Advertisement

Variability in invasive mediastinal staging for lung cancer: A multicenter regional study

Open ArchivePublished:February 09, 2018DOI:https://doi.org/10.1016/j.jtcvs.2017.12.138

      Abstract

      Objective

      Prior studies have reported underuse of—but not variability in—invasive mediastinal staging in the pretreatment evaluation of patients with lung cancer. We sought to compare rates of invasive mediastinal staging for lung cancer across hospitals participating in a regional quality improvement and research collaborative.

      Methods

      We conducted a retrospective study (2011-2013) of patients undergoing resected lung cancer from the Surgical Clinical Outcomes and Assessment Program in Washington State. Invasive mediastinal staging included mediastinoscopy and/or endobronchial/esophageal ultrasound-guided nodal aspiration. We used a mixed-effects model to mitigate the influence of small sample sizes at any 1 hospital on rates of invasive staging and to adjust for hospital-level differences in the frequency of clinical stage IA disease.

      Results

      A total of 406 patients (mean age, 68 years; 69% clinical stage IA; and 67% lobectomy) underwent resection at 5 hospitals (4 community and 1 academic). Invasive staging occurred in 66% of patients (95% confidence interval [CI], 61%-71%). CI inspection revealed that 2 hospitals performed invasive staging significantly more often than the overall average (94%, [95% CI, 89%-96%] and 84% [95% CI, 78%-88%]), whereas 2 hospitals performed invasive staging significantly less often than overall average (31% [95% CI, 21%-44%] and 17% [95% CI, 7%-36%]).

      Conclusions

      Rates of invasive mediastinal staging varied significantly across hospitals providing surgical care for patients with lung cancer. Future studies that aim to understand the reasons underlying variability in care may inform quality improvement initiatives or lead to the development of novel staging algorithms.

      Key Words

      Abbreviations and Acronyms:

      EBUS (endobronchial ultrasound-guided nodal aspirate), IMS (invasive mediastinal staging), NSCLC (non–small cell lung cancer), PET (positron emission tomography), SCOAP (Surgical Care Outcomes Assessment Program), STS (Society of Thoracic Surgeons)
      Figure thumbnail fx1
      Variation in hospital-level rates of invasive mediastinal staging.
      Rates of invasive mediastinal staging vary significantly across hospitals that provide surgical care to patients with lung cancer.
      Our study demonstrates what many surgeons suspect—lung cancer staging varies significantly across hospitals. This finding provides the critical first step in a line of investigation that seeks to better understand the reasons underlying this highly variable care. Better understanding variability in lung cancer staging may inform quality improvement initiatives or lead to novel staging algorithms.
      See Editorial Commentary page 2672.
      An American College of Surgeons Commission on Cancer study described substantially lower than expected rates of invasive mediastinal staging (IMS) for resected lung cancer (27%) in 2001.
      • Little A.G.
      • Rusch V.W.
      • Bonner J.A.
      • Gaspar L.E.
      • Green M.R.
      • Webb W.R.
      • et al.
      Patterns of surgical care of lung cancer patients.
      In addition, fewer than half of all patients who underwent mediastinoscopy had lymph node tissue submitted for pathologic evaluation. Reed
      • Little A.G.
      • Rusch V.W.
      • Bonner J.A.
      • Gaspar L.E.
      • Green M.R.
      • Webb W.R.
      • et al.
      Patterns of surgical care of lung cancer patients.
      said in her discussion at an annual meeting of the Society of Thoracic Surgeons (STS) that, “The results are an indictment of the present care of patients with non–small cell lung cancer.”
      Since then, numerous studies have reported apparent underuse of IMS in other lung cancer populations using other data sources
      • Little A.G.
      • Gay E.G.
      • Gaspar L.E.
      • Stewart A.K.
      National survey of non–small cell lung cancer in the United States: epidemiology, pathology and patterns of care.
      • Farjah F.
      • Flum D.R.
      • Ramsey S.D.
      • Heagerty P.J.
      • Symons R.G.
      • Wood D.E.
      Multi-modality mediastinal staging for lung cancer among medicare beneficiaries.
      • Lathan C.S.
      • Neville B.A.
      • Earle C.C.
      The effect of race on invasive staging and surgery in non-small-cell lung cancer.
      • Vest M.T.
      • Tanoue L.
      • Soulos P.R.
      • Kim A.W.
      • Detterbeck F.
      • Morgensztern D.
      • et al.
      Thoroughness of mediastinal staging in Stage IIIA non–small cell lung cancer.
      • Ost D.E.
      • Niu J.
      • Elting L.S.
      • Buchholz T.A.
      • Giordano S.H.
      Determinants of practice patterns and quality gaps in lung cancer staging and diagnosis.
      • Ost D.E.
      • Niu J.
      • Elting L.S.
      • Buchholz T.A.
      • Giordano S.H.
      Quality gaps and comparative effectiveness in lung cancer staging and diagnosis.
      • Faris N.
      • Yu X.
      • Sareen S.
      • Signore R.S.
      • McHugh L.M.
      • Roark K.
      • et al.
      Preoperative evaluation of lung cancer in a community health care setting.
      ; however, there are several challenges in drawing firm conclusions about the quality of care based on these studies. First, contemporary guidelines have shifted away from recommending routine IMS.
      National Comprehensive Cancer Network
      Non–small cell lung cancer (Version 5.2017).
      • Silvestri G.A.
      • Gonzalez A.V.
      • Jantz M.A.
      • Margolis M.L.
      • Gould M.K.
      • Tanoue L.T.
      • et al.
      Methods for staging non-small cell lung cancer.
      • Ettinger D.S.
      • Akerley W.
      • Borghaei H.
      • Chang A.C.
      • Cheney R.T.
      • Chirieac L.R.
      • et al.
      Non–small cell lung cancer, version 2.2013.
      Second, there are no known datasets with sufficient granularity to directly assess adherence with guideline-recommended IMS. Variation in care delivery is a commonly used surrogate measure for suboptimal care. Nonetheless, there may be other reasons for variability in care, such as concerns over the level of evidence supporting guideline recommendations or providers who perceive IMS to be time-consuming, inconvenient, or even dangerous. These speculative reasons form the basis of a long-standing suspicion of variability in rates of IMS. Variability in hospital-level rates of IMS has never been documented.
      Leveraging a unique data source of academic and community-based thoracic surgical practices in the Puget Sound region of Washington State, we sought to compare hospital-level rates of IMS among patients with resected non–small cell lung cancer (NSCLC). We hypothesized that there would be significant variation in rates of IMS across hospitals.

