Sex and Surgical Outcomes and Mortality After Primary Total Knee Arthroplasty: A Risk-Adjusted Analysis


  • Jasvinder A. Singh,

    Corresponding author
    • Birmingham VA Medical Center and University of Alabama at Birmingham, and Mayo Clinic College of Medicine, Rochester, Minnesota
    Search for more papers by this author
    • Dr. Singh has received consultancy fees, speaking fees, and/or honoraria (less than $10,000 each) from Allergan, Novartis, Regeneron, Saviant, and URL, and (more than $10,000 each) from Areda and Takeda.

  • C. Kent Kwoh,

    1. VA Center for Health Equity Research and Promotion and Philadelphia VA Medical Center, Philadelphia, and VA Pittsburgh Healthcare System, Pittsburgh, Pennsylvania
    Search for more papers by this author
  • Diane Richardson,

    1. VA Center for Health Equity Research and Promotion and Philadelphia VA Medical Center, Philadelphia, and VA Pittsburgh Healthcare System, Pittsburgh, Pennsylvania
    Search for more papers by this author
  • Wei Chen,

    1. VA Center for Health Equity Research and Promotion and Philadelphia VA Medical Center, Philadelphia, and VA Pittsburgh Healthcare System, Pittsburgh, Pennsylvania
    Search for more papers by this author
  • Said A. Ibrahim

    1. VA Center for Health Equity Research and Promotion, Philadelphia VA Medical Center, and University of Pennsylvania Perelman School of Medicine, Philadelphia, and VA Pittsburgh Healthcare System, Pittsburgh, Pennsylvania
    Search for more papers by this author

  • The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States Government.

University of Alabama at Birmingham, Faculty Office Tower 805B, 510 20th Street S, Birmingham, AL 35294. E-mail:



Total knee arthroplasty (TKA) is a widely utilized and effective treatment option for end-stage knee osteoarthritis (OA). Knee OA is more prevalent among women compared to men, but there are limited data on the sex differences in surgical outcomes after primary TKA.


Our sample consisted of all primary TKAs performed in Pennsylvania during the fiscal year 2002. We used International Classification of Diseases, Ninth Revision, Clinical Modification codes to identify major complications and surgical revision. We used mixed-effects logistic regression models to examine the associations between sex and all-cause mortality, readmissions, and major surgical complications. We used proportional hazards models to assess the risk of surgical revision after index arthroplasty. We adjusted for race, age, hospital teaching status, hospital procedure volume, insurance status, and risk of mortality.


In 17,994 primary TKAs, there were 46 and 220 deaths at 30 days and 1 year, respectively. Compared to women, men had higher adjusted odds of 1-year mortality (odds ratio [OR] 1.48 [95% confidence interval (95% CI) 1.13–1.94]) after primary TKA. The overall odds of most major 30-day complications did not differ by sex except for surgical wound infections, which were higher in men compared to women (OR 1.31 [95% CI 1.08–1.60]); 30-day readmission was higher in men (OR 1.25 [95% CI 1.10–1.43]). Men had significantly higher rates of revision of index knee arthroplasty at 5 years (hazard ratio 1.20 [95% CI 1.05–1.36]) compared to women.


The higher rates of mortality, hospital readmissions, revision surgery, and wound infections in men undergoing elective primary TKA compared to women indicate there is a sex disparity in these outcomes.


An estimated 670,000 total knee arthroplasties (TKAs) were performed annually in the US according to the 2009 Healthcare Cost and Utilization Project ([1]), with a higher prevalence of TKAs in women than men ([2]). Total joint arthroplasty (TJA) is associated with a decrease in pain, improved functional capabilities, and improved quality of life ([3-5]). The effectiveness of TKA in achieving these goals has made TKA among the most commonly performed elective surgeries in the elderly ([6]). In conjunction with the continued aging of the population, the prevalence of TJA is expected to increase over the coming decades, with an estimated 3.5 million TKAs to be performed annually by 2030 in the US ([7]).

