Length of stay in healthy women with mean baseline characteristics (mean body mass index and preoperative creatinine hematocrit levels) undergoing uncomplicated total hip and total knee arthroplasty of typical duration and discharged home. Diagonal lines indicate the incidence rate ratio; shaded area on either side of each line, 95% CI.
eTable. Expected Postoperative Length of Stay (Days) for an Otherwise Healthy 65-Year-Old Woman With Average Baseline Characteristics Undergoing Uncomplicated Total Joint Replacement of Average Duration
Hart A, Bergeron SG, Epure L, Huk O, Zukor D, Antoniou J. Comparison of US and Canadian Perioperative Outcomes and Hospital Efficiency After Total Hip and Knee Arthroplasty. JAMA Surg. 2015;150(10):990-998. doi:10.1001/jamasurg.2015.1239
The combination of an aging population, growing number of medical interventions, and surging economic burden of health care has renewed interest in reevaluating and exploring new health care models.
To compare the performance of the US and Canadian health care systems by assessing major complications following primary total hip arthroplasty (THA) and total knee arthroplasty (TKA) and to measure the efficiency of both health care models by comparing the postoperative length of stay.
Design, Setting, and Participants
With patients grouped according to the country where they underwent surgery, we queried the National Surgical Quality Improvement Program database to identify 55 335 Canadian and US patients who underwent primary elective THA or TKA between January 1, 2011, and December 31, 2012.
Main Outcomes and Measures
Differences in patient demographics, as well as the rate of 30-day major complications and length of stay, were compared between patients hospitalized in the United States and Canada.
Baseline characteristics were similar between the groups. Most US patients underwent general anesthesia (THA, 61.8%; TKA, 59.4%); Canadian patients received more regional anesthesia (THA, 78.7%; TKA, 81.0%). Patients in the United States received more transfusions postoperatively (THA, 9.0% more; TKA, 6.4% more; P < .001) and had shorter hospitalizations (THA, 1.4 days less; TKA, 1.3 days less; P < .001) with a greater proportion of patients discharged to rehabilitation facilities (THA, 21.6% more; TKA, 26.6% more; P < .001). With results reported as incidence rate ratios (95% CIs), after adjusting for all other variables, risk factors, and adverse outcomes, having surgery in Canada increased the postoperative length of stay by 57% (1.57 [1.53-1.61]) for THA and 49% (1.49 [1.46-1.52]) for TKA. With results reported as odds ratios (95% CIs), major complications were significantly more common in Canada following TKA (1.65 [1.31-2.07]) but not THA (1.09 [0.79-1.51]).
Conclusions and Relevance
The rate of major complications was significantly higher in Canada following TKA, but there was no significant difference following THA. Patients undergoing the procedures in US hospitals also had substantially shorter lengths of hospital stay, perhaps reflecting more efficient postoperative care and discharge planning in those facilities.
During the past decade, the combination of an aging population, growing number of medical interventions, and surging economic burden of health care have put health care reform into the economic and political limelight. Primary hip and knee arthroplasty are prime examples. More than 1 million hips and knees are being replaced yearly in the United States1 and Canada2; demand is steadily growing2,3 and associated health care costs are rising.4- 7 It is ever more apparent that a sustainable model for joint replacement surgery must emphasize value8- 10 without compromising patient outcomes.
The United States and Canada have adopted very different approaches to the funding of health care and provide a unique opportunity for comparing care models. Overall health expenditure per capita is almost 1.5 times higher in the United States compared with Canada,11 and in-hospital costs for hip and knee replacement are similarly disparate.12 To our knowledge, few studies have directly compared in-hospital complications and hospital efficiency between both health care models partly owing to a lack of standardized metrics focused on short- and long-term results.12- 14 However, new tools are emerging, such as the American College of Surgeons’ National Surgical Quality Improvement Program (NSQIP),15 which prospectively collects demographic, perioperative, and postoperative data from hundreds of sites across North America. The NSQIP database is being used increasingly in the field of orthopedics16- 19 and allows for a direct comparison of perioperative outcomes and complications from a large cohort of patients undergoing the same surgical procedures in the United States and Canada.