      Materials and Methods

       Study Design and Population

      We performed a retrospective cohort study of adult NSCLC patients who underwent resection between July 2011 and December 2013 (n = 514). Patients were excluded if they had a prior history of lung cancer (n = 43) or were treated with induction therapy (n = 59). Data were obtained from the medical record of hospitals/health systems in the Puget Sound region of Washington State participating in the Collaborative to Improve Native Cancer Outcomes study.
      Comparative Effectiveness Research Translation Network
      Lung cancer quality improvement collaborative.
      Of 13 Collaborative to Improve Native Cancer Outcomes study hospitals (Appendix E1), 7 contributed lung cancer cases. Two hospitals contributing fewer than 10 cases over the duration of the study were excluded from the analysis resulting in the exclusion of 6 additional patients. This study was not considered human subjects research by the University of Washington Institutional Review Board because both patient and hospital information were de-identified.

       Data Source

      Data were obtained from the Surgical Care and Outcomes Assessment Program (SCOAP). SCOAP is a physician-led quality improvement system in Washington State. The SCOAP database is a prospective, multicenter clinical registry that contains data on patients’ characteristics, perioperative and surgical details, as well as clinical outcomes. SCOAP is administered by the Foundation for Health Care Quality. Research and development in SCOAP is funded in part by grants to the University of Washington Surgical Outcomes Research Center and the Comparative Effectiveness Research Translation Network.
      • Devine E.B.
      • Capurro D.
      • van Eaton E.
      • Alfonso-Cristancho R.
      • Devlin A.
      • Yanez N.D.
      • et al.
      Preparing electronic clinical data for quality improvement and comparative effectiveness research: The SCOAP CERTAIN automation and validation project.
      • Flum D.R.
      • Alfonso-Cristancho R.
      • Devine E.B.
      • Devlin A.
      • Farrokhi E.
      • Tarczy-Hornoch P.
      • et al.
      Implementation of a “real-world” learning health care system: Washington state's comparative effectiveness research translation network (CERTAIN).
      In 2013, the Comparative Effectiveness Research Translation Network developed a thoracic surgery quality improvement module for use by hospitals that participate in SCOAP. Stakeholders include thoracic surgeons, pulmonologists, advanced practice providers, and hospital administrators.
      Comparative Effectiveness Research Translation Network
      Lung cancer quality improvement collaborative.
      This group determined the need to collect information about lung cancer staging and added disease-specific variables (eg, provider documented clinical stage, details of IMS [listed in the next paragraph], and use of noninvasive staging modalities) to the lung cancer module within SCOAP.

       IMS

      Abstractors recorded information from patients’ medical records on the type (eg, mediastinoscopy or mediastinotomy, endobronchial ultrasound-guided nodal aspirate [EBUS], or esophageal ultrasound-guided nodal aspirate) and timing (before resection, at the time of resection, or both before and at the time of resection) of IMS if it was performed. In addition, abstractors collected data on the number of unique mediastinal nodal stations sampled during IMS. Abstractors did not collect information on the indications for IMS, and therefore it is not possible to determine guideline adherence at the patient level.

       Analysis

      Patient characteristics were summarized by frequencies (categorical variables) and means ± standard deviation or medians (interquartile range) for normally and nonnormally distributed continuous variables, respectively. Univariate differences across hospitals are summarized by χ2 tests (categorical variable) and analysis of variance tests or Kruskal-Wallis tests for normally and nonnormally distributed continuous variables, respectively. All analyses were performed using Stata version 14.2 (StataCorp LLC, College Station, Tex).
      Confidence interval (CI) inspection was used to compare hospital-level rates of IMS to each other and the overall average rate for the study population. Because small sample sizes at any single hospital can influence the observed rate of IMS and erroneously inflate the magnitude of variation, we used an empirical Bayes shrinkage estimator.
      • Dimick J.B.
      • Staiger D.O.
      • Birkmeyer J.D.
      Ranking hospitals on surgical mortality: the importance of reliability adjustment.
      • Dimick J.B.
      • Ghaferi A.
      • Osborne N.H.
      • Ko C.Y.
      • Hall B.L.
      Reliability adjustment for reporting hospital outcomes with surgery.
      • MacKenzie T.A.
      • Grunkemeier G.L.
      • Grunwald G.K.
      • O'Malley A.J.
      • Bohn C.
      • Wu Y.X.
      • et al.
      A primer on using shrinkage to compare in-hospital mortality between centers.
      This approach, using a mixed effects logistic regression model, also allowed for adjustment for clinical stage IA tumors. Guidelines allow for omission of IMS in a subset of these patients,
      • Ettinger D.S.
      • Akerley W.
      • Borghaei H.
      • Chang A.C.
      • Cheney R.T.
      • Chirieac L.R.
      • et al.
      Non–small cell lung cancer, version 2.2013.
      • Detterbeck F.C.
      • Jantz M.A.
      • Wallace M.
      • Vansteenkiste J.
      • Silvestri G.A.
      Invasive mediastinal staging of lung cancer: ACCP evidence-based clinical practice guidelines (2nd edition).
      and the frequency of clinical stage IA disease may vary across hospitals; therefore, adjustment for clinical stage IA is imperative to avoid confounding bias in comparisons of hospital-level rates of IMS. Finally, this approach allowed for estimation of the proportion of variability explained by differences in the rates of clinical stage IA disease across hospitals (by calculating the ratio of random effects between models). CIs for point estimates following reliability adjustment were based on empirical Bayes variance estimates and on approximate normality of the posterior facility-level effects. A post hoc analysis calculated the median odds ratio as another measure of across-hospital variation in IMS.
      • Larsen K.
      • Merlo J.
      Appropriate assessment of neighborhood effects on individual health: Integrating random and fixed effects in multilevel logistic regression.
      Missing data analysis revealed no missing values among variables included in risk and reliability adjustment.
      A prespecified sensitivity analysis was conducted in 2 different subgroups in an attempt to evaluate hospital-level variability in IMS rates among populations more likely to be candidates for IMS. Specifically, 1 subgroup analysis was restricted to patients who underwent preoperative positron emission tomography (PET) on the presumption that the provider had a higher index of suspicion of NSCLC. Another subgroup analysis restricted to patients who underwent an anatomic resection on the presumption that surgeons preferentially perform an anatomic resection for cases that are certain to be NSCLC.