Although TJA is considered to be a relatively safe procedure ([3]), only 8.8% of men and 12.7% of women who are considered candidates for the procedure are willing to undergo a surgical intervention ([8]). This discrepancy is primarily due to patient uncertainty about the risks associated with surgery ([9]). Several known major complications of TJA include venous thromboembolism (VTE; a combination of deep venous thrombosis and pulmonary embolism), myocardial infarction (MI), and even death ([3, 10]). Potential predictive factors associated with an increased risk of surgical complications include age, race/ethnicity, medical comorbidities, body mass index, insurance type, hospital procedure volume, and sex ([11-14]).

Another important complication of TJA is early implant failure leading to early revision surgery. The surgical revision rate due to implant failure is ∼1% per year after TKA ([2, 15]). Callahan et al evaluated 130 studies assessing >9,000 patients who had undergone primary TKA and reported a surgical revision rate of 4% at a minimum of 4 years of followup ([15]). Women made up nearly two-thirds of all patients undergoing TJA in the US ([2]). Despite several studies examining the effect of sex on pain and functional outcomes after TJA, sex has largely gone underrecognized as it relates to surgical complications. In recent years, concerns have been raised regarding anatomic differences of the hip and knee between men and women ([16]). Such differences may have implications for surgical outcomes such as revision rates. In the present study, we used a large regional database consisting of >18,000 elective TKAs done at over 169 hospitals. Our objective was to determine whether, in risk-adjusted analyses, men and women undergoing primary TKA differed in their surgical outcomes and adverse events, as assessed by overall mortality, major complications, hospital readmissions, and surgical revision rates.

Box 1. Significance & Innovations

  • This study provides a risk-adjusted analysis of postarthroplasty complications from a regional sample of participants receiving an elective primary total knee arthroplasty (TKA).
  • Compared to women, men had a 48% higher adjusted odds of 1-year mortality after primary TKA.
  • Compared to women, men had higher adjusted odds of surgical wound infections (31% higher), 30-day readmission (25% higher), and 5-year revision hazard rate (20% higher), findings that add to the current knowledge.


Data sample and data collection

Our study cohort was identified using the Pennsylvania Health Care Cost Containment Council (PHC4) database, which includes demographic data from all patient discharges from 169 hospitals in the state of Pennsylvania. The cohort comprised all cases of primary TKA performed in the state of Pennsylvania during the fiscal year 2002 and outcomes were assessed throughout fiscal year 2007. The cases were identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 81.54 for primary TKA. All patients who had undergone primary hip or revision knee or hip replacement surgery during the same hospitalization were excluded from the analysis. We excluded patients who had a documented prior TKA using ICD-9-CM code v43.65. The database includes information on demographics for all patients who underwent TKA or total hip arthroplasty. The study cohort and methodology have previously been described in detail ([12]).

Study measures

The primary predictor of interest was sex. The other variables examined as potential confounders of surgical outcome included age, race or ethnicity, hospital teaching status (teaching or nonteaching), insurance status (categorized as none or unknown, Medicaid, Medicare/government, or private), and hospital procedure volume. We categorized hospital procedure volume into 2 levels (1–100 and >100 TKAs performed). This categorization was based on an a priori consideration of the potential impact of hospital volume balanced with a consideration of the relatively small number of very high volume facilities. Based on procedure volume categories in a previous study ([12]), we combined very low (≤25) with low (26–100) procedure volume categories, and we combined high (101–200) with very high (>200) procedure volume categories.

For surgical mortality risk adjustment, we used the 3M All Patient Refined-Diagnosis Related Group Risk of Mortality (APR-ROM) ([17]) subclass, as assigned in the PHC4 database. This validated risk assessment tool, developed from clinically based models tested with historical data, provides a categorical risk assessment based on patient age, principal diagnosis, the type of surgical procedure, and interactions with comorbid conditions and combinations of comorbidities. The APR-ROM assigns, for the specific surgical procedure, a minor, moderate, major, or extreme risk of death to each patient ([18-21]). We included 1 patient whose APR-ROM category was unknown with the largest risk group (the minor risk group, which included ∼80% of men and women) and labeled this group as minor/unknown.