The purpose of the present study was to use the NSQIP database to compare the US and Canadian health care systems following primary total hip arthroplasty (THA) and total knee arthroplasty (TKA). Our primary objective was to compare 30-day major complication rates for patients undergoing THA and TKA in either country. Our secondary objective sought to measure the efficiency of both health care models by comparing the postoperative length of stay for patients with similar baseline characteristics and receiving the same surgical procedure. Length of hospital stay has been used as a surrogate measure of efficiency.20 The null hypothesis for the present study was that both health care systems would show similar efficiency without any associated increase in postoperative major complications following THA and TKA.
The NSQIP database is the first nationally validated, risk-adjusted, outcome-based program collecting data on preoperative risk factors, intraoperative variables, and 30-day postoperative mortality and morbidity outcomes for patients undergoing major surgical procedures.21 The database is prospectively populated by surgical clinical reviewers at each participating institution.15 Variables are carefully defined,22 collected for 30 days even if the patient is discharged from the hospital, and internally audited; this procedure results in a disagreement rate of less than 1.8%.15 Cases included in the database are either a continuous collection of all patients or a random sample of patients for each tracked Current Procedural Terminology23 code (27130 for THA; 27447 for TKA) using an 8-day systematic sampling process. Inclusions are continuous and there is no filtering. Participation in the NSQIP is voluntary and open to all hospitals provided they adhere to the stringent quality controls of the database. There are currently more than 450 participating sites, with most situated in North America.24 The present ratio of US participating hospitals to Canadian hospitals is 13:1.24
The NSQIP data are published by the American College of Physicians and are compliant with the Health Insurance Portability and Accountability Act and therefore do not require institutional review board approval. The data are deidentified. We queried the NSQIP database to identify patients who underwent primary elective THA and TKA between January 1, 2011, and December 31, 2012, using Current Procedural Terminology codes 27130 and 27447, respectively. Emergency cases, revisions, hemiarthroplasties, bilateral single-stage procedures, and cases performed outside of the United States or Canada were excluded from our study. Furthermore, cases in which the operative time was less than 30 minutes or more than 300 minutes were considered outliers and excluded. A total of 55 335 primary joint replacement procedures in the NSQIP database met our search criteria and were included in the study.
Patient baseline characteristics were divided into demographic, preoperative comorbidities, preoperative laboratory values, and operative variables (Table 1). Although the NSQIP tracks more than 250 variables, only a subset is relevant to patients who undergo arthroplasty and are consistently reported in the database. Therefore, only variables germane to our study population were included in our analysis. Variables with a medical record completion rate of less than 80% were further excluded to avoid bias in the statistical analysis. Demographic characteristics included age, sex, and country where the surgery was performed (United States or Canada). A special request was made to the NSQIP to obtain the country of surgery since that is not part of the standard list of disclosed variables contained in the original database. Comorbidities comprised body mass index (calculated as weight in kilograms divided by height in meters squared), types 1 and 2 diabetes mellitus, smoking, regular use of corticosteroids (for a chronic condition), chronic obstructive pulmonary disease, congestive heart failure with dyspnea, hypertension, and American Society of Anesthesiologists classification. Laboratory tests included the hematocrit, platelet count, and creatinine. Operative variables included type of anesthesia (regional or general), duration of the surgical procedure, blood transfusion within 72 hours after surgery, and disposition after discharge from the hospital. Patients were considered to have been discharged home if they returned to their primary residence (house or nursing facility) where they were prior to undergoing THA or TKA. All other patients were discharged to a rehabilitation center, skilled or unskilled care facility, other facility not considered home, or separate acute care facility.
