%) 163 (10.4%) 1471

T HE J OURNAL OF B ONE &JOINT SURGERY d JBJS. ORG IN -HOSPITAL C OMPLICATIONS F OLLOWING ANKLE ARTHRODESIS VOLUME 99-A d N UMBER 17 d S EPTEMBER 6, 2017 VERSUS ANKLE ARTHROPLASTY

TABLE II Unadjusted Perioperative Complication Risks

Complication TAA Arthrodesis OR (95% CI) P Value

Major 84 (5.3%) 134 (8.5%) 1.63 (1.23-2.17) <0.001 Minor 93 (5.9%) 74 (4.7%) 0.70 (0.45-1.10) 0.143 Death 0 0 —— scores were slightly better and revision rates were lower after Using the Mayo Clinic greedy matching SAS macro, we statistically 17 TAA than after ankle arthrodesis. matched patients who underwent TAAwith those who had an arthrodesis .Once Little is known regarding the comparative perioperative a control is matched to a case in greedy matching, it cannot be matched to another case—i.e., each TAA-treated patient is matched to a unique arthrodesis-treated complication rates between the 2 procedures, which is useful patient, if one is available. The matching was based on age (within 5 years), sex information for shared decision-making and informed con- (male or female), race (white, black, Hispanic, Asian, Native American, or other), sent. With the relatively low complication rates associated with hospital location/teaching status (urban teaching, urban non-teaching, or rural), each of these procedures, very large sample sizes are required year of surgery (within 1 year), comorbidities (see Appendix), adjunctive foot/ to achieve adequate statistical power. Moreover, direct com- ankle procedures (yes or no), and surgical indication (osteoarthritis, trauma, parisons are problematic because of the inherent selection rheumatoid arthritis, or other). A successful match was between 2 patients who bias resulting from the surgical indications, which vary between had the same categorical values for all matching criteria, and a total of 1,574 3 (37.6%) of the patients who underwent TAA were successfully matched with a procedures. Jiang et al. compared national complication rates patient who underwent arthrodesis. Prior to matching, the average age at the between TAA and arthrodesis and, after controlling for demo- time of the arthrodeses was 56 years compared with 62 years at the time of the graphics and health status, found lower complication rates after TAAs (p < 0.001); after matching, it was 61.0 and 60.7 years, respectively (p = TAA. Adjusting for the effects of covariates does not quite create 0.442). (There was no significant difference in age distribution; p = 0.653.) an apples-to-apples comparison in the same way that statistical Following matching, we collapsed race into a dichotomous variable (white/ matching does15. The purpose of this study was to compare non-white). Descriptive statistics of the matched study sample are reported U.S. national rates of perioperative (in-hospital) complica- in Table I. Standard descriptive statistics, including measures of central tendency tions between statistically matched groups of patients treated and variation as well as frequencies and proportions, were calculated. At the with either ankle arthrodesis or TAA. bivariate level, the occurrence of complications was then compared between patient cohorts (i.e., ankle arthrodesis versus TAA). To facilitate comparison between patient groups, we constructed a series of 2-by-2 tables, comparing Materials and Methods the event (e.g., major complication) to the non-event category (i.e., no ata from the 2002 to 2013 releases of the Nationwide Inpatient Sample complications). Cochran-Mantel-Haenszel and Fisher exact tests were then D(NIS) were used to compare rates of perioperative complications between used to compare the frequency and relative risk of complications between fi ankle arthrodesis and TAA. Patient cohorts were identi ed using International patient groups. fi fi Classi cation of Diseases, 9th Revision, Clinical Modi cation (ICD-9-CM) Multivariate analyses were conducted with 3 separate logistic regression procedure codes for ankle arthrodesis (81.11) and TAA (81.56). We also used models to determine the adjusted rates and odds of (1) minor complications, (2) ICD-9-CM procedure codes to identify patients with concomitant foot and major complications, and (3) mortality. We utilized multivariate logistic re- ankle procedures and ICD-9-CM diagnosis codes to identify the indication for gression to control for and independently examine the impact of patient age, race the procedure (i.e., the primary musculoskeletal diagnosis), comorbidities, (white or non-white), surgical indication, comorbidities, additional procedures, fi and perioperative complications. Perioperative complications were classi ed time period of the procedure, and hospital location/teaching status on the fi as none, minor, major, or death based on a previously reported classi cation relative risk of complications between patient groups. — fi system developed using the Delphi method i.e., the classi cation was de- Full details of the matching criteria are reported in the Appendix, as veloped through consensus by a team of fellowship-trained arthroplasty sur- 16 are the results of the unmatched analyses. All analyses were unweighted and geons and approved by the senior authors (fellowship-trained foot and ankle thus represent only the patients contained in the NIS releases. We did not fi surgeons). Our classi cation system is limited to medical and orthopaedic attempt to estimate a national annual incidence of complications for the complications that are most directly related to surgery (see Appendix). Ad- 2 procedures. ditionally, we utilized the Uniform Bill discharge disposition, included in the NIS, to identify patients who had died during their hospital stay. Patients were fi — Results classi ed as having 1 of 4 categories of complications inpatient death, major Mortality and Major Complications complications, minor complications, or no complications—according to their most serious complication. For example, a patient with major complications here were no in-hospital deaths in either group. The rate who died in the hospital was classified only as an in-hospital death. Minor Tof major complications after the ankle arthrodeses (8.5%; complications (see Appendix) included phlebitis, venous embolism, pulmo- 134) was significantly higher (p < 0.001) than that after the nary insufficiency, emphysema from the procedure, hemorrhage, hematoma, TAA procedures (5.3%; 84) (Table II). Ankle arthrodesis was fi seroma, a nonhealing wound, and unspeci ed complications. Major complica- associated with a >1.6 times higher likelihood of a major tions (see Appendix) included cardiac arrest; vascular complication; pulmonary complication compared with TAA (unadjusted odds ratio embolism; ventilator-associated pneumonia; shock; wound disruption; retained = fi = surgical item; infected seroma; prosthetic mechanical complication, breakage, or [OR] 1.63, 95% con dence interval [CI] 1.23 to 2.17). dislocation; periprosthetic fracture or infection; and other another complication The results of the multivariate analysis are presented in involving the prosthesis. Table III. The risk of a major complication following ankle 1472