To assess complication rates, we used ICD-9-CM codes to identify medical and surgical complications such as MI, VTE, prosthetic device complication, and surgical wound infection (Table 1). The overall complication rate was defined as the occurrence of ≥1 of MI, VTE, prosthesis-related complication, or surgical wound infection within 30 days of the primary TKA. We identified 30-day readmissions based on the occurrence of any inpatient admission within 30 days of discharge from the index hospitalization. Deaths were identified based on the presence of a nonmissing value for the variable indicating the number of days from surgery to death. To assess surgical revision rates, the entire cohort was followed up prospectively for up to 5 years following the index surgery (Table 1). The study was approved by the University of Pittsburgh Institutional Review Board.

Table 1. Diagnostic codes for complications*
 ICD-9-CM codes
  1. ICD-9-CM = International Classification of Diseases, Ninth Revision, Clinical Modification; VTE = venous thromboembolism; MI = myocardial infarction; TKA = total knee arthroplasty.
VTE415.1, 415.11, 415.19, 451.11, 451.19, 451.2, 451.81, 451.9, 453.40, 453.41, 453.42, 453.8, 453.9
MI410.00, 410.01, 410.10, 410.11, 410.20, 410.21, 410.30, 410.31, 410.40, 410.41, 410.50, 410.51, 410.60, 410.61, 410.70, 410.71, 410.80, 410.81, 410.90, 410.91, additional code of 997 with any of the above for postoperative MI
Prosthetic device complication996.40, 996.41, 996.42, 996.43, 996.46, 996.47, 996.49
Surgical wound infection682.5, 682.6, 682.8, 682.9, 998.51, 998.59
Revision TKA surgery81.55, 0080, 0081, 0082, 0083, 0084

Statistical analysis

All comparisons of categorical variables between men and women were done using chi-square tests. Unadjusted comparisons of the complication rates between women and men who underwent primary TKA were done for each individual complication as well as the combined complications. In adjusted analyses, we evaluated sex differences for 30-day complications, 30-day readmission, 1-year mortality, and surgical revision within 5 years (5-year TKA revision rates). To compare the adjusted odds of complications between men and women, we fit mixed-effects logistic regression models that accounted for clustering at the hospital level with random hospital intercepts and included patient age, race/ethnicity, surgical risk of mortality, insurance status, and hospital teaching status and procedure volume as fixed effects. We incorporated age as a continuous and categorical predictor, and after confirming that the results were essentially identical, we reported results based on categorical age. The sex differences in mortality and 30-day readmission were similarly evaluated using mixed-effects logistic regression models, controlling for the same covariates. An interaction between sex and surgical risk was tested for inclusion in models for mortality, 30-day readmission, and overall complications. Since this interaction was not significant, we concluded that the effect of sex on these outcomes did not differ across the levels of surgical risk.

We estimated unadjusted revision rates using the Kaplan-Meier survival analysis, in which patients who died or were lost to followup were censored. To further examine sex differences after adjusting for the covariates listed above, we performed a time to event analysis using competing risks Cox proportional hazards regression. In our analysis, we treated death as a competing risk with respect to surgical revision rather than as a loss to followup because death precludes revision, and because it is unlikely that patients who die have the same risk of surgical revision as patients who do not die. We fitted a single competing risks proportional hazards model adjusted for all covariates and included death as a competing outcome in addition to surgical revision. In this way, we estimated the impact of sex on surgical revision and on death separately, but within the same model ([22]). We censored patients who did not experience either revision or death by the end of the 5-year followup period. We report hazard ratios (HRs) for surgical revision estimated from this model.

We assessed the validity of the proportional hazards assumption for each covariate by testing for a correlation between scaled Schoenfeld residuals with the rank of failure time and by inspecting time-based HR plots. All statistical analyses were performed using SAS, version 9.2.


Baseline demographic and clinical characteristics

We identified 17,994 primary TKAs done at 169 hospitals in the state of Pennsylvania during the fiscal year 2002 (Figure 1). The median age was 69 years for men and 69 years for women; 84% of the patients were identified as white, while 3% of men and 6% of women were identified as African American. Race was categorized as either unknown or other for 10.6% of men and 10.0% of women. Other demographic and clinical characteristics of the study cohort are shown in Table 2.

Figure 1.

Patient selection flow chart. PHC4 = Pennsylvania Health Care Cost Containment Council; LOS = length of stay.

