The incidence of patients with end-stage renal disease (ESRD) has increased significantly over the years, with 120,688 new cases of ESRD being reported in the United States in 2014 alone.1 Dialysis continues to be the predominant treatment modality for ESRD, and the number of dialysis-dependent patients has paralleled the rise in ESRD cases. Over the past decade, the dialysis-dependent population has increased 24%, reaching an all-time high of 113,944 incident cases in 2013.2 The increased prevalence of dialysis is particularly relevant to orthopaedic surgeons, given its negative associations with bone health, osteoarthritis, amyloidosis, and metabolic abnormalities that can increase the risk for fractures and decrease potential for healing after such injuries.3,4
Hip fracture among elderly patients is a common but dangerous occurrence, with more than 250,000 patients being admitted annually in the United States5 and a 1-year mortality rate between 20% and 35%.5,6 In addition, annual costs of caring for those with hip fractures are estimated to be as high at $15 billion per year, creating a significant burden on healthcare costs. The incidence and costs are expected to rise because of the aging population of the United States, with projections showing a quadrupling of patients older than 85 years by 2050.7,8
Hospitals and legislatures have recognized the need for optimizing patient outcomes as a method of cost minimization. For example, the Patient Protection and Affordable Care Act sought to decrease costs through the creation of the Hospital Readmission Reduction Program, which penalizes hospitals who do not meet pre-established markers for acceptable readmission rates.9 Furthermore, the Bundled Payments for Care Initiative (BPCI) created by the Center for Medicare and Medicaid Services (CMS) eschews traditional fee-for-service reimbursement in favor of increasing provider and hospital accountability for patient outcomes.10 To facilitate the optimization of patient outcomes, studies of complications after common procedures in high-risk, high-cost populations such as hip fracture repair in dialysis-dependent patients are necessary.
Currently, the literature regarding surgical outcomes in dialysis-dependent hip fracture patients is limited by small populations not exceeding more than 62 patients.11–20 Unfortunately, the limited statistical power of these studies constrains the utility and generalizability of the perioperative data. Furthermore, although these studies suggest that dialysis patients fare worse after hip fracture, there is little mention of whether this is due to dialysis as an independent risk factor or whether the negative outcomes are driven by overall health deficits in this population. In contrast, one larger study using the National Trauma Database contained larger numbers but the database is limited to the inpatient stay, and the data sets are not necessarily nationally representative, especially as higher-level trauma centers are excluded from the database outcomes.21,22
To address the above-noted limitations, the current study used the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) database. This database is a large, national database that provides 30-day outcomes and has been shown to be robust in performing risk-adjusted outcomes analysis.23,24 With these factors in mind, the aim of the current study is to compare perioperative outcomes and complications after hip fracture in dialysis-dependent and non–dialysis-dependent cohorts in the NSQIP database through multivariable regression analysis of coarsened exact matching (CEM) samples.
Data Source and Study Cohort
NSQIP contains information on >150 variables abstracted directly from patient charts by trained clinical reviewers at participating hospitals across the United States.25 As of 2016, it contained data from more than 600 member hospitals and included variables addressing patient demographics and comorbidities, perioperative factors, and outcomes (ie, complications, readmission, and mortality) reported within 30 postoperative days, regardless of discharge status. Interrater reliability for information abstraction from patient records is reported to be greater than 98%.25
The current study queried 2005 to 2016 NSQIP data for all patients aged 60 years and older undergoing isolated hip fracture repair (CPT codes: 27236, 27244, and 27245). Patients were excluded if they presented with an International Classification of Diseases, 9th or 10th revision, diagnosis code for infection, tumor, or emergency. Covariate information obtained from NSQIP for use in this study included age, sex, body mass index (BMI) (kg/m2), functional status before injury, and American Society of Anesthesiologists' (ASA) classification score. Receipt of dialysis, defined by NSQIP as any patient currently requiring or receiving dialysis before surgery, was the primary explanatory variable.