      Results

      Between 2011 and 2013, 406 patients met criteria for this study from across 5 hospitals in the Puget Sound region of Washington State. Clinical and treatment variables are summarized in Table 1. Preoperative patient factors that varied across hospitals included insurance case-mix (P < .001), American Society of Anesthesiologists class (P < .01), and clinical stage IA disease (P < .001). Age, sex, race, Charlson comorbidity index, and lung function did not vary significantly across hospitals (all P values > .05). Operative factors (extent and approach to resection) varied significantly across hospitals (P ≤ .001), as did pathologic stage (P = .008), but histology did not (P = .95). Nine percent of patients were ultimately diagnosed with pathologic N2 disease.
      Table 1Characteristics of patients undergoing resection for lung cancer across hospitals
      All (n = 406)Hospital A (n = 130)Hospital B (n = 134)Hospital C (n = 72)Hospital D (n = 40)Hospital E (n = 30)P value
      Mean age (y)68.2 (10.0)66.7 (10.6)68.8 (9.9)68.2 (10.0)69.7 (8.8)71.5 (8.5).12
      Women228 (57)70 (54)84 (64)38 (53)16 (40)20 (67).05
      Race.79
       White354 (88)109 (85)116 (87)64 (89)37 (95)28 (93)
       Black7 (2)4 (3)2 (2)1 (1)0 (0)0 (0)
       Asian26 (6)9 (7)10 (8)5 (7)0 (0)2 (7)
       Native American4 (1)3 (2)0 (0)0 (0)1 (3)0 (0)
       Pacific Islander3 (1)1 (1)2 (2)0 (0)0 (0)0 (0)
       Other8 (2)2 (2)3 (2)2 (3)1 (3)0 (0)
      Insurance<.001
       Commercial145 (36)39 (30)49 (37)44 (61)9 (23)4 (13)
       Medicare77 (19)9 (7)16 (12)27 (38)15 (38)10 (33)
       Medicaid13 (3)5 (4)5 (4)0 (0)2 (5)1 (3)
       Tricare (military)2 (<1)1 (1)0 (0)0 (0)1 (3)0 (0)
       Other government3 (1)2 (2)0 (0)0 (0)0 (0)1 (3)
       Uninsured2 (<1)1 (1)0 (0)1 (1)0 (0)0 (0)
       Government and commercial126 (31)63 (48)44 (33)0 (0)7 (18)12 (40)
       Mixed government38 (9)10 (8)20 (15)0 (0)6 (15)2 (7)
      Charlson comorbidity index.34
       0227 (56)73 (56)69 (51)47 (65)21 (53)17 (57)
       1141 (35)42 (32)50 (37)21 (29)15 (38)13 (43)
       228 (7)12 (9)11 (8)1 (1)4 (10)0 (0)
       3+10 (2)3 (2)4 (3)3 (4)0 (0)0 (0)
      ASA class.01
       II71 (17)12 (9)25 (19)20 (28)10 (25)4 (13)
       III312 (77)107 (82)99 (74)51 (71)29 (73)26 (87)
       IV23 (6)11 (8)10 (7)1 (1)1 (3)0 (0)
      Smoking status.13
       Former235 (60)78 (63)79 (60)42 (63)20 (53)16 (53)
       Current96 (25)31 (25)39 (30)11 (16)8 (21)7 (23)
       Never60 (16)15 (12)14 (11)14 (21)10 (26)7 (23)
      % Predicted FEV183 (69-96)81 (66-95)80 (66-94)84 (75-100)87 (73-100)84 (68-103).16
      % Predicted DLCO64 (52-77)64 (55-75)63 (49-80)66 (51-78)66 (62-83)60 (50-70).31
      Clinical stage<.001
       IA281 (69)74 (57)110 (82)54 (75)22 (55)21 (70)
       IB54 (13)15 (12)11 (8)15 (21)8 (20)5 (17)
       IIA33 (8)17 (13)6 (5)2 (3)4 (10)4 (13)
       IIB20 (5)10 (8)6 (5)0 (0)4 (10)0 (0)
       IIIA15 (4)11 (8)1 (1)1 (1)2 (5)0 (0)
       IIIB1 (<1)1 (1)0 (0)0 (0)0 (0)0 (0)
       IV2 (<1)2 (2)0 (0)0 (0)0 (0)0 (0)
      Preoperative PET scan371 (92)125 (96)120 (90)64 (89)38 (97)24 (83).049
      Extent of resection.001
       Wedge35 (9)19 (15)9 (7)5 (7)1 (3)1 (3)
       Segmentectomy7 (2)3 (2)2 (1)0 (0)2 (5)0 (0)
       Lobectomy272 (67)71 (55)102 (76)46 (64)28 (70)25 (83)
       Sleeve lobectomy2 (<1)1 (1)0 (0)0 (0)1 (3)0 (0)
       Lobectomy and wedge66 (16)19 (15)18 (13)17 (24)8 (20)4 (13)
       Bilobectomy14 (3)9 (7)1 (1)4 (6)0 (0)0 (0)
       Pneumonectomy10 (2)8 (6)2 (1)0 (0)0 (0)0 (0)
      MIS approach292 (72)84 (65)129 (96)41 (57)33 (83)5 (17)<.001
      Right sided resection247 (61)81 (62)81 (60)39 (54)28 (70)18 (60).75
      Pathologic stage.008
       05 (1)2 (2)1 (1)1 (1)0 (0)1 (3)
       IA201 (50)51 (39)79 (59)40 (56)20 (50)11 (37)
       IB84 (21)24 (18)28 (21)17 (24)6 (15)9 (30)
       IIA38 (9)15 (12)11 (8)5 (7)1 (3)6 (20)
       IIB32 (8)15 (12)8 (6)2 (3)5 (13)2 (7)
       IIIA41 (10)22 (17)5 (4)7 (10)7 (18)0 (0)
       IV5 (1)1 (1)2 (1)0 (0)1 (3)1 (3)
      Pathologic nodal stage.053
       pN0317 (79)93 (73)110 (83)55 (79)32 (80)27 (90)
       pN147 (12)18 (14)17 (13)7 (10)2 (5)3 (10)
       pN236 (9)17 (13)5 (4)8 (11)6 (15)0 (0)
      Histology.95
       Adenocarcinoma270 (67)82 (63)87 (65)49 (68)30 (75)22 (73)
       AIS10 (2)5 (4)3 (2)1 (1)0 (0)1 (3)
       Squamous cell carcinoma86 (21)30 (23)30 (22)12 (17)9 (23)5 (17)
       Large cell carcinoma7 (2)3 (2)2 (1)2 (3)0 (0)0 (0)
       NSCLC NOS22 (5)6 (5)9 (7)6 (8)0 (0)1 (3)
       Other11 (3)4 (3)3 (2)2 (3)1 (3)1 (3)
      Values are presented as mean ± standard deviation, n (%), or median (interquartile range). ASA, American Society of Anesthesiologists; FEV1, forced expiratory volume in 1 second; DLCO, diffusion capacity of carbon monoxide; PET, positron emission tomography; MIS, minimally invasive surgery (eg, video-assisted thoracoscopic surgery or robotic); AIS, adenocarcinoma in situ; NSCLC, nonsmall cell lung cancer; NOS, not otherwise specified.
      Overall, 268 patients underwent IMS (66% [95% CI, 61%-71%]). Mediastinoscopy accounted for a majority (85%) of IMS procedures, and a majority (64%) of patients underwent IMS on the day of resection. The median number of mediastinal lymph node stations sampled was 3 (range, 0-7). Seven percent of patients (95% CI, 4%-10%) underwent IMS but did not have lymph node tissue sampled based on pathologic evaluation. Of 18 patients who did not have lymph node tissue sampled, 11 underwent mediastinoscopy, 6 underwent EBUS, and 1 underwent both mediastinoscopy and EBUS. Table 2 summarizes patterns of IMS across hospitals in this study. The type of IMS modality varied across hospitals (P < .001), as did the timing of IMS (P < 0 .001). Unadjusted rates of IMS varied significantly across hospitals (P < .001). Hospital B performed IMS without any lymph node tissue sampled 2 times more often (14%) than the overall average.
      Table 2Patterns of invasive cancer staging for patients undergoing resection of lung cancer across hospitals
      All (n = 406)Hospital A (n = 128)Hospital B (n = 132)Hospital C (n = 70)Hospital D (n = 40)Hospital E (n = 30)P value
      Invasive mediastinal staging<.001
       None138 (34)11 (8)34 (25)56 (78)10 (25)27 (90)
       Mediastinoscopy/mediastinotomy only228 (56)96 (74)87 (65)13 (18)29 (73)3 (10)
       EBUS only9 (2)2 (2)5 (4)1 (1)1 (3)0 (0)
       EUS only0 (0)0 (0)0 (0)0 (0)0 (0)0 (0)
       EBUS and mediastinoscopy/mediastinotomy30 (7)21 (16)7 (5)2 (3)0 (0)0 (0)
       EBUS and EUS and mediastinoscopy/mediastinotomy1 (<1)0 (0)1 (1)0 (0)0 (0)0 (0)
      Any invasive mediastinal staging268 (66)119 (92)100 (75)16 (22)30 (75)3 (10)<.001
      Timing of invasive staging
      Calculated among patients who underwent invasive mediastinal staging.