The rate of 30-day major complications consisted of deep wound infection, pneumonia, pulmonary embolus, acute renal failure requiring dialysis, cerebrovascular accident, cardiac arrest, and myocardial infarction. Mortality within 30 days and return to the operating room for any reason were also included in the calculation of major complications. If a patient required preoperative admission, only the postoperative length of hospital stay was used to measure hospital efficiency. Patients who died postoperatively were excluded from the analysis of length of stay.
Baseline characteristics comparing patients from the United States and Canada were calculated using a 2-sample t test for continuous variables or a Fisher exact test for binary variables. A χ2 test was used to compare 3 or more categorical variables. Continuous variables are reported as mean (SD), and categorical variables are expressed as absolute values and percentages (Table 1). Unadjusted 30-day major complication rates and length of stay were calculated using the same statistical tests (Table 2).
We examined the simultaneous effect of a patient’s country of surgery on the 30-day major complication rate while holding all other variables constant using a logistic regression model. Independent variables included those listed in Table 1 since these were believed to be potential confounders in the relationship of country of surgery and 30-day major complication. A multivariable logistic regression model was created using a priori knowledge to assess the adjusted effect of the country of surgery on a patient’s risk of major complication while controlling for other risk factors. The baseline model included initial variables believed to be important confounders and worth controlling. Further models were then created by sequentially including other variables as well as interaction and nonlinear terms. Variables thought to be unlikely true confounders were ultimately excluded from the final model. Although the final models for both TKA and THA were similar, they were not exactly concordant given that they were generated from 2 independent data sets.
The postoperative length of hospital stay was analyzed using negative binomial regression to model count data while controlling for all other variables. The dependent variable in the model was postoperative length of stay and was found to have a negative binomial distribution. Initial screening of the independent variables was performed using a simple univariable model, and incidence rate ratios with 95% CIs are reported in Table 3. A multivariable negative binomial regression model was then created using methods similar to the logistic regression. Potential confounders affecting length of stay included baseline characteristics, perioperative outcomes, and major and minor complications. The maximum likelihood estimation of the different models was assessed by comparing the log likelihood and Schwartz Bayesian information criterion to previous models. The final model was chosen according to its ability to best predict postoperative length of stay following THA and TKA. A goodness-of-fit Pearson χ2 test of the final model confirmed the count data were not overdispersed when modeled with a negative binomial distribution. Statistical analyses were performed using SAS, version 9.3 (SAS Institute Inc).
In the United States and Canada, respectively, a total of 19 480 and 2039 THAs and 30 979 and 2837 TKAs were included in the analyses. A comparison of baseline and perioperative characteristics is provided in Table 1. Overall, clinical patient baseline characteristics were similar between the countries: US patients were slightly younger (THA, by 2.2 years; TKA, 1.9 years), heavier (THA, 1.1 higher body mass index; TKA, 0.9 higher body mass index), had a higher prevalence of hypertension (THA, 11% more patients; TKA, 10% more patients), and tended to have a higher American Society of Anesthesiologists score (P < .001). Most US patients underwent general anesthesia (THA, 61.8%; TKA, 59.4%; P < .001), and Canadian patients underwent more regional anesthesia (THA, 78.7%; TKA, 81.0%; P < .001). Length of surgery was significantly longer for US patients (THA, by 21 minutes; TKA, 18 minutes; P < .001). The US patients received more transfusions postoperatively (THA, 9.0% more; TKA, 6.4% more; TKA; P < .001). Disposition on discharge from the hospital was markedly different between groups, with a greater proportion of patients discharged to rehabilitation facilities in the United States (THA, 21.6% more; TKA, 26.6% more; P < .001).
A comparison of postoperative complications is provided in Table 2. For THA, US procedures had fewer deep wound infections (0.5% vs 0.8%; P = .03) but more postoperative pneumonias (0.3% vs 0.05%; P = .04). For TKAs, US procedures had fewer pulmonary emboli (0.6% vs 1.2%; P = .002). When analyzed using a multivariable model controlling for baseline and operative characteristics (Table 4), patients undergoing surgery in Canada had a significantly higher odds ratio (OR) of major complications after TKA compared with those in the United States (OR, 1.65; 95% CI, 1.31-2.07). Conversely, there was no significant difference in major complications following THA between countries after accounting for all other variables (OR, 1.09; 95% CI, 0.79-1.51).