Perioperative Adverse Events and Secondary Outcome Measures
The primary outcome variable for the study was defined as the occurrence of any perioperative any adverse event (AAE). The occurrence of adverse events was abstracted from NSQIP based on variables reported within 30-day of index operation.
Any adverse event (AAE) was noted if there was a reported occurrence of either a severe adverse event (SAE) or minor adverse event (MAE). SAE included documented occurrence of deep surgical site infection, sepsis, failure to wean from a ventilator within 48 hours, need for reintubation, renal failure, thromboembolic event (deep vein thrombosis/pulmonary embolism), cardiac arrest, myocardial infarction, or cerebrovascular accident (stroke). MAE included superficial surgical site infection, wound dehiscence, pneumonia, urinary tract infection, or postoperative renal insufficiency.
Additional secondary outcome measures were also assessed, including (1) the need for the patient to return to the operating room, (2) readmission (reported in NSQIP from 2011 onward), and (3) mortality (evaluated independent of other adverse events) within 30 days of index operation.
Primary and secondary outcome differences between patients who received and who did not receive dialysis were compared using chi-squared tests and unadjusted (univariate) logistic regression. Risk-adjusted differences were further compared using multivariable logistic regression, accounting for the influence of age, sex, BMI, functional status, and ASA classification.
To account for the relatively small sample size among dialysis patients and ensure the consistency of the reported multivariable model effects, risk-adjusted differences were also compared using logistic regression within CEM cohorts for each outcome. CEM, like other forms of matching, is designed to reduce underlying confounding differences between explanatory variable groups (ie, between patients who did and did not receive dialysis) by identifying patients who are otherwise similar with respect to the indicated covariates. CEM specifically functions by temporarily coarsening continuous covariates (eg, age) into predetermined set width groups and matching patients based on the coarsened bins, in addition to directly matching based on categoric variables (eg, sex). Thus, compared with other matching techniques, CEM is more invariant to measurement error, ensuring that reducing balance in one covariate has no effect on other included covariates, balancing nonlinearities, and being extremely computationally efficient.26–28 It does not require iterations for balance checking and rematching or use of a separate procedure for estimation. Moreover, because CEM is from the family of monotonic imbalance bounding methods, matching is precise on coarsened and included categoric variables, making further adjustment on the same variables redundant (ie, risk adjustment on matched covariates is not needed).
Akin to the risk-adjusted logistic regression models, patients in CEM cohorts were matched on differences in age, sex, BMI, functional status, and ASA classification. All patients in each explanatory group for each outcome meeting matching parameters were retained (1:1 matching for nondialysis versus dialysis receipt). To assess for the success of the CEM procedure, overall differences between matched and nonmatched cohort sets for each outcome were compared using the L1 statistics technique.26–28 The L1 statistic is a comprehensive measure of global imbalance between groups. Values close to one indicate complete separation (ie, the groups are not matched), while values close to zero indicate perfect balance (ie, no difference between groups). Successful CEM should result in a reduction of the before matching L1 value. A direct comparison of covariate distributions after matching between dialysis and nondialysis patients for the AAE cohort was also included and is presented in Table 1. Similar methodology has been previously used in surgical studies in other fields.29,30
Subanalyses compared differences in AAE based on when reported events occurred—before versus after discharge—between patients who received and did not receive dialysis. The timing of adverse events was determined in acute coronary syndrome NSQIP by comparing the number of postoperative days that an adverse event was recorded to the corresponding patient's reported index hospital length of stay.
All analyses and data management were conducted using Stata Statistical Software: Release 14.2 (StataCorp LP, College Station, TX). Two-sided P values <0.05 were considered statistically significant. Our institution's Human Investigation Committee (Institutional Review Board) determined the study exempt.
