GENERAL THORACIC 996 FARJAH ET AL Ann Thorac Surg SURGEON SPECIALTY AND LUNG CANCER OUTCOMES 2009;87: Among 221,208 patients diagnosed with lu

Transcription

1 J. MAXWELL CHAMBERLAIN MEMORIAL PAPER FOR GENERAL THORACIC SURGERY GENERAL THORACIC SURGERY: The Annals of Thoracic Surgery CME Program is located online at To take the CME activity related to this article, you must have either an STS member or an individual non-member subscription to the journal. GENERAL THORACIC Surgeon Specialty and Long-Term Survival After Pulmonary Resection for Lung Cancer Farhood Farjah, MD, MPH, David R. Flum, MD, MPH, Thomas K. Varghese, Jr, MD, Rebecca Gaston Symons, MPH, and Douglas E. Wood, MD Department of Surgery, University of Washington, Seattle, Washington Background. Long-term outcomes and processes of care in patients undergoing pulmonary resection for lung cancer may vary by surgeon type. Associations between surgeon specialty and processes of care and long-term survival have not been described. Methods. A cohort study (1992 through 2002, follow-up through 2005) was conducted using Surveillance, Epidemiology, and End-Results-Medicare data. The American Board of Thoracic Surgery Diplomates list was used to differentiate board-certified thoracic surgeons from general surgeons (GS). Board-certified thoracic surgeons were designated as cardiothoracic surgeons (CTS) if they performed cardiac procedures and as general thoracic surgeons (GTS) if they did not. Results. Among 19,745 patients, 32% were cared for by GTS, 45% by CTS, and 24% by GS. Patient age, comorbidity index, and resection type did not vary by surgeon specialty (all p > 0.10). Compared with GS and CTS, GTS more frequently used positron emission tomography (36% versus 26% versus 26%, respectively; p 0.005) and lymphadenectomy (33% versus 22% versus 11%, respectively; p < 0.001). After adjustment for patient, disease, and management characteristics, hospital teaching status, and surgeon and hospital volume, patients treated by GTS had an 11% lower hazard of death compared with those who underwent resection by GS (hazard ratio, 0.89; 99% confidence interval, 0.82 to 0.97). The risks of death did not vary significantly between CTS and GS (hazard ratio, 0.94; 99% confidence interval, 0.88 to 1.01) or GTS and CTS (hazard ratio, 0.94; 99% confidence interval, 0.87 to 1.03). Conclusions. Lung cancer patients treated by GTS had higher long-term survival rates than those treated by GS. General thoracic surgeons performed preoperative and intraoperative staging more often than GS or CTS. (Ann Thorac Surg 2009;87: ) 2009 by The Society of Thoracic Surgeons Lung cancer is responsible for more deaths than prostate, breast, and colorectal cancer combined [1]. Pulmonary resection provides optimal therapy with curative intent for those with early stage disease and select patients with locally advanced lung cancer. Lung resections are performed by surgeons with and without board certification in thoracic surgery. Two previous studies have shown that board certification is associated with lower operative mortality rates a marker of surgical safety [2, 3]. Five-year survival rates a marker of treatment effectiveness have not been described among patients under the care of different types of surgeons. It has been suggested, although not demonstrated empirically, that long-term outcomes might vary by surgeon type because of differential use of processes of care, such Accepted for publication Dec 1, Presented at the Forty-fifth Annual Meeting of The Society of Thoracic Surgeons, San Francisco, CA, Jan 26 28, Winner of the J. Maxwell Chamberlain Memorial Award for General Thoracic Surgery. Address correspondence to Dr Wood, University of Washington, Department of Surgery/Division of Cardiothoracic Surgery, Box , 1959 NE Pacific St, Seattle, WA 98195; dewood@u.washington.edu. as preoperative staging, extent of resection, and use of multimodality treatment [4]. Using the Surveillance, Epidemiology, and End- Results (SEER)-Medicare database, we examined variation in the use of processes of care and long-term survival by surgeon specialty among lung cancer patients. We hypothesized that patients under the care of boardcertified general thoracic surgeons (GTS) would have higher survival rates than those under the care of general surgeons (GS), and that differential use of processes of care might explain any observed variation in outcome. Material and Methods A cohort study was conducted of lung cancer patients diagnosed between 1992 and 2002 using the SEER- Medicare database. An overview of the SEER-Medicare database is available elsewhere [5]. To determine surgeon specialty, the SEER-Medicare database was linked to the American Board of Thoracic Surgery Diplomates list as described previously [2]. The University of Washington Institutional Review Board approved this investigation by The Society of Thoracic Surgeons /09/$36.00 Published by Elsevier Inc doi: /j.athoracsur