T HE J OURNAL OF B ONE &JOINT SURGERY d JBJS. ORG IN -HOSPITAL C OMPLICATIONS F OLLOWING ANKLE ARTHRODESIS VOLUME 99-A d N UMBER 17 d S EPTEMBER 6, 2017 VERSUS ANKLE ARTHROPLASTY

TABLE III Adjusted Minor and Major Complication Risks and Associated Factors

Complication* Minor Major

Time of surgery 2002-2005 —— 2006-2009 0.65 (0.35-1.22) 1.49 (0.91-2.41) 2010-2013 0.37 (0.21-0.67) 1.40 (0.89-2.19) Procedure TAA —— Arthrodesis 0.71 (0.45-1.13) 1.77 (1.32-2.39) Age group <48 yr 1.26 (0.54-2.94) 2.12 (1.28-3.54) 48-58 yr 1.09 (0.60-1.98) 1.54 (1.01-2.32) 59-67 yr 0.82 (0.46-1.45) 1.43 (0.96-2.12) >67 yr —— Race White —— Non-white 0.87 (0.50-1.52) 1.13 (0.80-1.60) Sex Male —— Female 1.07 (0.67-1.70) 1.33 (0.99-1.79) Hospital location/teaching status Urban teaching —— Urban non-teaching 1.52 (0.93-2.49) 1.02 (0.74-1.40) Rural 2.54 (0.73-8.81) 1.72 (0.60-4.93) Hospital region West —— Midwest 1.01 (0.55-1.86) 0.80 (0.52-1.21) Northeast 1.04 (0.49-2.20) 0.71 (0.43-1.16) South 0.75 (0.41-1.37) 1.04 (0.72-1.50) No. of diagnoses 1-3 —— 4-6 4.22 (2.04-8.74) 2.28 (1.56-3.33) 7-9 9.84 (4.57-21.19) 3.41 (2.21-5.27) >9 9.26 (3.39-25.34) 1.66 (0.80-3.46) Diabetes No —— Yes 0.84 (0.30-2.39) 0.71 (0.35-1.42) Indication Osteoarthritis —— Trauma 1.36 (0.75-2.46) 2.23 (1.53-3.26) Rheumatoid arthritis NS† 1.62 (0.61-4.26) Other 1.71 (0.84-3.48) 8.25 (5.79-11.74)

