Background: The relationship between smoking and complications after total joint arthroplasty is unclear. Prior studies have been limited by relatively small sample sizes or investigation of select cohorts. The purpose of this study was to investigate the association between smoking and readmission and/or reoperation within 90 days of total joint arthroplasty in a large, non-select cohort of patients.
Methods: Using our institutional database, we retrospectively identified patients who underwent primary total joint arthroplasty between 2000 and 2014. Patients were stratified into 1 of 3 groups: current smokers, former smokers, and nonsmokers. The association between smoking status and subsequent readmission and/or reoperation within 90 days was investigated using multivariate regression analysis.
Results: We retrospectively identified 15,264 patients (6,749 male and 8,515 female) who underwent 17,394 total joint arthroplasties during the study period. Of these patients, 1,371 (9.0%) were current smokers, 5,195 (34.0%) were former smokers, and 8,698 (57.0%) were nonsmokers. Former smokers reported a median of 22.2 years (range, 0.2 to 60 years) of abstinence prior to the surgical procedure. Current smokers were significantly younger (p < 0.001) at a mean age (and standard deviation) of 57.7 ± 10.3 years than nonsmokers at 63.2 ± 11.8 years. Current smokers were significantly more likely than nonsmokers to undergo reoperation for infection (odds ratio [OR], 1.82 [95% confidence interval (CI), 1.03 to 3.23]; p = 0.04), and former smokers were at no increased risk (OR, 1.11 [95% CI, 0.73 to 1.69]; p = 0.61). Packs per decade were independently associated with an increased risk of 90-day nonoperative readmission regardless of smoking status (OR, 1.12 [95% CI, 1.03 to 1.20]). Lastly, neither smoking status nor packs per decade were associated with aseptic or total reoperations.
Conclusions: This study, after controlling for confounding factors, demonstrated not only that current smokers have a significantly increased risk of reoperation for infection within 90 days of a surgical procedure compared with nonsmokers, but also that the amount that one has smoked, regardless of current smoking status, significantly contributed to increased risk of nonoperative readmission.
Level of Evidence: Prognostic Level III. See Instructions for Authors for a complete description of levels of evidence.
Total joint arthroplasty is a successful procedure for treating end-stage degenerative joint disease. The demand for primary total hip arthroplasty and total knee arthroplasty is expected to increase over the next decade in the United States1, and a simultaneous increase in revision total joint arthroplasty is also expected1,2.
Although the number of revision total joint arthroplasties is expected to increase, a reduction in overall complication rates is expected because of advancements in medical, surgical, and anesthetic management3. Although some important risk factors for complications such as age-associated comorbidities may not be modifiable4, other risk factors such as diabetic control, alcohol consumption, and cigarette smoking may be more responsive to perioperative intervention and management4,5.
The risks associated with cigarette smoking in the perioperative period have been previously evaluated in the total joint arthroplasty population. However, the relationship between smoking and complications is unclear. Evidence from previous studies suggests an association between smoking and poor postoperative outcomes; however, most studies are limited because of a lack of adequate sample size, absence of a multivariate analysis, or investigation of a select non-generalizable cohort such as the Veterans Administration Database, or because they have varying definitions of complications4-8. Khan et al.9 and Peersman et al.10 evaluated the effect of smoking status on reoperation for infection using a univariate analysis. Khan et al. reported that smokers had no increased risk of superficial infection (odds ratio [OR], 1.0 [95% confidence interval (CI), 0.5 to 1.8]) or deep infection (OR, 3.4 [95% CI, 0.7 to 17.2]) compared with nonsmokers. However, Peersman et al. noted that active smoking was associated with an increased overall risk of infection (OR, 1.47 [95% CI, 1.21 to 1.78]; p = 0.01)10,11.
Given these contrasting reports, the risk associated with smoking in patients undergoing total joint arthroplasty remains to be clearly defined. The enactment of the Affordable Care Act has placed hospital readmissions under increasing scrutiny, and the Centers for Medicare & Medicaid Services (CMS) National Quality Strategy has singled out hospital readmissions as one of the parameters requiring substantial improvement12. If smoking is associated with elevated perioperative risk of readmission and/or reoperation, then it may be reasonable to engage the patient in a smoking cessation program prior to total joint arthroplasty.