2 GENERAL THORACIC 996 FARJAH ET AL Ann Thorac Surg SURGEON SPECIALTY AND LUNG CANCER OUTCOMES 2009;87: Among 221,208 patients diagnosed with lung cancer, exclusions were made as follows: diagnosis at death or autopsy (n 5,109), unoperated patients (n 188,469), age younger than 66 years (n 2,376), diagnosis of another primary malignancy 3 months before and 6 months after lung cancer diagnosis (n 1,445), part A or B coverage only or concurrent health maintenance organization enrollment (n 2,208), patients who could not be linked to hospital or surgeon files (n 1,816), and those with histologic classification inconsistent with lung cancer (n 40). These exclusions were necessary to ensure the completeness and accuracy of using claims data for research purposes as discussed previously [6]. Three groups of patients were defined on the basis of whether their operation was performed by a GS, cardiothoracic surgeon (CTS), or GTS. Surgeons who could not be matched to the American Board of Thoracic Surgery list were considered GS. Board-certified thoracic surgeons who performed at least one coronary artery bypass graft or valve procedure among any patient in this cohort between 1992 and 2002 were classified as CTS. All other board-certified thoracic surgeons were considered to be GTS. The primary outcome was overall survival. Death attributable to any cause was ascertained using the Medicare Enrollment Database and follow-up was available through Secondary end points included lung cancer cause-specific survival, operative mortality, tracheostomy or reoperation within 90 days of resection, prolonged length of stay greater than 14 days (PLOS), discharge to an institutional care facility, and readmission within 30 days of index hospitalization. Causespecific death information was only available through the SEER registry with limited follow-up through Tracheostomy and reoperation were defined by their Healthcare Common Procedure Coding System billing codes (Appendix). Health-care utilization was ascertained using the Medicare Provider Analysis and Review file. Processes of care included the use of staging modalities, extent of resection, cardiac procedures, and multimodality therapy and were ascertained using billing codes within the Carrier and Outpatient claims files (Appendix). Patient-level covariates were mostly available through the SEER registry [7]. The Klabunde-modified Charlson comorbidity index was calculated using claims from the Carrier and Outpatient files in the year before diagnosis [8]. Hospital-level covariates were ascertained using the Medicare Healthcare Cost Report and Provider of Service survey. Teaching status was based on affiliation with a medical school or a resident-to-bed ratio of greater than 0.25 [9]. Hospital and surgeon annual average volumes were calculated among the SEER-Medicare cohort [10]. All statistical analyses were performed using STATA (Special Edition 9.2; Statacorp, College Station, TX). Differences in categorical variables were tested using the 2 test, and differences in means and medians were tested using analysis of variance and nonparametric equality-ofmedians test, respectively. Survival estimates were obtained using Kaplan Meier methods. Survival time was defined as the interval between date of resection and date of death or censoring. Because of the competing risk of Fig 1. (A) Changes as a function of time in the proportion of patients cared for by different surgeons. (B) Changes as a function of time in the proportion of surgeons by surgeon specialty. death attributable to other causes the cumulative incidence of death from lung cancer was used to estimate causespecific survival [11]. For similar reasons, the cumulative incidences of tracheostomy, reoperation, and readmission were estimated while censoring for deaths occurring within the period of outcome ascertainment (ie, 90 days for tracheostomy). Cox proportional hazards models were used to evaluate the unadjusted and adjusted relationship between surgeon specialty and overall and cause-specific survival, tracheostomy, reoperation, and readmission. Schoenfeld residuals were used to examine the proportional hazards assumption and extended (stratified) Cox models were fitted if this assumption was not met. Logistic regression was used to evaluate the relationship between surgeon specialty and processes of care and unadjusted and adjusted early deaths, PLOS, and discharges to ICFs. All regression models adjusted for clustering at the hospital level. Because 20% of patients had at least one missing covariate, all multivariate analyses were repeated using multivariate imputation by chained equations [12]. Two-sided probability values less than 1% were considered significant. Results Among 19,745 patients, 24% underwent operations by GS, 45% by CTS, and 32% by GTS. The proportion of

3 Ann Thorac Surg FARJAH ET AL 2009;87: SURGEON SPECIALTY AND LUNG CANCER OUTCOMES Table 1. Patient and Disease Characteristics by Surgeon Specialty Patient and Disease Characteristic All (n 19,745) GS (n 4,677) CTS (n 8,807) GTS (n 6,261) p Value Age (%) y y y y Mean SD (y) Median [range] (y) 75 [66 100] 75 [66 96] 75 [66 100] 75 [66 100] Male sex (%) Race (%) White Black Other Missing Low income (%) Missing Low education (%) Missing Unmarried (%) Missing Geography (%) West East Midwest South Area of residence (%) Metropolitan Urban Rural Prior malignancy (%) Comorbidity index (%) Histology (%) NSCLC SCLC NOS Other Stage (%) I II IIIA IIIB IV Missing GENERAL THORACIC CTS cardiothoracic surgeons; GS general surgeons; GTS general thoracic surgeons; NOS not otherwise specified; NSCLC non small cell lung cancer; SCLC small cell lung cancer; SD standard deviation. patients cared for by GS decreased with time from 28% to 21% whereas the percentage of patients cared for by CTS increased from 40% to 46% (Fig 1A). General thoracic surgeons cared for a relatively constant proportion of patients as a function of time. Of the 1,848 surgeons in the study, 42% were GS, 37% were CTS, and 21% were GTS. The distribution of surgeons did not change remarkably during the 11-year study period (Fig 1B).