      <.001
       Before resection97 (36)3 (3)87 (87)4 (25)1 (3)2 (67)
       At the time of resection147 (55)95 (80)12 (12)10 (63)29 (97)1 (33)
       Both before and at the time of resection24 (9)21 (18)1 (1)2 (13)0 (0)0 (0)
      Number of nodal stations sampled
      Calculated among patients who underwent invasive mediastinal staging.
       Mean ± standard deviation3.0 ± 1.23.2 ± 1.02.7 ± 1.42.9 ± 1.33.2 ± 1.42.7 ± 0.6.05
       Median (interquartile range)3 (0-7)3 (0-7)3 (0-5)3 (0-5)3 (0-5)3 (2-3).34
       No nodes sampled18 (7)1 (1)14 (14)1 (6)2 (6)0 (0).004
      Values are presented as n (%) unless otherwise indicated. EBUS, Endobronchial ultrasound-guided nodal aspirate; EUS, esophageal ultrasound-guided nodal aspirate.
      Calculated among patients who underwent invasive mediastinal staging.
      Adjusted rates of IMS varied widely ranging from 17% (95% CI, 7%-36%) at Hospital E to 94% (95% CI, 89%-96%) at Hospital A (Figure 1). Hospitals A, B, and D had markedly higher rates of IMS compared with Hospitals C and E. Hospital A and B performed IMS significantly more often than the overall average, whereas hospitals C and E performed IMS significantly less often than the overall average. Adjusting for clinical stage IA disease explained only 3% of the variability across hospitals, despite large variation in the frequency of clinical stage IA disease across hospitals. Our prespecified subgroup analyses revealed the same degree of variability in hospital-level rates of IMS among patients who underwent anatomic resection or underwent a PET scan (Figures 2 and 3).
      Figure thumbnail gr1
      Figure 1Rates of invasive mediastinal staging by hospital compared with population average. CI, Confidence interval.
      Figure thumbnail gr2
      Figure 2Rates of invasive mediastinal staging by hospital (among patients undergoing anatomic resection). CI, Confidence interval.
      Figure thumbnail gr3
      Figure 3Rates of invasive mediastinal staging by hospital (among patients with preoperative positron emission tomography [PET] scans). CI, Confidence interval.
      Several post hoc analyses were conducted. As an alternative measure of variability in hospital-level rates of IMS, we estimated the median odds ratio across hospitals to be 4.99 (95% CI, 1.56-7.24), indicating significant across-hospital variance. This value is interpreted as follows: if a randomly selected patient from 1 center were to transfer care to a hospital with a higher probability of performing IMS, then his or her risk of undergoing IMS would increase by 499%. Table 3 summarizes a post hoc analysis exploring potential differences in patient characteristics between those who did and did not undergo IMS. Compared with those who did not receive IMS, patients who underwent IMS had a different payer mix (P < .001) and were more likely to have a higher American Society of Anesthesiologists class (P = .002), smoke (P = .002), have a lower predicted forced expiratory volume in 1 second (P = .03), have undergone a preoperative PET scan (P < .001), and have a higher clinical stage (P = .001). A fully adjusted model did not change the conclusions of our primary analysis, and it revealed that patient-level variables accounted for a total of 5% of the variation in rates of IMS across hospitals. Another patient-level stratified analysis showed that patients with clinical stage IB or higher disease had higher rates of IMS compared with those with clinical stage IA (78% vs 61%; P < .001). Finally, hospitals with an exclusively thoracic surgical-based practice model (A, B, and D) had higher rates of IMS compared with hospitals with a mixed-specialty practice model (82% vs 19%; P < .001).
      Table 3Characteristics of patients receiving and not receiving invasive staging
      All (n = 406)Invasive staging (n = 265)No invasive staging (n = 137)P value
      Mean age (y)68.2 ± 10.068.0 ± 10.468.8 ± 9.3.48
      Women228 (57)150 (56)78 (57).95
      Race.33
       White354 (88)231 (87)123 (90)
       Black7 (2)6 (2)1 (1)
       Asian26 (6)17 (6)9 (7)
       Native American4 (1)4 (2)0 (0)
       Pacific Islander3 (1)3 (1)0 (0)
       Other8 (2)4 (2)4 (3)
      Insurance<.001
       Commercial145 (36)89 (33)46 (41)
       Medicare77 (19)35 (13)42 (30)
       Medicaid13 (3)11 (4)2 (1)
       Tricare (military)2 (<1)2 (1)0 (0)
       Other government3 (1)3 (1)0 (0)
       Uninsured2 (<1)1 (<1)1 (1)
       Government and commercial126 (31)97 (36)29 (21)
       Mixed government38 (9)30 (11)8 (6)
      Charlson comorbidity index.36
       0227 (56)148 (55)79 (57)
       1141 (35)90 (34)51 (37)
       228 (7)22 (8)6 (4)
       3+10 (2)8 (3)2 (1)
      ASA class.002
       II71 (17)34 (13)37 (27)
       III312 (77)217 (81)95 (69)
       IV23 (6)17 (6)6 (4)
      Smoking status.002
       Former235 (60)159 (62)76 (57)
       Current96 (25)70 (27)26 (19)
       Never60 (16)28 (11)32 (24)
      % Predicted FEV183 (69-96)80 (69-94)85 (72-100).03
      % Predicted DLCO64 (52-77)63 (51-76)66 (56-80).15
      Clinical stage.001
       IA281 (69)170 (63)111 (80)
       IB54 (13)36 (13)18 (13)
       IIA33 (8)25 (9)8 (6)
       IIB20 (5)20 (7)0 (0)
       IIIA15 (4)14 (5)1 (1)
       IIIB1 (<1)1 (<1)0 (0)
       IV2 (<1)2 (1)0 (0)
      Preoperative PET scan371 (92)255 (96)116 (85)<.001
      Extent of resection.02
       Wedge35 (9)20 (7)15 (11)
       Segmentectomy7 (2)6 (2)1 (1)
       Lobectomy272 (67)180 (67)92 (67)
       Sleeve lobectomy2 (<1)0 (0)2 (1)
       Lobectomy and wedge66 (16)40 (15)26 (19)
       Bilobectomy14 (3)12 (4)2 (1)
       Pneumonectomy10 (2)10 (4)0 (0)
      MIS approach292 (72)201 (75)91 (66).05
      Right-sided resection247 (61)163 (61)84 (61).77
      Pathologic stage.01
       05 (1)1 (<1)4 (3)
       IA201 (50)121 (45)80 (58)
       IB84 (21)56 (21)28 (20)
       IIA38 (9)26 (10)12 (9)
       IIB32 (8)26 (10)6 (4)
       IIIA41 (10)34 (13)7 (5)
       IV5 (1)4 (1)1 (1)
      Pathologic nodal stage.003
       pN0317 (79)198 (74)119 (89)
       pN147 (12)37 (14)10 (7)
       pN236 (9)31 (12)5 (4)
      Histology.049
       Adenocarcinoma270 (67)174 (65)96 (70)
       AIS10 (2)5 (2)5 (4)
       Squamous cell carcinoma86 (21)67 (25)19 (14)
       Large cell carcinoma7 (2)5 (2)2 (1)
       NSCLC NOS22 (5)10 (4)12 (9)
       Other11 (3)7 (3)4 (3)
      Values are presented as mean ± standard deviation or n (%). ASA, American Society of Anesthesiologists; FEV1, forced expiratory volume in 1 second; DLCO, diffusion capacity of carbon monoxide; PET, positron emission tomography; MIS, minimally invasive surgery (eg, video-assisted thoracoscopic surgery or robotic); AIS, adenocarcinoma in situ; NSCLC, non–small cell lung cancer; NOS, not otherwise specified.