The purpose of our study was to investigate the association of smoking with readmission and reoperation within 90 days of the surgical procedure at a high-volume, tertiary referral total joint arthroplasty center.
Methods and Materials
Following institutional review board approval, we retrospectively identified patients from our prospectively maintained institutional database who underwent either primary total hip arthroplasty or total knee arthroplasty from 2000 to 2014 at a single institution. Patients were stratified by self-reported smoking status into 1 of 3 groups: smoker, former smoker, or nonsmoker. Patients were excluded from our cohort if smoking status was not recorded or if they were cigar or pipe smokers. Our final cohort consisted of 15,264 patients who underwent a total of 17,394 procedures. Packs per day as well as years smoked were recorded for smokers and former smokers when these data were available. It should be noted that 3% of patients were excluded for either not having any comments about tobacco or, if there were comments about tobacco, it was noted as “pipe smoker” or “cigar smoker,” which did not provide quantifiable data for analysis. To calculate packs per decade, the packs per day were adjusted for every 10 years smoked.
Patient demographic characteristics such as age, sex, body mass index (BMI), race, joint (hip or knee), type of surgical procedure (simultaneous bilateral, staged bilateral, or unilateral), Charlson Comorbidity Index, smoking status (including packs per decade), and years of abstinence for former smokers were evaluated. Age was reported as age in decades to match the reported packs-per-decade variable. The cohort was evaluated for 90-day readmission and/or reoperation at the same institution. A manual chart review was performed for all patients who had readmission and/or reoperation after the original primary index surgical procedure date to determine the indication for readmission and/or reoperation.
Descriptive statistics were utilized to report the cohort demographic characteristics, incidence of reoperation, and reasons for readmission. First, to compare the direct effect of smoking, a univariate analysis was utilized to compare the rates of aseptic and septic reoperation as well as nonoperative readmission among smokers, nonsmokers, and former smokers. Next, to identify predictors associated with 90-day readmission and reoperation outcomes, we utilized a series of 3 stacked multiple regressions to evaluate the effect of smoking status on septic and aseptic reoperation as well as nonoperative readmission. Model 1 evaluated smoking status without accounting for the Charlson Comorbidity Index and packs per decade, Model 2 evaluated smoking status accounting for packs per decade but not the Charlson Comorbidity Index, and Model 3 evaluated smoking status accounting for both packs per decade and the Charlson Comorbidity Index. These separate models were used to eliminate the potential of colinearity among confounding independent variables within each model. Additionally, a variance inflation factor (VIF), a measurement to describe the magnitude of multi-colinearity (correlation among predictors) existing in a regression analysis, was calculated for every Model 3. No reported VIF was greater than 1.5. A VIF equal to 1 indicates no colinearity, a VIF of >1 and <5 is moderately correlated, and a value of ≥5 is highly correlated. This methodology allowed for the most accurate assessment of smoking status, packs per decade, and comorbidities with regard to septic reoperation, aseptic reoperation, and nonoperative readmission. All statistical analyses were performed with the use of R software (version 3.2; R Foundation for Statistical Computing).
Of the 17,394 total joint arthroplasties, there were 8,917 total hip arthroplasties (51.3%), of which 6,797 (76.2%) were unilateral total hip arthroplasties, 1,666 (18.7%) were staged bilateral total hip arthroplasties, and 454 (5.1%) were simultaneous bilateral total hip arthroplasties. There were 8,477 total knee arthroplasties (48.7%), of which 5,299 (62.5%) were unilateral total knee arthroplasties, 1,857 (21.9%) were staged bilateral total knee arthroplasties, and 1,321 (15.6%) were simultaneous bilateral total knee arthroplasties. Of the 15,264 patients, there were 1,371 current smokers (9.0%), 5,195 former smokers (34.0%), and 8,698 nonsmokers (57.0%). The mean age (and standard deviation) was 57.7 ± 10.3 years for current smokers, 65.0 ± 10.2 years for former smokers, and 63.2 ± 11.8 years for nonsmokers. Current smokers were significantly younger than nonsmokers (p < 0.001). There were 6,749 male patients (44.2%) and 8,515 female patients (55.8%) with a mean BMI of 30.1 kg/m2 (range, 12.1 to 57.1 kg/m2) and a mean Charlson Comorbidity Index of 0.46 (range, 0 to 9). Current smokers had a higher preoperative prevalence of chronic obstructive pulmonary disease (p < 0.001), myocardial infarction (p = 0.002), and peripheral vascular disease (p = 0.002) compared with nonsmokers. Current smokers reported a median of 233.0 packs per decade (range, 1 to 2,250 packs per decade), and former smokers reported both a median of 221.1 packs per decade (range, 0.24 to 1,800 packs per decade) and a median of 22.2 years of abstinence (range, 0.2 to 60 years) prior to the surgical procedure. The patient demographic characteristics are described in detail in Table I.