*The values are given as the OR, with the 95% CI in parentheses. †NS = sample not sufﬁcient—i.e., there were too few patients with rheumatoid arthritis to estimate the OR. arthrodesis was, for the most part, unchanged after adjusting dependent effects on the risk of a major complication. for patient and hospital factors (OR = 1.77, 95% CI = 1.32 to Younger patients (£67 years old) had a higher risk of a major 2.39). Patient age and health status had signiﬁcant and in- complication, as did patients in poor health, with the risk 1473

T HE J OURNAL OF B ONE &JOINT SURGERY d JBJS. ORG IN -HOSPITAL C OMPLICATIONS F OLLOWING ANKLE ARTHRODESIS VOLUME 99-A d N UMBER 17 d S EPTEMBER 6, 2017 VERSUS ANKLE ARTHROPLASTY

increasing nearly 2 to 3 times depending on the number of patients who had had either a TAA or an arthrodesis between comorbidities on record. 2002 and 2011. The overall complication rate was 9% for the arthrodesis group and 5% for the TAA group. After adjusting Minor Complications for the effects of patient demographics and comorbidities, The rates of minor complications were very similar (p = 0.14) the diabetic patients in their arthrodesis group had 2.7 times for the ankle arthrodesis (4.7%; 74) and TAA (5.9%; 93) the relative risk (p < 0.0001) of overall complications com- groups (Table II), but after adjusting for patient and hospital pared with the non-diabetic patients in that group18.Inour factors the ankle arthrodesis group had a 29% lower risk of study, the major and minor complication rates of 5.3% and a minor complication (OR = 0.71, 95% CI = 0.45 to 1.13) 5.9% after the TAAs and the major complication rate of 8.5% (Table III). The time period of the surgery had a significant after the arthrodeses were no different than the rates in the and independent effect on the risk of a minor complication study by Schipper et al. However, our minor complication (Table III), which was lower for the procedures performed rate of 4.7% after the arthrodeses was lower than the rate in between 2010 and 2013 than for those performed between their arthrodesis group. Interestingly, when we adjusted for 2002 and 2005. Also, as the number of comorbidities in- the effects of covariates that were similar to the ones that creased, the risk of minor complication risks increased—4to they used, we found arthrodesis to be associated with a 1.8 10 times depending on the number of associated diagnoses times higher likelihood of a major complication, which is recorded. smaller than the effect reported by Schipper et al. for diabetic patients. Discussion Using national administrative data from academic he purpose of our study was to compare the risks of centers rather than NIS data, Zhou et al.1 recently reported an Tminor and major inpatient complications as well as in- overall complication rate of 1.4% and a mortality rate of <1% patient mortality between patients who underwent TAA and after 2,340 TAA procedures performed between 2007 and those treated with ankle arthrodesis. Using a national dis- 2011. In a randomized clinical trial, Saltzman et al. reported charge database, the NIS, we identified 1,574 statistical matches an overall major complication rate of 6.2% after TAA and of TAA-treated and arthrodesis-treated patients (a total of 1.5% after arthrodesis6. Our major complication rate was 3,148 patients). This methodology is useful to eliminate, or similar to that reported by Saltzman et al., but our minor minimize, patient selection bias when random allocation is complication rate was higher. The difference in minor com- not possible and allows for an apples-to-apples comparison. plication rates is likely due to differences in our definitions Using our previously described complication schema, we found of complications. that the rate of major perioperative complications was sig- The differences between the complication rates that nificantly higher after ankle arthrodesis than after TAA. The we reported and those in similar NIS studies3,18 may be due rate of minor complications was very low—approximately to a number of factors, including varying definitions of 5% to 6%—regardless of the surgery type and, although it complications (i.e., the inclusion of different ICD-9-CM was not statistically significant, the ankle arthrodesis group codes). Furthermore, we differentiated minor complications, has a 29% lower adjusted risk. There were no in-hospital major complications, and mortality, which other studies3,18 deaths in either group. appear to have grouped together. Also, our data were ob- The NIS has been used previously to compare rates of tained 1 and 2 years after the last year of data collection in complications between TAA and arthrodesis, but caution the 2 previous NIS studies (i.e., in 2012 and 2013 rather than must be exercised when comparing these rates between 2011). It is interesting to note that the risk of minor com- studies because of varying definitions of complications and plications decreased over the years but the risk of major differences in sampling. Using NIS data from 2002 through complications increased. This effect of time may be due to a 2011, Jiang et al.3 reported the unadjusted rate of overall number of different factors. Generally, complications fol- complications to be 3.7% for 3,002 patients who underwent lowing TAA have decreased with improvements in instru- TAA and 6% for 12,250 who underwent arthrodesis. After ments and implants, patient selection, and surgeon learning adjusting for patient demographics and comorbidities, they curves. While we attempted to control for technological found the TAA group to have a 30% lower relative risk (p = improvements by using time as a proxy and for many patient 0.03) of overall complications3. Because we used different variables such as sex, age, and health status, there are other ICD-9-CM codes to define complications it is difficult to influential factors, including bone quality, degree of de- directly compare our findings with those of Jiang et al. formity, and prior ankle surgery, for which we could not However, our minor complication rate of 5.9% following TAA control. TAA is more likely to be performed in younger, (and 4.7% following arthrodesis) is slightly higher than the healthier patients with better bone quality and smaller defor- rate reported by Jiang et al. whereas the major complication mities. Ankle arthrodesis is more likely to be performed in rates in both our TAA and our arthrodesis group were con- sicker patients with more compromised bone quality and larger siderably higher. deformities. This could explain the differences in the risks of a Schipper et al.18 used NIS data to compare periopera- major complication. An additional factor may be the migration tive complication rates between diabetic and non-diabetic of a proportion of these surgical procedures to the outpatient 1474