A total of 288 dialysis and 16,392 nondialysis patients who underwent hip fracture surgery met the study inclusion criteria. The nondialysis group was older in age compared with the dialysis group (median age: 83 versus 77 years, P < 0.001). The nondialysis group also had a lower percentage of male patients (28.53% versus 51.04%, P < 0.001), lower ASA score (3 versus 4, P < 0.001), and a lower BMI (24 versus 25 kg/m2, P = 0.020) (Table 1). All results were considered statistically significant for P values less than 0.05 and subsequently bolded in all tables.
To control for differences in preoperative characteristics and group size, a CEM group of 288 nondialysis patients was chosen from the total population to generate a comparable cohort. After CEM, age, sex, BMI, functional status before injury, and ASA class were statistically similar in both cohorts (prematching AAE L1 = 0.547; postmatching AAE L1 = 0.247) (Table 1).
Rates of 30-day adverse outcomes of dialysis-dependent and nondialysis patients are presented in Table 2. On univariate analysis, dialysis-dependent patients were found to have a greater odds of AAE (odds ratio [OR] = 2.17, 95% confidence interval [CI] 1.71 to 2.75, P < 0.001), SAE (OR = 2.36, 95% CI 1.83 to 3.05, P < 0.001), return to the operating room within 30 days (OR = 2.04, 95% CI 1.16 to 3.60, P = 0.013), readmission to the hospital within 30 days of the operation (OR = 2.45, 95% CI 1.79 to 3.33, P < 0.001), and mortality within 30 days of the operation (OR = 2.29, 95% CI 1.59 to 3.32, P < 0.001) (Table 2).
Because of preoperative differences in cohort demographics, a multivariable logistic regression controlling for age, sex, BMI, preoperative functional status, and ASA class was performed. After controlling for these preoperative demographic and overall health factors, dialysis-dependent patients remained more likely to experience AAE (OR = 1.64 times, 95% CI 1.28 to 2.09, P < 0.001), SAE (OR = 1.72, 95% CI 132 to 2.09, P < 0.001), hospital readmission (OR = 1.89, 95% CI 1.37 to 2.60, P < 0.001), and mortality within 30 days of the operation (OR = 1.65, 95% CI 1.12 to 2.44, P = 0.011) (Table 2).
Finally, a multivariable comparison of dialysis patients and a selection of CEM nondialysis patients was performed. In this analysis, patients on dialysis were more likely to experience AAE (OR = 1.90, 95% CI 1.33 to 2.71, P < 0.001), SAE (OR = 1.77, 95% CI 1.20 to 2.61, P = 0.004), return to the OR (OR = 1.66, 95% CI 0.68 to 4.06, P = 0.271), and experience readmission within 30 days of the operation (OR = 2.48, 95% CI 1.46 to 4.22, P = 0.001), (Table 2 and Figure 1).
Pre- versus Post-discharge Outcomes
The timeline of postoperative complications in both patient cohorts was further analyzed. A sizable percentage of AAEs, SAEs, and MAEs occurred postdischarge. In nondialysis patients, 24.38% of AAEs occurred postdischarge (Table 3 and Figure 2, A), whereas 16.67% occurred postdischarge in dialysis patients (Table 3 and Figure 2, B). In general, SAEs were more likely than MAEs to occur before discharge, although several types of SAEs and MAEs had considerable postdischarge incident rate.
The dialysis-dependent population is continuing to rise in the United States. This is particularly relevant to orthopaedic surgeons because of an increased risk for fractures and decreased potential for healing after such injuries in this population.3,4 Although studies analyzing postoperative complications in hip fracture patients receiving dialysis have been conducted, they have typically focused on long-term outcomes and have been limited by sample size and/or an absent control group.13–19,21 To address these shortcomings, the current study assessed the association of dialysis dependence with perioperative adverse events after hip fracture surgery using the NSQIP database.