Table 2. Clinical and demographic characteristics of patients who underwent primary TKA (N = 17,994)*
 Women (n = 11,669)Men (n = 6,325)Pa
  1. Values are the percentage unless otherwise indicated. TKA = total knee arthroplasty; IQR = interquartile range.
  2. aP values are based on chi-square tests for categorical variables and Wilcoxon's rank sum test for numeric variables (age and volume).
  3. bSignificant.
Race  < 0.0001b
African American5.93.3 
Hospital teaching status  0.9386
Teaching hospital24.624.7 
Nonteaching hospital75.475.3 
Surgical risk of death  < 0.0001b
Major/extremely likely3.54.7 
Insurance  < 0.0001b
Hospital procedure volume, median (IQR)172 (111–279)182 (112–290)0.0028b
Hospital procedure volume  0.2053
Age, median (IQR) years69 (60–76)69 (60–75)0.0876
Age, years  0.6468

Surgical outcomes, unadjusted comparisons

The surgical outcomes are shown in Table 3. There were no differences in the rates of specific complications such as MI, VTE, or prosthesis failure by sex. The rates of 30-day wound infection complications were higher in men than women (2.77% and 2.13%, respectively [P = 0.007]). However, in terms of overall complications, men had higher rates than women of both in-hospital complications (2.04% and 1.67%, respectively [P = 0.076]) and complications within 30 days of the index TKA (6.18% and 5.26%, respectively [P = 0.010]). Men also had higher rates than women of 30-day readmission (6.74% and 5.44%, respectively [P = 0.0004]) and 1-year mortality (1.60% and 1.02%, respectively [P = 0.0008]). In addition, men had higher revision rates than women, both at 1 year (1.76% and 1.39%, respectively [P < 0.0005]) and at 5 years (6.23% and 5.35%, respectively [P = 0.010]).

Table 3. Surgical outcomes of primary TKA*
 Total (n = 17,994)Women (n = 11,669)Men (n = 6,325)P
  1. Values are the number (percentage) unless otherwise indicated. TKA = total knee arthroplasty; MI = myocardial infarction; VTE = venous thromboembolism.
  2. aSignificant.
  3. bMortality percentage calculated as the number of patients who died divided by the total number of patients.
  4. cRevision rates were estimated using the Kaplan-Meier method. P values are based on the log rank test.
In-hospital complications    
Overall complications324 (1.80)195 (1.67)129 (2.04)0.076a
MI60 (0.33)35 (0.30)25 (0.40)0.28
VTE184 (1.02)119 (1.02)65 (1.03)0.96
Surgical wound infection60 (0.33)29 (0.25)31 (0.49)0.007a
Prosthesis failure or device complication30 (0.17)16 (0.14)14 (0.22)0.186
30-day complications    
Overall complications1,005 (5.59)614 (5.26)391 (6.18)0.010a
MI89 (0.49)50 (0.43)39 (0.62)0.086
VTE482 (2.68)305 (2.61)177 (2.80)0.46
Surgical wound infection423 (2.35)248 (2.13)175 (2.77)0.007a
Prosthesis failure or device complication55 (0.31)33 (0.28)22 (0.35)0.451
Readmission and mortality    
30-day readmission1,061 (5.90)635 (5.44)426 (6.74)0.0004a
30-day mortalityb46 (0.26)24 (0.21)22 (0.35)0.071
1-year mortalityb220 (1.22)119 (1.02)101 (1.60)0.0008a
Revision ratesc    
1-year revision298 (1.57)162 (1.39)136 (1.76)< 0.0005a
5-year revision1,012 (5.66)627 (5.35)385 (6.23)0.01a

Thirty-day major complications from multivariable models

In the adjusted multivariable models, there were no statistically significant sex differences regarding 30-day overall complications, MI, and VTE (Figure 2). However, there were significantly higher 30-day surgical wound infection rates in men compared to women (odds ratio [OR] 1.31 [95% confidence interval (95% CI) 1.08–1.60, P = 0.007]).

Figure 2.

Adjusted odds ratios for complications within 30 days in men following primary total knee arthroplasty, with women as the reference category. Analyses were adjusted for age category, race, surgical risk of mortality category, insurance group, teaching status, and hospital procedure volume. 95% CI = 95% confidence interval.