The unadjusted length of stay following THA and TKA was significantly different between countries, with US patients undergoing shorter hospitalizations (THA, 1.4 days less; TKA, 1.3 days less; P < .001). After adjusting for all other variables, patients undergoing THA in Canada had a 57% longer length of stay compared with those in the United States receiving the same procedure (incidence rate ratio [IRR], 1.57; 95% CI, 1.53-1.61) (Table 3). Similarly, the adjusted length of hospital stay was 49% longer for patients having a TKA in Canada compared with those in the United States (IRR, 1.49; 95% CI, 1.46-1.52) (Table 3). The expected postoperative length of stay for an otherwise healthy woman with average baseline characteristics undergoing uncomplicated THA or TKA of mean duration is described in the Figure and eTable in the Supplement.
The introduction of quality improvement programs (eg, NSQIP) that prospectively collect standardized metrics on thousands of patients from different countries provides a powerful vehicle for hospitals and governments to improve health care. To date, few studies have compared outcomes in arthroplasty procedures performed in Canada and the United States,12,25 with the countries sharing similarities with respect to the surgery and patients. Registry data demonstrate that the indications for joint replacement, types of implants used, and surgical techniques are similar between Canada and the United States,2,26 therefore facilitating a direct comparison. In the present study, an analysis of more than 55 000 primary THAs and TKAs performed in the United States and Canada revealed similar baseline characteristics of patients in both countries. Nonetheless, the rate of major complications was slightly higher in Canada following TKA but there was no significant difference following THA. Patients receiving treatment in US hospitals also had substantially shorter lengths of hospital stay, with a greater proportion of patients being discharged to rehabilitation facilities.
The rate of major complications following primary THA and TKA is very low in both Canada and the United States, lending credence to the safety and success of these procedures. Our study exemplifies the ongoing decline in major complications, presumably owing to improvements in surgical technique, anesthesia administration, and perioperative medical management. For example, the 30-day mortality rate reported by Parvizi et al27 in a review of 30 000 THA procedures demonstrated a continual decline in the mortality rate from 0.94% in the 1970s to 0.27% in the 1990s and was further corroborated by the even lower rate of 0.19% shown in our study. Previously reported rates of cardiac events,28,29 pneumonia,29 and wound infections30,31 are similarly higher than those found in the present study. Notwithstanding the similarity in unadjusted major complication rates between patients receiving care in the United States and Canada (Table 2), the multivariable analysis identified a significantly higher overall rate of major complications in Canadian patients after TKA. It is unclear from the data collected in this study why Canadian patients undergoing TKA have a higher odds of major complication; therefore, this finding merits further investigation. One plausible explanation is that Canadian patients spend more time on waiting lists,25,32,33 which has been shown34 to result in progressive loss of mobility and health-related quality of life while waiting more than 6 months to undergo joint replacement. Although no direct association has been made between prolonged wait times and increasing postoperative complications, patients living with hip or knee osteoarthritis have a higher risk of death compared with the general population, with the risk increasing with the severity of walking disability.35
During the past decade, length of hospital stay has decreased substantially in both Canada2 and the United States28,36,37 as a result of fast-track surgical pathways,38 improved postoperative pain control,39 minimally invasive arthroplasty techniques,40 and level 1 evidence that home-based rehabilitation is as effective as inpatient rehabilitation.41 In the present study, patients in Canada remained hospitalized 57% and 49% longer than did those in the United States following THA and TKA, respectively (Table 3). The growing supply-demand mismatch in hospital beds available for total joint replacement,42 in addition to emerging fixed-payment hospital reimbursement programs in the United States, have intensified efforts to reduce length of stay. The prompter discharge of patients in US hospitals may therefore reflect this concerted effort with greater efficiency in postoperative care and discharge planning. Furthermore, the availability of rehabilitation facilities in Canada is limited43 and may represent an important obstacle to patient discharge. The proportion of patients in the United States discharged to a rehabilitation facility was more than 3-fold that in Canada.