      Discussion

      We hypothesized that there would be significant variation in hospital-level rates of IMS and aimed to compare hospital-specific rates of IMS to each other and to the overall population mean. The key finding of our study is that there is significant variation in rates of IMS across hospitals providing surgical care to patients with NSCLC. Patient-level factors, including clinical stage IA disease, explain very little of the variability. Post hoc analyses confirm our key finding. Hospitals with a purely thoracic surgical practice (as opposed to a mixed-specialty practice) were associated with higher rates of IMS.
      Variability in care is often considered a marker of poor quality care. Surgeons may not adhere to guidelines because they are unaware that they exist. Those with a practice solely dedicated to thoracic surgery may be more likely to know guideline recommended indications for IMS. Supporting this hypothesis is the observation that an exclusively thoracic surgical-based hospital practice model was associated with higher rates of IMS. However, a previous study examining the relationship between surgeon specialty and outcomes reported no differences in the use of mediastinoscopy among general, cardiothoracic, and thoracic surgeons.
      • Farjah F.
      • Flum D.R.
      • Varghese T.K.
      • Symons R.G.
      • Wood D.E.
      Surgeon specialty and long-term survival after pulmonary resection for lung cancer.
      Survey research may better elucidate the frequency of surgeon awareness of guideline recommendations and factors associated with both awareness and adherence.
      Another possible explanation for variation in IMS is that surgeons may purposefully choose to not adhere to guidelines because they do not accept the recommendations to be valid. A previous study showed that providers may deviate from guidelines when either the level of supporting evidence for guidelines is low or when they perceive the supporting evidence to not be generalizable to their patient population or practice.
      • Brouwers M.C.
      • Makarski J.
      • Garcia K.
      • Akram S.
      • Darling G.E.
      • Ellis P.M.
      • et al.
      A mixed methods approach to understand variation in lung cancer practice and the role of guidelines.
      Indeed the level of evidence for guideline recommended indications for IMS is 2.
      National Comprehensive Cancer Network
      Non–small cell lung cancer (Version 5.2017).
      • Silvestri G.A.
      • Gonzalez A.V.
      • Jantz M.A.
      • Margolis M.L.
      • Gould M.K.
      • Tanoue L.T.
      • et al.
      Methods for staging non-small cell lung cancer.
      • Ettinger D.S.
      • Akerley W.
      • Borghaei H.
      • Chang A.C.
      • Cheney R.T.
      • Chirieac L.R.
      • et al.
      Non–small cell lung cancer, version 2.2013.
      There are no randomized controlled trials comparing varying intensities of IMS (ie, routine vs selective vs highly selective IMS). It was recently shown that guideline-recommended IMS has a sensitivity and specificity of 100% and 35%, respectively.
      • Farjah F.
      • Backhus L.M.
      • Varghese T.K.
      • Manning J.P.
      • Cheng A.M.
      • Mulligan M.S.
      • et al.
      External validation of a prediction model for pathologic N2 among patients with a negative mediastinum by positron emission tomography.
      • Farjah F.
      • Madtes D.K.
      • Wood D.E.
      • Flum D.R.
      • Zadworny M.E.
      • Waworuntu R.
      • et al.
      Vascular endothelial growth factor C complements the ability of positron emission tomography to predict nodal disease in lung cancer.
      Clinically, these estimates mean that existing guideline recommendations perfectly select all patients with true nodal disease to undergo IMS, but they also select many patients without true nodal disease to undergo IMS. The high rate of negative IMS may provide the rationale for why some surgeons do not accept guideline recommendations. A high true negative rate of IMS may be perceived as futile, inefficient, or cost-ineffective,
      • Meyers B.F.
      • Haddad F.
      • Siegel B.A.
      • Zoole J.B.
      • Battafarano R.J.
      • Veeramachaneni N.
      • et al.
      Cost-effectiveness of routine mediastinoscopy in computed tomography- and positron emission tomography-screened patients with stage I lung cancer.
      although recent evidence suggests these perceptions may not be correct.
      • Czarnecka-Kujawa K.
      • Rochau U.
      • Siebert U.
      • Atenafu E.
      • Darling G.
      • Waddell T.K.
      • et al.
      Cost-effectiveness of mediastinal lymph node staging in non–small cell lung cancer.
      However, underlying the rationale for guidelines directing a majority of lung cancer patients to IMS is the view that the identification of mediastinal nodal disease (N2 or N3, Stage IIIA or IIIB) is important from a patient and treatment selection perspective. Even if the multimodality options for stage IIIA are equivalent with regard to long-term survival,
      • Martins R.G.
      • D'Amico T.A.
      • Loo Jr., B.W.
      • Pinder-Schenck M.
      • Borghaei H.
      • Chaft J.E.
      • et al.
      The management of patients with stage IIIA non-small cell lung cancer with N2 mediastinal node involvement.
      many patients and providers would likely agree that more accurate staging leads to more informed treatment decisions. Nonetheless, some providers may be skeptical of this view and believe that patients are better served by initial surgical resection.
      Better understanding the reasons underlying hospital-level variability in IMS is an opportunity to improve care. Qualitative research (eg, focus groups and key informant surgeon interviews) can help identify the reasons for variability in practice patterns. To the extent that a knowledge gap explains variability in IMS, hospital credentialing and privileging policies could be an avenue for evaluating and increasing the knowledge base of surgeons who do not have a dedicated oncologic practice and/or board certification in thoracic surgery.
      • Blackmon S.H.
      • Cooke D.T.
      • Whyte R.
      • Miller D.
      • Cerfolio R.
      • Farjah F.
      • et al.
      The Society of Thoracic Surgeons expert consensus statement: a tool kit to assist thoracic surgeons seeking privileging to use new technology and perform advanced procedures in general thoracic surgery.