Thirty-day readmission and mortality from multivariable models

After adjusting for age, race/ethnicity, surgical risk, insurance status, hospital teaching status, and hospital procedure volume, the odds of 30-day readmission were significantly higher for men compared to women (OR 1.25 [95% CI 1.10–1.43, P = 0.0006]) following primary TKA (Figure 3). Thirty-day mortality was rare (46 deaths for men and women combined); we did not fit a multivariable model because the number of deaths was too small to support our model with its 7 independent variables.

Figure 3.

Adjusted odds ratios (ORs) for 30-day readmission and 1-year mortality and adjusted hazard ratio (HR) for 5-year revision rate in men after knee arthroplasty, with women as the reference category. Analyses were adjusted for age category, race, surgical risk of mortality category, insurance group, teaching status, and hospital procedure volume. 95% CI = 95% confidence interval.

One-year mortality from multivariable models

After adjusting for age, race/ethnicity, surgical risk, insurance status, hospital teaching status, and hospital procedure volume, the odds of 1-year mortality were significantly higher for men compared to women (OR 1.48 [95% CI 1.13–1.94, P = 0.004]) following primary TKA (Figure 3).

Surgical revision at 5 years from multivariable models

After controlling for potential confounding variables in a Cox proportional hazards model for surgical revision with mortality as a competing risk, allowing HRs to be different for the 2 outcomes and incorporating 5 years of followup, the adjusted HR for surgical revision (comparing men to women) was 1.20 (95% CI 1.05–1.36, P = 0.006).


In this sample of patients from a large regional database, we found that men had significantly higher rates of 30-day surgical wound infection rates compared to women, but we found no significant sex differences regarding other major complications following TKA. Compared to women, men had higher odds of surgical revision during a 5-year followup after the index primary TKA. Clinically, what an HR of 1.20 means is that at any given time, 1.20 times as many men as women undergo surgical revision. In addition, the overall mortality 1 year after TKA was also higher for men compared to women. A major advance in our study compared to previous studies of surgical complications is that our analyses accounted for surgical risk with a measure that includes interactions of multiple comorbid conditions and age with the type of surgical procedure and the principal diagnosis (APR-ROM). This, and our use of competing hazards models to examine mortality and surgical revision, added methodologic and clinical rigor to our study. As a result, we believe that our findings might be less subject to bias due to confounding and modeling limitations and might be more generalizable.

Although studies examining sex differences in surgical outcomes after TKA are sparse in the literature, our findings are consistent with that of similar, albeit limited, studies ([15, 23]). A previous analysis found that the overall incidence of major surgical complications was low for both men and women; this supports the general view that TKA is a relatively safe procedure ([24, 25]). Previous studies found a higher risk of revision surgery 2 years after primary TKA ([23]) and 90-day mortality ([11]) in men compared to women, and another study found no difference in the overall in-hospital complication rate by sex ([24]); these findings are consistent with the results from our study. The risk of revision surgery after TKA in women compared to men was reported in a recent review that focused on sex differences ([26]). The potential reasons for higher revision rates in men include a higher rate of osteolysis ([27]) and/or higher rates of polyethylene wear and infection ([28]) compared to women. One may speculate that the higher rates of wear and osteolysis in men may be related to higher physical activity and/or differences in knee biomechanics. Our findings parallel a finding from a previous study that male sex was associated with a higher rate of complications after hip arthroplasty in the Medicare population ([29]). Exploratory analyses revealed that the relationship between sex and surgical revision differed between patients who were <65 years of age compared to patients ≥65 years. We incorporated age in the models as both a continuous and categorical predictor, and after confirming that the results were identical, we reported HRs adjusted for categorical age while noting that the APR-ROM class also incorporated an age adjustment.

The higher mortality in men versus women after TKA likely reflected the sex-specific cardiovascular disease mortality advantage for women, which is well described in the literature ([30-32]). Our study significantly adds to the current knowledge in this area by confirming these findings in risk-adjusted analyses of a large state-wide database, and also adds further knowledge regarding sex differences by providing complication rates by sex.