The current study identified 289 geriatric dialysis-dependent patients who had undergone hip fracture surgery. This cohort was significantly larger than those of previous studies examining dialysis dependence as an independent prognosticator of postoperative outcomes, in which populations have ranged from 8 to 62 patients.11–20 With the increased power of the current study, univariate, multivariate, and matched multivariate analyses were used to compare geriatric hip fracture patients with and without preoperative dialysis.
The univariate analyses demonstrated that dialysis dependence was significantly associated with an increased risk of adverse events. These findings are generally consistent with the notion that dialysis patients exhibit multiorgan dysfunction,13,31 increasing the risk for postoperative complications.32 Specifically, anemia, electrolyte/fluid imbalance, and cardiovascular events deriving from chronic renal failure escalate the risk of surgical mortality and morbidity. In addition, dialysis patients typically exhibit malnutrition, coagulopathy, and compromised immunity, increasing susceptibility to wound complications and prosthetic infection.33 Indeed, the current dialysis cohort had a significantly higher percentage of patients having an ASA score of three or greater, confirming greater disease burden.
To determine the role of dialysis dependence in perioperative outcomes, multivariate analysis controlling for demographics and overall patient health condition was conducted. The dialysis-dependent cohort was found to be at an increased risk for AAE, major adverse event, unplanned readmission, and mortality within 30 postoperative days. To strengthen the analysis and account for the disparate sizes of the dialysis-dependent and nondialysis cohorts, CEM was performed. The dialysis sample similarly showed an increased risk of all the aforementioned outcomes relative to the non–dialysis-matched sample with the exception of mortality.
These findings underscore the significance of dialysis as an independent predictor of postoperative complications and readmission, with dialysis-dependent patients being at increased odds of adverse events even when compared against a demographically and health status–matched nondialysis cohort. These results are particularly relevant to the evolving reimbursement system, as the CMS, established under the Patient Protection and Affordable Care Act of 2010, has been experimenting with various bundled payment schemes to minimize costs of orthopaedic procedures.34
In 2013, The Bundled Payments for Care Improvement (BPCI) Initiative was passed. This was a voluntary 3-year program, which demonstrated cost savings and subsequently led to the passage of the Comprehensive Care for Joint Replacement program (CJR) in 2016.35 The CJR is mandatory for all lower extremity joint arthroplasties in 67 geographical locations.36 Notably, although the CJR risk adjusts for nonelective fractures, research suggests that compensations for complex patient populations are likely inadequate.36–38 These higher-risk populations, as seen in the present analysis for dialysis patients, are far more likely to have an adverse event or readmission and therefore use more resources and costs for their treatment. As this risk factor is not modifiable through any known means, the result is that the patient's baseline health at the time of their presentation with an acute hip fracture has a tremendously negative impact, regardless of how well the surgery was performed.
More recently, the CMS approved the Surgical Hip and Femur Fracture Treatment Program (SHFFT), which began July 2017.36 However, the American Academy of Orthopaedic Surgeons has already expressed concern over inadequate risk adjustment and outcome measures.36 The current study brings meaningful data to the discussion of risk-adjusted reimbursement for management of higher-risk patients, as dialysis dependence alone was found to be associated with over a two-fold risk of 30-day readmission and a nearly two-fold risk for AAE or a major adverse event.
The current study does have limitations. Mostly, these are inherent to the limitations of the data set itself. NSQIP does not provide long-term follow-up or surgery-specific outcomes. In addition, NSQIP does not specify cumulative dialysis duration or type of dialysis (hemodialysis versus peritoneal dialysis), which could be pertinent to patient survival and complication rates.39
To our knowledge, this is the most comprehensive study analyzing the relationship between dialysis dependence and extensive perioperative adverse events in dialysis-dependent hip fracture patients. The strength of the current study derives from the quality of the NSQIP database, which contains prospectively collected, robust, validated data that is nationally representative and follows patients for 30 days.40
In summary, dialysis dependence is independently associated with perioperative adverse events in patients undergoing hip fracture surgery. These findings highlight the importance of preoperative patient counseling, surgical caution, and heightened surveillance throughout the perioperative period in this high-risk population. Furthermore, hospitals, physicians, and lawmakers should take these data into account when developing bundle payment reimbursement strategies.