The overall baseline characteristics of patients undergoing THA or TKA in the United States and Canada had no clinically important differences in terms of demographics, comorbidities, and preoperative laboratory test values. There were, however, several interesting differences in perioperative and postoperative care between countries. Most US patients underwent general anesthesia (THA, 61.8%; TKA, 59.4%), whereas most Canadian patients received regional anesthesia (THA, 78.7%; TKA, 81.0%). Numerous studies18,44- 47 have shown that regional anesthesia has several advantages over general anesthesia in patients undergoing arthroplasty, including lower rates of deep vein thrombosis, pulmonary embolism, intraoperative bleeding and transfusion, surgical site infection, and total operative cost. More recently, however, Harsten et al48 found that general anesthesia provided for shorter length of stay and better pain control 6 hours after surgery. Postoperative transfusion rates in US patients were almost double those of the rates in Canadian patients. The discrepancy likely represents differences in blood-conserving practices between the countries, such as the widespread administration of tranexamic acid, decreased use of autologous blood, and adoption of restrictive transfusion policies that have become commonplace among Canadian centers.49,50
Limitations of this study stem from inadequacies of the NSQIP database. Outcomes are not tracked beyond 30 days following surgery, and information on the type of implants and approach used during surgery are not recorded. Nonetheless, the database includes an exhaustive list of perioperative variables and complications as well as high-quality data, which are entered prospectively, standardized, and audited. In the present study, registry data were used to infer that the technical aspects and indications for surgery were similar between the countries. Second, arthroplasty-specific outcomes, such as revision rates and functional outcomes, are not captured in the NSQIP database. In addition, the NSQIP database represents a small subset of hospitals in Canada and the United States and may not be representative of each country. Furthermore, selection bias related to surgeon and hospital volume, type of hospital (academic vs community), and surgeon experience cannot be controlled for with the variables collected in the NSQIP database. Nevertheless, the fact that more than 450 hospitals are included in the database mitigates selection bias and makes the NSQIP the largest database of its kind. Another strength of the NSQIP is that it is one of the few databases that includes hospitals and patients from different countries allowing for a direct comparison between countries given similar methods and data collection.
Comparing perioperative complications of surgical procedures for primary joint replacement performed in US and Canadian centers provides a vehicle for implementing change based on the strengths and weaknesses of each system. In this study, we looked at more than 55 000 primary THAs and TKAs performed in the United States and Canada. Although patient baseline characteristics and procedures were similar, the rate of major complications was higher in Canada following TKA, but no significant difference was observed following THA. Length of stay was substantially longer in Canadian hospitals by approximately 50%. Gains in efficient patient discharge may be gleaned from US counterparts without increasing major complication rates.
Accepted for Publication: April 12, 2015.
Corresponding Author: Adam Hart, MSc, MD, Department of Orthopaedic Surgery, Jewish General Hospital, 3755 Côte-Sainte-Catherine Rd, Montréal, QC H3T 1E2, Canada.
Published Online: August 19, 2015. doi:10.1001/jamasurg.2015.1239.
Author Contributions: Dr Hart had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hart, Bergeron, Huk, Zukor, Antoniou.
Acquisition, analysis, or interpretation of data: Hart, Bergeron, Epure.
Drafting of the manuscript: Hart, Bergeron.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Hart, Bergeron, Epure.
Study supervision: Bergeron, Huk, Zukor, Antoniou.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Julia Makhlin, MSc, and Gina Ciccotosto, MSc (A) (Jewish General Hospital), helped with accessing the National Surgical Quality Improvement Program database and made the special request to access the country-specific variable in the database. There was no financial compensation.