Our study addresses a limitation observed in previous studies using a Medicare database; these studies only included patients ages >65 years. Patients ages <65 years make up ∼40% of all TJA recipients ([11]). Our study included all patients who underwent TKA in Pennsylvania; therefore, these findings are more generalizable than those from studies limited by focusing only on patients with a more limited age range, such as patients from a Medicare dataset. Our findings can be used to inform patients of their risk of complications and revision prior to TKA.

Understanding sex variations in surgical outcomes after TJA is of critical importance because the prevalence of hip and knee osteoarthritis (OA), as well as the primary indication for elective hip or knee arthroplasty, is greater among women than men ([8]). Although the age-adjusted rates of TJA are higher for women, the overall need for arthroplasty is 3 times greater for women compared to men ([8]). Orthopedic surgeons are more hesitant to recommend joint arthroplasty to women ([33]), possibly because of a concern that women may have worse outcomes after surgery than men. This concern stems from previous research suggesting that women, compared with men, delay surgery until approaching end-stage knee OA, which may have an effect on clinical and surgical success rates ([34]). Interestingly, sex-specific joint prostheses have not been shown to impact perioperative surgical complications, patient-reported outcomes such as pain and function, or revision rates following primary knee arthroplasty.

There is growing interest in sex-specific joint prostheses. Female-specific knee implants have been marketed to orthopedic surgeons, indicating the possibility of improved surgical outcomes for women despite a paucity of evidence-based peer-reviewed studies ([16, 35]). Most women do not receive sex-specific implants, which were recently introduced. Therefore, the use of these implants is unlikely to explain the sex-based differences. Although it is possible that sex-specific joint prostheses may impact patient-centered clinical outcomes such as postprocedure pain and function, we were not able to evaluate these outcomes in our database. We did not have sufficient information in our database to assess the possibility that sex-specific joint prostheses provide better long-term surgical outcomes for women compared with men (i.e., beyond the 5-year postoperative period). Since our study was not designed to assess the underlying reasons for poorer implant survival in men, more research is needed to further investigate the causes of implant failure in men. Our findings should not be overinterpreted to suggest the development of male-specific implants.

There are important limitations to consider when interpreting the results of this study. First, this analysis used a large administrative database that was designed for hospital performance assessment but contains inadequate information on key variables including body mass index; preoperative pain, function, and radiographic stage; and postoperative outcomes such as patient improvement in pain and function. There may also have been potential inconsistencies in documentation. However, the use of administrative databases such as the PHC4 database to evaluate patient outcomes has been supported by leading health care quality organizations such as the Agency for Healthcare Research and Quality and the Medicare Professional Review Organization ([36]). Second, our database lacked information regarding the utilization of a specific brand name of joint prostheses (i.e., the Gender Knee). Therefore, by using the PHC4 data, we could not accurately compare the different types of prostheses or assess patient satisfaction, pain, or functional status. It is possible that women may have worse patient-reported outcomes post TKA, such as less improvement in pain and function, possibly impacting their decision to undergo revision arthroplasty. Third, our sample consisted of only patients who underwent surgery in the state of Pennsylvania. It is possible that, for some patients, followup procedures such as surgical revisions may have been performed out of state. However, there is no specific reason to assume that state surgical outsourcing should vary by patient sex. Another limitation is that we did not have an accurate record of previous joint replacements; therefore, the revision TKA may be related to the index TKA or contralateral TKA done previously. However, this is the limitation common to all database studies and not specific to only our study. This misclassification likely biased our results toward the null and was not expected to differ by sex, so it is an unlikely explanation for the observed sex differences. Last, our conclusions may not be readily extrapolated to other regions or states of the country, because there is a significant regional variation regarding the rates of TKAs performed ([37]).

In conclusion, this analysis of nearly 18,000 patients who underwent primary knee arthroplasty in the state of Pennsylvania found that men had more complications, worse outcomes, and higher mortality compared to women. Future studies are needed to evaluate whether these findings can be generalized to the nation at large. Our findings showing a higher risk of complications and mortality in men compared to women add important knowledge to our understanding of sex differences in TKA. While women have a disadvantage in terms of care in several health conditions compared to men, it is reassuring to note that complications following TKA are in fact lower for women.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Singh had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Singh, Kwoh, Richardson, Ibrahim.

Acquisition of data. Ibrahim.

Analysis and interpretation of data. Singh, Richardson, Chen, Ibrahim.