Journal Logo

Research Article

The Impact of Resident Involvement on Postoperative Complications After Shoulder Arthroscopy: A Propensity-Matched Analysis

Gulbrandsen, Trevor R. MD; Khazi, Zain M. BS; Shamrock, Alan G. MD; An, Qiang MBBS, MPH; Duchman, Kyle MD; Marsh, J. Lawrence MD; Westermann, Robert W. MD; Wolf, Brian MD, MS

Author Information
JAAOS: Global Research and Reviews: September 2020 - Volume 4 - Issue 9 - p e20.00138
doi: 10.5435/JAAOSGlobal-D-20-00138

Abstract

Resident participation in surgical procedures is vital during orthopaedic training. Arthroscopy is especially important because previous studies have found that shoulder arthroscopy is the second most common procedure performed by orthopaedic surgeons taking the American Board Orthopaedic Surgery (ABOS) part II examination.1 The Residency Review Committee in Orthopaedic Surgery considers shoulder arthroscopy a core competency and requires residents to meet the benchmark shoulder arthroscopic case minimums before graduation.2 It is important to know whether the risk of postoperative complications is increased when residents participate in shoulder arthroscopy procedures.

Surgeon educators must assess both resident skill and surgical case complexity when determining how residents will participate in a surgical case. The duration of the procedure, patient safety, and cost effectiveness are all important considerations. General surgery, ophthalmology, neurosurgery, and urology have evaluated complications when residents are involved,3-8 with various procedures, but the impact on complications when residents participate in shoulder arthroscopy procedures has not been studied well.

The purpose of the current study was to investigate whether resident involvement in shoulder arthroscopic procedures affects short-term postoperative complication rates and surgical time in propensity score-matched cohorts. We hypothesize that resident involvement is not associated with increased postoperative complication rates or surgical time after shoulder arthroscopy.

Methods

Data Source

The American College of Surgeons National Surgical Quality Improvement Program database (NSQIP) was analyzed from 2006 to 2012 to identify patients undergoing shoulder arthroscopy procedures. The NSQIP prospectively collects over 300 preoperative and postoperative patient variables from over 600 private and academic hospitals across the United States.9 This information is collected by trained clinical reviewers for up to 30 days after surgery. Information collected in the NSQIP includes patient demographics, medical comorbidities, intraoperative information, additional surgeries, and postoperative complications within 30 days of surgery. Patient records were retrieved and analyzed using the Current Procedural Terminology (CPT) and International Classification of Diseases Ninth Revision codes. Cases are selected based on a Health Insurance Portability and Accountability Act compliant systematic sampling process with a reported interobserver disagreement rate of 2%.9,10 Owing to the data being acquired in a Health Insurance Portability and Accountability Act compliant manner, this study was granted exemption from the local institutional review board.

Patient Selection and Determining Resident Involvement

From 2006 to 2012, 15,857 shoulder arthroscopy cases were identified using relevant CPT codes (Table 1). The study period was limited to 2006 to 2012 because the NSQIP only collected information regarding resident involvement before 2013. Of the cases identified in the database, 5354 cases (33.8%) were initially excluded because of lack of information regarding resident involvement in the case. Furthermore, 425 cases (2.7%) were excluded because of concomitant open or miniopen procedures such as open rotator cuff repair, open biceps tenodesis, or mass excisions. In addition, 79 cases (0.5%) were excluded because of preoperative diagnosis of septic arthritis of the shoulder or osteomyelitis around the shoulder joint. As a result, before matching, only 9999 cases were further analyzed, of which 1798 (18%) were performed with a resident in the operating room (Figure 1).

Table 1 - Shoulder Arthroscopy CPT Codes
CPT code Description n, (%)
29806 Shoulder arthroscopy, capsulorrhaphy 755 (7.6)
29807 Shoulder arthroscopy, repair of SLAP lesion 988 (9.9)
29819 Shoulder arthroscopy, removal of loose or foreign body 48 (0.5)
29820 Shoulder arthroscopy, partial synovectomy 28 (0.3)
29821 Shoulder arthroscopy, complete synovectomy 55 (0.6)
29822 Shoulder arthroscopy, limited débridement 377 (3.8)
29823 Shoulder arthroscopy, extensive débridement 442 (4.4)
29824 Shoulder arthroscopy, distal claviculectomy (the Mumford procedure) 544 (5.4)
29825 Shoulder arthroscopy, lysis of adhesions with or without manipulation under anesthesia 159 (1.6)
29826 Shoulder arthroscopy, decompression of subacromial space with partial acromioplasty 2967 (29.7)
29827 Shoulder arthroscopy, rotator cuff repair 3532 (35.3)
29828 Shoulder arthroscopy, biceps tenodesis 118 (1.2)
CPT = Current Procedural Terminology code, SLAP = superior labral tear from anterior to posterior