1. Saran R, Robinson B, Abbott KC, et al.: US renal data system 2016 annual data report: Epidemiology of kidney disease in the United States. Am J Kidney Dis 2017;69:A7-A8.
2. 2015 USRDS Annual Data Report Volume 2: ESRD in the United States. United States Renal Data System, 2015.
3. Ponnusamy KE, Jain A, Thakkar SC, Sterling RL, Khanuja HS: Inpatient mortality and morbidity for dialysis-dependent patients undergoing primary total hip or knee arthroplasty. J Bone Joint Surg Am 2015;97:1326-1332.
4. Sunday JM, Guille JT, Torg JS: Complications of joint arthroplasty in patients with end-stage renal disease on hemodialysis. Clin Orthop Relat Res 2002;397:350-355.
5. Brauer CA, Coca-Perraillon M, Cutler DM: Incidence and mortality of hip fractures in the United States. JAMA 2009;302:1573-1579.
6. Rosen MK, Lubarsky DA, Lane JM: An economic evaluation of a systems-based strategy to expedite surgical treatment of hip fractures. J Bone Joint Surg Am 2011;93:1326-1334.
7. Cummings SR, Rubin SM, Black D: The future of hip fractures in the United States: Numbers, costs, and potential effects of postmenopausal estrogen. Clin Orthop Relat Res 1990;252:163-166.
8. Day J: Population projections of the United States by age, sex, race and Hispanic origin: 1995 to 2050, in Census UBot. Washington, DC, Government Printing Office, 1996.
9. Boozary AS, Manchin J, Wicker RF: The Medicare hospital readmissions reduction program: Time for reform. JAMA 2015;314:347-348.
10. Manchikanti L, Caraway DL, Parr AT, et al.: Patient protection and affordable care Act of 2010: Reforming the health care reform for the new decade. Pain Physician 2011;14:E35-E67.
11. Sano K, Ito K, Inahata Y, et al.: Operative treatment of hip fracture in haemodialysed patients. J Orthop Surg (Hong Kong) 2004;12:158-163.
12. Swift O, Ayub A, Mathavakkannan S, et al.: Outcomes following surgery for fractured neck of femur in dialysis patients: A 5-year review from a district general hospital in the United Kingdom. BMC Nephrol 2016;17:26.
13. Karaeminogullari O, Demirors H, Sahin O, et al.: Analysis of outcomes for surgically treated hip fractures in patients undergoing chronic hemodialysis. J Bone Joint Surg Am 2007;89:324-331.
14. Schaab PC, Murphy G, Tzamaloukas AH, et al.: Femoral neck fractures in patients receiving long-term dialysis. Clin Orthop Relat Res 1990:224-231.
15. Klein DM, Tornetta P, Barbera C, Neuman D: Operative treatment of hip fractures in patients with renal failure. Clin Orthop Relat Res 1998:174-178.
16. Tierney GS, Goulet JA, Greenfield ML, Port FK: Mortality after fracture of the hip in patients who have end-stage renal disease. J Bone Joint Surg Am 1994;76:709-712.
17. Blacha J, Kolodziej R, Karwanski M: Bipolar cemented hip hemiarthroplasty in patients with femoral neck fracture who are on hemodialysis is associated with risk of stem migration. Acta Orthop 2009;80:174-178.
18. Kalra S, McBryde CW, Lawrence T: Intracapsular hip fractures in end-stage renal failure. Injury 2006;37:175-184.
19. Sakabe T, Imai R, Murata H, et al.: Life expectancy and functional prognosis after femoral neck fractures in hemodialysis patients. J Orthop Trauma 2006;20:330-336.