4 GENERAL THORACIC 998 FARJAH ET AL Ann Thorac Surg SURGEON SPECIALTY AND LUNG CANCER OUTCOMES 2009;87: Table 2. Management Characteristics by Surgeon Specialty Management Characteristic All (n 19,745) GS (n 4,677) CTS (n 8,807) GTS (n 6,261) p Value Mediastinoscopy (%) PET (%) a Lymphadenectomy (%) b Neoadjuvant therapy (%) c None Radiation Chemotherapy Chemoradiation Type of resection (%) Open wedge VATS wedge b Segmentectomy VATS anatomic resection b Lobectomy Sleeve lobectomy Bilobectomy Pneumonectomy Sleeve pneumonectomy Completion pneumonectomy Extra-pleural pneumonectomy Not otherwise specified Adjuvant therapy (%) d None Radiation Chemotherapy Chemoradiation a Calculated among the cohort. b Calculated among the cohort. c Calculated among those with stage III or higher cancer. d Among patients who survived their index hospitalization. CTS cardiothoracic surgeons; GS general surgeons; GTS general thoracic surgeons; PET positron emission tomography; VATS video-assisted thoracoscopic surgery. Patient characteristics were largely similar by surgeon specialty (Table 1). Statistically significant differences in race, education, marital status, history of prior malignancy, and histology were observed. Compared with GTS, GS and CTS cared for a higher proportion of low-income and rural patients. More than half (53%) of the GS patients lived in the West in contrast to only 32% of GTS patients. Several aspects of management varied by surgeon specialty (Table 2). The use of positron emission tomography and lymphadenectomy were significantly higher among GTS compared with GS and CTS. General thoracic surgeons performed video-assisted thoracoscopic surgery segmentectomy/lobectomy more often than GS and CTS, and performed bilobectomy less often than GS and CTS. Among patients with stage III/IV cancer, the use of neoadjuvant therapy was higher among GTS. Provider characteristics varied by surgeon specialty (Table 3). General thoracic surgeons were higher-volume surgeons compared with CTS and GS. Cardiothoracic surgeons and GTS more often cared for patients at higher-volume centers compared with GS. General thoracic surgeons more frequently cared for patients at teaching hospitals compared with CTS and GS. Unadjusted outcomes varied by surgeon specialty (Table 4). Patients cared for by GTS had higher overall and cause-specific survival rates compared with those cared for by CTS and GS. Differences in survival were larger among patients with early-stage lung cancer (Fig 2). Thirty-day mortality was lower among GTS compared with CTS and GS. Tracheostomy, reoperation, PLOS, discharge to institutional care facility, and readmission rates did not vary by surgeon specialty. Multivariate analyses revealed significant variation in survival by surgeon specialty (Table 5). After adjustment for patient and disease characteristics, GTS patients had a 17% and 9% lower hazard of death compared with GS and CTS patients, respectively, and CTS patients had a 9% lower hazard of death than GS patients. Further adjustment for management and hos-

5 Ann Thorac Surg FARJAH ET AL 2009;87: SURGEON SPECIALTY AND LUNG CANCER OUTCOMES Table 3. Provider Characteristics by Surgeon Specialty Provider Characteristic All (n 19,745) GS (n 4,677) CTS (n 8,807) GTS (n 6,261) p Value Surgeon characteristics Terciles of volume (%) Low Medium High Hospital characteristics Terciles of volume (%) Low Medium High Teaching hospital (%) Cancer cooperative group (%) a NCI designated center (%) a GENERAL THORACIC a Among the 2002 cohort only. CTS cardiothoracic surgeons; GS general surgeons; GTS general thoracic surgeons; NCI National Cancer Institute. pital characteristics and surgeon volume revealed an 11% lower hazard of death for GTS versus GS patients, but differences in survival between GTS and CTS or CTS and GS patients were no longer significant. General thoracic surgeons patients had a 14% significantly lower adjusted hazard of cause-specific death compared with GS patients. Adjusted operative mortality, tracheostomy, reoperation, PLOS, discharge to institutional care facility, and readmission rates did not vary significantly by surgeon specialty. Several post hoc exploratory analyses were conducted. All multivariate analyses were repeated using multiple imputation to examine the influence of missing covariate data on the overall findings. The results were similar to that of the primary analysis. Because our method of defining surgeon specialty might have Table 4. Unadjusted Outcomes by Surgeon Specialty Outcome All (n 19,745) GS (n 4,677) CTS (n 8,807) GTS (n 6,261) p Value 5-Year survival (%) Overall All stages Stage I/II Stage III/IV Lung-cancer cause-specific All stages Stage I/II Stage III/IV Early mortality (%) In-hospital day Operative Tracheostomy (%) Reoperation (%) Length of stay Prolonged 14 days (%) Mean SD (days) Median [range] (days) 7 [1 233] 7 [1 233] 7 [1 153] 7 [1 136] Discharge to ICF (%) a Readmission (%) a a Among patients who survived their hospitalization. CTS cardiothoracic surgeons; GS general surgeons; GTS general thoracic surgeons; ICF institutional care facility; SD standard deviation.