Figure 1
Figure 1:
Flowchart of inclusion and exclusion criteria. NSQIP = National Surgical Quality Improvement Program

Preoperative Demographics and Comorbidities

Basic patient demographics including age, sex, and race (Caucasian, Black, or other) are recorded in the database and were compared between resident-involved and no resident involvement cohorts. Preoperative comorbidities such as obesity were calculated from each patients' height and weight. Other preoperative conditions including diabetes mellitus, smoking history, chronic obstructive pulmonary disease, and the American Society of Anesthesiologist (ASA) classification were compared between the two cohorts. Diabetes mellitus is categorized as insulin-dependent, noninsulin-dependent, or no diabetes in the database; however, for the purposes of this study, the patients were categorized as having diabetes mellitus (insulin-dependent or noninsulin-dependent) or not. Similarly, smoking status was categorized as current smoker, former smoker, or nonsmoker; however, in this study, patients were categorized as having a history of smoking tobacco (current or former smoker) or nonsmoker. In addition to preoperative comorbidities, certain preoperative laboratory values are also recoded in the NSQIP. For the purposes of this study, laboratory values for preoperative albumin, creatinine, hematocrit, international normalized ratio, platelet count, and white blood cell count were compared between resident-involved and nonresident cohorts (Table 2).

Table 2 - Comparison of Patient Demographics and Comorbidities Between Resident Involved and Nonresident Cohorts
Patient Characteristics Nonresident Resident Involved P Value
Unadjusted (n = 8705) Matched (n = 1737) Unadjusted (n = 1798) Matched (n = 1737) Unadjusted, P Value Matched, P Value
Demographics and comorbidities
 Age, mean (SD) 51.72 (15.02) 53.04 (14.9) 52.58 (14.85) 52.74 (14.76) 0.0219 0.5776
 Female sex 4491 (40.1) 708 (40.8) 731 (40.8) 711 (40.9) 0.5959 0.9175
Race/ethnicity
 Caucasian 6517 (74.9) 1312 (75.5) 1175 (65.4) 1137 (65.5) <0.0001 <0.0001
 Black 579 (6.7) 129 (7.4) 135 (7.5) 127 (7.3)
 Other 1609 (18.5) 296 (17.0) 488 (27.1) 473 (27.2)
Obesity, BMI >30 kg/m2, mean (SD) 29.7 (6.6) 29.4 (6.3) 29.6 (7.0) 29.5 (6.3) 0.0729 0.9070
Diabetes mellitus 1050 (12.1) 189 (10.9) 205 (11.4) 201 (11.6) 0.4318 0.5190
Smoking 1679 (19.3) 288 (16.6) 307 (17.1) 300 (17.3) 0.0316 0.5872
Chronic obstructive pulmonary disease 198 (2.3) 48 (2.8) 39 (2.2) 39 (2.3) 0.7840 1.3285
The American Society of Anesthesiologist classification
 1-No disturbance 1524 (17.6) 282 (16.2) 306 (17.0) 287 (16.5) 0.0634 0.3745
 2-Mild disturbance 5147 (59.3) 1057 (60.9) 1070 (59.5) 1046 (60.2)
 3-Severe disturbance 1972 (22.7) 389 (22.4) 403 (22.4) 386 (22.2)
 4-Life-threatening 42 (0.5) 9 (0.5) 18 (1.0) 18 (1.0)
Mean preoperative laboratory values
 Hematocrit, mean (SD) 41.5 (4.11) 41.6 (4.05) 41.13 (4.13) 41.17 (4.12) 0.0143 0.0645
 White blood cell count, mean (SD) 7.08 (2.3) 7.08 (2.83) 7.17 (2.46) 7.17 (2.47) 0.8947 0.4233
 Platelet count, mean (SD) 245.41 (66.67) 242.54 (64.3) 239.19 (67.91) 238.6 (67.38) 0.0039 0.1619
 INR, mean (SD) 1.05 (0.34) 1.04 (0.22) 1.12 (0.47) 1.09 (0.34) 0.3374 0.3139
 Albumin, mean (SD) 4.19 (0.43) 4.20 (0.43) 4.1 (0.40) 4.1 (0.39) 0.0001 0.0024
 Creatinine, mean (SD) 0.94 (0.45) 0.95 (0.26) 0.93 (0.43) 0.93 (0.42) 0.2489 0.0024
BMI = body mass index, INR = international normalized ratio
Bold: significant finding (P < 0.05).