For all 15,264 patients undergoing 17,394 procedures, the total 90-day readmission rate was 0.83% (n = 144) for nonoperative treatments and 2.6% (n = 445) for operative treatments. Of the 144 nonoperative readmissions, the top 2 reasons for 90-day readmission rates were wound complications (80, 55.6% of such readmissions) and venous thromboembolic events (28, 19.4% of such readmissions). Furthermore, with regard to the 90-day readmissions for reoperation, 0.71% (n = 124) were for septic reoperations and 1.8% (n = 321) were for aseptic reoperations. Further causes for reoperation and readmission are listed in Table II.
Of the 124 septic reoperations at 90 days, 42 (33.9%) consisted of placement of antibiotic-loaded spacers, 35 (28.2%) were limited to irrigation and debridement, 33 (26.6%) were irrigation and debridement with polyethylene exchange, and 14 (11.3%) were one-stage exchange arthroplasties. For the 321 aseptic reoperations, the most common indications were stiffness (212 procedures [66.0%]), fracture (47 procedures [14.6%]), and hematoma evacuation (20 procedures [6.2%]). All other causes for operative and nonoperative readmission are listed in Table II.
When assessing 90-day reoperation and nonoperative readmission using a univariate analysis, current smokers were significantly more likely than nonsmokers to undergo 90-day reoperation for infection (OR, 1.80 [95% CI, 1.10 to 2.18]; p = 0.03); however, smokers did not have increased risk of 90-day aseptic reoperation (p = 0.59) or 90-day nonoperative readmission (p = 0.48) compared with nonsmokers (Table III). Each additional pack-year for current smokers significantly contributed to the 90-day nonoperative readmission rate (OR, 1.01 [95% CI, 1.00 to 1.02]; p = 0.01).
In each of the stacked multiple regression models evaluating the independent predictive factors for septic reoperation, current smokers remained significantly more likely than nonsmokers to undergo septic reoperation (1.2% [18 of 1,531] compared with 0.56% [55 of 9,880]; p = 0.04). When smoking status, packs per decade, and Charlson Comorbidity Index were accounted for (Table IV), current smokers were 1.82 times more likely to undergo septic operation (OR, 1.82 [95% CI, 1.03 to 3.23]; p = 0.04). However, former smokers (OR, 1.11 [95% CI, 0.73 to 1.69]; p = 0.61) and packs per decade (OR, 1.02 [95% CI, 0.93 to 1.12]; p = 0.65) were not significant predictors of requiring a septic reoperation (Table IV).
When evaluating predictive factors for aseptic and total reoperation, no significant association (p > 0.05) was noted with regard to smoking status and packs per decade (Tables V and VI). Additionally, no significant association was reported between smoking status and nonoperative readmission (OR, 0.66 [95% CI, 0.32 to 1.35]; p = 0.25) (Table VII). The rate of nonoperative readmission for current smokers (0.65% [10 of 1,531]) was similar to those of both former smokers (0.87% [52 of 5,983], p = 0.50) and nonsmokers (0.87% [86 of 9,880], p = 0.48) (Table III); however, packs per decade were significantly associated with nonoperative readmission (OR, 1.12 [95% CI, 1.03 to 1.20]; p = 0.009). Furthermore, a pairwise proportion test was utilized to assess if packs per decade were significantly different between current smokers (233.0 packs per decade) and former smokers (221.1 packs per decade) in terms of predicting nonoperative readmission; however, there was no significant difference noted (p = 0.99).