T HE J OURNAL OF B ONE &JOINT SURGERY d JBJS. ORG IN -HOSPITAL C OMPLICATIONS F OLLOWING ANKLE ARTHRODESIS VOLUME 99-A d N UMBER 17 d S EPTEMBER 6, 2017 VERSUS ANKLE ARTHROPLASTY environment, with the patients at higher risk for complications effect of other surgical procedures, we excluded patients who remaining in the inpatient environment, which is what we had both an arthrodesis and a TAA performed during the studied. Finally, the statistical matching method followed by same encounter. the multivariate analysis may have captured a more accurate In conclusion, we found no in-hospital deaths or sta- effect of surgery type on perioperative complication rates than tistically significant differences in the risk of minor com- revealed in the previous studies3,18. plications in a matched cohort of 3,148 patients treated with A major strength of our study is that we utilized a sta- TAA or ankle arthrodesis. However, there was a 1.8 times higher tistically matched cohort of patients. Given the different in- risk of a major complication after ankle arthrodesis. Younger and dications for each procedure, selection bias is inherent in sicker patients may require additional counseling regarding the direct comparisons19 and such bias may have affected the re- complication risks regardless of the surgical procedure. Our sultsofpreviousstudies.Otherrelatively large and/or national findings are consistent with other studies3,18 that have shown studies3,4,9,18 showed that TAA-treated patients were older and TAA to be a comparatively safe procedure. had lower rates of medical comorbidities than those who underwent arthrodesis. While multivariate analyses can help Appendix to adjust for these differences between groups, matching al- Tables showing the ICD-CM-9 codes used to identify lows for a more direct, apples-to-apples comparison between complications, the matching criteria, and the results of procedures20. A multivariate analysis then further isolates the analyses of the unmatched cohort are available with the online independent effects of the covariates. Therefore, the statistical versionofthisarticleasadatasupplementatjbjs.org(http:// effects of the covariates are more likely to represent true ef- links.lww.com/JBJS/E208). n fects on the outcome rather than selection bias. However, even with these rigorous analytic methods to control for confounding, it is possible that residual confounding for which we were unable to control exerted statistical influence on the outcomes of our study. Susan M. Odum, PhD1,2 Bryce A. Van Doren, MPA, MPH1,2 This study also has a number of limitations. The NIS 3 data set contains only administrative data associated with a Robert B. Anderson, MD W. Hodges Davis, MD3 hospital discharge record. It does not include many clinical variables that are relevant to TAA or arthrodesis such as im- 1OrthoCarolina Research Institute, Charlotte, North Carolina plant details, radiographic data, implant positioning, etc. Also, during the study period, the database did not include a 2Health Services Research Program, College of Health & Human Services, linking variable that would have allowed inclusion of com- University of North Carolina at Charlotte, Charlotte, North Carolina plications that occurred after discharge or inclusion of read- 3OrthoCarolina Foot and Ankle Institute, Charlotte, North Carolina missions. As with any administrative database, coding errors are possible and care must be taken to examine the data for E-mail address for S.M. Odum: [email protected] systematic coding patterns that reveal coding errors. For example, to minimize the possibility that we were capturing the ORCID iD for S.M. Odum: 0000-0001-7769-4782