Outcomes

The NSQIP database tracks patients for any occurrence of various adverse events for 30 days postoperatively. Therefore, only 30-day postoperative complications were compared between the cohorts. The primary outcomes evaluated were surgical complications, medical complications, surgical time, and length of hospital stay. In this study, surgical complications were defined as any occurrence of superficial surgical site infection (SSI), deep SSI, wound dehiscence, intraoperative or postoperative blood transfusions, and neurologic deficits. Medical complications were defined as any occurrence of pulmonary embolism, deep vein thrombosis, urinary tract infection, renal insufficiency, myocardial infarction, >48 hours of ventilation, cerebrovascular incident, pneumonia, or sepsis.

Statistical Analysis

Retrospective analysis may be confounded by selection bias; therefore, a propensity score-matching algorithm was used to create similar groups for the resident-involved and nonresident-involved cohorts. In this study, a 1:1 propensity score match was used based on age, sex, body mass index, obesity, smoking history, and ASA classification to match cases with resident involved to no resident involvement cases. The matched cohorts were compared for preoperative patient demographics, comorbidities, laboratory values, and postoperative complications using Fisher exact test or Pearson Chi-square test for categorical variables and Students t-tests for continuous variables. In addition, a multivariate Poisson analysis was also performed to compare surgical time and length of hospital stay between the cohorts. Finally, a post hoc analysis was also performed to assess the power and the effect size of the study. All statistical analyses were performed using SAS version 9.4 (Cary, NC) with statistical significance set at P < 0.05.

Results

Patient Characteristics

After propensity score matching, 3474 cases (50% with resident involvement) were included in the study. Both matched cohorts were similar in age (P = 0.578), sex (P = 0.918), body mass index (BMI) (P = 0.907), diabetes mellitus (P = 0.519), chronic obstructive pulmonary disease (P = 0.329), smoking history (P = 0.588), and ASA classification (P = 0.375), which confirmed an appropriate match (Table 2). Both groups were similar in preoperative laboratory test results except for albumin levels (nonresident cohort: 4.20 ± 0.43; resident Involved: 4.10 ± 0.39, P = 0.002) and creatinine (nonresident cohort: 0.95 ± 0.26; resident involved: 0.93 ± 0.42, P = 0.006).

Postoperative Complication Rates

The overall incidence of 30-day complications was similar in the nonresident- (0.92%) and resident-involved (0.75%) groups (P = 0.576; Table 3). No significant difference was noted in postoperative surgical complications including SSI (P = 1.000), deep SSI (P = 1.000), neurologic deficit (P = 1.000), or blood transfusion (P = 1.000). Furthermore, no significant difference was observed in postoperative medical complications including pulmonary embolism (P = 0.2183), deep vein thrombosis (P = 0.2498), urinary tract infection (P = 1.000), pneumonia (P = 1.000), or sepsis (P = 1.000).

Table 3 - Comparison of 30-day Complication Rates Between Nonresident- and Resident-Involved Groups
Complications Nonresident Cohort, n = 1737(%) Resident, n = 1737 (%) P Value
Surgical time (min) 75.1 ± 40.5 75.9 ± 35.9 0.034
Overall complication rate 16 (0.92) 13 (0.75) 0.576
Surgical complications
 Superficial surgical site infection 3 (0.17) 2 (0.12) 1.000
 Deep surgical site infection 0 (0) 1 (0.06) 1.000
 Neurologic deficit 0 (0) 1 (0.06) 1.000
 Blood transfusion 1 (0.06) 2 (0.12) 1.000
Medical complications
 Pneumonia 1 (0.06) 2 (0.12) 1.000
 Urinary tract infection 3 (0.17) 2 (0.12) 1.000
 Sepsis 0 (0) 1 (0.06) 1.000
 Deep vein thrombosis 3 (0.17) 0 (0) 0.2498
 Pulmonary embolism 5 (0.29) 2 (0.12) 0.2183
Bold: significant finding (P < 0.05).