      Greater access to continuing medical education—such as the GAIN (Engaging an Interdisciplinary Team for NSCLC Diagnosis, Personalized Assessment, and Treatment) curriculum sponsored by the American College of Chest Physicians, National Comprehensive Cancer Network programs, and others—is another opportunity to address knowledge gaps.
      CHEST GAIN
      Regional non–small cell lung cancer summits.
      Finally, clinical registries (eg, STS and National Surgical Quality Improvement Program) can facilitate performance feedback in terms of process compliance (ie, benchmarked rates of IMS).
      American College of Surgeons
      NSQIP data collection, analysis, and reporting.
      • Grover F.L.
      • Shahian D.M.
      • Clark R.E.
      • Edwards F.H.
      The STS National Database.
      • Jacobs J.P.
      • Shahian D.M.
      • Prager R.L.
      • Edwards F.H.
      • McDonald D.
      • Han J.M.
      • et al.
      Introduction to the STS National Database Series.
      On the other hand, if the variability in IMS is due to widespread disagreement about the optimal indications for IMS, then more research will be needed to elucidate the optimal IMS strategy. One example of an alternative staging strategy is selective IMS using prediction models. Preliminary research suggests that prediction models can, similar to existing practice guidelines, maintain 100% sensitivity while achieving higher specificity.
      • Farjah F.
      • Backhus L.M.
      • Varghese T.K.
      • Manning J.P.
      • Cheng A.M.
      • Mulligan M.S.
      • et al.
      External validation of a prediction model for pathologic N2 among patients with a negative mediastinum by positron emission tomography.
      • Farjah F.
      • Lou F.
      • Sima C.
      • Rusch V.W.
      • Rizk N.P.
      A prediction model for pathologic N2 disease in lung cancer patients with a negative mediastinum by positron emission tomography.
      Clinically, this means that the use of a prediction model is expected to lead to fewer IMS procedures without compromising the ability to detect true nodal disease before first treatment. This hypothesis could also be tested through a randomized controlled trial. Still other alternative staging strategies may arise from qualitative interviews with thoracic surgeons and pulmonologists.
      One unexpected finding was that the rate of IMS in this study (66%) was higher than previously reported (27%).
      • Little A.G.
      • Rusch V.W.
      • Bonner J.A.
      • Gaspar L.E.
      • Green M.R.
      • Webb W.R.
      • et al.
      Patterns of surgical care of lung cancer patients.
      Differences across studies may reflect differences in the quality of data collection or delivery of care over time. Another possibility is the Hawthorne effect, given that all hospitals agreed to participate in a regional quality improvement collaborative. However, this explanation is unlikely because the decision to measure use of IMS was made by our regional collaborative in 2013, the study period spans 2011 to 2013, and there were no interventions intended to influence rates of IMS. Furthermore, rates of IMS for the entire study population were lower than expected population rates of IMS derived from prior studies (77%-79%).
      • Farjah F.
      • Backhus L.M.
      • Varghese T.K.
      • Manning J.P.
      • Cheng A.M.
      • Mulligan M.S.
      • et al.
      External validation of a prediction model for pathologic N2 among patients with a negative mediastinum by positron emission tomography.
      • Farjah F.
      • Madtes D.K.
      • Wood D.E.
      • Flum D.R.
      • Zadworny M.E.
      • Waworuntu R.
      • et al.
      Vascular endothelial growth factor C complements the ability of positron emission tomography to predict nodal disease in lung cancer.
      • Mazzone P.J.
      • Vachani A.
      • Chang A.
      • Detterbeck F.
      • Cooke D.
      • Howington J.
      • et al.
      Quality indicators for the evaluation of patients with lung cancer.
      In addition, although rates of failed lymph node sampling (7%) were much lower than that previously reported (∼ 47%),
      • Little A.G.
      • Rusch V.W.
      • Bonner J.A.
      • Gaspar L.E.
      • Green M.R.
      • Webb W.R.
      • et al.
      Patterns of surgical care of lung cancer patients.
      opportunities for quality improvement remain given that 1 hospital failed sample lymph node tissue in up to 14% of patients undergoing IMS—a rate double that of other hospitals. Failing to submit or obtain adequate lymph node tissue is an outcome measure that could be monitored by a clinical registry such as the STS General Thoracic Surgery Database.
      • Grover F.L.
      • Shahian D.M.
      • Clark R.E.
      • Edwards F.H.
      The STS National Database.
      • Jacobs J.P.
      • Shahian D.M.
      • Prager R.L.
      • Edwards F.H.
      • McDonald D.
      • Han J.M.
      • et al.
      Introduction to the STS National Database Series.
      There are important limitations to this study. There may be concerns about generalizing regional results to the rest of the nation. By design, this multicenter study included mostly community hospitals and only a single academic center in an attempt to maximize generalizability. We expect hospital-level variability in IMS to be found across the nation, and several members of our study team are currently testing this hypothesis using the STS General Thoracic Surgery Database (although this database also has generalizability concerns). Because we were unable to directly access medical records to determine the indications for or against IMS, we could not measure adherence with guideline recommendations at the patient level. Further, because we were unable to measure central versus peripheral tumors, some of the variation in IMS may be explained by differences in case-mix (defined by central tumors) across hospitals. Neither did we have information on the provider who performed or decided against performing IMS, and therefore we cannot better understand the contribution of individual providers or provider-specialty to variation in care. We may have underestimated rates of IMS. Data abstractors may not have been able to record IMS occurring at a referring hospital/health system (ie, not participating in our collaborative). However, given the fact that rates of IMS are higher than previously described in community practice, it is unlikely that we significantly underestimated rates of IMS. Finally, we were unable to elucidate the reasons for variable rates of IMS across hospitals. As discussed earlier, we plan to conduct qualitative and mixed-methods research to gain a better understanding of the determinants of variability in IMS.