6 GENERAL THORACIC 1000 FARJAH ET AL Ann Thorac Surg SURGEON SPECIALTY AND LUNG CANCER OUTCOMES 2009;87: Fig 2. (A) Unadjusted survival rates by surgeon specialty among patients with stage I or II lung cancer. (B) Unadjusted survival rates by surgeon specialty among patients with stage III or IV lung cancer. (General Thoracic Surgeon [solid line]; Cardiothoracic Surgeons [dashed line]; General Surgeons [dotted line].) been susceptible to misclassification bias, we conducted an analysis based on membership to the General Thoracic Surgery Club (GTSC). The GTSC patients had a 17% lower risk of death compared with patients cared for by surgeons without board certification in thoracic surgery (hazard ratio, 0.83; 99% confidence interval, 0.78 to 0.94), and a 14% lower risk of death compared with patients cared for by surgeons with board certification in thoracic surgery but not enrolled in the GTSC (hazard ratio, 0.86; 99% confidence interval, 0.78 to 0.94). Because others have described a relationship between surgeon specialty and operative mortality [2, 3], we explored whether time trends might explain the null findings in the current study. Between 1992 and 2002, operative mortality rates for GTS decreased from 7.9% to 4.4% (adjusted p trend 0.029) whereas those for GS were stable (3.1% to 6.6%; adjusted p trend 0.41). Declining rates of tracheostomy (3.0% to 1.1%; adjusted p trend 0.001) and PLOS (14.4% to 8.1%; adjusted p trend 0.001) as a function of time were also observed among GTS but not GS (tracheostomy 1.0% to 3.3%; adjusted p trend 0.11; and PLOS 13.0% to 14.2%; adjusted p trend 0.86). Comment This study evaluated long-term outcomes and care among a nationally representative cohort of lung cancer patients treated by different types of surgeons. After adjustment for several well-known prognostic factors for survival, patients under the care of GTS had an 11% lower risk of death compared with those treated by GS. General thoracic surgeons used preoperative and intraoperative staging procedures more often than GS or CTS, more often applied video-assisted thoracoscopic techniques, and less often performed bilobectomy. Differences in patient selection and therapeutic decision-making as a result of more extensive cancer staging might explain some differences in patient outcomes by surgeon specialty. Accurate staging is believed to result in more appropriate use of cancer therapy and therefore better outcomes. The use of positron emission tomography is associated with greater diagnostic accuracy [13, 14], and the effect of intraoperative mediastinal node dissection on survival is currently being evaluated by an American College of Surgeons Oncology Group trial [15]. In the current study, GTS more frequently used positron emission tomography and performed lymphadenectomy. Variation in outcomes by surgeon may also be related to aspects of patient selection and clinical decisionmaking not measured in our study. For instance, forced expiratory volume, performance status, and extent of resection are associated with survival [16 18] and may vary across patients cared for by different surgeons. Predicted postoperative forced expiratory volume may be used to determine eligibility for resection [19], and some believe that select patients with limited pulmonary function may benefit from sublobar resections [20]. The role of performance status in determining eligibility for resection is unclear, but providers tend to underestimate performance status (relative to patient self-report) when measured by the Eastern Cooperative Oncology Group scale [18]. It is unknown whether different types of surgeons assess physiologic eligibility for resection differently, and we were unable to evaluate this hypothesis further because the SEER-Medicare database does not record these variables. Although the extent of resection did not vary by surgeon specialty, the appropriateness of resection could not be determined without knowledge of other clinical factors. Other explanations for differences in patient outcomes by surgeon specialty include stage migration, lead-time bias, and confounding. Stage migration also known as the Will-Rogers phenomenon describes a situation in which more accurate stage classification leads to better observed survival rates [21]. Because GTS used staging modalities more often, it is possible that better survival differences as compared with other surgeons patients are explained in part by stage migration. Lead-time bias occurs when the onset of observation is different for different groups of patients. The average interval between presentation and therapy is approximately 90 days [22]. If GTS work in systems of care that more expeditiously work up, refer, and treat patients, then it is

7 Ann Thorac Surg FARJAH ET AL 2009;87: SURGEON SPECIALTY AND LUNG CANCER OUTCOMES Table 5. Adjusted Outcomes by Surgeon Specialty Adjusted Outcome CTS vs GS GTS vs GS GTS vs CTS HR [99% CI] p Value HR [99% CI] p Value HR [99% CI] p Value Overall survival Unadjusted 0.94 [ ] [ ] [ ] Add patient characteristics 0.93 [ ] [ ] [ ] Add disease characteristics 0.91 [ ] [ ] [ ] Add management characteristics 0.92 [ ] [ ] [ ] Add hospital characteristics 0.94 [ ] [ ] [ ] Add surgeon volume a 0.94 [ ] [ ] [ ] Cause-specific survival Unadjusted 0.91 [ ] [ ] [ ] Adjusted a 0.90 [ ] [ ] [ ] Operative mortality Unadjusted 0.88 [ ] [ ] [ ] Adjusted a 0.92 [ ] [ ] [ ] Tracheostomy Unadjusted 1.24 [ ] [ ] [ ] Adjusted a 1.32 [ ] [ ] [ ] Reoperation Unadjusted 1.06 [ ] [ ] [ ] Adjusted a 1.06 [ ] [ ] [ ] Prolonged length of stay ( 14 days) Unadjusted 1.01 [ ] [ ] [ ] Adjusted a 1.08 [ ] [ ] [ ] Discharge to institutional care facility b Unadjusted 0.97 [ ] [ ] [ ] Adjusted a 1.12 [ ] [ ] [ ] Readmission b Unadjusted 1.09 [ ] [ ] [ ] Adjusted a 1.08 [ ] [ ] [ ] GENERAL THORACIC a Patient characteristics included age, sex, race, income, education, marital status, geography, area of residence, history of prior malignancy, and comorbidity index. Disease characteristics included cancer histology and stage. Management characteristics included mediastinoscopy, neoadjuvant therapy, and extent of resection. Hospital characteristics included hospital volume and teaching status. Adjusted refers to models adjusting for patient, disease, management, and hospital characteristics and surgeon volume. b Among patients who survived their hospitalization. CI confidence interval; CTS cardiothoracic surgeons; GS general surgeons; GTS general thoracic surgeons; HR hazard ratio. possible that our results suffer from lead-time bias because we measured survival time based on the date of resection. Nevertheless, if this is the reason for survival differences, then our results would still be consistent with the notion that GTS (and the systems in which they work) provide more expeditious and higher quality care than other surgeons. Finally, as discussed earlier, an inability to measure variables such as forced expiratory volume, performance status, and the severity of comorbidities might have resulted in confounding bias. If so, our results might be explained by sicker patients being preferentially referred to GS. Yet, it is more likely that specialists care for sicker patients. If true, then the association between surgeon specialty and long-term outcomes might, in fact, be underestimated. Several possibilities may explain variability in practice patterns across surgeons. Like other fields of surgery, there is little level 1 evidence to support the use of specific processes of care in the operative management of lung cancer patients. As a result, practice patterns more likely depend on knowledge, experience, attitudes, and beliefs. Surgeons may differ in this regard based on presence or absence of specialty training in cardiothoracic surgery, a broad scope of practice versus a narrower scope of subspecialty practice, participation in continuing education activities specific to thoracic surgery, and networking within groups of similarly trained and practicing surgeons. Alternatively, practice patterns may be influenced by factors unrelated to surgeons such as barriers to accessing care. For instance, patients under the care of GS and CTS more commonly lived in rural areas. It is possible that technologies such as positron emission tomography are less likely to be available to these patients and surgeons. However, limited access to care is unlikely to fully explain differences in care patterns because procedures such as mediastinoscopy and lymphadenectomy are dependent on the surgeon rather than the system in which they work.

8 GENERAL THORACIC 1002 FARJAH ET AL Ann Thorac Surg SURGEON SPECIALTY AND LUNG CANCER OUTCOMES 2009;87: Although we found no evidence of variation in adverse events by surgeon specialty, we did observe decreasing trends in operative mortality, tracheostomy, and PLOS as a function of time, but only among patients cared for by GTS. Pulmonary complications are the leading cause of early death after lung resection [23, 24], and they were recently shown to be the most common reason for PLOS [25]. Declining trends in these end points with time suggest that GTS might have increasingly adopted processes of care that mitigate the risk of pulmonary complications after lung resection. Examples of processes of care include the use of diffusion-limited capacity of carbon monoxide for risk stratification and patient selection [26], a video-assisted thoracoscopic surgery approach to pulmonary resection [27], thoracic epidural catheters [28], and preoperative pulmonary rehabilitation [29]. This investigation had limitations. There is no standard definition for a GTS. Our use of a major cardiac surgery procedure to define CTS with a mixed cardiac and general thoracic practice was purposefully conservative in an effort to identify dedicated GTS. Groups such as the GTSC recognize GTS as those with a noncardiac thoracic surgery caseload of at least 50% [30]. When using GTSC membership as the criteria for defining GTS, survival differences between surgeons were even more dramatic, suggesting that our a priori definition for surgeon specialty was at least conservative. However, using a professional society membership may not be generalizable if there are subsets of GTS who do not seek membership in an organization such as the GTSC. Moreover, although American Board of Thoracic Surgery certification is the best available surrogate for specialty training in cardiothoracic surgery in the United States, there are likely to be foreign-trained CTS performing lung cancer resections in the United States who do not have American Board of Thoracic Surgery certification and would have been categorized as GS in our analysis. One would expect this misclassification to elevate the outcomes of nonspecialty surgeons in this study and is another reason why the difference between GS and GTS may actually be underestimated in our analysis. We made no attempt to differentiate outcomes between academic and nonacademic GTS. Although a higher proportion of teaching institution affiliation was observed in the GTS group, the final analysis adjusted for hospital teaching status to allow for comparisons of outcomes that were independent of academic status. Our findings may not be generalizable because of our exclusion criteria, particularly limiting the study to Medicare beneficiaries. Given that the median age of operated on lung cancer patients is 67 years [31] and that Medicare provides coverage for 97% of elderly patients, our findings are at least relevant to a majority of