The purpose of this study was to determine the association between smoking and 90-day revision surgical procedures and readmissions following primary total joint arthroplasty using a single institutional database. When controlling for confounding factors using a multivariate analysis, our primary result indicated that current smokers had a significantly higher rate of septic reoperation compared with nonsmokers. Furthermore, each additional pack-year significantly contributed to total reoperations. Additionally, there was a trend toward significance (p = 0.053) in the difference between former smokers and current smokers with regards to any 90-day reoperation or nonoperative readmission. These findings are in concurrence with several prior studies that have shown minimal differences in total complications between former smokers and current smokers7,13,14.
Although previous studies have shown associations between tobacco smoking and postoperative complications, results have been controversial because of inadequate sample size and implementation of a univariate analysis. Given that the reported rates of periprosthetic joint infection and unplanned readmission have been reported as low as 0.7% for infection and 5.3% for unplanned readmission15,16, it is crucial to evaluate a large patient cohort to effectively assess the true risk of tobacco smoking in the population undergoing total joint arthroplasty. Møller et al. evaluated the risk of complications after total joint arthroplasty associated with smoking status, and Espehaug et al. evaluated the risk of complications after total hip arthroplasty associated with smoking status; however, both studies evaluated fewer than 1,000 current smokers. Møller et al. reported that smokers were at a significantly higher risk of any complications (OR, 3.2 [95% CI, 1.6 to 4.7]) after total joint arthroplasty, but Espehaug et al. noted that neither current nor former heavy smokers were at increased risk of revision total hip arthroplasty (OR, 2.7 [95% CI, 1.4 to 3.8])17,18. Small cohorts of patients undergoing total joint arthroplasty may explain the inconsistent results that persist throughout the smoking literature6,19-22. In this present study, 17,394 cases were evaluated, which provided a more robust analysis given the narrower confidence intervals.
Univariate analysis also has limitations. When using a univariate analysis to assess smoking status and postoperative complications, inherent comorbidity and demographic differences that may exist in the patient population are not accounted for. Within our cohort, current smokers were significantly younger than nonsmokers and had higher preoperative prevalence of chronic obstructive pulmonary disease, myocardial infarction, and peripheral vascular disease compared with nonsmokers. These findings were similar to those of Duchman et al.11, who reported that the mean age of current smokers was much lower than those of former smokers and nonsmokers (58.9 years compared with 68 years, p < 0.001). Additionally, studies have shown that cigarette smokers have increased cardiovascular disease23 and a higher risk of recurrent myocardial infarction24,25. Studies in colorectal and abdominal surgery indicate that patients with congestive heart failure have a higher risk of developing surgical site infections26,27. Thus, we performed a series of stacked multiple regressions to adjust the weighted effect of smoking status, packs per decade, and comorbidities, given the potential for colinearity among confounding variables. This analytic approach provides a more accurate, independent association of smoking status, packs per decade, and associated comorbidities with postoperative reoperations and readmissions.
To overcome the limitations of other studies with small sample sizes and univariate analysis, our large-scale study was conducted in a single institution and both univariate and multivariate analyses were performed. Other studies in the literature, such as those by Singh and Duchman et al., utilized national databases, including the national Veterans Affairs Surgical Quality Improvement Program (VASQIP) and the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP), to assess the association of smoking status in elective total joint arthroplasty with 30-day postoperative complication rates including surgical site and other infections, pneumonia, stroke, myocardial infarction, mortality, and other complications. In line with our current findings, Singh7 and Duchman et al.11 reported that current smokers were significantly more likely than nonsmokers to experience postoperative complications, including surgical site infection. Furthermore, pack-years were also significantly associated with total postoperative complications. Additionally, there are fundamental limitations when using hospital databases. First, the mean age in the VASQIP cohort was 64 years, which indicates that a large portion of the patient cohort qualified for both Veterans Affairs (VA) treatment and Medicare coverage. Data on veterans who are dually eligible for VA and Medicare suggest that anywhere from 17% to 54% use both VA and fee-for-service care28-30. Linking Medicare or private-sector data to VASQIP is not possible; therefore, the reported rates of readmission and complications may be underreported. This differs from our cohort, as we utilized unique patient identifiers to evaluate readmission and reoperation at a large, single institution that is a referral center, allowing for more accurate follow-up and manual chart review. Second, 95% of the VASQIP cohort was older men, so these reported smoking risk factors are applicable to neither women nor a younger patient demographic category. Additionally, the Centers for Disease Control and Prevention have reported that those who serve in the military have a higher prevalence of cigarette smoking than civilians31. Within our cohort, as well as the ACS NSQIP, a much more equal distribution of men and women was present, which allows for a more generalizable risk of smoking status for patients undergoing elective total joint arthroplasty.