Poisson Regression Model

Analysis of surgical time and length of hospital stay was performed using a Poisson regression model. Shoulder arthroscopy cases with resident involvement had significantly longer surgical times (resident involved: 75.9 ± 35.9 minutes; surgeon only: 75.1 ± 40.5 minutes, P = 0.034) when compared with cases performed without a resident (Table 4). No difference was observed in the length of hospital stay for cases with a resident versus cases performed without a resident (resident involved: 0.23 ± 2.42 days; surgeon only: 0.24 ± 2.79 days, P = 0.295).

Table 4 - Poisson Regression Analysis for Surgical Time and Length of Hospital Stay
Variable Nonresident Cohort Resident Involved P Value
n = 1447 n = 1447
Surgical time, min (SD) 75.1 ± 40.5 75.9 ± 35.9 0.0340
Length of hospital stay, days (SD) 0.24 ± 2.79 0.23 ± 2.42 0.2951
Bold: significant finding (P < 0.05).

Discussion

The number of shoulder arthroscopic procedures being performed in the United States is increasing.11 Therefore, it is essential for orthopaedic residents to have adequate involvement and training with these procedure. This study compared matched patient cohorts, from the NSQIP database, who underwent shoulder arthroscopy with a resident involved compared with the procedure solely being performed by practicing surgeon without resident involvement. We identified an overall low 30-day complication rate (<0.92%) with no notable difference in surgical or medical complications between the two groups. A statistically significant, however, clinically insignificant difference was observed in surgical time (P = 0.034), with longer procedures performed in the resident-involved cohort. Overall, this study demonstrates that resident involvement does not increase perioperative morbidity after shoulder arthroscopy.

The low complication rate demonstrated in this study is similar to previous studies investigating the overall complication rate of shoulder arthroscopy. Using the NSQIP database, Shields et al.12 similarly identified a short-term complication rate of 1% with a 0.57% rate of major complications and 0.53% rate of minor complications after shoulder arthroscopy. Similarly, Rubinstein et al.13 reported a 1.6% incidence of adverse events after shoulder arthroscopy in patient aged 60 years and older. Shin et al.14 examined complications associated with shoulder arthroscopy in recently trained surgeons by using the ABOS database. They performed a thorough study by evaluating 27,072 shoulder arthroscopy procedures performed by surgeons who underwent the ABOS part II examination from 2012 to 2016. They reported the rates of surgical, medical, and anesthetic complications to be 7.9%, 2.2%, and 1.0%, respectively in arthroscopic shoulder procedures. Although the surgical and medical complications were higher than other studies, including the present one, they specifically evaluated less experienced surgeons and used a larger database that included complications that the NSQIP does not, including postoperative stiffness/arthrofibrosis. In addition, the ABOS database complications are self-reported by the surgeons. Overall, the literature demonstrates a low incidence of complications after shoulder arthroscopy, and the findings of this study provide further support.

Proper resident education and training is essential for the future of healthcare. With the combination of work hour restrictions and increased diversity of orthopaedic procedures and subspecialization, residency programs must seek optimal procedural exposure for their residents to prepare them for independent practice. It is crucial for residents to be involved in core orthopaedic surgical cases to develop these skills. The impact of resident involvement in surgical procedures has been explored in various surgical and orthopaedic procedures.15-22 Haughom et al.21 performed a retrospective analysis on the impact of residents involved in total hip arthroplasty20 and total knee arthroplasty cases. These two studies did not identify resident involvement as a risk factor for 30-day morbidity or mortality in these selected cases. Cvetonovich et al. performed a propensity score analysis for patients undergoing total shoulder arthroplasty and reported no evidence of increased 30-day complications or mortality with associated resident involvement.19 More recently, Lebedeva et al.15 confirmed these findings of low complication rates with no impact of resident involvement in patients undergoing anterior cruciate ligament reconstruction.

Using the NSQIP database, Basques et al.23 investigated resident involvement in shoulder arthroscopy. They identified a 1.09% overall complication rate with resident involvement not being associated with complications, adverse events, or 30-day readmission. No reported data were observed on the differences of the length of hospital stay. They did not determine a notable difference in surgica; times between cases involving residents and those without a resident. This study also identified a low overall complication rate of 0.92% and did not find notable differences in postoperative complications, length of hospital stay, or adverse events between shoulder arthroscopies with and without residents involved. By contrast, this study determined a statistically significant difference in surgical times with resident involved cases being marginally longer on average. However, this finding is likely clinically insignificant because the resident involved cohort took 45 seconds longer on average. This minimal difference can be due to various reasons including referral of complex cases to academic centers, which can be difficult to control for using the database.