      Conclusions

      We describe—for the first time—that there is significant variability in rates of IMS across hospitals (Video 1). We speculate that this variability might be due to gaps in the quality of care, disagreement over the best indications for IMS, or both. Planned qualitative and mixed-methods studies will likely reveal the reasons underlying variability and lead to either quality improvement and educational interventions and/or trials comparing novel IMS strategies to guide recommended staging.
      Figure thumbnail fx2
      Video 1Dr Farhood Farjah explains the significance of the research findings. Video available at: http://www.jtcvsonline.org/article/S0022-5223(18)30311-8/fulltext.

       Conflict of Interest Statement

      Authors have nothing to disclose with regard to commercial support.
      The authors thank the Surgical Care and Outcomes Assessment Program for providing data for this research. The authors also thank Rebecca G. Symons; Megan Zadworny; Hao He, PhD; and Katherine Odem-Davis, PhD, at the Surgical Outcomes Research Center for helping with data acquisition and statistical consultation.

      Appendix E1. Participating hospitals

      MultiCare Allenmore Hospital
      MultiCare Good Samaritan Hospital
      Harborview Medical Center
      Peace Health St Joseph Medical Center
      Providence Regional Medical Center Everett
      St Joseph Medical Center
      St Francis Hospital
      St Anthony Hospital
      St Clare Hospital
      St Elizabeth Hospital
      MultiCare Tacoma General Hospital
      University of Washington Medical Center
      Virginia Mason Medical Center