The NSQIP databases are prospectively maintained and account for nearly 10% of all registered United States hospitals32. However, this database was not designed to evaluate orthopaedic-specific conditions. Therefore, orthopaedic etiologies and diagnoses such as periprosthetic joint infection and specific aseptic complications may not be available for analysis11. Additionally, the NSQIP database does not identify the laterality of the limb requiring reoperation; therefore, generalized assumptions must be used as to which surgical procedures postoperative complications should be attributed28. In contrast, our readmission and reoperation records were identified via manual chart review, which allowed for specific and accurate identification of true septic complications, aseptic complications, and associated reoperation.
The present study evaluates 90-day septic and aseptic reoperation and nonoperative readmission rates using an institutional database, but our results need to be viewed in light of certain limitations. First, patients’ smoking status, packs per decade, and years of abstinence were retrospectively identified and were self-reported; therefore, recall bias could skew the accuracy of the collected data33,34. Morales et al. noted that self-reporting by patients newly diagnosed with cancer may not accurately assess smoking status33. These findings are also supported by Alberg et al., who noted that in patients who underwent a surgical procedure for head and neck cancer, self-reported recent smoking yielded a high proportion (39%) of false negatives34. Second, packs per decade and years of abstinence were treated as independent variables. We realize that it is important to estimate not only how much cumulative risk to the patient can accrue from smoking, but also how much of that damage can be recovered over time once the patient quits smoking. Addressing these issues properly may require more precise data than what is available in this study. Studies have suggested implementing biochemical nicotine tests as well as preoperative smoking cessation protocols to objectively define thresholds of both nicotine levels and cessation periods can better guide orthopaedic surgeons on how to effectively manage smokers in need of total joint arthroplasty35-39. Third, our patient population was from a single institution; therefore, rates of smoking as well as complication rates may not be applicable to other regions. Fourth, data on other potential risk factors for surgical revision for infection such as activity level, occupation, personality characteristics, depression, and alcohol use were not included. Finally, there may have been patients readmitted to a hospital outside the Thomas Jefferson University Hospital network, so it was possible that the calculated readmission rate was slightly deflated. However, the loss to follow-up was expected to be evenly distributed among smokers, former smokers, and nonsmokers; our results were still a reasonable reflection of the risk of operation and readmission among the different cohorts.
In conclusion, this study, after controlling for confounding factors, demonstrates that current smokers have a significantly increased risk of reoperation for infection and that packs per decade, regardless of current smoking status, is associated with unplanned nonoperative readmission. In an era in which there is substantial focus on value-based patient care and bundle payments, CMS has placed a strong emphasis on minimizing unplanned readmissions. Readmission rates are determined by access, regulatory policies, age-related comorbidities, and social determinants12. As tobacco use remains the single largest preventable cause of death and disease in the United States40, it is crucial to identify the true effect of cigarette smoking on postoperative total joint arthroplasty complications and to counsel patients on potential risks after a surgical procedure. Further research also needs to be conducted with regard to the effect of smoking cessation programs on complications after total joint arthroplasty.
Investigation performed at the Rothman Institute at Thomas Jefferson University, Philadelphia, Pennsylvania
A commentary by William G. Hamilton, MD, is linked to the online version of this article at jbjs.org.
Disclosure: There was no source of external funding 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.
- Copyright © 2017 by The Journal of Bone and Joint Surgery, Incorporated