20. Tosun B, Atmaca H, Gok U: Operative treatment of hip fractures in patients receiving hemodialysis. Musculoskelet Surg 2010;94:71-75.
21. Belmont PJ Jr, Garcia EJ, Romano D, et al.: Risk factors for complications and in-hospital mortality following hip fractures: A study using the national trauma data bank. Arch Orthop Trauma Surg 2014;134:597-604.
22. Alluri RK, Leland H, Heckmann N: Surgical research using national databases. Ann Transl Med 2016;4:393.
23. Kim L, Mabry C, Klimberg VS: Quality of benchmarks for assessment of care will influence outcome. Ann Surg 2007;245:672-673.
24. Weiss A, Anderson JE, Chang DC: Comparing the national surgical quality improvement program with the nationwide inpatient sample database. JAMA Surg 2015;150:815-816.
25. ACS-NSQIP. User Guide for the 2015 ACS NSQIP Participant Use Data File (PUF). Chicago, IL, 2015.
26. Haider AH, David JS, Zafar SN, et al.: Comparative effectiveness of inhospital trauma resuscitation at a French trauma center and matched patients treated in the United States. Ann Surg 2013;258:178-183.
27. Iacus SM, King G, Porro G: Causal inference without balance checking: Coarsened exact matching. Polit Anal 2012;20:1-24.
28. Blackwell M, Iacus S, King G, et al.: CEM: Coarsened exact matching in Stata. Stata J 2010;9:524-46.
29. Savitch SL, Shah PC: Closing the gap between the laparoscopic and open approaches to abdominal wall hernia repair: A trend and outcomes analysis of the ACS-NSQIP database. Surg Endosc 2016;30:3267-3278.
30. Augustin T, Moslim MA, Brethauer S, et al.: Obesity and its implications for morbidity and mortality after cholecystectomy: A matched NSQIP analysis. Am J Surg 2017;213:539-543.
31. Himmelfarb J: Hemodialysis complications. Am J Kidney Dis 2005;45:1122-1131.
32. Erkocak OF, Yoo JY, Restrepo C, et al.: Incidence of infection and inhospital mortality in patients with chronic renal failure after total joint arthroplasty. J Arthroplasty 2016;31:2437-2441.
33. Li WC, Shih CH, Ueng SW, et al.: Uncemented total hip arthroplasty in chronic hemodialysis patients. Acta Orthop 2010;81:178-182.
34. Clement RC, Derman PB, Kheir MM, et al.: Risk adjustment for Medicare total knee arthroplasty bundled payments. Orthopedics 2016;39:e911-e916.
35. Siddiqi A, White PB, Mistry JB, et al.: Effect of bundled payments and health care reform as alternative payment models in total joint arthroplasty: A clinical review. J Arthroplasty 2017;32:2590-2597.
36. Carter Clement R, Bhat SB, Clement ME, Krieg JC: Medicare reimbursement and orthopedic surgery: Past, present, and future. Curr Rev Musculoskelet Med 2017;10:224-232.
37. Clement RC, Kheir MM, Soo AE, et al.: What financial incentives will Be created by Medicare bundled payments for total hip arthroplasty? J Arthroplasty. 2016;31:1885-1889.
38. Ellimoottil C, Ryan AM, Hou H, et al.: Medicare's new bundled payment for joint replacement may penalize hospitals that treat medically complex patients. Health Aff (Millwood) 2016;35(9):1651-1657.
39. Sinnakirouchenan R, Holley JL: Peritoneal dialysis versus hemodialysis: Risks, benefits, and access issues. Adv Chronic Kidney Dis 2011;18:428-432.
Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Orthopaedic Surgeons.
40. Ottesen TD, Zogg CK, Haynes MS, Malpani R, Bellamkonda KS, Grauer JN: Dialysis patients undergoing total knee arthroplasty have significantly increased odds of perioperative adverse events independent of demographic and comorbidity factors. J Arthroplasty 2018;33:2827-2834.