      Supplementary Data

      References

        • Little A.G.
        • Rusch V.W.
        • Bonner J.A.
        • Gaspar L.E.
        • Green M.R.
        • Webb W.R.
        • et al.
        Patterns of surgical care of lung cancer patients.
        Ann Thorac Surg. 2005; 80: 2051-2056
        • Little A.G.
        • Gay E.G.
        • Gaspar L.E.
        • Stewart A.K.
        National survey of non–small cell lung cancer in the United States: epidemiology, pathology and patterns of care.
        Lung Cancer. 2007; 57: 253-260
        • Farjah F.
        • Flum D.R.
        • Ramsey S.D.
        • Heagerty P.J.
        • Symons R.G.
        • Wood D.E.
        Multi-modality mediastinal staging for lung cancer among medicare beneficiaries.
        J Thorac Oncol. 2009; 4: 355-363
        • Lathan C.S.
        • Neville B.A.
        • Earle C.C.
        The effect of race on invasive staging and surgery in non-small-cell lung cancer.
        J Clin Oncol. 2006; 24: 413-418
        • Vest M.T.
        • Tanoue L.
        • Soulos P.R.
        • Kim A.W.
        • Detterbeck F.
        • Morgensztern D.
        • et al.
        Thoroughness of mediastinal staging in Stage IIIA non–small cell lung cancer.
        J Thorac Oncol. 2012; 7: 188-195
        • Ost D.E.
        • Niu J.
        • Elting L.S.
        • Buchholz T.A.
        • Giordano S.H.
        Determinants of practice patterns and quality gaps in lung cancer staging and diagnosis.
        Chest. 2014; 145: 1097-1113
        • Ost D.E.
        • Niu J.
        • Elting L.S.
        • Buchholz T.A.
        • Giordano S.H.
        Quality gaps and comparative effectiveness in lung cancer staging and diagnosis.
        Chest. 2014; 145: 331-345
        • Faris N.
        • Yu X.
        • Sareen S.
        • Signore R.S.
        • McHugh L.M.
        • Roark K.
        • et al.
        Preoperative evaluation of lung cancer in a community health care setting.
        Ann Thorac Surg. 2015; 100: 394-400
        • National Comprehensive Cancer Network
        Non–small cell lung cancer (Version 5.2017).
        (Available at:)
        • Silvestri G.A.
        • Gonzalez A.V.
        • Jantz M.A.
        • Margolis M.L.
        • Gould M.K.
        • Tanoue L.T.
        • et al.
        Methods for staging non-small cell lung cancer.
        Chest. 2013; 143: e211S-e250S
        • Ettinger D.S.
        • Akerley W.
        • Borghaei H.
        • Chang A.C.
        • Cheney R.T.
        • Chirieac L.R.
        • et al.
        Non–small cell lung cancer, version 2.2013.
        J Natl Compr Cancer Netw. 2013; 11: 645-653
        • Comparative Effectiveness Research Translation Network
        Lung cancer quality improvement collaborative.
        (Available at:)
        • Devine E.B.
        • Capurro D.
        • van Eaton E.
        • Alfonso-Cristancho R.
        • Devlin A.
        • Yanez N.D.
        • et al.
        Preparing electronic clinical data for quality improvement and comparative effectiveness research: The SCOAP CERTAIN automation and validation project.
        EGEMS (Wash, DC). 2013; 1: 1025
        • Flum D.R.
        • Alfonso-Cristancho R.
        • Devine E.B.
        • Devlin A.
        • Farrokhi E.
        • Tarczy-Hornoch P.
        • et al.
        Implementation of a “real-world” learning health care system: Washington state's comparative effectiveness research translation network (CERTAIN).
        Surgery. 2014; 155: 860-866
        • Dimick J.B.
        • Staiger D.O.
        • Birkmeyer J.D.
        Ranking hospitals on surgical mortality: the importance of reliability adjustment.
        Health Serv Res. 2010; 45: 1614-1629
        • Dimick J.B.
        • Ghaferi A.
        • Osborne N.H.
        • Ko C.Y.
        • Hall B.L.
        Reliability adjustment for reporting hospital outcomes with surgery.
        Ann Surg. 2012; 255: 703-707
        • MacKenzie T.A.
        • Grunkemeier G.L.
        • Grunwald G.K.
        • O'Malley A.J.
        • Bohn C.
        • Wu Y.X.
        • et al.
        A primer on using shrinkage to compare in-hospital mortality between centers.
        Ann Thorac Surg. 2015; 99: 757-761
        • Detterbeck F.C.
        • Jantz M.A.
        • Wallace M.
        • Vansteenkiste J.
        • Silvestri G.A.
        Invasive mediastinal staging of lung cancer: ACCP evidence-based clinical practice guidelines (2nd edition).
        Chest. 2007; 132: 202S-220S
        • Larsen K.
        • Merlo J.
        Appropriate assessment of neighborhood effects on individual health: Integrating random and fixed effects in multilevel logistic regression.
        Am J Epidemiol. 2005; 161: 81-88
        • Farjah F.
        • Flum D.R.
        • Varghese T.K.
        • Symons R.G.
        • Wood D.E.
        Surgeon specialty and long-term survival after pulmonary resection for lung cancer.
        Ann Thorac Surg. 2009; 87: 995-1006
        • Brouwers M.C.
        • Makarski J.
        • Garcia K.
        • Akram S.
        • Darling G.E.
        • Ellis P.M.
        • et al.
        A mixed methods approach to understand variation in lung cancer practice and the role of guidelines.
        Implement Sci. 2014; 9: 36
        • Farjah F.
        • Backhus L.M.
        • Varghese T.K.
        • Manning J.P.
        • Cheng A.M.
        • Mulligan M.S.
        • et al.
        External validation of a prediction model for pathologic N2 among patients with a negative mediastinum by positron emission tomography.
        J Thorac Dis. 2015; 7: 576-584
        • Farjah F.
        • Madtes D.K.
        • Wood D.E.
        • Flum D.R.
        • Zadworny M.E.
        • Waworuntu R.
        • et al.
        Vascular endothelial growth factor C complements the ability of positron emission tomography to predict nodal disease in lung cancer.
        J Thorac Cardiovasc Surg. 2015; 150: 796-803
        • Meyers B.F.
        • Haddad F.
        • Siegel B.A.
        • Zoole J.B.
        • Battafarano R.J.
        • Veeramachaneni N.
        • et al.
        Cost-effectiveness of routine mediastinoscopy in computed tomography- and positron emission tomography-screened patients with stage I lung cancer.
        J Thorac Cardiovasc Surg. 2006; 131: 822-829
        • Czarnecka-Kujawa K.
        • Rochau U.
        • Siebert U.
        • Atenafu E.
        • Darling G.
        • Waddell T.K.
        • et al.
        Cost-effectiveness of mediastinal lymph node staging in non–small cell lung cancer.
        J Thorac Cardiovasc Surg. 2017; 153: 1567-1578
        • Martins R.G.
        • D'Amico T.A.
        • Loo Jr., B.W.
        • Pinder-Schenck M.
        • Borghaei H.
        • Chaft J.E.
        • et al.
        The management of patients with stage IIIA non-small cell lung cancer with N2 mediastinal node involvement.
        J Natl Compr Canc Netw. 2012; 10: 599-613
        • Blackmon S.H.
        • Cooke D.T.
        • Whyte R.
        • Miller D.
        • Cerfolio R.
        • Farjah F.
        • et al.
        The Society of Thoracic Surgeons expert consensus statement: a tool kit to assist thoracic surgeons seeking privileging to use new technology and perform advanced procedures in general thoracic surgery.
        Ann Thorac Surg. 2016; 101: 1230-1237
        • CHEST GAIN
        Regional non–small cell lung cancer summits.
        (Available at:)
        • American College of Surgeons
        NSQIP data collection, analysis, and reporting.
        (Available at:)
        https://www.facs.org/quality-programs/acs-nsqip?
        Date accessed: September 29, 2016
        • Grover F.L.
        • Shahian D.M.
        • Clark R.E.
        • Edwards F.H.
        The STS National Database.
        Ann Thorac Surg. 2014; 97: S48-S54
        • Jacobs J.P.
        • Shahian D.M.
        • Prager R.L.
        • Edwards F.H.
        • McDonald D.
        • Han J.M.
        • et al.
        Introduction to the STS National Database Series.
        Ann Thorac Surg. 2015; 100: 1992-2000
        • Farjah F.
        • Lou F.
        • Sima C.
        • Rusch V.W.
        • Rizk N.P.
        A prediction model for pathologic N2 disease in lung cancer patients with a negative mediastinum by positron emission tomography.
        J Thorac Oncol. 2013; 8: 1170-1180
        • Mazzone P.J.
        • Vachani A.
        • Chang A.
        • Detterbeck F.
        • Cooke D.
        • Howington J.
        • et al.
        Quality indicators for the evaluation of patients with lung cancer.
        Chest. 2014; 146: 659-669