The findings of this study differ slightly from the findings by Basques et al. because of differences in statistical analysis methods. Basques et al. performed a multivariate Poisson regression analysis for all 15,774 patients to compare the rates of postoperative complications between the two cohorts. By contrast, this study had a more narrow inclusion criteria and used a propensity score analysis to match patients in resident involved and nonresident cohorts.

Propensity score matching was introduced in 1983 and is a complementary method when randomization is not possible.24 The propensity score is the probability of the assigned treatment that is conditional on certain baseline characteristics. Through this technique, a balancing score is created, and therefore, the baseline covariates between the two comparable cohorts will be similar. This allows for one to perform a nonrandom observational study that is designed to contain certain analytic characteristics of a randomized controlled trial. Although propensity score balancing methods do not allow causal conclusions, the elimination of various confounding variables can be performed, whereas regression-based approaches do not provide this same assurance.25,26 With concerns of the risk of selection bias, because Basques et al. cohort consisted only 12.3% cases involving a resident, we elected to perform propensity score analysis. Through this matching algorithm, the cohort was narrowed down to 3474 cases (50% with resident involvement). An analysis confirmed appropriate matching because both groups were similar in age (P = 0.578), sex (P = 0.918), BMI (P = 0.907), diabetes mellitus (P = 0.519), chronic obstructive pulmonary disease (P = 0.329), smoking history (P = 0.588), and ASA classification (P = 0.375). Further analysis in this matching cohort demonstrated safe and timely patient outcomes when residents are involved in shoulder arthroscopy procedures.

There were several limitations with this study. First, the retrospective analysis limits causal conclusions and only associations can be ascertained. In addition, several confounders are observed that cannot be accounted for, even with propensity score matching. Academic teaching hospitals are often tertiary centers and treat patients with complex orthopaedic problems, which alters both the operative and postoperative care of the patients. The NSQIP does not report orthopaedic-relevant issues including case-specific factors, previous surgery, or the degree of injury and only uses general CPT codes. Therefore, we cannot perform propensity match analysis based on case complexity. Furthermore, the NSQIP data have its own limitations, including the inability to assess and control for specific radiographic findings or soft-tissue injuries. In addition, NSQIP does not detail the extent of resident involvement during the surgery and only if the resident was listed during the case. Only 30-day complication rates are reported; therefore, this study is unable to assess medium- and long-term complications, patient-reported outcomes, or functional status of the patient postoperatively. This includes orthopaedic-specific outcomes. Finally, the NSQIP is the largest available database with information regarding resident involvement during surgical procedures. Therefore, the lack of short-term adverse impact of resident involvement after shoulder arthroscopy may be secondary to a type II error. According to our post hoc analysis, the power of this study was low at 0.09. However, the calculated effect size for resident involvement and adverse events was very low, suggesting that if a relationship exists, then this finding is unlikely to be clinically significant. Despite these limitations, the NSQIP database and the statistical analysis performed provide a sound analysis on the impact of resident involvement on postoperative complications and surgical times for shoulder arthroscopy.

Conclusions

Resident involvement in shoulder arthroscopy procedures is not associated with increased risk for medical or surgical 30-day postoperative complications. Overall, this study demonstrated no increased perioperative morbidity associated with resident surgical involvement.