References

1. Zhou H, Yakavonis M, Shaw JJ, Patel A, Li X. In-patient trends and complications 7. Saltzman CL, Kadoko RG, Suh JS. Treatment of isolated ankle osteoarthritis with after total ankle arthroplasty in the United States. Orthopedics. 2016 Jan-Feb;39(1): arthrodesis or the total ankle replacement: a comparison of early outcomes. Clin e74-9. Epub 2016 Jan 5. Orthop Surg. 2010 Mar;2(1):1-7. Epub 2010 Feb 4. 2. Pugely AJ, Lu X, Amendola A, Callaghan JJ, Martin CT, Cram P. Trends in the 8. Krause FG, Windolf M, Bora B, Penner MJ, Wing KJ, Younger AS. Impact of use of total ankle replacement and ankle arthrodesis in the United States complications in total ankle replacement and ankle arthrodesis analyzed with a Medicare population. Foot Ankle Int. 2014 Mar;35(3):207-15. Epub 2013 validated outcome measurement. J Bone Joint Surg Am. 2011 May 4;93(9): Oct 31. 830-9. Epub 2011 Apr 15. 3. Jiang JJ, Schipper ON, Whyte N, Koh JL, Toolan BC. Comparison of perioperative 9. Daniels TR, Younger AS, Penner M, Wing K, Dryden PJ, Wong H, Glazebrook complications and hospitalization outcomes after ankle arthrodesis versus total M. Intermediate-term results of total ankle replacement and ankle arthrode- ankle arthroplasty from 2002 to 2011. Foot Ankle Int. 2015 Apr;36(4):360-8. Epub sis: a COFAS multicenter study. J Bone Joint Surg Am. 2014 Jan 15;96 2014 Oct 30. (2):135-42. 4. Raikin SM, Rasouli MR, Espandar R, Maltenfort MG. Trends in treatment of 10. Gibson JN, Thomson CE. Arthrodesis or total replacement arthroplasty for advanced ankle arthropathy by total ankle replacement or ankle fusion. Foot Ankle hallux rigidus: a randomized controlled trial. Foot Ankle Int. 2005 Sep;26(9): Int. 2014 Mar;35(3):216-24. Epub 2013 Dec 19. 680-90. 5. Haddad SL, Coetzee JC, Estok R, Fahrbach K, Banel D, Nalysnyk L. Interme- 11. Ling JS, Smyth NA, Fraser EJ, Hogan MV, Seaworth CM, Ross KA, Kennedy JG. diate and long-term outcomes of total ankle arthroplasty and ankle arthrodesis. Investigating the relationship between ankle arthrodesis and adjacent-joint A systematic review of the literature. J Bone Joint Surg Am. 2007 Sep;89(9): arthritis in the hindfoot. A systematic review. J Bone Joint Surg Am. 2015 May 1899-905. 6;97(9):e43. 6. Saltzman CL, Mann RA, Ahrens JE, Amendola A, Anderson RB, Berlet GC, 12. Easley ME, Vertullo CJ, Urban WC, Nunley JA. Total ankle arthroplasty. J Am Brodsky JW, Chou LB, Clanton TO, Deland JT, Deorio JK, Horton GA, Lee TH, Mann JA, Acad Orthop Surg. 2002 May-Jun;10(3):157-67. Nunley JA, Thordarson DB, Walling AK, Wapner KL, Coughlin MJ. Prospective con- 13. Ellington JK, Gupta S, Myerson MS. Management of failures of total ankle trolled trial of STAR total ankle replacement versus ankle fusion: initial results. Foot replacement with the agility total ankle arthroplasty. J Bone Joint Surg Am. 2013 Dec Ankle Int. 2009 Jul;30(7):579-96. 4;95(23):2112-8. 1475