patients. The SEER-Medicare provider volume insufficiently captures true provider volume, although the resulting misclassification tends to bias the volume-outcome relationship toward the null hypothesis [10]. Covariate data may not have been missing completely at random, in which case both our case-complete and imputed data analyses may have been biased. Analyses of cause-specific death data may not be valid because cause of death information was based on death-certificate data [7]. Finally, these data span an 11-year period and are at least 6 years old. The SEER-Medicare database was the most contemporaneous database available to test our hypothesis, and because the primary end point was long-term survival, the data were necessarily outdated. Because the fields of thoracic oncology and surgery have changed rapidly during this time, and because specialists may adopt changes earlier than generalists, our data may not provide the most contemporaneous comparison of outcomes across surgeons. Limitations notwithstanding, our results are relevant to debates about the best method of improving the quality of thoracic surgical care in the United States. One option might be to encourage referral of potentially resectable lung cancer patients to GTS. Any such policy will have to address potential barriers to accessing care, particularly among low-income patients and those who live in rural areas. Additionally, any such policy would have to ensure an adequate workforce of GTS. Although our results show that patients are increasingly being referred to CTS rather than GS over time, GTS cared for a similar proportion of patients as a function of time, and the distribution of surgeons has not changed. Attempting to improve outcomes by preferential referral to GTS without addressing infrastructure and workforce concerns might lead to the unintended consequence of depriving a large majority of patients care for an otherwise uniformly fatal disease. An alternative policy might link reimbursement to performance as a means of encouraging best practices among all types of surgeons practicing in a variety of settings. Performance would have to be linked to process measures known to affect long-term outcomes because measuring long-term survival is not feasible outside a research setting. Another measure of performance with a powerful impact on outcomes is participation in a large national or regional database with inclusion of all patients and frequent provider feedback with comparisons to peer norms. For instance, on August 5, 2008, the National Quality Forum endorsed participation in a systematic national database for general thoracic surgery [32]. However, any such pay-for-performance intervention would have to provide the resources to fund database efforts or other processes of care found to be beneficial to lung cancer patients, rather then simply instituting an unfunded mandate. In other settings, quality improvement initiatives have been speculated to be linked to cost-savings, and thus the expense of such interventions would be expected to be largely or completely offset a bonus beyond the intended benefit of improved long-term outcomes for lung cancer patients [33, 34]. Finally, while quality improvement policies are debated, general and cardiothoracic residency programs might wish to increase exposure to the general thoracic surgical service with an eye toward teaching important processes of care and clinical situations that require referral to a dedicated thoracic surgical oncologist. In conclusion, patients under the care of GTS had higher survival rates than those under the care of GS. General thoracic surgeons performed more extensive preoperative

9 Ann Thorac Surg FARJAH ET AL 2009;87: SURGEON SPECIALTY AND LUNG CANCER OUTCOMES and intraoperative cancer staging compared with CTS and GS. A better understanding of the reasons underlying these differences will likely help develop interventions aimed at mitigating unnecessary variation in care and outcomes. Patients, surgeons, physicians, and other stakeholders should recognize the potential risks and benefits associated with regionalization of care, pay-for-performance, and other policy-level interventions aimed at improving the quality of thoracic surgical care. This study used the linked SEER-Medicare database. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, Centers for Medicare & Medicaid Services; Information Management Services, Inc; and the SEER Program tumor registries in the creation of the SEER-Medicare database. Farhood Farjah was supported by a Cancer Epidemiology and Biostatistics Training Grant (T32 CA ) and National Research Service Award (F32 CA ) from the National Cancer Institute. The interpretation and reporting of these data are the sole responsibility of the authors. The views expressed in this article do not necessarily represent the official views of the National Cancer Institute, National Institutes of Health, Centers for Medicare & Medicaid Services, American Board of Thoracic Surgery, or the University of Washington. References 1. Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer statistics, CA Cancer J Clin 2007;57: Goodney PP, Lucas FL, Stukel TA, Birkmeyer JD. Surgeon specialty and operative mortality with lung resection. Ann Surg 2005;241: Silvestri GA, Handy J, Lackland D, Corley E, Reed CE. Specialists achieve better outcomes than generalists for lung cancer surgery. Chest 1998;114: Kohman LJ. What constitutes success in cancer surgery? Measuring the value of specialist care. Chest 1998;114: Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care 2002;40(8 Suppl):IV Farjah F, Wood DE, Varghese TK Jr, Symons RG, Flum DR. Trends in the operative management and outcomes of T4 lung cancer. Ann Thorac Surg 2008;86: Bach PB, Guadagnoli E, Schrag D, Schussler N, Warren JL. Patient demographic and socioeconomic characteristics in the SEER-Medicare database applications and limitations. Med Care 2002;40(8 Suppl):IV Klabunde CN, Potosky AL, Legler JM, Warren JL. Development of a comorbidity index using physician claims data. J Clin Epidemiol 2000;53: Ayanian JZ, Weissman JS. Teaching hospitals and quality of care: a review of the literature. Milbank Q 2002;80:569 93, v. 10. Hollenbeck BK, Hong J, Zaojun Y, Birkmeyer JD. Misclassification of hospital volume with surveillance, epidemiology, and end results Medicare data. Surg Innov 2007;14: Marubini E, Valsecchi MG. Analyzing survival data from clinical trials and observational studies. Chichester, UK: John Wiley & Sons, Ambler G, Omar RZ, Royston P. A comparison of imputation techniques for handling missing predictor values in a risk model with a binary outcome. Stat Methods Med Res 2007; 16: Lardinois D, Weder W, Hany TF, et al. Staging of non-smallcell lung cancer with integrated positron-emission tomography and computed tomography. N Engl J Med 2003;348: Pieterman RM, van Putten JW, Meuzelaar JJ, et al. Preoperative staging of non-small-cell lung cancer with positronemission tomography. N Engl J Med 2000;343: Allen MS, Darling GE, Pechet TT, et al. Morbidity and mortality of major pulmonary resections in patients with early-stage lung cancer: initial results of the randomized, prospective ACOSOG Z0030 trial. Ann Thorac Surg 2006;81: Birim O, Kappetein AP, van Klaveren RJ, Bogers AJ. Prognostic factors in non-small cell lung cancer surgery. Eur J Surg Oncol 2006;32: Jazieh AR, Kyasa MJ, Sethuraman G, Howington J. Disparities in surgical resection of early-stage non-small cell lung cancer. J Thorac Cardiovasc Surg 2002;123: Lilenbaum RC, Cashy J, Hensing TA, Young S, Cella D. Prevalence of poor performance status in lung cancer patients: implications for research. J Thorac Oncol 2008;3: Zeiher BG, Gross TJ, Kern JA, Lanza LA, Peterson MW. Predicting postoperative pulmonary function in patients undergoing lung resection. Chest 1995;108: Pettiford BL, Schuchert MJ, Santos R, Landreneau RJ. Role of sublobar resection (segmentectomy and wedge resection) in the surgical management of non-small cell lung cancer. Thorac Surg Clin 2007;17: Feinstein AR, Sosin DM, Wells CK. The Will Rogers phenomenon. Stage migration and new diagnostic techniques as a source of misleading statistics for survival in cancer. N Engl J Med 1985;312: Quarterman RL, McMillan A, Ratcliffe MB, Block MI. Effect of preoperative delay on prognosis for patients with early stage non-small cell lung cancer. J Thorac Cardiovasc Surg 2003;125: Deslauriers J, Ginsberg RJ, Piantadosi S, Fournier B. Prospective assessment of 30-day operative morbidity for surgical resections in lung cancer. Chest 1994;106(6 Suppl):329 30S. 24. Rostad H, Strand TE, Naalsund A, Talleraas O, Norstein J. Lung cancer surgery: the first 60 days. A population-based study. Eur J Cardiothorac Surg 2006;29: Wright CD, Gaissert HA, Grab JD, O Brien SM, Peterson ED, Allen MS. Predictors of prolonged length of stay after lobectomy for lung cancer: a Society of Thoracic Surgeons General Thoracic Surgery Database risk-adjustment model. Ann Thorac Surg 2008;85: Ferguson MK, Little L, Rizzo L, et al. Diffusing capacity predicts morbidity and mortality after pulmonary resection. J Thorac Cardiovasc Surg 1988;96: Cattaneo SM, Park BJ, Wilton AS, et al. Use of video-assisted thoracic surgery for lobectomy in the elderly results in fewer complications. Ann Thorac Surg 2008;85: Licker MJ, Widikker I, Robert J, et al. Operative mortality and respiratory complications after lung resection for cancer: impact of chronic obstructive pulmonary disease and time trends. Ann Thorac Surg 2006;81: Jones LW, Peddle CJ, Eves ND, et al. Effects of presurgical exercise training on cardiorespiratory fitness among patients undergoing thoracic surgery for malignant lung lesions. Cancer 2007;110: General Thoracic Surgery Club. Available at Accessed August 16, Little AG, Rusch VW, Bonner JA, et al. Patterns of surgical care of lung cancer patients. Ann Thorac Surg 2005;80: National quality forum endorses national consensus standards promoting accountability and public reporting. Available at Accessed October 16, Englesbe MJ, Dimick JB, Sonnenday CJ, Share DA, Campbell DA, Jr.The Michigan surgical quality collaborative: will a statewide quality improvement initiative pay for itself? Ann Surg 2007;246: Flum DR, Fisher N, Thompson J, Marcus-Smith M, Florence M, Pellegrini CA. Washington State s approach to variability in surgical processes/outcomes: Surgical Clinical Outcomes Assessment Program (SCOAP). Surgery 2005;138: GENERAL THORACIC

10 GENERAL THORACIC 1004 FARJAH ET AL Ann Thorac Surg SURGEON SPECIALTY AND LUNG CANCER OUTCOMES 2009;87: Appendix Billing Codes Resection HCPCS Chemotherapy HCPCS C9017 J0182 J8510 J8530 J8560 J8610 J899 J9000 J9001 J9010 J9045 J9060 J9062 J9070 J9080 J9090 J9091 J9092 J9093 J9094 J9095 J9096 J9097 J9170 J9180 J9181 J9182 J9190 J9201 J9206 J9208 J9230 J9250 J9260 J9265 J9280 J9290 J9291 J9350 J9360 J9370 J9375 J9380 J9390 J9999 Q0083 Q0084 Q0085 Q0125 Q0127 Q0128 Q0129 S0178 S0182 S9329 S9330 S9331 ICD-9 V58.1 V66.2 V RCC Radiation therapy HCPCS C1716 C1717 C1718 C1719 C1720 C1790 C1791 C1792 C1793 C1794 C1795 C1796 C1797 C1798 C1799 C1800 C1801 C1802 C1803 C1804 C1805 C1806 C2616 G0126 G0173 ICD-9 V58.0 V66.1 V PET HCPCS G0125 G0126 G0210 G0211 G212 G Mediastinoscopy/-otomy HCPCS Mediastinal lymphadenectomy HCPCS Reoperation HCPCS Tracheostomy HCPCS Cardiac procedures HCPCS HCPCS Healthcare Common Procedural Coding System; ICD-9 International Classification of Diseases, 9th ed; PET positron emission tomography; RCC Revenue Center Code.