References

1. Garrett WE Jr, Swiontkowski MF, Weinstein JN, et al.: American Board of Orthopaedic Surgery Practice of the Orthopaedic Surgeon: Part-II, certification examination case mix. J Bone Joint Surg Am 2006;88:660-667.
2. Milestones: Accreditation Council for Graduate Medical Education. https://www.acgme.org/acgmeweb/tabid/430/ProgramandInstitutionalAccreditation/NextAccreditationSystem/Milestones.aspx. Accessed January 2020.
3. Lee NJ, Kothari P, Kim C, et al.: The impact of resident involvement in Elective posterior cervical fusion. Spine (Phila Pa 1976) 2018;43:316-323.
4. Zuckerman SL, Lim J, Lakomkin N, et al.: Effect of resident and fellow involvement in adult spinal deformity surgery. World Neurosurg 2019;122:e759-e764.
5. Kim RB, Garcia RM, Smith ZA, Dahdaleh NS: Impact of resident participation on outcomes after single-level anterior cervical diskectomy and fusion: An analysis of 3265 patients from the American College of Surgeons National Surgical Quality Improvement Program Database. Spine (Phila Pa 1976) 2016;41:E289-E296.
6. Tayanithi P, Pungpapong K, Siramput P: Vitreous loss during phacoemulsification learning curve performed by third-year residents. J Med Assoc Thai 2005;88(suppl 9):S89-S93.
7. Codaccioni JL, Jean R, Depieds R, Jaquet P, Pianeta C, Lebreuil G: Clinical, histopathological and biological data for thyroiditis diagnosis [in French]. Ann Endocrinol (Paris) 1972;33:404-416.
8. Yong H, Chen Q, Yoo E, et al.: How does resident participation alter the outcome of surgery for pectus excavatum? J Surg Educ 2019;77:150-157.
9. Pugely AJ, Martin CT, Harwood J, Ong KL, Bozic KJ, Callaghan JJ: Database and registry research in orthopaedic surgery: Part 2: Clinical registry data. J Bone Joint Surg Am 2015;97:1799-1808.
10. Karhade AV, Larsen AMG, Cote DJ, Dubois HM, Smith TR: National databases for neurosurgical outcomes research: Options, strengths, and limitations. Neurosurgery 2018;83:333-344.
11. Weber SC, Abrams JS, Nottage WM: Complications associated with arthroscopic shoulder surgery. Arthroscopy 2002;18:88-95.
12. Shields E, Thirukumaran C, Thorsness R, Noyes K, Voloshin I: An analysis of adult patient risk factors and complications within 30 days after arthroscopic shoulder surgery. Arthroscopy 2015;31:807-815.
13. Rubenstein WJ, Pean CA, Colvin AC: Shoulder arthroscopy in adults 60 or older: Risk factors that correlate with postoperative complications in the first 30 days. Arthroscopy 2017;33:49-54.
14. Shin JJ, Popchak AJ, Musahl V, Irrgang JJ, Lin A: Complications after arthroscopic shoulder surgery: A review of the American Board of orthopaedic surgery database. J Am Acad Orthop Surg Glob Res Rev 2018;2:e093.
15. Lebedeva K, Bryant D, Docter S, Litchfield RB, Getgood A, Degen RM: The impact of resident involvement on surgical outcomes following anterior cruciate ligament reconstruction. J Knee Surg 2019 [Epub ahead of print].
16. Acheampong DO, Paul P, Guerrier S, Boateng P, Leitman IM: Effect of resident involvement on morbidity and mortality following thoracic endovascular aortic repair. J Surg Educ 2018;75:1575-1582.
17. Slopnick EA, Hijaz AK, Henderson JW, Mahajan ST, Nguyen CT, Kim SP: Outcomes of minimally invasive abdominal sacrocolpopexy with resident operative involvement. Int Urogynecol J 2018;29:1537-1542.
18. Mitchell PM, Gavrilova SA, Dodd AC, Attum B, Obremskey WT, Sethi MK: The impact of resident involvement on outcomes in orthopedic trauma: An analysis of 20,090 cases. J Clin Orthop Trauma 2016;7:229-233.
19. Cvetanovich GL, Schairer WW, Haughom BD, Nicholson GP, Romeo AA: Does resident involvement have an impact on postoperative complications after total shoulder arthroplasty? An analysis of 1382 cases. J Shoulder Elbow Surg 2015;24:1567-1573.
20. Haughom BD, Schairer WW, Hellman MD, Yi PH, Levine BR: Resident involvement does not influence complication after total hip arthroplasty: An analysis of 13,109 cases. J Arthroplasty 2014;29:1919-1924.
21. Haughom BD, Schairer WW, Hellman MD, Yi PH, Levine BR: Does resident involvement impact post-operative complications following primary total knee arthroplasty? An analysis of 24,529 cases. J Arthroplasty 2014;29:1468-1472.e1462.
22. Krell RW, Birkmeyer NJ, Reames BN, et al.: Effects of resident involvement on complication rates after laparoscopic gastric bypass. J Am Coll Surg 2014;218:253-260.
23. Basques BA, Saltzman BM, Mayer EN, et al.: Resident involvement in shoulder arthroscopy is not associated with short-term risk to patients. Orthop J Sports Med 2018;6:2325967118816293.
24. Rosenbaum PR, Rubin DB: The central role of the propensity score in observational studies for causal Effects. Biometrika 1983;70:41-55.
25. Austin PC: An introduction to propensity score methods for reducing the Effects of confounding in observational studies. Multivariate Behav Res 2011;46:399-424.
26. Inacio MC, Chen Y, Paxton EW, Namba RS, Kurtz SM, Cafri G: Statistics in brief: An introduction to the use of propensity scores. Clin Orthop Relat Res 2015;473:2722-2726.
Copyright © 2020 The Authors. Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Orthopaedic Surgeons.