T HE J OURNAL OF B ONE &JOINT SURGERY d JBJS. ORG IN -HOSPITAL C OMPLICATIONS F OLLOWING ANKLE ARTHRODESIS VOLUME 99-A d N UMBER 17 d S EPTEMBER 6, 2017 VERSUS ANKLE ARTHROPLASTY

14. SooHoo NF, Zingmond DS, Ko CY. Comparison of reoperation rates following research/division-biomedical-statistics-informatics/software/locally-written-sas- ankle arthrodesis and total ankle arthroplasty. J Bone Joint Surg Am. 2007 Oct;89 macros. Accessed 2015 April 14. (10):2143-9. 18. Schipper ON, Jiang JJ, Chen L, Koh J, Toolan BC. Effect of diabetes mellitus 15. Klungel OH, Martens EP, Psaty BM, Grobbee DE, Sullivan SD, Stricker BH, on perioperative complications and hospital outcomes after ankle arthrodesis Leufkens HG, de Boer A. Methods to assess intended effects of drug treatment and total ankle arthroplasty. Foot Ankle Int. 2015 Mar;36(3):258-67. Epub in observational studies are reviewed. J Clin Epidemiol. 2004 Dec;57 2014 Nov 20. (12):1223-31. 19. Krause FG, Schmid T. Ankle arthrodesis versus total ankle replacement: 16. Odum SM, Troyer JL, Kelly MP, Dedini RD, Bozic KJ. A cost-utility analysis how do I decide? Foot Ankle Clin. 2012 Dec;17(4):529-43. Epub 2012 comparing the cost-effectiveness of simultaneous and staged bilateral total knee Sep 21. arthroplasty. J Bone Joint Surg Am. 2013 Aug 21;95(16):1441-9. 20. Rose S, Laan MJ. Why match? Investigating matched case-control 17. Bergstrahl E, Kosanke J. gmatch. Locally written SAS macros 2003; http:// study designs with causal effect estimation. Int J Biostat. 2009 Jan 6; www.mayo.edu/research/departments-divisions/department-health-sciences- 5(1):1.