11 Ann Thorac Surg FARJAH ET AL 2009;87: SURGEON SPECIALTY AND LUNG CANCER OUTCOMES DISCUSSION DR CAROLYN E. REED (Charleston, SC): Dr Chitwood, Dr Wood, members and guests. This well-presented paper adds to the growing literature confirming that pulmonary resection for lung cancer performed by surgeons adequately trained and practicing general thoracic surgery will have better outcomes. The findings are very similar to those presented by Dr Paul Schipper at the Southern Thoracic Surgical Association in November. The authors have appropriately adjusted for a number of variables and addressed a number of potential limitations. I rise not to dispute the findings, which I firmly believe. I rise to again point out some astounding deficits in our own thoracic practices. In 2005 at the STS meeting in Tampa, Dr Alex Little presented Patterns of Surgical Care of Lung Cancer Patients. In 2001, only 27% of patients underwent mediastinoscopy, with about half not submitting any lymph node tissue to pathology, PET scans were obtained in 26%, and 42% of patients had no mediastinal staging at the time of operation. In 2007, at the AATS annual meeting, Dr Daniel Boffa reported on the STS general thoracic database spanning from 1999 to Thankfully, the rate of mediastinal lymph node evaluation was 65%. This study reported that lymphadenectomy was only used in 33% of patients by general thoracic surgeons. Now, granted, that is better than 11% by general surgeons, but, folks, we have a problem. This disturbing figure was not compensated by use of PET, 36%, or mediastinoscopy, 18%. Two conclusions: The word needs to get out to the public that lung cancer should be treated by proper surgeons; second, we need to continue to do better staging in our patients, both clinically, to offer them the correct treatment, and pathologically to ascertain need for adjuvant therapy and to counsel patients about their prognosis. I have a couple of questions. Did the rate of lymphadenectomy recorded include node sampling? Please tell us that more than one-third of the patients had some attempt at identifying pathologic N2 disease? Second, the ACGME residency case data reports that the national average of lung cancer resections done by general surgical residents during five years of training is six. Do you feel that we have now accumulated enough evidence to publicly advocate that all lung cancer patients should see a board-certified CT surgeon or equivalent? Thank you. I very much enjoyed your paper. DR FARJAH: Thank you, Dr Reed, for reviewing our work and your comments. Your first question was about whether our rate of lymphadenectomy included lymph node sampling, and the answer is that it might not have. Lymphadenectomy rates were based on a CPT code for lymphadenectomy. There is no CPT code for lymph node sampling. So I would agree with you that intraoperative N2 staging is important and that these low rates are concerning, especially among specialists. Until it can be verified that rates of intraoperative N2 staging are higher, I think that your concern is very appropriate and I absolutely agree with it. Your second question was about regionalization essentially. I think from the perspective of thoracic surgeons, the most likely policy conclusion that one would reach from this data is regionalization, and that very well may be the answer. Viewed from the perspective of a policy maker, however, there is always the concern about unintended consequences of short-sighted policies. Though I think it is true of any policy, for the purposes of discussion, we can use regionalization and our data as an 1005 example. If there is a strict policy of mandatory regionalization that does not address transportation and housing of patients and families, then what is going to happen to those patients? These are patients that may have received suboptimal care in the past but may subsequently not be able to receive any care, and again, we are talking about lung cancer, a uniformly fatal disease when untreated. Let us say the policy addresses that issue but patients refuse to travel. How will that policy be helpful if a small fraction of patients are no longer getting any therapy for lung cancer? And let us say that patients are willing to travel and that the infrastructure issues are addressed, are there enough general thoracic surgeons to accommodate the increase in volume? If not, will there be delays in care, and if there are delays in care, will that adversely affect outcomes? I do not know the answers to these questions and one could play out these adverse scenarios with any policy. So I think that our study answers a question that was posed a decade ago, and in the process of doing so it also raises many other questions. I think the things that can be done right now include, one, participation in either a national or a regional quality improvement initiative. The STS general thoracic database is one example of such an initiative, and NSQIP is another. The second thing that can be done is this information can be disseminated among surgical colleagues, nonsurgical colleagues, the public, your patients. There is no reason why this information should not be disseminated and there are no reasons why there cannot be passionate debates about the right answers. DR RICHARD J. SHEMIN (Los Angeles, CA): I rise not to question the added quality that a general thoracic surgeon brings to the care of patients with lung cancer. In the year 2000, the STS/AATS Taskforce on Practice and Access published a study reporting that about 70% of adult cardiac surgeons perform general thoracic surgery in their practice. So one question is if Board certified Thoracic Surgeons who primarily identify themselves as adult cardiac surgeons provide the same quality as our general thoracic colleagues? My second point is that the Taskforce worked with David Wenberg at Dartmouth. We studied regional variations in pulmonary resections and what type of surgeon performed the pulmonary resection. We found a significant misdistribution of general thoracic surgeons around the country. They were mostly located in large metropolitan areas. When we looked at general surgeons performing general thoracic procedures, they were in areas where there was poor access to a board-certified thoracic surgical specialist. The general surgeons, in general, performed less pulmonary resections than general thoracic surgeons. Thus, in part we are dealing with an access issue. You have partially addressed this issue with your comments on regionalization. However, the fact is unless there is an improved distribution of thoracic surgeons, non-thoracic surgeons will continue to be called upon to perform these procedures. Patients will not travel to regional centers. I am interested in your comments. DR FARJAH: Thank you for the questions, Dr Shemin. We found no evidence of a difference between cardiothoracic surgeons and general thoracic surgeons. So there is no evidence to suggest that cardiothoracic surgeons should not do these operations. GENERAL THORACIC